Tobacco Documents Online

, $ OSCAR AUEA$A-?-'H-T-RICHAIM J-STNG-T HI;NR4 W. BLACKBURN + FRED G. BOCK + JOSEF BROZEK + ANTONIO CIOCCO + J. H. COMROE, JR. + PHILIP COOPER + THOMAS ~ R. DAWBER + HAROLD F. DORN + C. M. FLETCHER + .S LAWRENCE GARFINKEL + WILLIAM HAENSZEL + E. i CUYLER HAMMOND + D. H O F F M A N N + DANIEL HORN + HAROLD A. KAHN + WILLIAM B. KANNEL + CHARLES J. KENSLER + ANCEL KEYS + PAUL KOTIN + LEIV KREYBERG + MORTON L. LEVIN + CECILIE L E U C H T E N B E R G E R+ RUDOLF LEUCHTENBERGER + ARTHUR JAMES LINDSEY + CHARLES C. McARTHUR + GEORGE E. MOORE + JAY NADEL RICHARD DOUGLAS PASSEY + WALTER REDISCH + STANLEY P. REIMANN + GRACE M. ROTH + ROBERT R. SECR'EST + RICHARD M. SHICK + ARTHUR PURDY STOUT + HENRY L. TAYLOR + HAROLD K. TSUJI + BENJAMIN L. 4

TOBACCO AND HEALTH It is generally agreed that there is amarkedassocia- tion between lung cancer and cigarette smoking. More recently significant evidence appeara to link c i g a r e t t e smoking with deaths attributed to coro- nary heart disease, peptic ulcer, bronchitis, and pulmonary fibrosis. In search for proof or lack of -proof of this causal re- lationship, a penetrating, scientific r e v i e w of the entire question of tobacco and health was cponsored jointly by the New York Academy of Medicine and the New York State Acad- emy of Preventive Medi- cine. Discussants included a significant n u m b e r of research investigators who have won recognition for evidence they have coin- piled on one or the other side of the controversy. NO SCIENTIFIC POSITION ON THE QUESTION WAS OMITTED. The discussions recorded here cover ... • Chemistrv and Pharma- colo of Tobacco Smoke • Experimental Pathology of Tobacco Smoke • T h e Interpretation of Statistical and Epidemio- logical Data Relating Smok- ing to Disease • Smoking and Lung Can- cer T200931 • Other Diseases Relating ~ t., c..,..~;...,

TOBACCO AND HEALTH T200932

~ ~ ~ TOBACCO" ~ 0 AND HEALTH ~ Edited by GEORGE JAMES, M.D. Deputy Coritrnissioner, Department of Health Adjunct Associate Professor, Public Health Practice Columbia.University School of Public Health and Administrative Medicine New York City and THEODORE ROSENTHAL, M.D. Assistant Commissioner, Department of Health Clinical Professor of Preventive Medicine New York University College of Medicine New York City CHARLES C THOMAS • PUBLISHER Springfield • lllinois • U.S.A. T200933

i CHARLES C THOIMIAS , PUBLISHER BArtrrMSi'orrE Hovsn 301-327 East Lawrence Avenue, Springfueld, Illinois, U.S.A. This book is protected by copyright. No part of it may be reproduced in any manner with- out written permission from the publisher. © 1.962, by CHARLES C TI-tO1viAS • PUBLISHER Library of Congress Catalog Card Number: 61-15382 With THOMAS BOOKS careful attention is given to all details of manufacturing and design. It is the Publisher's desire to present books that are satisfactory as to their physical qualities and artistic possibilities and appropriate for their particular use. THOMAS BOOKS will be true to those laws of quality that assure a good name and good will. G13, S Printrd in the United States of America CONTRIBUTORS OscAm AvmACx, M.D., Senior Medical Investigator, Veterans Admin- istration Hospital, East Orange, New Jersey; Associate Pro f essor of Pathology, New York Medical College, New York, New York RieIIAnD J. BINC, M.D., Professor and Chairman, Department of Medi- cine, Wayne State University College o f Medicine, Detroit, Michigan HENRY W. BLAcICSVaN, M.D., Assistant Pro f essor, Laboratory o f Ph ysiologieal Hygiene, University o f Minnesota, Minneapolis, Minnesota FIt>rD G. Bocs, Ph.D., Associate Cancer Research Scientist, Biological Station, Boswell Park Memorial Institute, Springville, New York JosEF BaozFnc, Ph.D., Chairman, Department of Psychology, Lehigh University, Bethlehem, Pennsylvania ANTONIO CIOCCO, Sc.D., Head, Department o f Biostatistics, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania J. H. Commor, JR., M.D., Director, Cardiovascular Research Institute, Universit y o f Cali f ornia Medical Center, San Francisco, Cali f ornia PHILIP COOPEn, M.D., Clinical Pro f essor o f Surgery, Albert Einstein College of Medicine, New York, New York; Chief of the Surgical Services, Veterans Administration Hospital, Bronx, New York TnomAs R. DAwwIt, M.D., Member, National Heart Institute, Public Health Service, Department of Health, Education and Welfare, Bethesda, Maryland HAROLD F. DOAN, Ph.D., Chief, Biometrics Research Branch, National Heart Institute, National Institutes of Health, Bethesda, Maryland C. M. FLt+;rcln;R, M.D., F.R.C.P., Senior Lecturer in Medicine, Post- graduate Medical School o f London, Physician to Hammersmith Hospital, London, England LAw$F,NC>; GARFINICEL, M.A., Chie f o f Fieid and Special Projects, Statistical Research Section, American Cancer Society, New York, New York v T2.00934

vi Tobacco and Health WILLIAM HAENSZEL, M.A., Associate Chief, Biometry Branch, National Cancer Institute, Bethesda, Maryland E. CuYLER HAMMOND, Sc.D., Director o f Statistical Research Section, American Cancer Society, New York, New York D. HOFFMANN, Ph.D., Section of Epidemiology, Division of Preventive Medicine, Sloan-Kettering Institute, New York, New York DANIEL HoRN, Ph.D., Director of Program Evaluation, American Cancer Society, New York, New York HAROLD A. KAHN, M.A., Member, National Heart Institute, National Institutes of Health, Public Health Service, Department of Health, Education and Welfare, Bethesda, Maryland WILLIAM B. KANNEL, M.D., Member, National Heart Institute, National Institutes of Health, Public Health Service, Department of Health, Education and Welfare, Bethesda, Maryland CHARLES J. Kr•.NSLER, Ph.D., Professor and Chairman, Department of Pharmacology and Experimental Therapeutics, Boston University School o f Medicine, Boston, Massachusetts ANCeI. KEYS, Ph.D., Laboratory of Physiological Hygiene, University of Minnesota, Minneapolis, Minnesota PAUL KOTIN, M.D.. Department of Pathology, School of Medicine, University of Southern California; Los Angeles County General Hospital, Los Angeles, Cali f ornia LEIv KREYI3ERC, i<1.D., Professor, Universitetes, Institutt for Generell Og, Eksperimentell Patologi, Rikshospilalet, Oslo, Norway MoRTON L. LEVJN, M.D., Professor, Department of Epidemiology, Roswell Park Memorial Institute, Buffalo, New York CECILIF LFUCHTF.NISI•',RCER, Ph.D., Senior Biologist and Cytochemist, Children's Cancer Research Foundation, Boston, Massachusetts RUDOLF LuUCHTrNnERCER, M.D., Senior Experimental Pathologist, Children's Cancer Research Foundation, Boston, Massachusetts ARTHUR JAMES LINDSI:Y, D.Sc., Ph.D., Head of Department of Chem- istry, Sir John Cass College, London, England CIIARLES C. MCAITrIIUR, Ph.D., Psychologist to the University Health Services, Ilarvnrd University, Cambridge, Massachusetts GEoRCr•. E. MooRI;, Department of Surgery, Roswell Park Memorial Institute, Buffalo, New York JAY NADEL, M.D., Associate Staff Member of the Cardiovascular Re- search Institute, University of California Medical Center, San Francisco, Caiiforn.ia Tobacco and Health vii RIeHARD DoucLAs PASSEY, M.C., M.D., D.P.H., Emeritus Pro f essor o f Experimental Pathology, Leeds University; Research Fellow, Chester Beatty Research Institute, London, England WALTER REVISCII, M.D., Associate Pro f essor o f Medicine, New York University School o f Medicine; Attending Physician, The Third Medical Division, Bellevue and University Hospitals; Research Associate, New York University Research Service, Goldwater Memorial Hospital, New York, New York STANLEY P. REIIvIANN, M.D., Director Emeritus, The Institute f or Cancer Research, Philadelphia, Pennsylania GRAcE M. RoTH, Ph.D., Department of Medicine, Lovelace Clinic, Albuquerque, New Mexico; Section o f Medicine, Ma yo Clinic, Rochester, Minnesota ROBERT R. SECREST, M.D., Department of Medicine, Lovelace Clinic, Albuquerque, New Mexico; Section of Medicine, Mayo Clinic, Rochester, Minnesota RIcHARD M. SHIcx, M.D., Department o f Medicine, Lovelace Clinie, ' Albuquerque, New Mexico; Section o f Medicine, Mayo Clinic, Rochester, Minnesota ARTHUR PURDY STOUT, M.D., Pro f essor Emeritus o f Surger y, College o f Physicians and Surgeons, Columbia University HENRY L. TAYLOR, Ph.D., Laboratory of Physiological Hygiene, Uni- versity of Minnesota, Minneapolis, Minnesota HAROLD K. Tsv1I, M.D., Chief Resident, Thoracic Surgery, Bronx Veterans Administration, Bronx, New York BENJAMIN L. VANDwRErI, Sc.D., Assistant Pro f essor, Institute o f Industrial Medicine, New York University Medical Center, New York, New York J. EDwIN WOOD, M.D., Associate Pro f essor o f Medicine, Medical Col- lege of Georgia; Director, Georgia Heart Association Laboratory f or Cardiovascular Research, Augusta, Georgia ERNESr L. WYNDER, M.D., Head, Section of Epidemiology, Sloan- Kettering Institute; Associate Pro fessor o f Preventive Medicine, Sloan-Kettering Division, Cornell Medical College, New York, New York J. YERUSHALMY, Ph.D., Professor of Biostatistics, University of Cali- fornia, Berkeley, California WILLIAM J. ZUKEL, M.D., Member, National Heart Institute, National Institutes of Health, Public Health Service, Department of Health, Education and Welfare, Bethesda, M^rifla.nd T200935

FOREWORD SHORTLY after the New York State Academy of ~ Preventive Medicine was founded, there was much discussion about its particular mission for the advancement of public health science and practice. The Academy established a number of committees to explore some of the crucial issues in these fields. One of these committees was asked to investigate the problem of tobacco and health. Although impressed with the wealth of evidence which links cigarette smoking etiologically with lung cancer and other diseases, the Committee observed the reservations with which several respected members of the scientific community viewed this evidence. It also recognized the paucity of data avail- able on the epidemiology of the smoking habit itself and on the measures which might be taken to curtail it. In order to establish a firm baseline beneath any further studies of the smok- ing problem and any recommendations for the control of smok- ing, the Committee urged the New York State Academy of Preventive Medicine to sponsor a two-day symposium which might provide the forum for a brief but penetrating and sober scientific review of the entire question of tobacco and health. The New York State Academy of Preventive Medicine dis- covered that the New York Academy of Medicine had an equal interest in sponsoring a scientific review of this subject and a joint Planning Committee was established. Financial support was received from the Rockefeller Foundation, the New York City Cancer Committee, and the New York State Cancer Society, the latter two being affiliates of the American Cancer Society. The session was held at the New York Academy of Medicine, - September 26 and 27, 1960, with attendance limited to the panel- ists and the members and guests of the two sponsoring organiza- tions. ix T200936

x Tobacco and Health It is generally agreed that there is a marked association be- tween lung cancer and cigarette smoking. Most epidemiologists who have studied the problem believe that the fourteen retro- spective studies, the three large prospective studies, the evidence from pathological, chemical and experimental studies, establish the causall relationship of cigarette smoking to lung cancer be- yond a reasonable doubt. Others believe so strongly that addi- tionall avenues of research must be explored that they see no reason to justify a decrease in the cigarette smoking habit. More recently significant evidence appears to link cigarette smoking with deaths attributed to coronary heart disease, peptic ulcer, bronchitis and pulmonary fibrosis. The public health problem was put succinctly by Dr. Hollis S. Ingraham, conference chairman, who pointed out in his opening remarks: "Should this excess mortality due to cigarette smoking prove to have a causal relationship, it would indicate that cigarettes cause more deaths than any other single recognized lethall agent, including alll the known bacteria, known viruses, bullets, wild animals, chemical poisons or even the American automobile." In its search for the proof or lack of proof of this causal relationship, the Committee selected as discussants a significant number of research investigators who have won recognition for the evidence they have compiled on one or the other side of the controversy. The Planning Committee did its best to insure that each major significant portion of evidence be presented and that no scienti ftc position on the question be omitted. Each partici- pant's address was followed by a lively discussion period, which regretfully cannot be reproduced as part of this volume. The editors wish to express their great appreciation to each of the speakers for his willingness to prepare a careful presentation based primarily upon his own research, and to subject his evi- dence and his views to that persistent probing and ruthless criti- cism which charsacterize the search for scientific truth among men with imbom,ded curiosity and hope for still more ultimate conquests against the ravages of disease. The editors wish to acknowledge their debt to all of their colleagues of both Academies who served on the Planning Com- mittee and most especially to Dr. Harry Kruse, Executive Secre- Tobacco and Health xi tary, Committee on Public Health, New York Academy of Medicine, Drs. Henry Aranow and McKeen Cattell of the New York Academy of Medicine, and to Dr. Morton L. Levin under whose Presidency of the New York State Academy of Preventive Medicine the Committee on Tobacco and Health was born. GEORGE JAMES, M.D. President, New York State Academy of Preventive Medicine, 1960-1961 Chairman, Committee on Tobacco and Health, 1958-1960 TxEonoRE RosENTHAL, M.D. Chairman, Committee on Tobacco and Health of the New York State Academy of Preventive Medicine, 1960-1961 T200937

CONTENTS Foreword ' ix ................................•-..._....-•----..........._.........-•..-..............._....... . ~g SESSION I CHEMISTRY AND PHARMACOLOGY OF TOBACCO SMOKE Chapter 1. The Pharmacology of Tobacco Smoke Effects of Chronic Ex- posure by CxAnm J. KiNSLm ................................................... 5 2. Some Observations Upon the Chemistry of Tobacco Smoke by A. J. LiNnsEx .......................................................................... 21 3. Some Aspects of the Chemistry of Tobacco Smoke by BEN- JAMIN L. VAN DWIIEN ................................................................ 33 4. The Effect of Cigarette Smoking on Coronary Blood Flow and Cardiac Work in Normal Subjects and Patients with Coronary Disease by RicHAnn J. &NC .................................... 48 5. Tobacco in Health and Disease Effect on Peripheral Circula- tion by J. Enwmt WOOD ............................................................ 51 SESSION 11 EXPERIMENTAL PATHOLOGY OF TOBACCO SMOKE 6. The Role of Skin Neoplasia in Tobacco Carcinogenesis b y E. L. Wnanm and D. HorrmAN .................................................... 61 7. The Significance of Mouse Skin Tests of Cigarette Smoke Condensate by FxEV G. Bocx and GEOxcE E. Moom : ............ 72 ~ 8. A Correlated Histological, Cytological and Cytochemical Study of the Major Bronchi From MYce Exposed to Cig- arette Smoke by CEcmE LEVcnroNSERcEa and RvnoLF LEVCa3TENSERcEa ----...--••-• ............................................................. 87 xui T200938

xiv Tobacco and Health Chapter Page 9. Neoplasia: Lungs by RIcFrAxn DovcLAs PASSEY ---------------------•-_ 105 1-~'10. Bronchial Pathology in Man by STANLEY P. REINIANN .............. 123 11. Changes in the Tracheobronchial Tree in Connection with Cigarette Smoking by OscAlr AuzRBAcH, AsTFIUIi PvanY STOVT, E. CInYLER HAMMOND and LAwRENCE GAHr•IrIIEL .... 136 12. Histological Types in Relation to Etiological Factors and Trends in Lung Cancer by LEiv KxEYBEnc .............................. 144 SESSION III THE INTERPRETATION OF STATISTICAL AND EPIDEMIOLOGICAL DATA RELATING SMOKING TO DISEASE 13. Smoking and Cancer: Retrospective Studies and Epidemio- logical Evaluation by MORTON L. LEViN .................................. 163 14. Death Rates and Causes of Death of Smokers and Nonsmokers by HAROLD F. Donrr ..........................................................••-----..• 172 15. Statistical Considerations and Evaluations of Epidemiologic Evidence by ANTONIO CIOCCO .................................................... 191 ~" 16. Statistical Considerations and Evaluation of Epidemiological Evidence b yJ. Yr•.BVSrIALNIY ...................................................... 208 SESSION IV SMOKING AND LUNG CANCER 17. The Effect of Smoking and Nicotine on Respiration by J. H. CoArnor, Jn. :,nd JAY NADEL . ........................................................ 233 18. Environmental Factors and Lung Cancer by PAVI. KoTIN ........ 244 19. Quantitative Evaluation of the Etiologic Factors in Lung Cancer bi/ WILLIAM HAENSZEL ........................................... _.-- 264 20. The Personal and Social Psychology of Smoking by CxAIU,Es C. McAnzTlml ............................................................•----......-....--.. 291 21. Smoking Habits of High School Students by DANIEZ. HORN •... 300 I! Tobacco and Health xv SESSION V OTHER DISEASES RELATING TO SMOKING Chapter Page 22. The Effect of Smoking on the Peripheral Circulation of Man: Studies on the Site of Action of bTicotine by GsACE M. ROTFI, Ph.D., RICxABn M. Sxlcx and RoBEBT R. SECREsT .... 311 23. Cardiovascular and Related Characteristics in Habitual Smok- ers and Non-Smokers b y HENRY W. BLACmmN, Josr.r' BaozEx, HENRY L. TAYLOR and ANCEL Ki.ys .......................... 323 24. Tobacco Allergy and Vascular Responses by WALTn Rmiscx.. 352 25. Coronary Heart Disease Incidence and Mortality Among Smokers and Non-smokers by Wu.i.IAm J. Zvm, TxomAs R. DAWBER, WILLIAM B. KANNEL, HAROLD A. KAxN and HAROLD F. DORN ............................•--•---.............--•--..................... 360 26. Chronic Bronchitis, Smoking and Air Pollution by C. M. FLETCHER 380 27. Interrelationship of Smoking and Peptic Ulcer by PxII,Ir Coorm and HAROLD K. Tsull .................................................... 402 T200939

TOBACCO AND HEALTH T200940

Session I CHEMISTRY AND PHARMACOLOGY OF TOBACCO SMOKE T200941

_ F--i ~ THE PHARMACOLOGY OF TOBACCO SMOKE EFFECTS OF CHRONIC EXPOSURE CHAar..ES J. KExsLEA, Ph.D. DESPJTE the lack of definitive information, it ap- pears reasonable that smokers choose to smoke tobacco products because of the pleasant aroma and sensations evoked, as well as for psychologic and emotional factors which vary in importance among smokers. The effects of smoking in the pharmacologic sense are due to the presence of a variety of chemical com- ponents in the smoke, which produce local effects at the point of contact and systemic effects when absorbed. The components of pharmacologic interest and studies on their activity have been reviewed by Haag and Larson (1944); Kensler (1954 ); and Comroe (1960) and Kensler (1960) in the New York Academy of Sciences Symposium on the effects of tobacco smoke in the cardiovascular system, held earlier this year. The chemistry of tobacco smoke has rec,ently been reviewed by Bentley and Berry (1959) and Johnstone and Plim3ner (1959) and will be covered by Drs. Lindsay and Van Duuren in this Symposium. As has been pointed out previously, pharmacologic effects are dependent on the dose, the dose schedule and period of exposure, and the degree of susceptibility or responsiveness of the indi- viduals and species studied. Pharmacologically active compounds also have thresholds or levels of doses and exposures below which no detectable response is elicited. This general principle has been challenged in the area of chemical carcinogenesis, but the problem, as discussed in National Research Council Report No. 5 T200942

6 Tobacco and Health 749, 1959, in this area is really the difficulty of determining threshold levels, and not the absence of such levels. Tobacco smoke is a complex mixture of chemicals which are present either in the gas phase or in the liquid aerosol or particu- late phase. Studies by Larson (1960) indicate that the amount (based on measurement of nicotine ) of particulate phase materials retained by smokers varies from 10 percent when it is puffed without inhalation to over 90 percent with inhalation. Thus, those who inhale the smoke from four cigarettes a day would get approximately the same dose of particulate components as those who merely puff two packages of cigarettes a day. Data of this type are not available on gas phase components, but the specialized nature of the lung makes it probable that the ratio would be even more in favor of retention by inhalers. Among the components of the gas phase of tobacco smoke, which, because of their established toxicity at definite exposure levels and their systemic action, are of interest to our discussion, are carbon monoxide and hydrogen cyanide. Although carbon monoxide accounts for approximately 4 per- cent of the gas phase of tobacco smoke (Philippe and Hobbs, 1956), measurement or carboxyhemoglobin levels in smokers ( IIanson and 1-Iastings, 1933) indicates that smoking raises the blood levels from the 1 to 2 percent in the non-smoker to 3 to 4 percent in the smoker. In normal individuals, carbon monoxide blood saturation up to 30 percent are usually without effect, whereas levels of 70 percent of saturation and higher are fatal. The small increase in carboxyhemoglobin levels in smokers would thus appear to represent no hazard to normal individuals, although McFarland has found some reduction in light sensi- tivity of the eye at 5 percent carboxyhemoglobin levels in non-smokers when they vigorously inhale cigarette smoke. The observation that only small increases in carboxyhemoglobin are found in smokers, despite the significant level of CO in tobacco smoke, is presamably due to the reduction in the concentration of CO when tihe puff (35 nil.) is diluted in the lungs and to the intermittent nature of the exposure. In the case of hydrogen cyanide, the amount contained in a 35 ml. puff of tobacco smoke is about 50 to 75 micrograms. At j Effects of Chronic Exposure 7 these levels of exposure, which amount to approximately one microgram per kilogram per 10 to 15 minutes for a chain smoker, the activity of detoxifying systems._ (largely conversion to thio- cyanate ) and excretion prevent the accumulation of significant concentrations of cyanide. The oxides of nitrogen, NO and NOz, have been found in cigarette, cigar and pipe smoke. Continuous exposure of man to NO2 has been reported to cause irritation and cough at 100 ppm. and severe irritation of the nose and larynx at 200 ppm. (Gray, 1959). The importance of the presence of NOZ in tobacco smoke, where exposure is of short duration, is not known, but as the American Standards Association has adopted 5 ppm. as the maximal acceptable concentration for continuous exposure, further study of this problem is needed. f At the site of contact in the trachea and lungs, cigarette smoke . has been reported to produce an inhibition of ciliary motility in in vitro experiments (Hilding, 1956; Falk, Tremer and Kotin, 1959) and to increase pulmonary airway resistance (Nadel, Tierney and Comroe, 1959). Ciliary motility has been shown to be inhibited by a great variety of materials in vitro, including ozonized gasoline fumes and components therein (Falk, Tremer, and Kotin, 1959) and many drugs affecting autonomic nervous system activity, including nicotine. In vitro dose response studies indicate that these agents at a low concentration actually stimulate ciliary motility and only at higher doses inhibit. Effects of the type that I am discussing are illustrated in the next two figures. In the first, it can be seen that low concentrations of ,epinephrine increase motility in vitro, and higher doses depress activity. When the higher doses are studied, even these result in an initial stimulation. As in the in vitro systems, normal actively metabolizing and excreting mechanisms may not be operating, extrapolation of in vitro data to man would require dose response studies to determine if at levels of use tobacco smoke produces a net increase or decrease in ciliary motility in the in vivo situation. This has not as yet been done. The effect of increased airway resistance (50%) has been observed in man, but is small in comparison to the increases (400-500%) required to produce dyspnea. It has been suggested T200943

$ Tobacco and Health EpinePhrine - C'iliary IUlovemenf a X-CONTROL 9- EPINEPHRINE oL_ EPIMEPNRINE(MomcaMC) VO-' !0"6 10"6 to' 60' Fig. 1. The effect of epinephrine on ciliary motility of rabbit tracheal preparations in airo. Each point represents the equilibrium motility after the administration of the drug in the concentration indicated. C'i I OrLI Iti/l ovemn t Fig. 2. The effect of epinephrine 10-' molar on ciliary motility as a function of time. Effects o f Chronic Exposure 9 by Comroe (1959) that this effect of tobacco smoke may be due to the physical properties of the aerosol, rather than to its pharmacologically active components. _ The particulate phase, which is impacted in the respiratory tract and then either absorbed or mechanically removed, is a complex chemical mixture. It has been shown to contain many aliphatic and aromatic hydrocarbons, alcohols and esters, sterols, aldehydes and ketones, organic acids, phenols and polyphenols, alkaloids (chiefly nicotine ) and other bases, amino acids and vitamins, inorganic elements and free radicals. Many of these compounds have low intrinsic pharmacologic activity and most are probably present in subthreshold doses, although definite information on these matters is lacking. Nico- tine, the major alkaloid in tobacco smoke, is retained by inhalers in amounts approximating 2 mg. per cigarette (Haag and Larson, 1944). If absorbed promptly, this would provide a 70 kilogram man with a dose of nicotine of the order of three micrograms per kilogram per minute for 10 minutes. The well-established toxic effects of nicotine, i.e., convulsions and neuromuscular blockade, are achieved in most species with doses of the order of several milligrams per kilogram. Lower doses, 10 to 30 micrograms per kilogram, however, produce effects on the cardiovascular and respiratory systems presumably by stimulating various chemo- receptors and ganglion cells and initiating various reflex responses (Comroe, 1960). Effects on circulation will be discussed by Drs. Bing and Wood later this morning and on respiration by Dr. Nadel tomorrow. From the studies of Burn (1960), it is apparent that many of the effects of nicotine on the circulatory system are mediated through the release of sympathometic amines as pre- treatment with reserpine diminishes or abolishes the response. The action of nicotine on the central nervous system is com- plex, and varies with the dose administered and also with experi- mental conditions. There is little doubt that nicotine is a potent central nervous system stimulant (Longo et al., 1954), although slight depressant activity has also been reported (Salama and Wright, 1952). Knapp and Domino (19,60) have recently studied the effects of 10 to 40 micrograms/kilogram in rabbits, cats, dogs, and monkeys. All of these an~-^^1- -rQ -hi-tPd to T200944

10 Tobacco and Health general surgery under diethyl ether or local procaine anesthesia and mid-pontine pre-trigeminal brain transections were per- formed. Such preparations show either high voltage delta waves or more characteristically spindle bursts. Following the intra- venous administration of nicotine, such a "sleep-like" EEG pattern converts quickly to the low voltage fast wave pattern of an awake animal. A minimal effective dose is in the order of 10 ,ug/kg. Generally, 20 to 40 pug/kg of nicotine are extremely effec- tive amounts. These arousal effects occur in animals that are insensitive to epinephrine arousal (such as the dog and the monkey ) as well as catechol depleted animals due to chronic reserpine administration. Preliminary evidence obtained on these animals with the first three cranial nerves transected in addition suggests that this is a direct effect of nicotine on the brain itself. The action of nicotine on the emetic center (Borison & Wang, 1953) appear to be direct central effects, although the respiratory center and posterior pituitary effects would appear to be due to peripheral chemoreceptor stimulation. Chronic EfFecf s Numerous investigators have studied the effects of chronic exposure of animals to tobacco smoke and nicotine, particularly with respect to effects on the cardiovascular system and the production of hyperplastic or neoplastic responses. These experi- ments have generally been done using doses or exposures which are far in excess of the most extreme case of human use of these materials. This is the usual approach to a study of the toxicology of a compound or material in the laboratory, where the investi- gators include doses which permit survival just long enough to allow tissue damage to become apparent by morphologic alter- ation. Needless to say, effects have been obtained with large doses which are not obtainable with small ones. The chronic administration of excessive (i.e., convulsant or just sllhconvlllsant) doses of nicotine have, for example, been reported to impair growth, decrease fertility, and occasionally produce vascular lesions. Ilueper (1943), for example, has re- ported that in rats, convulsant doses of nicotine produced degenerative changes in arteries and arterioles, which were Effects of Chronic Exposure 11 consistent with an anoxic type response. Thienes (1960), how- ever, working with just subconvulsant doses, failed to detect any vascular lesions, but did note ~ a small effect on weight changes which were associated with a decrease in body fat. At the present time, it is not established that the chronic administration of nicotine has produced any lesions which are characteristic or specific for nicotine. A number of experiments have been conducted in which mice or rats have been exposed to high concentrations of cigarette smoke at regular intervals over prolonged periods of their life span. One of the best of these studies is that by the Leuchten- bergers (1960 ), who exposed CF, hybrid mice to a lethal (some of the mice died during exposure ) concentration of cigarette smoke. The survivors of the early exposures were continued in the experiment, and it was found that morphologic changes had been produced in the lungs of these animals. The changes ob- served included atypical hyperplasia of the epithelial cells and the appearance of cell groups or clusters, described as "carcinoma in situ." These lesions appeared early, in the first few months, and despite lifetime or nearly lifetime exposure, did not progress to cancer. No one has produced cancer in experimental animals by exposure to tobacco smoke as such. Another type of experiment has been the application of tobacco smoke condensates in concentrated solution (usually 50% in acetone ) to the clipped skin of mice and rabbits. The most extensive series of experiments of this type are those published by Wynder and his associates (1953) in which it has been reported that the application of approximately 40 mg. of con- densate three times a week to the skin of mice has resulted in a modest incidence of papillomas and carcinomas at the site of application in lifetime experiments. Mortality and weight changes were also recorded, but systematic pathologic examina- tion of other tissues and organs has not been included in Wyn- der's or other publications of this type. Consequently, it is perhaps of interest to present some of our experiments of this type, which were done in the laboratories of Arthur D. Little, Inc., Cambridge, Massachusetts. In brief, cigarettes of several brands and cigars have been smoked in T200945

f 12 Tobacco and Health CR-F, MICE SURVIVAL OF EXPERIMENTAL ANIMALS Effects of Chronic Exposure 13 CA-F1 MICE PAPILLOMA INCIDENCE PRODUCED BY VARIOUS CIOARETTE CONDEN'SATES WNTHS OF PAINTING Fig. 3. The survival of animals painted with acetone or 50 per cent solutions of various cigarette smoke condensates in acetone. Over 100 animals were used per group. W. G. 1 and 2 and the Acetone W. G. are data previously published by Wynder, Graham and Croninger. mechanical smoking machines. The smoke condensate was col- lected at low temperatures, dissolved in acetone and applied to the backs of mice. Fifty percent solutions (by weight) in acetone were applied to the backs of CAF, hybrid mice or Swiss mice three times a week throughout their lives. The area covered was approximately eight sq. cm. as opposed to the 50 sq. meters available in a human respiratory tract. The amount applied was sufficiently high, in these as in Wynder's experiments, to kill a few of the mice with a single application. W'bt ~_ A ~- C --- ---~, t C 0 , i i , , , , , , ~ . ~ I . Fig. 4. Figure 4 illustrates the papilloma incidence produced by various cigarette smoke condensates on the skin of mice painted three times a week with approximately 100 milligrams of 50 per cent solutions of various smoke condensates in acetone. No papillomas were found in the acetone control group, although a low incidence of papillomas has been observed in similar controls of other experiments run in this laboratory. As in Figure 3, W. G. 1 and 2 are data previously published by Wynder, Graham and Croninger. T200946

14 Tobacco and Health CFa-F, MICE CARCINOMA INCIDENCE PRODUCED BY VARIOU'S CIGPRE,TTE CONDEN5ATES Effects o f Chronic Exposure P(-1PILLOMIa INCIDENCE SWISS MICE 30 s a YO 12 14 N6 VD 20 22 24 MONTHS OF PAINTING Fig. 5. The incidence of carcinomas on the skin of mice. These are the same animals included in Figure 3 and 4 The results of the lifetime experiment will be summarized in the next figures. In Figure 3, the effect of the application of the condensates or acetone alone on survival are shown. The appli- cation of tho large amounts of, condensate, as would be expected, decreased the survival time. The incidence of skin papillomas are shown in Figure 4. The incidence of skin carcinoma (con- firmed histo:ogically) is indicated in Figure 5. In view of atten- tion focusedd by the epidemiological studies on cigarette smoke condensate as opposed to cigar smoke condensate, cigar smoke condensate has also been examined. As is shown in Figure 6, the Fig. 6. Two cigarette and one cigar smoke condensate are compared in this figure in terms of their ability to produce papillomas on the skin of Swiss mice when painted three times a week in a 50 per cent. acetone solution. incidence of papillomas produced by cigar smoke condensate is no different from that of the cigarette smoke condensate. Simi- larly, there is no difference between cigar and cigarette smoke condensates when carcinoma incidences are compared, as in Figure 7. Dose-response studies carried out in our laboratories on tobacco condensates have yielded a significant incidence of skin cancer only at the highest dose level. From this and other data, it is clear that cigarette and cigar smoke condensates, when applied in large amounts, can produce papillomas and carcinomas on the skin of mice, but not as many or as rapidly as they would be produced by similar applications of as little as 5 micrograms of 3, 4-benzpyrene. This compound in much higher amounts has not been found to produce cancer in monkeys, and its activity for man is unknown. T200947 15 ~ cNI"mECQMD[NS1TE E ~ fNM'ETTE CpiQKNS1YTE 0 J'l G --.V_ U4NR CONDENSWTE J 4 4 E . I 4

16 Tobacco and Health CARCINOMA INCIDENCE SWISS MICE 50 Ck/PWETTE CONS[NJ1iTE E ~_ C46/WtC7iE COMDENSN7E 0 C/OJ99Q CON'DEN9ATE J O\ Effects o f Chronic Exposure 17 ~ TABLE I DI8'TRIBUTION OF OUTOROWTHS BY .SLX OF ANIMAL ~ AllOutgrowths Carcinomas O -i F Experiment Male Female Male Female u ~ Condensate A 10 35 2 12 Condensate B 17 29 8 5 ~ Condensate C 23 32 6 17 Condensate D 15 27 1 6 65 123 17 40 TABLE II LUNG TUMOR INCIDENCE Total Incidence ^ 6 8 10 12 16, 16 113 MONTHS OF PAINTING 22 Fig. 7. Two cigarette and one cigar smoke condensate are compared in this figure in terms of their ability to produce carcinomas on the skin of Swiss mice. These are the same animals as shown in Figure 6. Both Wynder's and our experiments have indicated a sex difference in mouse skin response to cigarette smoke condensates. As is shown in Table 1, twice as many female mice developed papillomas and carcinomas as did an equal number of males. In man, the incidence of bronchogenic carcinoma occurs much more frequently in males. Careful gross and histological study of the lungs of the CAF, mice used in the cigarette smoke condensate experiment, pre- sented in Table II, indicates that the spontaneous incidence of lung adenomas was not significantly different among the con- densate-trcated and control groups. Similarly, careful study of other organs and tissues, such as brain, heart, spleen, liver, kid- neys, adrenals, stomach, urinary bladder, testes or ovaries, mammary glands, large and small intestines, showed no lesions Effective Tolal No. Studied Lung Tumora Test Material (Lung) Histologically No. % None 25 25 10 40 Acetone 27 25 14 56 52 50 24 48 Condensate A 89 87 36 41 Condensate B 95 92 51 55 Condensate C 100 94 46 49 Condensate D 93 88 34 39 377 361 167 46 significantly associated with smoke condensate application, other than the skin effects reported above. A few tumors, such as sarcomas, lymphomas, leukemia, adenocarcinomas, occurred randomly in all groups. The combined incidence of these tumors in the tobacco condensate-treated group was about half that of the controls: 20 percent incidence in the controls and 6 to 11 percent in the condensate-treated groups. Another area of interest in the chronic administration of pharmacologically active materials are the associated problems of tolerance, habituation and addiction. The use of tobacco has generally been classified as a habit by those who deal with the

18 Tobacco and Health problems of addiction in the medical or pharmacologic sense (Seevers, 1958), in that it appears to involve primarily psycho- logic or emotional dependence, rather than physical dependence. Addiction testing experiments, such as are routinely done in monkeys with narcotics, have not been carried out, to my knowl- edge, with cigarette smoke or nicotine. Some animal experiments indicate that tolerance to small doses develops. Thienes (1960) most recently has reported that in very young rats, there was a marked development of tolerance to small doses, causing hind- leg weakness, but little or no tolerance to convulsive or fatal doses. When nicotine administration was begun after adult size had been attained, only slight tolerance developed. Finnegan and associates (1945) have attempted to clarify the relationship of nicotine to the tobacco habit by substituting low nicotine (1/10) tobacco for that usually smoked. Twenty-four habitual smokers (all inhalers ) were selected, each of whom felt he could not easily forego the habit. Six of the 24 subjects experienced no change in the pleasure they derived from smok- ing; 6 experienced an initial lack of the satisfaction they cus- tomarily desired from smoking; three definitely missed the nicotine, but became adapted to the change in one to two weeks; nine definitely missed the nicotine and continued to do so throughout the trial period of one month. The results of this small-scale experiment may or may not be representative of response of smokers in general, but it suggests that perhaps more than half mi ;ht be satisfied with low nicotine cigarettes. In conclusion to this brief survey of the pharmacology of tobacco smoke, I would like to emphasize that insofar as the available data go, the responses of animals to tobacco smoke or its components vary as a function of the dose, that the mere identification of the presence of a compound in this complex mixture does not mean that it is present in sufficient quantities to prodnce an effect, or for that matter, that the production of an effect in one species will be qualitatively or quantitatively dupli- cated in othvr species. The diversity of pharmacologic activity of tobacco smoke, however, makes it appear likely that excessive use of tobacco, like excessive use of alcohol or overeating, will carry with it some risk to susceptible or diseased individuals. E f f ects o f Chronic Exposure 19 REFERENCES Bentley, H. R, and I:. G. N. Berry (1959): The constituents of tobacco smoke. To- bacco Manufacturers' Standing Committee. London, England. Borison, H. L. and S. C. Wang (1953): PhysioloB7r .and pharmacology of vomiting. Pharmacol. Rev., 5:193. Burn, J. H. (1960): Action of nicotine on the peripheral circulation. Ann. New York Acad. Sci., 90, Art. 1:81. Comroe, J. (1960): Pharmacological actions of nicotine. Ann. New York Acad. Sci., 90, Art. 1:48. Falk, H. L., H. M. Trerner and P. Kotin (1959): Effect of cigarette smoke and its constituents on ciliated mucus-secreting epithelium. J. Nat. Cancer Inst., 23:999. Finnegan, J. K., P. S. Larson and H. B. Haag (1945): The role of nicotine in the cigarette habit. Science, 102:94. Gray, E. L. (1959): Oxides oE nitrogen: their occurrence, toxicity and hazard. A.M.A. Arch. Ind. Health., 19:479. Haag, H. B. and P. S. Larson (1942): Urinary excretion of nicotine by smokers. Federation Proc., 1:610. Haag, H. B. and P. S. Larson (1944): Recent advances in the pharmacology of nicotine. Virginia M. Monthly, 71:235. Hanson, H. B. and A. B. Hastings (1933): The effect of smoking on the carbon monoxide content of blood. J. A. M. A., 100:1481. Hilding, A. C. (1956): On cigarette smoking, bronchial carcinoma and ciliary action. New England J. lbi., 254:1155. Hueper, W. C. (1943): Chronic nicotine poisoning in rats and dogs. Arch. Path., 35:846. Johnstone, R. A. W. and J. R. Plimmer (1959): The chemical constituents of to- bacco and tobacco smoke. Chem. Revs., 59:885. Kensler, C. J. (1955): Pharmacology. In Wynder, E. L, The Biologic Effects of Tobacco, Little, Brown and Company, Boston, Ch. 2. Kensler, C. J. (1960): Components of pharmacologic interest in tobacco smoke. Ann. New York Acad. Sci., 90, Art. 1:43. Knapp, D. and E. F. Domino (1960): Personal communication. Larson, P. S. (1960): Absorption of nicotine under various conditions of tobacco use. Ann. New York Acad. Sci., 90, Art. 1:31. Leuchtenberger, R., C. Leuchtenberger, W. Zebrun and P. Shaffer (1960): A corre- lated histological, cytological, and cytochemical study of the tracheobronchial tree and lungs of mice exposed to cigarette smoke. Cancer, 13:956. Longo, V. G., G. P. Von Berger and D. Bovet (1954): The action of nicotine and of the "ganglioplexiques centraux" on the electrical activity of the brain. J. Pharmacol. Exper. Therap., 111:349. ' McFarland, R. A. (1952): Anoxia: Its effect on the physiology and biochemistry of the brain and on behavior. In: The Biology of Mental Health and Disease, Paul B. Hoeber Inc., New York, Ch. 22. T200949

20 Tobacco and Health Nadel, J. A., D. F. Tierney and J. H. Comroe (1959) : Pulmonary response to aerosols. Presented to the Third Air Pollution Medical Research Conference. Los Angeles, Calif. National Academy of Sciences-National Research Council (1959): Problems in the evaluation of carcinogenic hazard from use of food additives. Publication 749. Philippe, R. and M. Hobbs (1956): Some components of gas phase of tobacco smoke. Anal. Chem., 28:2002. Salama, S. and S. Wright (1952): Influence of various drugs on the action of curare on the central nervous system of the cat. Brit. J. Pharmacol., 7:14. Seevers, M. (1958): Drug Addictions. In Drill, V. A., Pharmacology in Medicine, McGraw- Hill, New York, Ch. 19. Thienes, C. H. (1960): Chronic nicotine poisoning. Ann. New York Acad. Sci., 90, Art. 1:239. Wynder, E. L., E. A. Graham and A. B. Croninger (1953): Experimental production of carcinoma with cigarette tar. Cancer Research 13:855. 2 SOME OBSERVATIONS UPON THE CHEMISTRY OF TOBACCO SMOKE A. J. LmsEY, D.Sc., Ph.D. VERY many analytical investigations have been carried out on tobacco smoke and about 270 organic compounds and 15 elements as inorganic compounds have been reported present in it.l," These include 16 carcinogens,* 14 alkaloidst and a number of other compounds with known biological activity. Only a fraction of the total constituents have been the subject of quantitative work, and because of the variable composition and intractable nature of the smoke it is often found difficult to repeat both qualitative and quantitative findings recorded in the litera- ture. A great deal more analytical work of both kinds is needed to attain more precise knowledge of the composition of tobacco smoke. Ideally quantitative analysis of all constituents should * Carcinogens in Tobacco Smoke Arsenious oxide, 1:2-benzanthracene, 3-4-benzfluoranthene, 10:11-benz8uoranthene, 11:12-benzfluoranthene, 1:12-benzperylene, 1:2-benzpyrene, 3:4-benzpyrene, Chrysene, 6:7-cyclopenteno-l:2-benzanthracene, 1:2-5:6-benzanthracene, 3:4-8:9-dibenzpyrene, 3:4-9:10-dibenzpyrene, 3-methyl-pyrene (or another alkyl pyrene), 2-naphthol. 21 t Alkaloids in Tobacco Smoke Anabasine, Anatabine, Cotinine, 2:5-dipyridyl, Metanicatine, Myosmine, n-Methylnicotinamide, Nicotine, Nicotyrine, Nornicotine, Poikiline, 3-Pyridyl-methql-ketone, 3-Pyridyl-ethyl-ketone, 3-Pyridyl-n-propyl-ketone. T200950

22 Tobacco and Health account for 100 per cent of the smoke, but in practice the present state of knowledge accounts for far less. There are a number of difficulties in attaining this ideal and of these the variable nature of the original material, tobacco, and the many variables of the smoking process are the most obvious. However, by using a single batch of original material conditioned to be as homogeneous as possible and carefully planned experi- mental procedures it is possible to obtain consistent results in a series of investigations. It is much more difficult to obtain quanti- tative agreement between different laboratories. Further progress in the chemistry of tobacco smoke must proceed on quantitative developments and it is proposed to outline some important quantitative problems in this paper and to indicate how these may be solved and also their possible significance in the tobacco- health problem. Subsidiary matters raised by this problem could also be of value in directing chemical research towards an early under- standing of the biological effects of tobacco smoke; the more important of these matters are raised in the following paragraphs. Many known constituents of tobacco smoke have not been tested for physiological action. This should be done on the purest material obtainable preferably synthetic and so not isolated from tobacco. This would avoid erroneous conclusions due to unsuspected trace components in the material tested. The results of such tests could well lead to important knowledge in the tobacco-health problem and also guide chemical researches in the most appropriate direction. Very little attention has been paid to the composition and physiological action of the gaseous/vapour phase of smoke. This also should be investigated chemically and biologically, and is especially important now because of the recent custom of testing biologically only the less volatile part of the disperse phase; a direct consequence of determining the amount of condensate on a dry basis. A study of side-stream smoke is very important. It has been neglected on the quite unfounded notion that since it is not deliberately drawn into the mouth during the smoking process, it can have no effect on the smoker. Side-stream smoke has not I Observations upon Chemistry of Tobacco Smoke 23 been seriously investigated chemically since Wenusch` and Pyriki.s The suggestion is now made that since it is continuously- breathed in fair concentration by the heavy smoker it may have a serious health effect, especially as it can represent the major proportion of the products of the smoking process. A full investi- gation of side-stream smoke is therefore proposed as a worth- while study. GENERAL REVIEW OF THE QUANTITATIWE RELATIONSHIPS IN TOBACCO SMOKE STUDIES Depending upon the smoking technique the ratio of main- stream smoke to side-stream smoke may vary. Generally for cigarettes with a process approximating to normal human smok- ing about 4 per cent of the weight of the tobacco consumed ap- pears in the main-stream smoke as condensible material and rather less appears in the side-stream smoke in similar form. These represent partial combustion products, produced by dis- tillation, thermal decomposition and rearrangement of pyrolytic products. The main changes occurring during smoking are more complete combustion producing carbon dioxide, carbon monoxide and large amounts o£water from the various tobacco constituents. The composition of typical flue cured cigarette tobacco is given in Table 1. The very large number of compounds reported in tobacco smoke must therefore originate from the changes undergone by these few compounds in the circumstances of smouldering com- bustion and, bearing in mind the large variation in the molecular weight of the smoke constituents, it follows that thermal decom- position is responsible for a number of labile fragments, possibly radicals, which later combine to form these varied products. The total smoke contains permanent gases, readily volatile compounds and less volatile compounds. And it is towards these latter compounds that greatest attention has been given because among them are to be found a number of physiologically active substances. The major gaseous constituents, are listed in Table 2. In order to give easily reproducible results for comparisons of amounts of condensate produced experimentally, the dry weight basis was introduced. This involves heating the condensate to T200951

u TABLE I COMPOoiTION OF FLUE-CURED VIRGIN7A CIGARETTE TORACCO Name of Substanc'e Quantity (%) Name of Substance Quantity (%) (continued) Ash (see (ii) below) - 13.2 Nitrogen compounda - Cellulose - -- - - 9.0 (see (i) below) - - 11.1 Citric acid - - - 0.6 Oxalic acid - - - 1.0 Fructose - - - - 7.8 Pectins - - - - 10.7 Glucose - - - - 11.0 Polyphenols - - - 2.0 Hemicelluloses -- - - 1.6 Resins (total) - - - 9.8 Lignin - -- - - 3.5 Starch - - - - 4.0 Malic acid - - - 10.1 Sucrose - - - - 4.2 Total - - - 99.6 i. Composition of total nitrogen compounds Amino nitrogen - Ammonia and amidcs Nicotine - - - 0.2 0.7 1.7 0.5 (continned) Nornicotine Protein - Unidentified 0.3 7.1 0.6 Nitrate - -- Total Oxides and carbonates of metals: Aluminum Arsenic* I3arium* Boron* Calcium Chromium* Cobalt* Copper* Iron 1,ead* Lithium* Magnesirm Manganese* Mercurey* Other componeuts: Chloride Phosphatc ii. Composition of ash * Present as traces Molybdenum* Nickel* Potassium Itubidium* Selenium* Silver* Sodium Strontium* Thallium* Tin* Titaninm* Vanadium* Zinc* Silicates Siilphates Observations upon Chemistry of Tobacco Smoke 25 constant weight at 100°, sometimes after solvent extraction and much of the whole smoke is lost in the attempt to give a constant result for competitive argument between manufacturers who wish to produce what they call a "low tar" content. Another effect is that trace materials are often irretrievably lost in the less tractable dry tar if this is used later for their determination. There is, of course, very sound reason to quote condensate on a dry basis because the main constituent is water and this is very variable in amount. If a dry weight basis is chosen for recording condensate amounts accurately it is best to weigh total con- densate after electrostatic precipitation and then to determine the water in it by the Karl-Fischer technique. OUTLINE OF SMOKING PROCESS FOR CIGARETTES The smoking process has been outlined in a number of publi- cations.e" The generally accepted process, although feasible, is in part, conjecture.® It is important to reconsider it in the light of experiments on side-stream smoke. During intervals between puffs the temperature gradient be- tween the hot end and the tobacco rod is very steep; according to Harlow® about 6000 in 4 mm. It might, therefore, be assumed that distillation of various volatile constituents including water would occur here and that the products might well condense a little further back in the cooler zones. In the regions a little nearer the glowing end, thermal decomposition, especially of the less volatile constituents, occurs and some of the products of this process pass into the side-stream smoke while other portions, it might be assumed, could also pass into the cooler zones of the to- bacco and there condense. The presence of ammonia and other volatile bases in the side-stream smoke is said to be due to the de- composition of amides, alkaloids and proteins and gives the well known alkaline reaction to side-stream smoke from a bright tobacco, which normally has an acidic main-stream smoke. Dur- ing suction the rapid combustion gives rise to a stream of hot gases and steam, and rapid steam distillation, thermal distillation and decomposition occur. The products condense into droplets which are partially deposited in the tobacco rod, and the "smoke" which passes through is an aerosol consisting of droplets a few T200952

26 Tobacco and Health TABLE 2 MA70R GAREOUS CONSTITUENTS OF MAIN-STREAM SMOKE Gas Per Cent by Yotume Nitrogen 75 Oxygen 13 Carbon dioxide 8 Carbon monoxide 2.5 Observations upon Chemistry of Tobacco Smoke 27 microns in diameter. During sucessive quiescent periods it is pos- sible that some of the deposited smoke is thermally decomposed and that this and other volatile products may distil further into the tobacco rod. Certainly redistillation occurs during the suction periods and material condensed in later stages of the smoking process is richer in heavier molecular weight materials because of a certain Amount of fractionation.® The condensible matter in main-stream smoke measured for various stub lengths was mea- sured previously® and Figure 1 contains some results from this work. Both gravimetric and spectrometric methods were em- ployed to measure the total smoke condensate for each stub length. In the smoking sequence there is thus no doubt that there is a steep rise in condensible material with decreasing residual stub length. It seemed desirable to attempt to determine the contribution made to re-volatilisation by each of the very distinct processes in the smoking cycle, that is by suction and by smouldering combustion. 98.5 MG so 40 ® 30 20 ® rv Q Io Fig. J. Fig. 2 T200953 I

28 Tobacco and Health A STUDY OF DISTILLATION PROCESSES DURING PRODUCTION OF SIDE-STREAM SMOKE Cigarettes equilibrated to uniform humidity were weighed individually in stoppered weighing bottles. They were then lighted by drawing a single puff and then supported horizontally. Quiescent combustion was allowed to proceed to marked points on the paper. The cigarette was then stubbed out into the weigh- ing bottle, which was also used to collect all the ash. The bottle was stopped up, cooled and weighed. The loss in weight repre- sents the total loss through side-stream smoke production. No adjustment was made for changes in weight due to the difference in chemical form of the ash constituents before and after the smoking, but t his represents only a small difference. Figure 2 shows the losses in weight plotted against stub length. It will be noted that thee relationship is linear. If noteworthy amounts of material had distilled from tobacco adjacent to the hot end into cooler zones it would be expected that the losses experienced for the shortest stubs would have been greater than for corresponding lengths earlier in the process. It is, therefore, inferred that very little distillation into the tobacco rod occurs during continuous production of -;ide-stream smoke by smouldering combustion. It was also considered useful to find the effect of quiescent combustion oi7 smoke condensate already present in cigarettes and, to accomplish this, the smoke from a conditioned cigarette was drawn through another for the usual smoking sequence. The second cigarette, containing deposited smoke, was then submitted to smouldering conibustion as described above. The results were almost the same as those of Figure 2. The very considerable movement of material by vaporization and condensation during the normal smoking process is thus due solely to the deposition during suction of ma.in-stream smoke droplets on the tightly packed tobacco in regions behind the hot end. RAPID METHODS OF DETERMINATION OF SMOKE CONDENSATE IN MAIN-STREAM SMOKE The very great interest shown in recent years in the amounts of smoke condensate in main-stream smoke from cigarettes of varying type, and the need for determining the efficiency of fil- ters has made a rapid determination of smoke condensate impor- Obseriaations upon Chemistry of Tobacco Smoke 29 tant. In Figure 1 some of the ordinates were obtained by spectro- photometric measurement of the optical density of a diluted solution of the smoke in acetone. Two methods were used; either the determination of the optical density at fixed frequency or the measurement of the area under the absorption curve over the entire visible range of frequencies, and these methods gave iden- tical results. Their validity is dependent upon the absorption of the smoke solution being proportional to the mass of material present and so in turn upon the constancy of the composition of the material so far as light absorbing properties are concerned. The validity has now been checked by determining the amounts gravimetrically using the Cambridge Filter, obtainable from Phipps and Bird, Inc., 303, S. Sixth Street, Richmond, Virginia, U.S.A., and also by electrostatic precipitation and subsequent weighing. The latter experiments were made independently in . another laboratory and are indicated° in Figure 1. It is impor- tant to standardize the method gravimetrically and to make the spectrophotometric measurements on fresh materials; smoke condensates darken rapidly with age. A POSSIBLE MECHANISM FOR THE FORMATION OF THE POLYCYCLIC AROMATIC HYDROCARBONS This class of compound excites special interest partly because it contains a number of powerful carcinogens and has been proved to be the most active fraction of tobacco smoke in bio- logical painting experiments'° and also because it is universally associated with partial combustion. Not only have the polycyclic aromatic hydrocarbons been found in tobacco smoke, but in pyrolysis products from all the major constituents of vegetable matter" and in a number of other pyrolysis products as well as in partial combustion materials such as soots and tars. At one time hopes were held that specific precursors of this group of com- pounds could be removed from tobacco and thus prevent or greatly reduce their amounts in the smoke.'o.'2,18 By experimenting with flames of simple fuels some indication of the mode of formation of polycyclic hydrocarbons has been obtained: ' Thus it has been shown that in diffusion flames of pure hydrocarbon fuels of low molecular weights, partial combus- T200954

r 30 Tobacco and Health tion products, which must be assumed to be intermediate stages in the full combustion of the fuel, can be isolated by inserting cooled probes into the flames. The carbonaceous deposits on the probes contain a large number of polycyclic aromatic hydrocar- bons in addition to carbon. Methane, ethane, propane, n-butane, ethylene, acetylene, benzene, toluene, o-xylene, p-xylene, ethyl, benzene, p-cymene, cumene and styrene were all treated in this way and it was hoped that by studying the yields of the poly- cyclic aromatic hydrocarbons from various fuels it might be pos- sible to formulate a likely mechanism for their formation. The most important result is that of all these fuels, methane, the simplest, gives the highest yield of polycyclic aromatic hydrocarbons; 5.6 per cent by weight of the dried carbonaceous deposit. Spectroscopic study of the flames shows the presence of CH radicals and it seems reasonable that these radicals, which are the smallest possible, unite to produce the considerable array of polycyclic aromatic hydrocarbons. Many of the simple fuels used are known constituents of tobacco smoke'K and could thus he stages in the formation of this interesting class of compound. There is, of course, no reason why larger radicals produced from the more complex fuels should not also he respon sible for the pyrosynthesis, e.g. rra"Hj_~J c THE AMOUNTS OF BENZPYRENE IN MAIN-STREAM CIGARETTE SMOKE AT DIFFERENT STUB LENGTHS Cigarettes were smoked to two different stub lengths (1.5 and 3.5 cm. ) and the whole smoke and all the stubs were analyzed for 3:4-benzpyrene.$ The results are shown in Figure 3, where, against the appropriate stub-length histograms are shown with total benzpyrene from 100 cigarettes and the amount in the stubs ( black ). From the same number of cigarettes four times as much 3:4-benzpyrene was found in the whole smoke when short stubs (1.5 cm. ) were lef t as when long stubs (3.5 cm. ) were Observations upon Chemistry of Tobacco Smoke 31 discarded this is in harmony with statistical studies of lung cancer mortality and stub length in smoking populations.'° These results are not necessarily incompatible with others recently reported and based upon the benzpyrene proportion per gram of dry con- densate in early and late smoking experiments. 2.5 2.0 1.5 1.0 0.5 CM 1 2 3 STUB LENGTH Fig. 3 4 5 6 SUMMARY A large number of compounds are known to be present in tobacco smoke but only a fraction of these have been quantita- tively determined. Only a fraction too have been tested for their biological action. It is suggested that the vapour phases of both main-stream and side-stream smoke are worthy of further chemi- cal study and that the condensible part the side-stream smoke is T200955

32 Tobacco and Health also worth fuher examination. Recent studies have shown that distillation processes occurring during the smoking cycle are confined to the suction stages, that spectrophotometry can give reliable determinations of total condensate in main-stream smoke and that polycyclic aromatic compounds could originate from very small radicals such as CH, produced at the relatively high temperature of the combustion process. REFERENCES 1. Johnstone, R. A. W. and Plimmer, J. R.: Chemical Reviews, 1959, 59, 885. 2. Bentley, H. R. and Berry, E. G. N.: Tobacco Manufacturers Standing Committee Research Paper, No. 3, London, 1959, also 1st Supplement, 1960. 3. Kosak, A. I.: Experientia, 1954, 10, 69. 4. Wenusch, A.: Der Tabakrauch, Arthur Geist Verlag, Bremen, 1939. 5. Pyriki, C.: Z. Lebensm. Unters., 1940, 80, 42, 1948, 88, 254. 6. Wenusch, A.: Op. cit., p. 18. 7. Smyth, C. N.: Brit. med. J., 1959, 1, 506. 8. Lindsey, A. J.: Brit. J. Cancer, 1959, 13, 195. 9. Harlow, E. S.: Science, 1956, 123, 226. 10. Wynder, E. L.: Brit. med. J., 19'59,1, 317. 11. Gilbert, J. A. S. and Lindsey, A. J.: Brit. J. Cancer, 1957, 11, 398. 12. Wynder, E. L.: Brit. mcd. J., 1957, 1, 1. 13. Wynder, E. L. and Wright, G.: Cancer, 1951, 10, 255. 14. Lindsey, A. J.: Combustion and Flame, 1960, 4,261. 15. Hobbs, M. E. et al.: Anal. Chem.. 1956, 28, 211 and 1956, 28, 2002. 16. Hammond, E. C.: Brit. med. J., 1958, 2, 649. I t- 00 ~ ~ ~ ~ O ~ ~ 3 E~ SOME ASPECTS OF THE CHEMISTRY OF TOBACCO SMOKE BENJANmv L. VAN DuvRm, ScaD. Tzm chemistry of tobacco smoke has been studied extensively and several comprehensive reviews on this subject have appeared.', 14 Instead of reviewing all that work here I will attempt to present to you some of the recent studies carried out in our laboratories and, particularly, to discuss sonte potentially important new avenues of chemical research in this field. The work in our laboratory has been concerned with those compounds which are either known or suspected carcinogenic or cocarcinogenic agents. The polynuclear aromatic hydrocar- bons, a notorious group of carcinogens, have received a con- siderable amount of attention in various laboratories and I will summarize only very briefly our own findings in this atea. Roffo,Q` as early as 1939 described the identification of benzo ( a) pyrene from the pyrolysis of tobacco. Subsequent work by Lindsey and coworkerse,®,",'° confirmed and extended Roffo's work to the identification of a series of aromatic hydrocarbons in the smoke of cigarettes, cigars and pipe tobacco. Some dispute existed over the identification of these compounds by ultraviolet spectra alone. We undertook an examination of the column chromatographe'd neutral fraction and more recently the non-volatile basic fraction of cigarette tar by paper chromatography.2o, zs. Ps.'o By this pro- cedure the aromatic compounds could be ' purified to such an extent that only minimal amounts of impurities remained. Fur- thermore, the RF values of the individual compounds served as 33 T200956

34 Tobacco and Health x Some Aspects o f Chemistry o f Tobacco Smoke 35 Fig. 2 Fig. 1 an additional criterion for identification purposes. The separa- tions obtained by paper chromatography is exemplified in Figure 1 which shows pyrene, fluoranthene and a column chromato- graphed neutral cigarette tar fraction. For the large scale separa- tion of these compounds the materials were applied on the paper as bands to give the banded separations shown in Figure 2. The eluted spots or bands were subjected to examinations by ultra- violet spectroscopy as shown in Figure 3. In Figures 4 and 5 the fluorescence excitation and emission spectra of the same two materials are shown. Certainly, this establishes perfect agreement and together with the Rr values constitute four different criteria for identification of these materials. These methods have been extended by other workers for the examination of the aromatic compounds in city air, automobile exhaust products, etc. By the use of these analytical methods we were able to estab- lish the presence of 23 tetracyclic or higher polynuclear aromatic compounds, shown in Table 1. We made no attempt to isolate the smaller aromatic hydrocarbons since these have in most cases been shown to have no carcinogenic activity. In this table are included 19 hydrocarbons, one oxygen heterocyclic, 1,8,9- perinaphthoxanthene, and three nitrogen heterocyclics, 7H- dibenzo ( c,g )carbazole, dibenz( a,j ) acridine and dibenz-( a,h )- T200957

36 Tobacco and Health I Some Aspects of'Chemfstry of Tobacco Smoke 37 r F- U) z w 0 J Q f> F-a 0 240 I I I I I I l I: UNKNOMVN; 2: FLUORANTHENE, SOLVENT CYCLOHEXANE. 260 280 300 320 340 360 380 WAVELENGTH, mju ULTRAVIOLET ABSORPTION SPECTRA: Fig. 3 0 250 300 350 400 WAVELENGTH, mp FLUORESCENCE EXCITATION SPECTRA,ANA- LYSIS AT 465 mp; SENSITIVITY 0.10; i: UNKNOWN; 2: FLUORANTHENE, I mq,/LITER, SOLVENT CYCLO- HEXANE. Fig. 4 acridine. The latter two compounds were isolated from the basic fraction of cigarette tar.90 Of these 23 compounds 10 have shown carcinogenic activity in laboratory animals. These compounds, and the concentrations in which they are present are shown in Table 2. In addition to these ten compounds several others have been reported by other workers.'°, °Z Since there are inevitable losses in the purification of these compounds, the concentrations listed in the table are to be taken as minimal values only and their true concentrations may be three or four times that given in the Table. The formation of the aromatic hydrocarbons in the burning cigarette has been discussed by various workers and some possi- ble mechanisms of their synthesis was studied extensively by T200958

TABLE 1 HIGHER POLYNUCLEAR AROMATIC C014IPOUNDB IN CIOARETfE SMOICE CONDENSATE -y/100 cigaref.lev 7/100 cigarettes Benz(m, n, o)Fluoranthene 0.10 Pyrene 5.0 Benzo(j)Fiuoranthene .. 4-Methylpyrene 5.0 Benzo(k)Fl.uorantuene .. Fluoranthene 1.0 1, 2-Benzfluorene . . Methylfluoranthene Dibenz(a, h)Anthracene 0.05 Alkylfluoranthene Benzo(c)Phenanthrene .. Chrysene 0.06 Dibenzo(a,i)Pyrene 0.002 Alkylchrysene 0.04 Dibenz(a,j)Acrid:ne 0.27 Benzo(a)Pyrene 0.50 7H-Dibenzo(c, g)Carbazole 0.07 Methylbenzo(a)Pyrene 0.01 Dibenz(a, h)Acridine 0.01 Benzo(e)Pyrene 0.30 Perinaphthoxanthene .. Perylene Benzo(g, h, i)Peryleno 400 450 500 WAVELENGTH, Mp 550 FLUORESCENCE EMISSION SPECTRA, EXCITATION AT 364 m}I, SENSITIVITY 0.10; 1: UN- KNOWN; 2: FLUORANTHENE, 1.0 mq./LITER, SOLVENT CYCLOHEXANE. Fig. 5 Some Aspects o f Chemistry o f Tob TABLE 2 acco Smoke 39 CARCINOOENIC AROIdATIC COMPOUNDS IN CIGARET rE SMOEE CONDENSATE Compound Con y/i0 cenlralion 0 cigaretlea (Z ~ Benzo(a)pyrene 0.50 ~ (3, 4-Benspyrene) Benzo(e)pyrene 0.30 rl ~ (1, 2-Benzpyrene) ® Dibenz(a,j)aeridine 0.27 (1, 2, 7, 8-Dibenzacridine) ~ 7H-Dibenzo(c, g)carbazole 0.07 (3, 4, 5, 6-Dibenzearbazole) Chrysene 0.06 Dibenzo(a, h)anthracene 0.05 (1, 2, 5, 6-Dibenzanthracene) Dibenz(a, h)acridine 0.01 (1, 2, 5, 6-Dibenzacridine) Dibenzo(a,i)pyrene 0.002 (3, 4, 9, 10-Dibenzpyrene) Benzo(c)phenanthrene Benzo(j)fluoranthene (7, 8-Benzfluoranthene or 10,11-Benzfluoranthene) Badger and coworkers.' Of interest to us was the origin of the two carcinogenic dibenzacridines. Our results are summarized in Chart I. We found that when either pyridine, I, or nicotine, II, are pyrolysed at 7500 C, i.e. at a temperature approximating that in a burning cigarette, both the dibenzacridines III and IV are produced. Another aspect of the origin of the aromatic hydrocarbons which was studied by us concerned the aromatization of sterols. The sterol fraction obtained in our work constituted approxi- mately 5% of whole tar and from this we isolated, in an earlier study, stigmasterol." It occurred to us that the expected dehydro- genation products of stigmasterol, such as methylcyclopenteno- ~ phenanthrene, V, (Diels' hydrocarbon ), and related compounds should occur in tobacco tars. None of these compounds have yet been reported. Fa1k'1 has shown that another sterol, choles- T200959

40 Tobacco and Health I Pyrolysis (7500C) + IV terol, on pyrolysis at 360° C gave considerable quantities of V and related compounds. i ~ I _~ 1_511 Y In our experiments, pyrolysis of stigmasterol at 750° C did not yield cyclopelttenophenanthrene-type compounds but did give pyrene and bonzo ( a) pyrene. From these observations one can conclude that when the pyrolysis temperature is above a certain level direct aromatization by dehydrogenation becomes less prominent and that complete breakdown occurs to small active fragments, i.e., free radicals, from which the aromatics are re- synthesized. t t should be possible to establish the temperature Some Aspects of Chemistry of Tobacco Smoke 41 CH2 CH2 CH3 CH3 CH2 VIII C C\ 0 `2 C;2 ~C----CH 2 ,)- ' o----C 0 0 Chart II VII or temperature range in which the balance between these two competing processes is reversed. This result may be of signifi- cance in determining the effect of lowering the combustion tem- perature on the yield of aromatic carcinogens. Now, it is generally agreed that the concentrations in which these aromatic compounds are present in cigarette tars, although very significant, are not sufficient to account for the observed biological activity.20,91 This means that, in spite of all the work on the polynuclear aromatic compounds, we are still in the position of having to speculate about and search for additional carcinogenic and cocarcinogenic agents. There are several other components and potential components of tobacco tars which can be implicated. Let us examine a few of these prospective culprits; firstly carcinogenic agents: acetone, a known component of cigarette tar, gives ketene, VI, on pyrolysis as shown in Chart II. The latter compound readily dimerizes to diketene so that diketene is a potential component of cigarette T200960 C---CH2

42 Tobacco and Health smoke. There has been for some time uncertainty about the structure of diketene but the methylene-P-propiolactone struc- ture, VII, was established recently by nuclear magnetic resonance studies.z Diketene is of interest as a possible carcinogen because of its structural similarity to P-propiolactone, VIII. P-Propiolac- tone has been shown by the work of Roe and Glendenning" to be carcinogenic. These workers observed tumors by painting P-propiolactone on mouse skin. Because of its high reactivity it would be very difficult, if not impossible, to establish the presence of diketene in cigarette smoke. Further chemical studies on the possiblity of its occurrence in cigarette smoke will depend on the outcome of experiments now underway in our laboratory in which the carcinogenic activity of this compound is being examined. * Another type of compound which we consider of considerable interest as carcinogens are olifinic hydroperoxides. Squalene, IX, H3C\ C^CH--CHZ CHZ C- CH - CHZ CHZ C= CH-CHZ CH3 CH3 H3C/ I =CH-CHZ-CHZ C= CH-CNZ CHZ i= CH-CHZ 111 3 CH3 Squalene - IX was isolated by us from cigarette tars in an earlier study.27 The presence in cigarette smoke of a number of smaller olefinic mole- cules such as isoprene, butadiene and propylene have been re- ported by Hobhs and coworkers.lt These olefins are known to be oxidized quite readily by air to yield hydroperoxides and, possi- bly, other oxygenated derivatives. The presence of such hydro- peroxides in cigarette smoke thus becomes a distinct possibility. Fieser'Z has reported carcinogenic activity for an unsaturated sterol hydroperoxide, the partial structure of which is shown in Chart III, structure X. Kotin" has postulated carcinogenic ac- tivity for hydroperoxides and epoxides which he suggested are * All biological tcsting experiments described in this work are carried out under the direction of Di. L Orris of the Institute of Industrial Medicine, New York University Medical Center. Some Aspects of Chemistry of Tobacco Smoke 43 CH2=CH--CH::!=CH2 I H00 XI XII 02 OOH x CHART III XIII formed in the ozonization of olefins in polluted air. We have examined the carcinogenic activity of one such compound, vinyl- cyclohex-3-ene hydroperoxide, XIII, Butadiene, XI, on dimeriza- tion gives vinylcyclohex-3-ene, XII, and this by exposure to oxygen gives a small yield of the hydroperoxide, XIII. The struc- ture and preparation of this hydroperoxide was described by Brill.® In both Fieser's carcinogenic hydroperoxide and in vinyl- cyclohexene hydroperoxide, the hydroperoxide group is a to a double bond, in fact that is the only feature which these two molecules have in common. Using Brill's procedure we obtained a product which contained one-half of one percent of hydro- peroxide. This material was diluted in 1 to 1 proportion in ben- zene and tested by skin painting on mice. In one experiment, still underway, seven out of a total of 28 animals have developed tumors (4 cancers and 3 papillomas ). The presence of com- pounds of this type in cigarette smoke has not yet been estab- lished, but the observed carcinogenic activity coupled with the T200961

T 44 Tobacco and Health likelihood of their presence merits further investigation along these lines. Let us turn now to cocarcinogenic or potentiating agents. Phenol and substituted phenols have been shown by the work of Boutwell and others' to be capable of promoting the appear- ance of skin tumprs in mice following the application of single initiating doses of carcinogens. Phenols occur in both tobacco and in its smoke, and has been reported on by many workers', 14 Clemo,7 for example, reports a yield of approximately 1 mg. of phenols from the smoke of one cigarette. The studies of Gell- horn" suggest that whole tar has some cocarcinogenic activity. Furthermore, lloe29 recently reported on the potentiating action of a phenol fraction of cigarette tar by long-term painting on mice. Wynder" in a recent report obtained similar suggestive results from short term experiments. On the basis of these results it would seem likely that phenol and substituted phenols play a role in promoting the initiating activity of the carcinogens in cigarette smoke. Earlier work in our laboratory2° established the presence of long chain fatty acids in cigarette tars. These acids of chain length G., to GA (including both those with even numbers and with odd numhers of carbon atoms ) constitute approximately 1 percent of whole tar. Furthermore, slightly less than one-third of these acids occur in the cigarette tar as esters and/or amides. The esterifying moieties were not identified in our work. Their identification would be of interest because of the reported po- tentiating action of sorbitol and related esters of long chain acids, described by Setald." The implication that acids and phenols are cocarcinogenic or potentiating agents in cigarette smoke finds further support in an examination of the benzo(a)pyrene content of the tars obtained from cigarettes, cigars and pipe tobaccos. This work, carried out by Lindsey" ,;howed that cigar smoke and pipe tobacco smoke contained, respectively, three and eight times as much benzo(a)- pyrene as does cigarette smoke. Yet, in spite of this higher con- tent of benzo ( a) pyrene the tars from cigars and pipe tobacco gave the same yield of cancers by skin painting on mice.10 In addition, epidemiological data gathered by various investigators suggested a lower incidence of lung cancer deaths among pipe Some Aspects of Chemistry of Tobacco Smoke 45 and cigar smokers than among cigarette smokers 10 The notice- able chemical difference between cigarette smoke and cigar and pipe tobacco smoke lie in the strongly acid nature of cigarette smoke and the basic character of cigar and pipe tobacco smoke. These differences are probably associated with the methods of curing of the various tobaccos as well as with the differences in combustion temperatures. The significance of these acids and phenols in contributing to the biological activity of tobacco tars, earlier alluded to by Clemo,' would seem to justify further quan- titative chemical studies in this area. I have attempted to describe here what I believe to be promis- ing areas for further chemical research into the nature and carcinogenicity of tobacco smoke. Let us survey now what has been accomplished and what can conceivably be done to remove these deleterious components, assuming that people will continue to smoke in spite of the health hazards. The various possibilities which have been discussed and experimentally examined are: effective filtration, modification of the pyrolytic process by additives and, finally, removal of pre- cursors. We can discard immediately, the last proposal since all the organic material in the burning cigarette contribute to the production of tar and the aromatic carcinogens it contains. The possibilities of effective filtration appear to be promising. It is generally agreed that in filter cigarettes now on the market there is no selective removal of carcinogens. Certainly, it would appear a relatively simple matter to remove through a filter some of the acidic and phenolic components which I have discussed earlier. The third possible preventive measure i.e. modification in the pyrolytic process has been studied in some detail by the use of various additives. Thus Alvord and Cardon' found that the treat- ment of cigarette paper with ammonium sulfamate decreases the benzo(a)pyrene content of the cigarette smoke. Other workers have used yet other compounds to produce the same effect. The practical utility of such additives remain to be established. While it may be impossible to develop a safe cigarette, it is to be hoped that with the evidence gathered for them by scien- tists around the world, the tobacco concerns will in the foresee- able future come up with a product from which at least some of the harmful components have been remove - T200962

I 46 Tobacco and Health ACKNOWLEDGMENTS The work carried out at the Institute of Industrial Medicine of New York University Medical School was supported by a grant d' O~ V-4 ~ from the American Cancer Society. The author is indebted to Dr. Norton Nelson for encouragement, advice and interest in ~ this work and to Messrs. J. A. Bilbao, C. A. Joseph and F. L. ® Schmitt for technical assistance 14. REFERENCES 1. E. T. Alvord and S. Z. Cardon: The Inhibition of Formation of 3, 4-Benzpyrene in Cigarette Smoke, Brit. J. Cancer, 10, 498-503, 1956. 2. A. R. Bader, H. S. Gutowsky, G. A. Williams and P. E. Yankwich: The Proton Magnetic Resonance Spectrum and Structure of Diketene, J. Am. Chem. Soc., 78, 2385-2387, 1956. 3. (a) G. M. Badger, R. G. Buttery, R. W. L. Kumber, G. E. Lewis, A. G. Moritz and 1. M. Napier: The Formation of Aromatic Hydrocarbons at High Tem- peratures. Part I. Introduction, J. Chem. Soc., 2449-2452 (1958); (b) G. M. Badger ar.d R: G. Buttery: The Formation of Aromatic Hydrocarbons at High Temperatures. Part IV. The Pyrolysis of Styrene, 1. Chem. Soc., 2458-2463, 1958. 4. H. R. Bentley and E G. N. Berry (Editors): The Constituents of Tobacco Smoke: An Annotated Bibliography. Tobacco Manufacturer's Standing Com- mittee. Research Papers No. 3, London, 1959, 49 pp. 5. R. K. Boutwell and D. K. Bosch: The Tumor-Promoting Action of Phenol and Related Compounds for Mouse Skin. Cancer Research, 19, 413-424 (1959). (Work of previous investigators cited by Boutwell.) 6. W. F. Brill: Preparation of Hydroperoxide by the Autoxidation of 4-Vinycyclo- hexene. J. Org. Chem., 24, 257-259, 1959. 7. G. R. Clemo: Some Aspects of the Chemistry of Cigarette Smoke. 1. Tetra- hedron, 3, 168-174, 1958. 8. B. T. Commins, R. L. Cooper and A. J. Lindsey: Polycyclic Hydrocarbons in Cigarette Smoke. Brit. J. Cancer, 8, 296-302, 1954. 9. R. L. Cooper and A. J. Lindsey. 3, 4-Benzpyrene and other Polycyclic Hydro- carbons ia Cigarette Smoke. Brit. 1. Cancer, 9, 304-309, 1955. 10. Reviewed by D. F. Davies: A Review of the Evidence on the Relationship between Smoking and Lung Cancer. J. of Chronic Diseases, 11, 579-614, 1960. 11. lf. L. Falk, S. Goldfein and P. E. Steiner: The Products of Pyrolysis of Choles- terol at 'K0° C and their Relation to Carcinogens. Cancer Res., 9, 438-447, 1949. 12. L. F. Fieser: A Carcinogenic Oxidation Product of Cholesterol. J. Am. Chem. Sac., 77, 3928-3929, 1955. 13. A. Gellhorn: Cocarcinogenicity of Cigarette Tobacco Tar. Proc. Am. Assoc. ' Cancer .Res., 2, 109, 1956. Some Aspects of Chemistry of Tobacco Smoke 47 14. R. A. W. Johnstone and J. R. Plimmer: The Chemical Constituents of Tobacco and Tobacco Smoke. Chemical Reviews, 50, 885-936, 1959. 15. J. A. S. Gilbert and A. J. Lindsey. Polyclic Hydrocarbons in Tobacco Smoke: Pipe Smoking Experiments. Brit. J. Cancer, 10, 646-648, 1956. 16. E. Kennaway and A. J. Lindsey: Exogenous Factors in Lung Cancer. Brit. Med. Bulletin, 14, 124, 1958. 17. A. L Kosak, J. S. Swinehart, D. Taber and B. L. Van Dutut.m: Stigmasterol in Cigarette Smoke. Science,125, 991-992, 1957. 18. P. Kotin: The Role of Atmospheric Pollution in the Pathogenesis of Pulmonary Cancer. A Review. Cancer Research, 16, 375-393, 1956. 19. M. J. Lyons and H. Johnston: Chemical Investigations of the Neutral Fraction of Cigarette Smoke Tar. Brit. J. Cancer, 11, 554-562, 1957. 20. L. Orris, B. L. Van Duuren, A. I. Kosak, N. Nelson and F. L Schmitt: The Carcinogenicity for Mouse Skin and the Aromatic Hydrocarbon Content of Cigarette-Smoke Condensates. J. Nat. Cancer Inst., 21, 557-561, 1958. 21. J. S. Osborne, S. Adamek and M. E. Hobbs: Anal. Chem., 28, 211-215, 1956. 22. F. J. C. Roe and O. M. Glendenning: The Carcinogenicity of P-Propiolactone for Mouse Skin. Brit. J. Cancer, 10, 357-362, 1956. 23. F. J. C. Roe, M. H. Salaman and J. Cohen: Incomplete Carcinogens in Cigarette Smoke Condensate: Tumor-Production by a Phenolic Fraction. Brit. J. Cancer, 13, 623-633 1959. 24. A. H. Roffo: Carcinogenic Benzpyrene Obtained from Tobacco Tar. Zeits. f iir 1Crebs/orschung, 49, 588-597, 1939. 25. H. Setal'a: Tumor-Promoting and Cocarcinogenic Effects of.Some Non-fonic Lipophylic-hydrophilic (surface active) Agents. Acta path. et microbiol. Scandinau., Suppl. No. 115, 1956, 93 pp. 26. B. L Van Duuren and A, I. Kosak: Isolation and Identification of Some Com- ponents of Cigarette Smoke Condensate. J. Org. Chem., 23, 473-475, 1958. 27. B. L. Van Duuren and F. L. Schmitt. Isolation and Identification of Squalene from Cigarette Smoke Condensate. Chem. and Ind., 1006-1007, 1958. 28. B. L. Van Duuren. Identification of Some Polynuclear Aromatic Hydrocarbons in Cigarette Smoke Condensate. 1. Nat. Cancer Inst., 21, 1-16, 1958. 29. B. L Van Duuren: The Polynuclear Aromatic Hydrocarbons in Cigarette-Smoke Condensate 11. J. Nat. Cancer Inst., 21, 623-630, 1958. 30. B. L Van Duuren, J. A. Bilbao and C. A. Joseph: The Carcinogenic Nitrogen Heterocyclics in Cigarette-Smoke Condensate. J. Nat. Cancer Inst., 25, 53-61, 1960. 31. E. L Wynder, L. Fritz and N. Furth: Effect of Concentration of Benzpyrene in Skin Carcinogenesis. J. Nat. Cancer Inst., 19, 361-370, 1957. 32. E. L. Wynder and D. Hoffmann: A Study of Tobacco Carcinogenesis VII. The Role of Higher Polycyclic Hydrocarbons. Cancer, 12, 1079-1085, 1959. 33. E. L. Wynder and D. Hoffmann: Studies in Tobacco Carcinogenesis. Proc. Am. Assoc. Cancer Res., 3, 164, 1960. T200963

Effect of Smoking on Normal and Coronary Disease Patients 49 4 THE EFFECT OF CIGARETTE SMOKING ON CORONARY BLOOD FLOW AND CARDIAC WORK IN NORMAL SUBJECTS AND PATIENTS WITH CORONARY DISEASE RICHARD BING, M.D THC pharmacologic effects of nicotine are manifold and diverse. The alkaloid is known to stimulate and then depress sympathetic and parasympathetic ganglia. Its cardiac effect could result from direct action on the myocardium, from its effect on the extrins?c nerve supply to the heart, or on the neural structures witbin the heart; nicotine also could exert a direct action on the coronary vessels. The experimental literature on the effect of nicotine on coronary blood flow has been reviewed in preceding papers from this laboratory. Animal experiments have suggested that in general small quantities of nicotine in- crease coronary flow in dogs and rabbits while large doses cause vasoconstriction. Prior to the advent of coronary sinus catheteri- zation, the effect of nicotine on the coronary circulation of man was studied by means of the electrocardiogram or ballistocardio- gram. These results are difficult to interpret, because many of these changes could have resulted from the action of nicotine on cardiac work and rate, rather than from its effect on the coronary circulation. There has been no direct evidence from these studies to support the concept that nicotine causes coronary vasoconstric- tion in normal man. The advent of catheterization of the coronary sinus has made possible an investigation of the effect of tobacco smoke with nicotine as its main pharmacologic constituent, on the coronary blood flow in man. The first paper on this subject, published in Work supported by U. S. Public Health Service Grant No. H-5043, The Life Insurance Medical Research Fund, "I'he American Heart Association, The Michigan Heart Association, and the Tobacco Industry Research Committee. 48 1957, demonstrated that the smoking of one cigarette caused a significant rise in coronary blood flow and a significant decline in coronary vascular resistance and myocardial extraction of oxygen and glucose. The results obtained indicated that cigarette smok- ing in normal subjects did not result in constriction of the coro- nary blood vessels. Further studies on the effect of smoking on the coronary circulation of man were recently undertaken to study the effect of smoking of several cigarettes on the coronary circulation, heart rate, cardiac output and left ventricular work of normal individuals and of patients with coronary heart disease. The subjects used in this study had either no evidence of heart disease or had suffered a myocardial infarction at least one year prior to this study. The diagnosis of myocardial infarction was based on the clinical course and the electrocardiogram. All but one of these patients experienced angina pectoris. There was no cardiomegaly and arterial hypertension was absent. Six subjects served as control. All of the patients were habitual smokers. Coronary blood flow was measured by the nitrous oxide method. Heart rate, arterial pressure, and left ventricular lead were moni- tored during the test. The results of these studies showed relatively little difference in the response of normal individuals and of patients suffering from coronary artery disease. Both groups had a significant acceleration of heart rate and a uniform rise in arterial pressure. These hemodynamic changes were somewhat more pronounced in the coronary group. Patients with coronary heart disease had initially a lower cardiac index than normal individuals, but the cardiac output increased in both groups. The increment of left ventricular work during smoking was also more pronounced in the coronary group. No significant changes in coronary flow occurred either in the normal individuals or in the patients with prior myocardial infarction. As myocardial oxygen extraction also remained con- stant, myocardial oxygen usage did not change. In view of the possible importance of myocardial oxygen usage to the sympto- matology of coronary artery disease, the ratio of left ventricular work to its oxygen consumption was calculated. Because of a distinct increase of left ventricular work in presence of unchanged T200964

#, . 50 Tobacco and Health myocardial oxygen consumption, the ratio of left ventricular work to left ventricular oxygen consumption rose; the change was of similar magnitude in both normal individuals and in patients with coronary heart disease. The change in ratio work/Oz con- sumption was not attended by electrocardiographic evidence of ischemia either in the normal individuals or in the patients with myocardial infarction. It has been shown by several workers that augmentation of coronary blood flow may usually be anticipated when there is an increase in heart rate, systemic arterial pressure, cardiac output and left ventricular work. In this manner, the apparently greater oxygen requirements of the myocardium can be served. In the individuals of this series, this relationship could not be obtained. Although left ventricular work rose, myocardial oxygen consump- tion remained constant. Several explanations for this finding may be considered. The "coronary vascular resistance" may have been fixed in the patients with coronary heart disease. This is unlikely since normal individuals had the same response as the patients with coronary heart disease. The second possibility, that smoking resulted in a release of catechol amines which then caused a disproportionate increase in work of the heart is also unlikely. It has been shown that small amounts of catechol amines induce changes in coronary flow before increments in heart rate and pressure become apparent. The third possibility is that antidiuretic hormone, released during smoking, prevented the nonnal response of the coronary circulation to increased cardiac work. It is well known that this hormone is a potent coronary vasoconstrictor. This possibly cannot be excluded, but remains at this moment of doubtful significance. Experiments from this laboratory carried out in 1957 and dur- ing the last year have therefore demonstrated that cigarette smok- ing either increases or has no effect on coronary flow and myo- cardial oxygen consumption. The increase in the ratio left ventri- cular work/myocardial oxygen consumption suggests the possi- bility that the untoward effect of smoking may be due to the change in this ratio. This, in coronary subjects, could presumably produce ischemic symptoms. Clinical evidence, however, suggests that this is an uncommon occurrence. 11, 5 TOBACCO IN HEALTH AND DISEASE EFFECT ON PERIPHERAL CIRCULATION J. Enwua WooD, M.D TEE effects of smoking that are potentially harmful to normal individuals would seem to be of first importance in a discussion of tobacco and health. Secondly, and possibly just as. important, the effects of smoking that are noninjurious to the normal individual but potentially hazardous to the patient with a specific disease should be considered. The known acute effects of tobacco on the peripheral blood vessels can be reproduced by the dose of nicotine received during smoking. Further, other constituents of tobacco have little effect upon the blood vessels in the concentrations ordinarily present in smoke (1). Thus we may give primary attention to the pharmacologic effects of nicotine in considering the acute effects of tobacco upon the peripheral circulation. PHARMACOLOGICAL EFFECTS OF NICOTINE UPON PERIPHERAL BLOOD VESSELS It has been amply demonstrated that nicotine results in peri- pheral vasoconstriction. Experiments on humans have indicated that the vasoconstriction in response to smoking is dependent upon normal function of the sympathetic nerves supplying the blood vessels. In the absence of local sympathetic nervous ac- tivity, the vasoconstrictor response does not occur (2-5). More recently, the observations of Burn and his :co-workers have sug- gested that nicotine causes release of norepinephrine from chro- maffin tissue near the blood vessels, thus causing them to con- strict. The presence of stored norepinephrine in these chromaffin 51 T200965

52 Tobacco and Health cells seems to be dependent upon an intact sympathetic nervous system (6). While these findings add greatly to our understand- ing of the mechanism of action of the sympathetic nervous 3ystem, they do not alter the observation that sympathetic nerves are necessary to the vasoconstrictor action of small doses of nico- tine in humans. Direct stimulation of smooth muscle of blood vessels by nico- tine is a well-known property of this drug. Haimovicci observed in frogs that vasoconstriction could be produced in the absence of sympathetic nervous activity, thus indicating that direct stimu- lation of the smooth muscle had occured (7). More recent studies of nicotine injections into the brachial artery of a sympathectom- ized limb showed that this is not true of humans (8). Further, the observation that sympathectomized limbs do not show evi- dence of vasoconstriction during smoking also indicates that direct stimulation of smooth muscle by nicotine is not important in humans. Circulating catechol amines released either from the adrenal medulla or from chromaffin tissues elsewhere following a general- ized syrnpatho-adrenal discharge might also account for vaso- constriction observed during smoking. Although smoking does result in an increase in blood and urinary catechol amines, the quantity present would not account for the degree of vasocon- striction observed (9, 10). Again, the results of studies of locally sympathectomized limbs support the idea that circulating cate- chols are not important as vasoconstrictors in smoking. While the great majority of investigations directed toward the problem of acute circulatory effects of tobacco have dealt with the problem of diminished blood flow as a pharmacological action of nicotine, allergy to this or some other constituent of tobacco might likewise he considered one of its acute deleterious effects. This sahjcet is discussed in detail later in the conference by Dr. Redisch. THE EFFECT OF TOBACCO UPON BLOOD FLOW OF THE EXTREMITIES DURING REST IN A WARM ENVIRONMENT The most stz iking vasoconstrictor effects of smoking are seen in the digits although it is equally possible to demonstrate dimin- Effect on Peripheral Circulation 53 ished blood flow in the foot or in the skin of more proximal portions of an extremity (4, 11, 12). It would appear that blood. flow of skeletal muscle is either unaltered importantly or actually increased by smoking as has been demonstrated by Montgomery's group (12). BLOOD FLOW OF THE SKIN DURING SMOKING The above studies further define the area of interest with re- gard to the patient's health and suggest that primary concern should be in the blood flow of the skin during smoking. Repeated measurement of blood flow of the foot, which is primarily com- posed of skin, shows that the blood flow diminishes within a few seconds after smoking is initiated. Flow may begin to return to normal even as the subject continues to smoke and particularly if he smokes two cigarettes in succession. Cooling the environ- ment to 68"F. causes greater constriction of blood vessels of the skin than smoking in a warm environment. However, smoking in a cool environment results in an additive vasoconstrictor effect (11). If the stress of local ischemia is placed upon the tissues, the hyperemia that follows this period of arterial occlusion is diminished by smoking (5). The results of this experiment at least suggest that even blood flow to a part in special need of oxygen is rnodified by smoking. It may be of importance to recognize that a subject develops tolerance to the central nervous system emetic effect of smoking, but he does not develop tolerance to the vasoconstrictor effects of smoking (11). Thus, occasional tobacco use does not cause relatively excessive vasoconstriction, nor does heavy smoking result in an abatement of the vasoconstriction. The degree of reduction of blood flow in the skin appears in- consequential especially when compared with that induced by a cool environment. Undoubtedly the normal individual supplies his local oxygen needs in these circumstances by an increased extraction of oxygen from blood passing through the capillaries with a resultant increase in A-V oxygen difference. The question arises however, as to whether or not a patient with restricted blood flow to the skin might not suffer some ill effect from even a slight reduction in blood flow. Studies of patients who continue T200966

54 Tobacco and Health to smoke after an operative procedure on the vessels of an extremity compared with those who discontinue the use of tobacco following a similar procedure indicate that the prognosis for the limb is impaired as a result of smoking (13). This suggests that the mild vasoconstrictor effect of nicotine is of sufficient degree to tip the balance. Whether or not sympathectomy or the use of drugs that reduce sympathetic nervous activity would alter this situation for the patient with vascular disease is a subject that awaits study. THE ACUTE EFFECTS OF NICOTINE AND OF SMOKING UPON PERIPHERAL VEINS The acute effect of tobacco upon the peripheral veins is much less well studied than that upon the arterioles, because suitable methods for study of these vessels in humans have been available in the past. Eckstein and Horsley have found that small intra- venous doses of nicotine result in constriction of the peripheral veins of the forearm with movement of blood from these veins to more central veins (14). The results of other studies indicate that patients with congestive heart failure react to exercise with much greater constriction of the veins than normal individuals, and that this results in a rise in the venous pressure and an in- crease in central blood volume in association with the decrease of peripheral blood volume (15). The inference of these experi- ments is that a-natient with severe congestive heart failure who smokes might induce a further shift of blood to the central cir- culation with a resultant increase in dyspnea. In fact an occa- sional patient will complain that during the height of his heart failure smoking did seem to burden him with further dyspnea. Recent studies in our laboratories indicate that when patients with severe congestive heart failure smoke, peripheral venous pressure does rise from 2 to 4 cm. of water, presumably as a result of the venoconstrictor activity of nicotine. CHRONIC EFFECTS OF SMOKING AND NICOTINE UPON PERIPHERAL BLOOD VESSELS Two of the most important questions with regard to peripheral blood vessels, tobacco, and human health are the relationship Effect on Peripheral Circulation 55 between chronic use of tobacco and the rate of progression of atherosclerosis, and the relationship between use of tobacco and the day to day level of blood pressure especially in hypertensive patients. These two questions remain largely unanswered. TOBACCO AND ESSENTIAL HYPERTENSION Transient small rises of blood pressure may be observed in normal individuals while they are smoking. Greater rises of blood pressure are seen in so called hyper-reactors or in patients with hypertension during smoking (16). Studies of long-term use of tobacco versus control periods without its use have not been performed. Thomas has obtained data related to this question in her studies of family histories of Johns Hopkins Medical Stu- dents relative to hypertension and the responses of these students' blood pressures to smoking (17). Keys and his group have ob- tained information on the incidence of hypertension in smokers versus non-smokers in Finland, the results suggesting that no causal relationship exists between the use of tobacco and the presence of hypertension (18). The results of the Framingham study seem to lead to a similar conclusion (19). Also, present knowledge of the mechanism of essential hypertension would not suggest a causal relationship between this disease and the use of tobacco so that this question does not appear to be of immediate importance. The clinical observation that some patients seem to do better with regard to control of elevated blood pressures if they discontinue the use of tobacco indicates that a systematic investigation of this matter should be carried out. TOBACCO AND ATHEROSCLEROSIS It has been observed that administration of nicotine will result in an acceleration of the development of atherosclerosis induced in rabbits (20). Population studies have suggested that the use of tobacco was associated with an excessive incidence of coronarv artery disease, but the objection has been raised that individuals who elect to smoke might be independently prone to the disorder. Thus, the question of a mechanism whereby chronic effects of smoking might actually induce or accelerate atherosclerosis has been required. Gofman et al. have found that smokers have T200967

56 Tobacco and Health higher levels of serum lipoproteins and cholesterol than non- smokers (21), Lut Page and co-workers found that there was no acute change in the level of cholesterol or lipoproteins during smoking of two cigarettes (22). Hamlin has observed that the blood nonesterified fatty acids increase when sympathetic nervous activity is induced by tilting the subject (23). Whether or not mobilization of nonesterified fatty acids represents a heightened tendency for the development of atherosclerosis is a moot ques- tion, but there is this relationship between tobacco, the sympathe- tic nervous svstenl and lipid metabolism. SUMMARY The mechani,m of the vasoconstrictor action of nicotine is well defined especially as it applies to the dose and rate of administra- tion of this material to humans during smoking. The acute peripheral vascalar effects with special regard to the blood vessels of the skin are apparently innocuous in the normal human. There is somc reason to believe that these relatively minor effects may be detrimental in patients with obstructive vascular disease of the extremity. The chronic effects of repeated sympa- thetic ncrvotts activity induced lay nicotine or other more obscure effects of smoking are difficult to evaluate. It would seem especially important to stndy the long-term effects of smoking upon the rate of development of atherosclerosis and the long-term effect upon the level of the hlood pressure of patients with essen- tial hypertens.icm. REFERENCES 1. Roth, G. M., and R. M. Schick: Effect of smoking on the cardiovascular system of man. Cirenlatiroa, 17,443, 1958. 2. Maddock, W. C;., and Coller, F. A.: Peripheral vasoconstriction by tobacco and its relation to tltromho-angiitis obliterans. Ann, of Surg., 98, 70, 1933. 3. Roth, C. M., MacDonald, J. B., and Sheard, C.: The effect of smoking cigarettes and of intravenniGs administration of nicotine on the electrocardiogram, basal" metabolic rate, cutaneous tetnperature, blood pressure and pulse rate of normal persons. J. A. M. A., 125, 761, 1944. 4. Rappaport, S. I., Frank, IT. A., and Massell, T. B.: The effect of smoking upon blood flow in the sympathectrnnized limb. Circulation, 2, 850, 1950. 5. Cofrman, J. 11., Wood, J. E., and Wilkins, R. W.: Effect of cooling and of smoking tnbacrn upon the blood flow of reactive hyperemia of the foot. Circulation, 28, 177, 19iR. 01 6. ~ ~ ~ 7. ~ O 8. ~ 9. ~ N 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. iJffect on Peripheral Circulation 57 Burn, J. H., Leach, E. H., Rand, M. J., and Thompson, J. W.: Peripheral effects of nicotine and Acetylcholine resembling those of sympathetic stimulation. J. of Physiol., 148, 332, 1959. Haimovici, H.: Postganglionic site of action of nicotine with special references to its direct action on blood vessels. Proc. Soc. of Exp. Biol. and Med., 68, 516, 1948. Stromlad, B. C. R.: Effect of intra-arterially administered nicotine on the blood flow in the hand. British Med. J., 1, 484, 1959. Kiser, J. C., Booler, W. T., and Watts, D. T.: Blood epinephrine levels in dogs following intravenous administration of nicotine. Arch. Internal. Pharm. Dyn., 104, 403, 1955. Watts, D. T., and Bragg, A. D.: Effect of smoking on the urinary output of epinephrine and norepinephrine in man. J. Appl. Physiol., 9, 275, 1956. Eckstein, J. W., Wood, J. E., and Wilkins, R. W.: Comparative vasoconstrictor effects of inhaling tobacco smoke in warm and cool environments and before and after abstinence from tobacco. Am. Heart J., 35, 455, 1957. Rottenstein, H., Peirce, G. Russ, E., Felder, D., and Montgomery, H.: New York Acad., of Sci., 1960, in press. Juergens, J. L., Barker, N. W., and Hines, E. A., Jr.: Arteriosclerosis obliterans: Review of 520 cases with special reference to pathogenic and prognostic factors. Circulation, 21, 188, 1960. Eckstein, J. W., and Horsley, A. W.: Responses of the peripheral veins in man to the intravenous administration of nicotine. New York Acad. Sci., 1960, in press. Wood, J. E.: The mechanism of venous pressure elevation with exercise in human congestive heart failure. J. Clin. Inoest., 38, 1055, 1959. Roth, G. M.: Effects of smoking of tobacco on the cardiovascular system of normal persons and patients with hypertension. 1. Am. Geriatric Soc., 2, 270, 1954. Thomas, C. B., Murphy, E. A.: Observations in some possible precursors of essential hypertension and coronary artery disease. VI. Comparison of the circulatory reactivity to the cold pressor test and to the smoking test. Ann. Int. Med., 50, 970, 1950. Karvonen, M., Orma, E., Keys, A., Fidonza, F., and Brozek, B.: Cigarette smoking, serum cholesterol, blood pressure and body fatness. Lancet, 1, 492, 1959. Dawber, T. R., Kannel, W. B., Nicholas, R., Stokes, J., Kangan, A., and Gordon, T.: Some factors associated with the development of coronary heart disease. Am. J. of Public Health, 49, 1349, 1959. Wenzel, D. G., Turner, J. A., and Kissil, D.: Effect of nicotine on cholesterol . . . ., , , . 23. Hamlin, J. T., Hickler, R. B., and Hoskins, R. G.: Free fatty acid mobilization by neuroadrenergic stimulation in man. 1. Clin. Inoest., 39, 606, 1960. induced atherosclerosis in the rabbit. Circ. Rsch., 7, 256, 1959. Gofman, J. W., Lindgren, F. T., Strisower, B., deLalla, 0., Glazier, F., and Tamplin, A.: Cigarette smoking, serum lipoproteins, and coronary heart dis- ease. Geriatrics, 10, 349, 1955. Page, I. H., Lewis, L A., and Moinuddin, M.: Efl'ec6 of cigarette smoking on serum cholesterol anl lipoprotein concentrations. J A M A 171 1500 1959 T200968

Session 11 EXPERIMENTAL PATHOLOGY OF TOBACCO SMOKE , T200969

.. 6 THE ROLE OF SKIN NEOPLASIA IN TOBACCO CARCINOGENESIS E. L. WYNnER, M.D. and D. HOFFMANN, Ph.D. IN the light of present knowledge, laboratory ex- periments cannot prove that a given substance produces cancer . in man. Such proof can only come from epidemiological and statistical studies. These studies have established by standards regarded as beyond any reasonable doubt that smoking, particu- larly of cigarettes, is one of the causative factors of lung cancer. 'I While the positive animal experiment can underline the human ~ epidemiological findings, its basic significance lies in permitting a ~ study of the mechanism of carcinogenesis. Diferemcea iro routes and animal apeciea. Searching for a ' proper experimental tool, one attempts to design an experiment ~ which permits the application of the material in question with ~ relative ease, receive results with speed while at the same time using a tissue which is similar to the tissue involving man. For ~ this reason we eliminated the use of subcutaneous tissues, since ~ . the production of sarcomas seems less specific than carcinomas. ~ Using tobacco smoke condensate Druckrey has obtained sarcomas in rats (1). Thus, though we regard this test as less specific, tobacco smoke condensate can produce sarcomas. The tissue most similar to that of human broncho epithelium is ~ obviously the bronch0 epithelium in animals. The problem is that the human manner of smoking cigarettes cannot possibly be ~ duplicated in an experimental animal. It must be apparent, even ~ to the layman, that exposing mice in a cage to tobacco smoke is quite different than the human manner of smoking. Peacock in 61 T200970

62 Tobacco and Health Glasgow is conducting experiments where fowls inhale cigarettes. He encounters the problem that he cannot have them smoke a sufficient number of cigarettes per day (2). Blacklock induced cancer in the hilum region of rats by injecting tobacco smoke condensate together with tuberculi bacilli (3). Graham, Cronin- ger and Wynder applied tobacco smoke condensate to the bronchial-pleural fistula of dogs (4). They encountered the problem of infection and were unable to keep these animals alive longer than three years, a period too short to expect any neoplastic changes. Rockey applied tobacco smoke condensate to the trachea of dogs and though he found what may be considered pre-cancerous changes, the experiments were not carried out long enough to expect cancer to develop (5). These points must be considered when choosing the route and the type of experimental animal to be used. The reason we have exclusively used the mouse skin lies in its ease of application, similarity to the bronchial epithelium, particularly after meta- plasia has set in and its much greater specificity than the con- nective tissues. As said before, these animal studies are con- ducted to define the mechanism of tobacco carcinogenesis. If we succeed in obtaining a type of tobacco product which upon combustion would be followed by a marked reduction of carcino- genic activity to these animals, then we would certainly agree that, provided this product is practical and not toxic, it should be introduced to standard manufacturing processes. But only a follow-up study of smoking such a tobacco product could prove that we have reduced the carcinogens and co-carcinogens respon- sible for human lung cancer. Experiments. The procedures we are following in this field demand a closely coordinated program between the biologist and the chemist. It is for this reason that we are reviewing both biological and chemical studies. It is well established that cigarette smoke condensate is carcinogenic to skin of mice and rabbits. Seven different labora- tories have now agreed on this point, including the studies by Kensler (6-12). The differences in the biological activity are due to a variety of factors-the major one of which seems to be a difference in the dose given to the animals and the way the Role of Skin Neoplasia in Tobacco Careinogenesis Fig. 1 63 animal skin is treated. Our standard procedure is to apply tobacco smoke condensate three times weekly in a 50 per cent acetone solution to mouse skin for 15 months. The back of the mouse is shaved whenever necessary with an electric shaver. The animals receive about ten grams of tobacco smoke condensate a year. We are surprised to find from time to time comparisons to our studies by different investigators which did not consider the important points of total dose and skin preparation (13). In our experiments if the animals receive less than five grams of tobacco smoke condensate, no cancers are produced and if they receive less than three grams per year, no 'papillomas are pro- duced (14). This is to be expected and is similar to the dose response relationship found for man (15). Other factors to be T200971

64 Tobacco and Ilealth C/GARETTE TAR FRACTIONAT/ON- RELAT/VE BIOLOGICAL ACTlV1TY WHOLE TAR 100q. ++++ FMBCI21 NSOL.14q. Me C I2 SO'L. 86 9. t H2 04 ACID + NEUTRAL BASIC TAR 60q. 149. ++++ I I NICOTINE FREE BASIC C0 N0 3 2 i NEUTRAL TAR ACIDIC TAR 439. 129 +++ + HEXANE CARBON TET. BENZENE ETHYL METHANOL PYRI'DI'NE ACETATE 6Aq. 1.7g. 6.09. 21.5q. 5.2q. 1.00. + . +++ - - - - Fig. 2 considered are the method of tar preparation and application, the specific animal strain used, as well as the total area over which the tar has been applied. Little has criticized that too much tobacco smoke is applied to the animals compared to the amount of exposure to man and refers to the immcnse difference between the total lung surface and the small mousc skin area. It is known, of course, to every student of hmt; cancer that the majority of cancers do not develop in the large alveolar areas of the lung but in the bronchi. When one dissects and opens a bronchial tree of an adult, one can compare the area of the major bronchial tree to the area of a mouse skin, w1iich is at best 6 times larger (Fig. 1). The average lung cancer patient who is a heavy smoker (30 cigarettes average per day) is exposed to approximately thirty times more smoke condensate per year than a mouse receives in this time span (16). 1 a } n-C6H14 B (a) p ?50 pp'm ST +++ DISTR IBUTION` cyclohexane nitromethane i \4 BIIa 0.35% n-C6H141CCI4 B(a/p - ST - Fig. 3 Bllb 0.15% CCI 4 I BIII 0.9% ~ ~ B (alp 750 ppm ST +++ Having established that tobacco smoke condensate produces cancer in mice, the next question is how this is accomplished. The fractionation study published with Wright in 1957 indicated that the major carcinogenic fraction is the neutral fraction of the tar, particularly its column chromatography subfraction eluted with carbon tetrachloride (Fig. 2) (17). The neutral portion thus yields a fraction 1.7 per cent of the total smoke condensate with relatively high activity. Of course, since we have also limited activity in some of the other fractions, we,know that we either lost some of the carcinogens of the neutral portion, or that there are other types of carcinogens in the other fractions. We are now attempting to further identify the carcinogenic material in the T200972 CCI4 ELUATE FROM S ILICl1 GEL CHROMIATOGRAM OF THE NEUTRAL PORTION Role of Skin Neoplasia in Tobacco Carcinogenesis 65 B B(a) p80ppm ST +++ Silica Gel ~V-'

66 Tobacco and Health carbon tetrachloride subfraction. On the basis of short term tests we have now isolated a fraction containing only .15 per cent of the total tar which has the greatest relative activity (Fig. 3). When cigarettes are smoked under standard conditions (1 puff per minute, 2 seconds duration, puff volume 35 cc. ), benzo ( a)- pyrene is present in the condensate in 1.2 parts per million, in 70 ppm. in the carbon tetrachloride fraction and 750 ppm. in the subfraction of the carbon tetrachloride fraction. It is important to emphasize that the polynuclear hydrocarbons so far identified cannot account for, by themselves, more than 3 per cent of the biological activity of the total smoke condensate and not more than 10 per cent of the carbon tetrachloride fraction (18). This means that either other polynuelear hydrocarbons so far unidenti- fied must be present in tobacco smoke condensate, or that other types of carcinogens or co-carcinogens must be present in the condensate. We regard the aromatic polynuclear hydrocarbons as the major initiating carcinogens in tobacco smoke condensate, and suggest that if we could successfully reduce these initiators that, regardless of other carcinogens or co-carcinogens, the bio- logical activity would he significantly reduced. In one experi- ment we havc rcmove.d the carbon tetrachloride fraction, and only 6 pcr cci,t papillonlas were found compared to the usual 30 to 40 per cent, and no cancers were observed in 50 animals on which this condensate was applied for 15 months. With additional experiments in this field we intend to establish this result fin•ther. So far we do not believe that the relatively nega- tive results arc due to the loss of carcinogens during the chemical separation be.cause in a previous experiment we separated the basic, acidic and nclltral fraction, which were then combined, and no biological activity was lost (17). In our search for other carcinogens and co-carcinogens we studied di(fcrent condensate fractions by the short term seba- ceous gland tml. When comparing the basic, neutral and phenolic portions the latter two were particularly active and upon removal of the phenolic fraction from the tobacco smoke condensate, much of the short term activity was lost. This was of interest because it confirmed short term experiments with polvnuclear hydrocarbons showing that the positive short term results could INITIATOR + PROMOTER -110,CANCER Polycyclic Hydrocarbons Phenols + + Unknowns COMPOSITION OF STANDARD 14.0 CIGARETTE SMOKE CONDENSATE Unknowns (in Neutral, Basic and Acidic Fractions) Fig. 4 9.3 1.3 PPM •.7 47.2 NICOT + PHENOLS+ INSOLU- NEUTRAL NORNIC NAPHTHOLS ABLE PORTION 4NICOTIN- +PHENOLIC PORTION % FREE P°% % BASIC P % Fig. 5 T200973

T 68 Tobacco and Health not be due to the polynuclear hydrocarbons in the concentration as present in tobacco smoke condensate. We place particular emphasis on phenol and its derivatives since Boutwell and Bosch had already shown them to be co-carcinogens (19) and since Roe had produced evidence that the phenolic fraction of cigarette smoke condensate was co-carcinogenic to 9.10-dimethylbenz( a)- anthracene (20). Our long term studies using .005 per cent benzo ( a)pyrene applied three times a week as an initiator con- firmed the short term tests in that they show the promoting activity for the phenolic and acidic fractions but none for the neutral fraction. The fractions were all applied in 10 per cent acetone solution twice a week. Thus, the phenolic fraction tends to be co-carcinogenic in the same concentration that it is in tobacco smoke condensate. By itself, a 10 per cent phenolic fraction produced no tumors. We are now engaged in dividing the phenolic fraction into three subfractions consisting of 2.5, 9.0 and 7.5 per cent of the total phenolic portion and a residue of 70 per cent. The greatest short term activity was found in sub- fraction II which contains 0.8 per cent phenols from the whole condensate, while 55 per cent of this fraction is phenol itself. These studies confirm that phenol and its methyl and dimethyl derivatives are the main promoting material in the phenolic fraction. The biological activity of tobacco condensate seems a c.onsequcnce of initiating and promoting carcinogens (Fig. 4). In an attempt to predict the biological activity of a cigarette smoke condensate we should carry out the chemical analysis of several fractions and subfractions (Fig. 5). Preventive Measures. In which direction points the present experimental data? As far as preventive measures are concerned, there is first the question of reduction in smoke condensate. Ob- viously, the animal experiment, in line with human evidence, shows that the lower the exposure to smoke condensate, the lower the wmor incidence. Short of giving up smoking alto- gether, which would be the most desirable step, we must search for more practical aspects. One of these is to reduce the amount of cigarettes consumed, another to smoke cigarettes with as low a tar content as practical. The latter can be achieved by effective filter cigarettes (16). It is also established that the less one Role of Skin Neoplasia in Tobacco Carcinogenesis 69 inhales, the less smoke condensate will be retained in the respira- tory tract. The non-inhaler absorbs only about 10 per cent in contrast to the deep-inhaler who retains up to 90 per cent. One should also not smoke the cigarette to the very end, since we have found that the last half of the cigarette contains 42 to 44 per cent more particulates than the first half (16). It is advisable, therefore, to throw a cigarette away after it is half-smoked. We must attempt to reduce the initiating carcinogenic aromatic hydrocarbons. The formation of 'polynuclear hydrocarbons is known to be a function of incomplete combustion. Our studies have shown that the side stream of the smoke contains 4 times as much benz(a)pyrene as that of the main stream smoke (21). Research must continue for precursors of polynuclear hydro- carbons. Since nearly all organic substances upon incomplete combustion yield, among other combustion products, polynuclear hydrocarbons, a significant reduction in these components seems to be unlikely (22, 23). A reduction of the polyphenols in tobacco, which we regard as main precursors for the phenols, may be a more promising effort. Selective - filtration, though theoretically feasible, does not seem possible, at least not for the polynuclear hydrocarbons. Therefore, we must concentrate in our studies on promoting substances, particularly the phenols themselves. Preliminary work indicates that through the use of certain additives, par- ticularly copper (II) nitrate and calcium carbonate, we can obtain a significant reduction in the phenolic fraction of the smoke condensate. We have a program now on the way in which five different additives have been used for the preparation of 100,000 cigarettes each and their condensates are now being applied to the backs of 100 mice each. This type of crash program which we hope to continue and expand in cooperation with other research groups in this country and abroad and in which we hope tobacco research laboratories will join, as in fact some already have, will in our opinion yield a practical smoking product with reduced biological activity of the condensate to animal skin. We end this discussion with the question that we raised in the beginning. What is the purpose of the animal investigation? Its purpose is not to establish that cigarette smoking is causing I~IB" I! R~! '~e ~ itute, Inct T200974

70 Tobacco and Health lung cancer in man-a proof at present only given through statistical and epidemiological investigations-but rather to de- fine the mechanism of carcinogenesis, which we must then with care apply to the human data; just as in chemotherapy we first experiment with the animals and if successful, proceed to man, so we must proceed in the field of experimental carcinogenesis. Precedent is large in that if a substance has been shown to cause cancer in man and a variety of experimental animals, the specific agents responsible for this action are of similar nature. We would predict, therefore, that if we can find a safer smoking product for the experimental animal and then introduce this product to the human population that after a follow-up study, a reduction of lung cancer will occur. The ultimate proof of success rests, of course, on such a follow-up study. We are sure that most scientists in cancer research will agree that this is a logical procedure and that on the basis of existing knowledge, we can look with some hope in this direction as perhaps the most practical manner of at least partially solving the smoking-lung cancer problem. In view of the thousands of lives lost throughout the world from lung cancer and in view of the fact that so little progress is made to cure this type of disease, we must continue with our utmost effort to work towards the prevention of this disease. We believe that along the ways outlined here, at least a partial solution of the problem lies at hand. 4 , J . ~ 11. Bock, F. G., and Moore, G. E.: Carcinogenic activity of cigarette-smoke con- H densate. I. Effect of trauma and remote X-irradiation. J. Nat. Cancer Inst., E.•I 22:401-411, 1959. 12. Kensler, C. J.: Personal communicatiori. 13. Hamer, D., and Woodhouse, D.: Biological tests for carcinogenic action of tar from cigarette smoke. Brit. J. Cancer, 10:4'9-53, 1957. 14. Wynder, E. L., Kopf, P., and Ziegler, H.: A study of tobacco carcinogenesis. II. Dose-response studies. Cancer, 10:1193-2000, 1957. 15. Hammond, E. C., and Horn, D.: Smoking atid death rates-Report on 44 months of follow-up of 187, 783. J.A.M.A., 166:1159-1294, 1958. 16. Wynder, E. L., and Hoffmann, D.: Some practical aspects of the smoking- cancer problem. New England J. Med., 262:540-545, 1960. 17. Wynder, E. L., and Wright, G.: A study, of tobacco carcinogenesis. I. The primary fractions. Cancer, 10:255, 1957. 18. Wynder, E. L, and Hoffmann, D.: A study of tobacco carcinogenesis. VII. The role of higher polycyclic hydrocarbons. Cancer, 12:1079-1086, 1959. i 19. Boutwell, R. K., and Bosch, D.: The tumor-promoting agent of phenol and related compounds for mouse skin. Cancer Res., 19:413-424, 1959. i 20. Roe, F. J. C., Sallaman, M., Cohen, J., and Ber ; gman, J.: Incomplete carcinogens in cigarette smoke condensate: Tumor-promotion by phenolic fraction. Brit. J. Cancer, 13:623-633, 1959. 21. Wynder, E. L, Hoffmann, D.: The present status of laboratory studies of to- REFERENCES bacco carcinogenesis. Acta path. et microbiol. Scandinavica (In press). l. II.: Personal correspondence. Druckrey 22. Lindsey, A. J., Persand, K., and Candeli, A.: Reduction of benzpyrene in to- , bacco smoke. Brit. J. Med., 11: 821, 1959. 2. Peacock P. R.: Experimental cigarette smoking by domestic fowls. Brit. J. , 9:461-463 1955. Cancer ' 23. Wynder, E. L, Wright, G., and Lam, J.: A study of tobacco carcinogenesis. , , V. The role of pyrolysis. Cancer, 11:1140-1148, 1958. 3. Blacklock, J. W. S.: The production of lung tumors in rats by 3:4 benzpyrene, methylcholanthrene and the condensate from cigarette smoke. Brit. J. Cancer, 11:181-191, 1957. 4. Graham, E. A., Croninger, A. B., and Wynder, E. I..: Unpublished data. 5. Rockey, E. E., Kuschner, M., Kosak, A. I., and Mayer, E.: The effect of tobacco tar on the bronchial mucosa of dogs. Cancer, 11:466-472, 1958. 6. Wynder, E. L., Graham, E. A., and Croninger, A. B.: Experimental production of carcinoma with cigarette tar. Cancer Res., 13:855-864, 1953. 7. Wynder, S. L., Lupberger, A., and Grener, C.: Experimental production of cancer with cigare{te tar: Strain dil£erences. Brit. J. Cancer, 10:507, 1956. Role of Skin Neoplasia in Tobacco Carcinogenesis 71 8. Sugiura, K.: Experimental production of carcinoma in mice with cigarette smoke tar. Gann, 47:243, 1956. 9. Orris, L, Van Duuren, B., Kosak, A. I., Nelsom, N., and Schmitt, F.: Carcino- genicity of mouse skin and aromatic hydrocarbon content of cigarette smoke condensates. J. Nat. Cancer Inst., 21:557-561, 1958. 10. Engelbreth-Holm, J., and Ahlmann, J.: Production of carcinoma in ST/Eh mice with cigarette tar. Acta path. et microbiol. Scandinavica, 41:267-272, 1957 T200975

7 THE SIGNIFICANCE OF MOUSE SKIN TESTS OF CIGARETTE SMOKE CONDENSATE FaLO G. Bocu, Ph.D. and GroaGE E. MoonE SKIN studies have served two major functions in the experimental evaluation of tobacco smoke. First, they were undertaken to test the hypothesis that cigarette smoke could cause cancer ( l). Second, skin testing has been used in order to characterize the active agents and to assay the effectiveness of steps taken to roduce the potency of tobacco smoke. Studies with these aims are being conducted in several laboratories, and ex- amples of their application have been described by Wynder (2, 3). The results of these investigations, although not conclusive, suggest that identification of the mouse-skin carcinogens of ciga- rette smoke can be achieved within a reasonable future. As was expected, questions have been raised concerning the relevancy of mnuso skin tests as a measure of the hazard to the human lung. These questions fall into three major categories that may be summarized as follows: (1) Is the "tar" that smoking machines produce really the same as that in the cigarette smoke inhaled by man? (2) Are the positive results actually due to substances present in the tar, and not just to the method of skin painting? (3) Are mouse skin and human lung sufficiently similar in response that extrapolation from such experiments is justified? From the scientific point of view, it would be best to test a carcinogen in the tissue and species under consideration. Since this cannot be done, some objections to animal studies are un- 72 Signi fic¢nce o f Mouse Skin Tests 73 Fig. 1. Analytical Smoking Machine. Twenty cigarettes are smoked in ten minutes, each being puffed individually. The smoke is collected in coils immersed in a mixture of dry ice and ethanol. Provision can be made to measure the volume of any puff or to make a kymographic tracing of puff pressure vs. time. likely ever to be fully answerable. Nevertheless, such considera- tions are actually peripheral, and the main issues are well worth discussing. THE VALIDITY OF MACHINE SMOKING In the past seven years, the question of the validity of machine smoking has been answered by the weight of accumulated data. Smoking machines are designed to duplicate various standards of normal smoking behavior. In general, they fall into two groups: analytical and manifold. The analytical instruments smoke each cigarette individually in a highly controlled way. Where the smoke from a few cigarettes will suffice for the collection of data, the analytical machines give a reproducible and relatively accurate smoke. However, they T200976

74 Tobacco and Health Fig. 2. Production Smoking Machine. Six hundred cigarettes are smoked in ten minutes. The machine loads, lights, and puffs the cigarettes, and ejects the butts. Thirty cigarettes are puffed simultaneously. require a large amount of labor for a small yield of tar. The machine shown in Figure 1 can smoke about 500 cigarettes a day (4), and the amount of tar recovered in that time is about 10 gm. The manifold type of machine smokes a large number of cigarettes simnltaneously. In any group of cigarettes smoked at one time, some will be puffed more vigorously than the average and others less, because of differences in the air resistance of the individual cigarettes. This type of machine yields a large amount of smoke condensate for the labor expended, although the product is not as well controlled. Even so, the variations in puff volume, burning temperature, and butt length are certainly comparable to those occurring among human smokers. The machine in Figure 2° routinely smokes 20,000 cigarettes a day. Within each class of smoking machines, differences exist in the nature of the puff itself. For example, most machines use constant * Manufactured by Process & Instruments, Brooklyn, N. Y. I Signi ficance o f Mouse Skin Tests 75 suction during a puff. Others have been designed to start the puff with a vigorous suction that gradually falls off in a manner comparable to that of some human subjects (5). Additional factors that have been varied include the duration of the puff, the interval between puffs, and the method of lighting the ciga- rettes. Positive results have been obtained with a large variety of manifold machines using constant suction and variable suction and with puff frequencies ranging from -four per minute to one per minute (1,6,7,8,9,10,11). Puff duration has ranged from one second to five seconds. The cigarettes have been lit by gas flame and by hot wire. In this laboratory, tumors have developed in 25 per cent of mice painted with the products of an analytical smoking machine, in 27 per cent of mice painted with tar from a manifold machine using decreasing suction, and in 33 per cent of mice painted with tar from a manifold machine using constant suction. It is note- worthy that each of these machines, with the same duration and frequency of puff, produced tars of essentially the same potency. In contrast, Orris and his colleagues (9) found that the tumor- igenic potency of two different tars differed by a factor of about 2, and the concentration of polycyclic hydrocarbons differed by a factor of 3 or 4. The machines that produced the two tars differed, however, chiefly in frequency and duration of puff. Thus, the results appear to be affected more by differences in puff frequency and duration than by differences in the volume and suction characteristics. The important point, however, is that carcinogenic tars have been obtained with many machines, all of which were designed to duplicate smoking patterns of one or more segments of the smoking public. No unique character- istics of the machines have been observed that might be accused as the cause of the positive experimental observations. EFFECTS OF THE METHOD OF APPLICATION The question of effects due to the method of skin painting has also been answered during the past seven years. When several workers had trouble in duplicating certain of the results obtained by Wynder et al., it was natural that some question should be T200977

I 76 Tobacco and Health 01 G N ~ ~ ao~ 0 ~--I ~ ~ ~ ® ~ 304 \ ~ 3 ~ w V ~ ~ \I ~ ~ LL zo-~ 0 ~I \ \ E"+ i-- I ~ z w \ \ \ U \ I \ ~ Ix io w a \ ~ ~ ~ \ \ \ ~ \ \ II ~u TZ s YE sz 3M 3. Inflnciicr of tramna iipon carcinogenic action of cigarette smoke. (See trxt for details.) raised concerning the manipulations of the mice involved. Many of the experiments were conducted by application of tar solutions using a camel's hair brush. It could be argued that some effect of the brushing on the mouse hair might have contributed to the results, particularly in view of the toxic nature of the tar solu- tions. Since hair was clipped repeatedly from the experimental animals, there was an additional possibility that microscopic cuts contrihuted .fin•ther to the results. In order to test such hypotheses, we set up an experiment in which mouse skin was subjected to various degrees of trauma and was then painted with tobacco tar in acetone. In all cases, paint- ing was donc by allowing the tar solution to flow over the back, using a tuherculin syringe to deliver a measured dose without irritation. Lxperimental trauma ranged from shaving to abrasion with sandpaper. Figure 3 illustrates the results of this study. 'The micc in Group I were shaved with an electric clipper every second week, immediately before painting with crude tar. Those in Group II were handled the same as those in Group I, except that acetone .vas used as a solvent control. The animals in Group Signi ficance o f Mouse Skin Tests 77 III were shaved once a month, with a three-day period between shaving and the next painting. The mice in Group IV were shaved whenever the growth of hair made it appear desirable. (This procedure is the one customarily used in our laboratory. ) Group V consisted of animals that were "sandpapered" imme- diately before painting with tobacco tar, and Group VI served as their acetone controls. Finally, the mice in Groups VII and VIII were never shaved, but were painted with tar or acetone, respectively, without disturbing the skin in any way. With the exception of Group IV, each of the painted groups received the smoke from 6.7 cigarettes a day in two applications, morning and night. Group IV received only one painting daily, and thus only half of the full dose. In each painted group, 27 to 30 mice lived for a period of six months or more. No tumors were induced in the absence of tar painting, but significant numbers were developed in each tar-painted group. No significant difference in tumor rate occurred among the treated groups, even though they were subjected to a wide range of trauma. The general agreement between these results and those from experiments using a brush for tar application indicates that tumors do not arise from peculiarities of the method of application. Additional support for this conclusion is provided by the fact that comparable results have been obtained from other laboratories using various application techniques. THE RELEVANCE OF MOUSE SKIN STUDIES TO THE HUMAN LUNG The final question relates to the justification of studies using mouse skin when what we are really interested in is the human lung. This question cannot, of course, be answered directly by measuring the relative sensitivities of mouse skin and human lung to the carcinogenic materials present in cigarette smoke. Nevertheless, it can be answered indirectly, by comparing the sensitivity of the skin with that of other tissues in the same species, and then comparing the sensitivity of mouse tissues with that of tissues in other species. Most of the recent studies dealing with the relative sensitivities of various tissues to carcinogenic materials have involved only T200978

78 Tobacco and Health aR 50 I 0 40 / / ~ I I D 1 x 1 _ ~ 3 30 M 10 0 33 6.7 13.3 _X- 01WHOLE TAR+ -0- HEPTANE SOLUBLE •-+-• BENZENE ELUATE DOSE tCfGS. PER DAY) Fig. 4. Carcinogenic potency of cigarette tar fractions. (See text for details of preparation. ) single compounds or single types of compounds. Therefore, in order to decide what bearing these studies have on the question under consideration, it is necessary to identify the carcinogenic material present in cigarette smoke. Some evidence regarding its nature is already at hand. In a systematic shidy of fractions of smoke, Wynder and Wright (2) found the activity primarily in a carbon tetrachloride eluate of a neutral fraction which had been absorbed on a silica gel column. This is the fraction which contained the polycyclic hydrocarbons. Unfortmiatcly, different numbers of cigarettes were represented by the fractions applied to the different groups of mice, so that the yield of carcinogen in the active fraction cannot be estimated. Recently, Roe et al. (10) have suggested that most of the tumorigenic activity of cigarette smoke condensate is due to the promoting action of its phenols, acting in concert with initiators which may or may not be found in smoke. Their observations are Signi ficance o f Mouse Skin Tests 79 quite impressive, and have been substantiated by Wynder and HofFmann (3). In an attempt to assess the role of polycyclic hydrocarbons, we have compared the action of three fractions, each representing a different degree of purity of the polycyclics. Figure 4 illustrates the results of this study. Smoke condensate, washed from the condensers with acetone, was evaporated under reduced pressure, providing about 2.8 gm. of whole tar per carton. The more polar constituents were pre- cipitated by adding heptane to an acetone solution of the whole tar. This provided 0.7 gm. of heptane-soluble material per carton. The heptane-soluble fraction, when absorbed on silica gel and eluted, first with heptane and then with benzene, provided a benzene eluate containing 0.16 gm. per carton. An examination of this latter fraction showed only traces of phenolic and acidic . components, totaling about 15 micrograms for each cigarette smoked. This amounts to less than one percent of the acidic fraction that other workers (3,13) have reported to be present in the original tar. To obtain a measure of quantitation, each fraction was applied to mice in dosages differing by two-fold steps according to the number of cigarettes from which the fraction was obtained. In this way, each fraction could be compared with the others at the same dose level or at twice or one half that level. Use was made of one very high dose of the benzene eluate, in the expectation that this highly refined fraction might have very little activity. The data show an increase in tumorigenic effect, as a result of the removal of polar materials. It can be seen that the heptane- soluble fraction was more potent than the whole-tar fraction, but not twice as potent. The benzene eluate was about half as potent as the heptane-soluble fraction from which it was derived. This recovery is quite good, in view of the tendency of polycyclic hydrocarbons to undergo degradation when exposed to light. If phenolic materials were responsible for the original activity, the benzene eluate should have been almost;completely inactive. Thus, these data do not support the suggestion that phenols play an obligatory role in skin tumor formation, but rather indicate that much of the activity of the whole tar, if not most of it, is T200979 I

80 Tobacco and Health due to the benzene eluate. This eluate contains the bulk of the polycyclic hydrocarbons. If the polycyclic hydrocarbons in cigarette smoke are in fact responsible for some or all of the tumors that smoke condensate induces in mouse skin, we are in a position to examine the validity of tests made with mouse skin. Even at first glance, such tests seem eminently reasonable. After all, polycyclic hydrocarbons are known to produce tumors by direct contact with the target tissue. Complete justification of these tests, however, requires proof that the minimum carcinogenic dosage of polycyclic hydrocar- bons is not significantly lower for mouse skin than for the human lung. It has been thought that skin is particularly sensitive to lipid carcinogens because it is covered by a lipid film and because the hair shaft and sebaceous gland provide a path of entry and a reservoir for carcinogen storage in the vicinity of the target cells. Nevertheless, in studies using the duck foot, the hamster cheek pouch, and the mouse forestomach, many workers (14, 15, 16) have demonstrated that neither a lipid film nor a pilosebaceous apparatus is necessary for tumor formation by polycyclic hydro- carbons. As a matter of fact, several experiments (17,18,19,20,21) have demonstrated that bronchial epithelium is also sensitive if the carcinogen is applied in an appropriate manner. In our laboratory (22), a group of 47 mice were painted once with dimethylbenzanthracene. Forestomach tumors appeared in 13, but only eight developed skin ttunors, in spite of the fact that the carcinogen was applied to the skin and reached the stomach only through licking. The induction of neoplasms by direct injec- tion of carcinogens into subcutaneous tissues, muscle, the stomach, and the brain, has also been demonstrated (23). From experiments like these, we may conclude that the skin is not a hypersensitive tissue which might provide falsely positive results. A more important aspect of this general problem is the possi- bility that carcinogens are handled differently by the cells of different species. Species variations in skin carcinogenesis have been known for some time. Among laboratory animals, the rat, the guinea pig, and perhaps the monkey are particularly resistant to hydrocarbons, (24,25,26), whereas the mouse, the hamster, and Significance of Mouse Skin Tests 81 the rabbit are sensitive. In the case of man, occupational contact with sources of polycyclic hydrocarbons seems almost certainly a cause of skin cancer (27). In .contrast, coal tar as used in dermatologic practice does not appear to give rise to tumors. The reason for this paradox is, at present, unexplained. Differences in skin sensitivity may be due to differences in biochemical mechanisms. In this case, species specificity of skin cells may reflect specificity of other cell types as well. One example of biochemical specificity appears to be the azo dye carcinogenesis, since rats are particularly sensitive and show a marked tendency for the formation of protein carcinogen com- plexes in the liver. In contrast, species differences may also be due to anatomical factors which limit the amounts of carcinogen that reach the target cell. In the latter case, differences in skin sensitivity would not indicate parallel responses by other tissues. . Anatomical variation in rodent skin is well known. Mouse skin is very thin. When the hair is not growing, the epidermis con- sists of only two cell layers. In contrast, guinea pig skin is quite thick and has a multi-celled epidermis. Recent work in our labor- atory has been directed toward an analysis of the relationship of these anatomical differences to carcinogenesis. Although this work is still in the early stages of development, the preliminary results are pertinent to the present discussion. When the skin of various rodents is examined after "painting," differences in carcinogen concentration are found. The data are shown in Figure 5. In this experiment, * the backs of the animals were treated with a 1 per cent solution of the hydrocarbon in a solvent consisting of 1 per cent mineral oil in benzene. After two hours, when the skin concentration of the carcinogen was at a peak, the treated skin was carefully washed with benzene, and was analyzed for polycyclic hydrocarbons. Although the results depended upon the compound tested, the over-all pattern is quite apparent. Generally much less of the carcinogen was found in the skin of the guinea pig than in that of more sensitive species. For example, only about a fifth as much carcinogen was found in guinea pig skin as in mouse skin. * Complete details will be published elsewhere. T200980

, 82 Tobacco and Health 25-1 12~ ® RABBIT ® MOUSE ~ GUINEA PIG I~ HAMSTER IO~ 7 -Z 9 v+ Z Y N ® 1 ~. A 8 16 6 15 88 16 6 15 DMBA BP 44 16 6 15 MC Fig. 5. Penctr;ition of carcinogens into skin of various rodents. The numbers of animals used are given at the bottoms of the respective bars. DMBA = 7,12-dimethylbenz(a)anthracene. BP = benzo(a)pyrene. MC = 3-methvlcholanthrene. This situation is further complicated by the nature of hair growth in rodents. Mice, hamsters, and rabbits are characterized by cyclic waves of hair growth, and our studies were conducted when the hair of these rodents was resting. Guinea pigs have a mosaic pattern of hair growth, so that growing hair can always be found scattered over the entire back. In mice, growing hair is associated with a rapid clearance of the carcinogens from the skin (28). This is demonstrated by the data of Figure 6. In this experiment, hair growth was stimulated by plucking the club hairs of a previous cycle. A marked effect of hair growth upon clearance of the hydrocarbon is apparent. If the effective dose of carcinogen is considered to be the 2 Signi fccance o f Mouse Skin Tests 83 N ~ -X- RESTING HAIR N ~ '4 day' GROWING HAIR ~ ~ , O ~ z ~ ~ ~ • H k ~ _ ~ ~X .% . ~ o ~- 48 T1tNE AFTER PWNTPNG (Ixa) Fig. 6. Effect of hair growth on clearance of benzo(a)pyrene from "Painted" skin of mice. "Resting Hair" was used 22 days after plucking. concentration times the duration of exposure, the relative doses are proportional to the areas under the curves of Figure 6. Mea- sured in this way, the integrated dose for skin with resting hair is nearly three times as great as that for skin with growing hair. A comparable analysis of the persistence of carcinogen in guinea pig skin has not been completed at this time, but should prove of considerable interest. It is clear that the responses of different species can be ade- quately compared only if the tests are run under conditions such that the target cells receive equal doses of carcinogen. No such studies have yet been made, and a valid comparison of the responses of the target cells for different species is therefore impossible at this time. That being the case, the mouse skin data cannot be rejected merely because of differences in the responses of different species. On the contrary, in view of the history of occupational coal-tar tumors, continued use of mouse skin for assays of human carcinogens seems appropriate. T200981

84 Tobacco ancl Health Signi fccanee o f Mouse Skin Tests 85 CONCLUSIONS In studies of cigarette smoke carcinogenesis, it would appear en r'I 8. Gu6rin, M., and Cuzin, J. L.: Action Carcinog&e du Goudron de Fumee de Cigarette sur la Peau de Souris. Bull. Assoc. Franc. pour 1'etude du Cancer, 44:387-408 1957 that the following facts have been established: N , . 9. Neukomm, S.: Recherches Expi~rimen£ales sur le Pouvoir Canc6ri&e de Ia 1. The induction of mouse skin tumors is not due to artifacts ~ Fum6e du Tabac et d'autres Polluants de 1'atmosph&e. Oncologia 10:137-156 of smoke collection. ~ , , 1957. 2. Neither are the results due to the method of applying the ® 10. Orris, L., Van Duuren, B. L., Kosak, A. I., Nelson, N., and Schmitt, F. L: smoke concentrate. 3. Fractions of cigarette smoke obtained by procedures that concentrate polycyclic hydrocarbons are carcinogenic to mouse ~ The Carcinogenicity for Mouse Skin and the Aromatic Hydrocarbon Content of Cigarette-Smoke Condensates. 1. Nat. Cancer Inst., 21:557-561, 1958. 11. Roe, F. J. C., Salaman, M. J., and Cohen, J.: Incomplete Carcinogens in Cigarette Smoke Condensate: Tumour-Promotion by a Phenolic Fraction. Brit. J. Cancer 13:623-633 1959 skin. It is not es-Cablished that these fractions are carcinogenic for , , . 12. Sugiura, K.: Experimental Production on Carcinoma in Mice with Cigarette Smoke Tar. Gann, 47:243-24'4, 1956. the humai bronchial epithelium. Nevertheless, pure polycyclic hydrocarbons can produce tumors in rodent bronchial epithelium. There is evidence that, under some conditions, polycyclic hydro- carbons prodt:ce tumors in human skin, and there is no unequiv- ocal evidence that, given satisfactory exposure, the cells of any species are inherently resistant to these agents. In summary, studies using mouse skin seem to be worth con- tinucd attcntion. Using these tests, considerable progress has been made in the seven years since the report of Wynder et al. We may reasonably expect this progress to continue. 13. Burgan, J. G.: The Production of Cigarette Smoke Condensate. Appendix to ref. .# 10. 14. Rigdon, R. H.: Tumors Induced in Skin without Follicles An Experimental Study in the Duck. Cancer Re.s.,16:804-807, 1956. 15. Salley, J. J.: Histologic Changes in the Hamster Cheek Pouch During Early Hydrocarbon Carcinogenesis. J. Dental Res., 36:48-55, 1957. 16. Berenblum, I., and Haran, N.: The Influence of Dose of Carcinogen, Emptiness of Stomach, and Other Factors on Tumor Induction in the Forestomach of the Mouse. Cancer, Res., 15:504-509, 1955. 17. Andervont, H. B.: Pulmonary Tumors in Mice. IV. Lung Tumors Induced by Subcutaneous Injection of 1:2:5:6-Dibenzanthracene in Different Media and by Its Direct Contact with Lung Tissues. Pub. Health Rep., 52:1584-158J, 1937. REFERENCES 1. Wynder, E. L., Graham, E. A., and Croninger, A. B.: Experimental Production 18. Blacklock, J. W. S.: The Production of Lung Tumours in Rats by 3:4 Benz- pyrene, Methylcholanthrene and the Condensate from Cigarette Smoke. Brit. J. Cancer, 11:181-191, 1957. of Carcinoma with Cigarette Tar. Cancer Rcs., 13:855-865, 1953. 2. Wynder, E. L., and Wright, G.: A Study of Tobacco Carcinogenesis. I. The Primary Frat•ti ins. Cancer, 10:255-271, 1957. 3. Wynder, E. I.,, and Hoffmann, D.: Studies in Tobacco Carcinogenesis. Proc. Am. Assnc. for Cancer Res., 3:164, 1960. 4. Schur, M. C)., and Rickards, J.: Design and Operation of a Multiple Cigarette Smoking Machine. Ecusta Paper Corp., Pisgah Forest, N. C. 5. Bradford, J. A., IIarlan, W. R., and Hamner, H. R.: Nature of Cigarette Smoke. Technic cif Experimental Smoking. Ind. Eng. Chem., 28:836-839, 1936. 19. Della Porta, G., Kolb, I.., and Shubik, P.: Induction of Tracheobronchlal Carci- nomas in the Syrian Golden Hamster. Cancer Res., 18: 592-597, 1958. 20. Kuschner, M., Laskin, S., Cristofano, E., and Nelson, N.: Experimental Carcinoma of the Lung. Proc. Third National Cancer Conf., pp. 485-495, 1956. 21. Niskanen, K, 0.: Observations on Metaplasia of the Bronchial Epithelium and Its Relation to Carcinoma of the Lung; Pathoanatomical and Experimental Researches. Acta Path. et ll4icroblot. Scandinao., Suppl., 80:1-180, 1949. 22. Bock, F. G., and King, D. W.: A Study of the Sensitivity of the Mouse Fore- stomach Toward Certain Polycyclic Hydrocarbons. J. Nat. Cancer Inst., 23: 833-839 1959. 6. Bock, F. G., and Moore, G. F: Carcinogenic Activity of Cigarette-Smoke Con- dcnsate. 1. Effcct of Trauma and Remote X Irradiation. J. Nat. Cancer Inst., 22:401-411, 1959. , 23. Hartwell, J. L: Survey of Compounds which have been Tested for Carcinogenic Activity, 2nd ed., (U. S. Gov. Printing Office, Washington, D. C., 1951). 7. Engelhreth-Holm, J., and Ahlmann, J.: Production of Carcinoma in ST/Eh Mice with Cigarette Tar. Acta Path. et 114icrobiol. Scand., 41:267-272, 1957. Shubik, P., and Hartwell, J. L: Survey of Compounds which have been Tested for Carcinogenic Activity, Supp. 1(U. S. Gov. Printing Office, Washington, D. C., 1957). T200982 11

86 Tobacco and Health 24. Berenblum, I.: The Carcinogenic Action of 9,10-Dimethyl-1,2-Benzanthracene on the Skin and Subcutaneous Tissues of the Mouse, Rabbit, Rat, and Guinea Pig. 1. Nat. Cancer Inst., 10:167-174, 1949. 25. Pfeiffer, C. A., and Allen, E.: Attempts to Produce Cancer in Rhesus Monkeys with Carcinogenic Hydrocarbons and Estrogens. Cancer Res., 8:97-127, 1948. 26. Sugiura, K., Smith, W. E., and Sunderland, D. A.: Experimental Production of Carcinoma in Rhesus Monkeys. Cancer Res., 16:951-955, 1956. 27. Hueper, W. C.: Environmental and Occupational Cancer. Pub. Health Rep., Supp. 209, pp. 1-69, 1948. 28. Berenblum, I., Haran-Ghera, N., and Trainin, N.: An Experimental analysis of the Hair Cycle Effect in Mouse Skin Carcinogenesis. Brit. J. Cancer, 12:402- 413, 1958. ACKNOWLEDGMENT The authors wish to acknowledge the invaluable assistance of Miss Mary Burnham, Mr. Paul Clark, and Miss Shirley Crouch. ~ r-04 N ~ ~ ~ ~ ~ 8 ~ H A CORRELATED HISTOLOGICAL, CYTOLOGICAL AND CYTOCHEMICAL STUDY OF THE MAJOR BRONCHI FROM MICE EXPOSED TO CIGARETTE SMOKE CECILIE LEUCHTENBERGER, Ph.D. and RUDOLF LEUCHTENBERGER, M.D. I INTRODUCTION IN recent years the question of a causal relationship between cigarette smoking and human lung cancer has become a widely discussed issue. Although statistical, clinical and histo- pathological data on humans, as well as experimental observations on animals, have been interpreted by some investigators in favor of such a relationship, the topic is still controversial, and addi- tional research is urgently needed. The experimental approach to this problem is by no means an easy task, not only because our present-day knowledge of carcinogenesis in general is so scanty, in other words, because we have so little information concerning factors and the pathway by which tumors are produced, but particularly because cigarette smoking is an exclusively human habit which has not been dupli- cated on a large scale in animals, although, for show purposes, a few chimpanzees have been taught to smoke. The most desirable experimental approach would certainly be if one could develop the smoking habit in anthropoid apes and then study, under controlled conditions, the sequence of events produced in the tracheobronchial tree of the animals. Needless to say, the setting This study was supported by a grant from the Tobacco Industry Research Com- mittee. 87 T200983

88 Tobacco and Health up of such a study, as important as it would be, is extremely difficult. Large numbers of anthropoid apes and proper housing facilities, which would be needed to permit studies over the whole life span of the animals, would involve considerable expense. Furthermore, one would have to search for specially trained personnel able to handle the anthropoids and teach them to smoke. Since, with the facilities at our disposal, such a project was out of the question, it was felt that some fruitful information might also be obtained from studies on smaller animals exposed to cigarette smoke, if provision wete made that the animals would inhale the smoke. In view of the fact that the majority of human lung cancers (nearly 90%) take their origin in one of the major bronchi of the tracheobronchial tree, with only secondary invasion to the lung, and since the columnar epithelial cells which line the lumina of the bronchi are more directly exposed to inhaled cigarette smoke than are the f arther distant epithelial cells of bronchioles and alveoli in the hing itself, studies of the traeheobronchial tree, and particularly of the major bronchi, from animals exposed to cigarette smoke seemed especially pertinent. Surprisingly enough, to the best of our knowledge, no systematic studies on the tracheobronchial tree fi•om cigarette-smoke-exposed animals had been pnblished prior to our own studies, which we started several years ago. Our work, as presented here, is a common effort in which, be- sides Dr. Rudolf Leuchtenberger as pathologist, several young associates, namely Drs. Doolin and Zebrun and Miss Shaffer, participated in various phases of the work (1, 2, 3, 4). 11 METHODS AND ANIMAL EXPERIMENTS In order to cover several aspects of this complex problem, we attempted to study the sequential changes in the major bronchi of mice not only from a morphological but also from a chemical point of view. Since the cell occupies a fundamental position in the living organism and is probably one of the primary targets hit by injurious agents, the seduence of events occurring in cells was assessed from a morphological and from a chemical point of view, with special emphasis on the behavior of the DNA proteins. The significance of these intracellular substances for normal cell Correlated Study of Bronchi from Exposed Mice 89 life, growth and genetic continuity is now fully recognized and hardly needs to be stressed any more. A quantitative study of the DNA proteins seemed pertinent, since tumor cells are considered to be associated with abnormal DNA metabolism. A. Cytochemical Methods For the exploration of the DNA proteins we used the special quantitative cytoehemical techniques of microspectrophotometry and interference microscopy. Although we. cannot discuss any methodical details, we should like to say a few words about these methods, especially about microspectrophotometry, because these techniques are relatively recent and have opened completely new pathways for the study of the nucleoproteins. The development of these methods is based largely on work by Caspersson (5), who demonstrated that, by extending the optical properties of the microscope into the analytical sphere, the microscope can be used not only as the conventional tool for the morphological study of tissues and cells, but simultaneously as an instrument for the chemical analysis of the cell structure. The basic principle of microspectrophotometry is actually very simple and closely resembles that so frequently applied in analytical chemistry. As in the photometric chemical analysis of solutions, the amount of light absorbed by a cell structure at a specific wavelength is used as a basis for the qualitative and quantitative analysis of the in- tracellular substances. Therefore a microspectrophotometer such as the one we used in our own laboratories is actually nothing else than a microscope combined with a photometric device which permits light absorption measurements in single cells. The unique feature of this method lies not only in the possibility that, for example, a DNA quantity as small as 2.5x10'® mg. can be determined in a single nucleus, but that this analysis can be done in microscopic sections in situ. In other words, since the archi- tecture of the cells and their relationship within a tissue are preserved, a direct comparison between cell morphology and DNA content from cell to cell can be made directly under the microscope. Thus the DNA behavior can be explored under normal and abnormal conditions during such dynamic processes as growth, mitosis and meiosis (6, 7). The conventional biochemical analysis is obviously not suitable T200984

90 Tobacco ancl Health for the exploration of these problems, because the biochemical DNA determinations can be carried out only on relatively large cell populations and after the cells and tissues are destroyed and extracted. The DNA value, therefore, is, of course, only an average value. While such computed average value may be representative for the cell if a cell suspension is analyzed in which each cell has the same uniform DNA content, it is actually an erroneous value if suspensions of cells with varying DNA content, such as is the case during cell division and, particularly, under abnormal conditions, are analyzed. The possibility of a direct correlation between physiological stage of cell, it.% microscopical appearance and its chemical com- position assumes particular significance when one takes into consideration that, in the final analysis, a change in physiological stage and morphology of a cell is more or less the expression of preceding chetriical alterations. The detection of intracellular changes in the DNA proteins before structural alterations mani- fe.st themselves is indeed one of the intriguing advantages of microspectrophotometry and interference microscopy, and there- fore we felt that a simultaneous study of morphological cell strue- ture and of these important intracellular chemical constituents might help greatly in assessing the biological effect of cigarette smoke. B. Animal Experiments To carry out the investigation discussed in this report, an attempt was made to standardize as far as possible the conditions iinder which the experiments were conducted. The 600 mice which we used for this study were CF, female mice three months of age. Upon arrival, the mice were placed in cages located in a specially designed room which was used exclusively for these smoking experiments. To insure that the mice were exposed to smoke only when they were in the smoking chamber and that the control mice were not exposed to smoke at all, no smoking was allowed in this room. For each experimental mouse a control mouse of the s~ime age and weight was used. In order to study the effect of cigarette smoke, mice were exposed to smoke in a chamber which is shown in Figure 1. This T200985

92 Tobacco and Health apparatus permits the housing of both the exposed and the con- trol mice at the same time under the same standardized condi- tions. The only exception is, of course, that the compartment containing the control mice receives no smoke, while, as can be seen, the compartment containing the experimental mice receives smoke from cigarettes mounted on a mechanism which lights them at required intervals. Mice were exposed to cigarette smoke every day, with the exception of weekends and holidays. If a mouse appeared sick, and especially if weight loss was noted, it obtained a rest from smoking. After the daily smoking period, the mice were taken from the smoking chamber and placed in their cages, with food and water, wlu•re they stayed until the next smoking period. Regular weighiTig of the mice showed that mice exposed to cigarette smoke frequently either lost weight or did not gain any, as compared to the nonexposed control mice. At certan periods after exposure to cigarette smoke, experi- mental mice and their controls were sacrificed. Blocks taken from the tracheohronchial tree and lung tissues were immediately fixed and proccssed under the same standardized conditions. III RESULTS A. Histologicaj and Cytological Changes Turning now to the experimental observations, it should be pointed out that one of the most striking results of our study is the variability in response of the major bronchi from mice after exposure to cigarettce smoke. The implication of this differential response, whicli was observed regardless of whether mice were exposed to cigarette smoke for a short or a long period, will be discussed later. Suffice it to say here that the microscopic findings in the bronchi can be divided into three groups: no significant changes, mild rhanges and severe changes. Table 1 ilh:strates the variability in response of bronchi, demonstrating the three findings in mice after exposure to cigarette smoke: Group I Normal-undistinguishable from the control. Correlated Study of Bronchi from Exposed Mice 93 Table 1 SIJMMAR`r OF cQRRELATEU t-0ISra'ATFfOLOGM AJWO crTOCFEMM ALTERATIONS IN TNE LEFT rAAJOR ®ROrM FROM MiucE EXPOSED TO CIGARETTE SM0KE 111STOPATFqLOGICAL AND CY1DL4CICAL ALTEWJITIONSI CYiOCFEMlCJiL ALTENATMt IN NUCLEI PFqJFEPATME ATY'RSM UN(:NEASE INCREASE INCfdEASE I~NCREASE BRONCHIT5 EPNNEUAI M9TOSIS OF IN IN IN IN l7iAMGEB CELLS NIXOLEIJi SIZE PNOPEINS ONY MASS UNA EROR I NOT FNESENT NOT NOT NOT Sp.ETIMES SMETkIES Sq.ETOMES NpT OR SIKiNT FAESENT PRESENT PRESENT PRESENT pqE$ENT PRESENT PFF-SENT GROUP 1I PRESENT MR.D ~ FPoEIX1ENf FREpUENi FNEQLENr NOT PRESENT ~ ~~NT OR >LXHT GI70UP III PRESENT MAR1fEp PNESENf FREQSIENT FFEpi1EMT FFdEOUENT FREQUENT FREQUENT If ONE OR MORE OF THESE FEATURES MAY BE PRESENT IN pFFERENT AREAg OF THE SAME BRONCNUS Group II Mild bronchitis and peribronchitis; epithelium swol- Group III len. Severe bronchitis, peribronchitis, association with marked proliferation of lining epithelium. There appears to be a stepwise sequence of lesions in the epithelium, from a normal resting one to a swollen one to a proliferating one and, finally, to an atypical proliferating one. It should be mentioned that the changes are often multifocal along the bronchi, and that, although bronchitis was sometimes noted without proliferation of the epithelium, there was never any proliferation of the epithelium without bronchitis. Figure 2 represents the characteristic appearance of a cross- section of a bronchus from a nonexposed control mouse or from an exposed mouse-undistinguishable from a control. Figures 3a, 3b and 3c display severe changes such as bronchitis, peribronchitis, associated with atypical proliferation of lining epithelium after exposure to cigarette smoke. The latter changes are very similar to those which Auerbach and coworkers (8) reported in bronchi of humans who are heavy cigarette smokers. This similarity certainly justified a cautious optimism as to the validity of our experimental approach, because it appears that exposure of animals to cigarette smoke is a useful T200986

Fig. 2. Cross section of the left major bronchus from control mouse. Note the regularity of the lining epithelium and the cartilage ( C) . H. and E. X 100 Fig. 3a. Cross section of the left major bronchus from mouse exposed to smoke of 132 cigarettes within an experimental period of 38 days. Note the severe bronchitis and peri- bronchitis, the many-layered lining epithelium, and the pus in the lumen (P). H. and E. X 100 TIMN 011621 8 - m vt

96 Tobacco and Health Fig. 3c. Sama as Tigmr. 3b at a higher magnification. Note marked dysplasia of thc rhithelium with loss of polarity of cells and disparity in the miclcar and cellular size. H. and E. X 970 Correlated Study of Bronchi from Exposed Mice 97 approach, in spite of the realization that it is different from the actual process of cigarette smoking. The fate of this dysplasia in the bronchial epithelium of mice seemed, of course, of particular interest, even regardless of its possible causation by exposure to cigarette smoke. The obvious questions which had to be answered are: Is this dysplasia a pre- cancerous state; in other words, will it progress to invasive Carcinoma, especially when the exposure to cigarette smoke is continued during the whole life span? Will the alterations remain stationary, or will they revert back to normal, especially when exposure is stopped. Although the extensive study which we have started along these lines is still under way, we can already state that in none of the mice which have been exposed to smoke from up to 1,500 cigarettes over a period of nearly two years has an invasive Carcinoma been found. On the other hand, micee which had been exposed to cigarette smoke for three months or more, and where exposure to smoke has then been stopped for more than three months, have frequently shown nearly normal bronchial epithelium, strongly suggesting reversibility of the changes, at least in some of the cases observed so far. Of course, we are aware of the pitfalls encountered when interpreting data indicating varying responses. Nevertheless, a decrease in incidence of the lesions is also suggested by our microscopical studies, in which we attempted to assess the se- quence of this reversibility and where we have found that there seems to be a gradual desquamation of the atypical epitheliiun, followed by restoration of the normal one. B. Cytochemical Changes Turning now to the cytochemical studies, the characteristic results concerning the behavior of the DNA proteins are shown in Figure 4. These data are based on microspectrophotometric analysis of approximately 20,000 individual nuclei from 200 mice. On the left the DNA data are grouped, while the protein data are on the right. The correlated cytochemical studies of DNA and proteins in individual cells of bronchial epithelium of mice similarly did not reveal significant differences between mice ex- posed for two months and those exposed to cigarette smoke for T200988

98 Tobacco and Health THE DESOXYNUCLEOPROTEIN CONTENT IN INDIVIDUAL NUCLEI OF BRONCHIAL EPITHELIAL CELLS FROM CONTROL MICE AND MICE AFTER EXPOSURE TO CIGARETTE SMOKE (DATA BASED ON MICROSPECTROPHOTOMETRIC ANALYSES OF APPROX 20.000 NUCLEI FROM 200 MICE ) 60 40 20 Ro 60 40 20 CONTROL MICE, EPITHELIUM EXPOSED MICE, EPITHELIUM NO SIGNIFICANT HISTOPATHOLOGICAL CHANGES EXPOSED MICE, EPITHELIUM PROLIFERATION AND ATYPISM ~ Im __W - EXPOSED MICE, EPITHELIUM NORMAL EXPOSURE STOPPED FOR OVER 3 MONTHS 4 8 12 16 0 2 4 6 8 10 12 AMOUNT OF DNA IN IO'"MGM AMOUNT OF,PROTEIN IN ARBITRARY UNITS Fig. 4. The clcsoxynucle'oprotein content in individual nuclei of bronchial epithelial cells from control mice and mice after exposure to cigarette smoke. (Data h.ised on microspectrophotometric analyses of approximateh20,000 nuclei from 200 mice.) Correlated Study of Bronchi from Exposed Mice 99 a year or longer. Regardless of dose or time of exposure, three types of cytochemical changes are found: Some mice which showed no significant histopathological lesions contained a normal protein and DNA content, indistinguishable from that of nonexposed controls (upper histogram); some mice, in spite of the absence of histopathological changes, revealed an increase in proteins (second histogram on right ) but maintained a normal DNA content (second histogram on left ), and, in the third type of change, which was always present whenever epithelial prolifer- ations was observed (third histogram ), both DNA and protein content were increased in the nuclei of the bronchial epithelium. Cessation of exposure to cigarette smoke (bottom histogram ) was accompanied by DNA and protein values similar to those of the nonexposed controls. Sometimes, after cessation of exposure, DNA and protein values were even slightly lower than those in nuclei from unexposed controls, but whether the difference is of biological significance or not cannot be answered at present. One may speculate that the lesser values are an expression of under- function of such epithelial cells. IV DISCUSSION OF THE RESULTS On the basis of the data presented so far it thus appears that exposure to cigarette smoke will evoke in the bronchi of some mice alterations the essential features of which are here sum- marized briefly: 1. Bronchitis associated with marked proliferation of epithe- hum Gradual change of normal epithelium to hyp'erplastic, dysplastic epithelium Multifocal changes No invasive Carcinoma 2. Early increase in nuclear proteins (before microscopical alterations ) followed by gradual increase in the DNA content 3. After exposure to cigarette smoke is stopp'ed (3 months or more ) microscopic and cyto'chemical findings normal, suggesting reversibility of bronchitis and dysplasia Number of mice: 200 Number of cells analyzed: 20,000 T200989

100 Tobacco and Health Table 2 FREQUENCY OF HPSTOPATHDLOG6CAL FINDINGS IN LEFT MAJOR BRONCHI OF MICE AFTER EXPOSURE TO SMOKE FROM VARYING NUMBERS OF CIGARETTES AND FOR VARYING LENGTHS OF TIME LENGTH OF NUMBER OF NUMBER NUMBER OF MICE IN RELATION TO BRONCHIAL FNlDIRlGS EXPOSURE TO SMOKE IN MONTHS CIGARETTES OF MICE HO 81SNIFICANT C,{AH,pES N6LD D0.GRCHfTlO W7THOUT ~'~THMeLDEPttHELIAL rHULOFEHATibN t~EV'EH£ MPONCHITIO AND PET/lNDNCHITlf 1Y7TH ATYPICAL EMTHELIAL VROLAFE/4ATIaN I- 3 100 - 200 36 20 ' .. . _ 9 - 7 4-6 200-500 36 19 10 7 9- 23 600 - 1600 34 1 9 7 8 Although, on the evidence of these findings, one may be tempted to conclude that there is a direct cause-effect relationship between cigarette smoke and the epithelial lesions, there are mainly two findings in our study which impose caution against drawing such a conclusion. The first one concerns the striking difference of response en- countered from mouse to mouse, which, as mentioned previously, occurred regardless of whether mice were exposed to relatively low (loses of cigarette smoke for a brief period or to a relatively high dose of cigarette smoke for long periods. In each of the groups, over 50 per cent of the mice did not show any bronchial lesions, while approximately 25 per cent showed mild and 20 per cent dysplastic lesions. The second finding concerns the absence of a relationship be- tween dose and length of exposure to cigarette smoke and frequency and severity of lesions produced, an example of which is given in Table 2. It is evident that, in spite of the considerable increase in length of exposure and number of cigarettes, the relationship between mice showing no significant changes and mice showing mild and severe changes in the bronchial tree did not alter apprcciably, whereby it is indeed astonishing that, among the last group, no invasive bronchogenic Carcinoma has been observed, although these mice had been exposed to Correlated Stud y o f Bronchi f rom Exposed Mice 101 enormous quantities of cigarette smoke, and, whenever possible, for nearly their whole life span. On the other hand, the difference in response of mice to cigarette smoke, including the considerable number of completely refractory cases, is not so surprising, but is in accordance with experience obtained in humans. Examination of the tracheo- bronchial tree of heavy smokers, as carried out by Auerbach and his associates (8), demonstrated also a wide range of alterations, including cases without lesions. Furthermore, statistical data on the relationship between bronchogenic Carcinoma and cigarette smoking have shown that, even among heavy smokers, only a relatively small percentage develop bronchogenic Carcinoma. The wide spectrum of findings and the independence of dose and duration of exposure to cigarette smoke would seem to detract from the importance of cigarette smoke as an injurious agent invariably affecting the major bronchi and would, rather, point to contributing factors characteristic for the individual animal. Thus it appears that, in cases in which a harmful effect on the tracheobronchial tree has been observed after cigarette smoke, the cigarette smoke may not be solely responsible,.but that there may be other elements within the host which may render the tissue susceptible; in other words, may create a special condition- ing for the injurious action of the cigarette smoke. While it is impossible, at present, to relate any particular factor or factors of the host to the resistance or susceptibility of the tracheobronchial tree to cigarette smoke, the investigation of agents possibly carried by the host which may contribute to the injury of the bronchi would seem of great interest. Among the many possible host factors, viruses deserve special consideration -the more so since they occur with a certain frequency in mice, sometimes in a latent form, and are capable of producing respira- tory lesions. In view of the increasing recognition of the dual role of viruses as either cytolytic agents or stimulators of cell proliferation, which may lead to tumor formation, the investigation of viruses in combination with exposure to cigarette smoke seems to us particularly worthwhile. There are two main lines which we are T200990

102 Tobacco and Health FREQUENCY OF PULMONARY ADENOMATOUS TUMORS AT DIFFERENT AGES IN 166 CF, FEMALE CONTROL MICE AND IN 231 CF, FEMALE MICE AFTER EXPOSURE TO CIGARETTE SMOKE 60 ; 52 Z 44 >- 36 U Z W M a W D: LL 281 / ~ yti 20 12 4 ~ J 45 39 M • MK'E EXPOSL~O 7r1 CICARErrE SMqrE O• COIVrR17LS (MOrEkPOSEDI AA/MBERS • Nt1M®E/r OF MICE ,a -.- 1-7 8-10 11-14 15-18 19-28 AGE OF MICE IN MONTHS Fig. 5. Frequency of pulmonary adenomatous tumors at different ages in 166 CF, female control mice and in 231 CFz female mice after exposure to cigarette smoke. investigating at present. The first is concerned with the detection of latent viruses in mice prior to exposure to cigarette smoke. It will be of interest to learn whether mice carrying a latent virus will exhibit a different response in the bronchi from that of mice free of virus. The second line is concerned with the effect of exposure to vit•us in addition to the exposure to cigarette smoke. If our concept is correct that viruses may act as co-factors in the production of the bronchial lesions, then mice which carry latent viruses prior to the exposure to smoke or which are infected with viruses should show frequent and perhaps more severe lesions, while mice free of virus should either be refractory or show mild lesions. Although tbis is about all that we have to say on the biological Correlated Study of Bronchi from Exposed Mice 103 effect of cigarette smoke on the bronchi of mice, we should like to discuss here data (Fig. 5) which are concerned with the effect of cigarette smoke in the incidence of spontaneously occurring little nodules in lungs of mice: so-called adenomatous tumors, which are quite different from bronchogenic Carcinoma. Several years ago, Essenberg (9), using relatively small groups of mice, reported that there was a striking increase in these adenomatous tumors in mice after exposure to cigarette smoke. However, as can be seen from this graph, under our experimental conditions, controls and exposed mice reveal a very similar frequency, which increases in both groups with age. They differ at only one point, which perhaps suggests a slightly earlier occurrence in the ex- posed mice, but actually the number of mice at this point is too small to draw a definite conclusion. In closing, we should like to say that we realize that our. studies cover only a small aspect of the many facets involved in the problem of the biological effect of cigarette smoke. Never- theless, the resemblance of bronchial lesions of mice exposed to cigarette smoke to those observed in human cigarette smokers justifies a cautious optimism as to the validity of our experimental approach in probing further into the complex problem of the relationship between cigarette smoking, environmental f actors, host factors and development of lung cancer. REFERENCES 1. Leuchtenberger, C., Leuchtenberger, R., and Doolin, P. F.: A correlated his- tological, cytological, and cytochemical study of the tracheobronchial tree and lungs of mice exposed to cigarette smoke. 1. Bronchitis with atypical epithelial changes in mice exposed to cigarette smoke. Cancer, 11:490-506, 1958. 2. Leuchtenberger, C., Leuchtenberger, R, Zebrun, W., and Shaffer, P.: A correlated histological, cytological, and cytochemical study, of the sequence of events in the bronchial epithelium from mice exposed to cigarette smoke. Acta Union Internationale contre le. Cancer, 15:632-638, 1959. 3. Leuchtenberger, C., Leuchtenberger, R., Zebrun, W., and Shaffer, P.: A correlated histological, cytological, and cytochemical study of the tracheobronchial tree and lungs of mice exposed to cigarette smoke. IL Varying responses of bronchi to cigarette smoke, absence of bronchogenic Carcinoma after prolonged ex- posure, and disappearance of bronchial lesions after cessation of exposure. Cancer, 13:721-732, 1960. T200991

104 Tobacco and Health 4. Leuchtenberger, R., Leuchtenberger, C., Zebrun, W., and Shaffer, P.: A correlated histological, cytological, and cytochemical study of the tracheobronchial tree and lungs of mice exposed to cigarette smoke. III. Unaltered incidence of grossly visiblre adenomatous lung tumors in female CF, mice after prolonged exposure to c:garette smoke. (In press for publication in Cancer, 1960.) 5. Caspersson, T.: Cell Growth and Cell Function; a Cytochemical Study. W. Norton and Co., New York, 1950. 6. Leuchtenberger, C.: Quantitative determination of DNA in cells by Feulgen microspectrophotometry. In: Standard Cytochemical Methods. J. F. Danielli, ed., Cambridge University Press, London, 1957. 7. Leuchtenberger, C.: Quantitative cytochemistry (microspectrophotometry), a fruitful approach to the study of disease. J. Mount Sinai Hosp., 24:971-982, 1957. 8. Auerhach, 0., Gere, J. B., Forman, J. B., Petrick, T. G., Smolin, H. J., Muehsam, G. E., Kassouny, D. Y., and Stout, A. P.: Changes in bronchial epithelium in relation to smoking and cancer of lung: report of progress. New England 1. Med., 256:n7-104, 1957. 9. Essenberg, J. M.: Cigarette smoke and incidence of primary neoplasm of lung In albino mouse. Science, 116:561-562, 1952. 9 NEOPLASIA: LUNGS RicxAnD Douct.As PASSEy, M.C., M.D., D.P.H. IN this short paper I shall attempt to convince you that respiratory cancer is a sequel to tissue damage of a non- specific nature rather than, as is widely believed, to the action of carcinogens. Let us begin by accepting that smokers and town dwellers have more lung cancer than non-smokers and country- men. We know that there are carcinogens in tobacco smoke and in the atmosphere of our large towns. It is natural, therefore, to suppose that cancer of the lung is the result of these carcinogens. Is this necessarily so? What I have in mind is that lung cancer may follow any damage, and I mean any damage, chemical, physical or bacterial, which can permanently disturb the compli- cated functioning of some part of the respiratory tract, such as: (1) Chronic respiratory lesions established by infection or other- wise such as chronic catarrh, chronic bronchitis, pulmonary tuberculosis, bronchiectasis, lung abscess and the damage resulting from recurrent pneumonias and pleurisy. (2) Chronic damage inflicted by tobacco or other smoke, or by the atmospheric contamination associated with certain in- dustries. Why is there any doubt that carcinogens are not responsible for the lung cancer we see around us? First of all, there are certain paradoxes which cannot in any way be explained on the assumption that lung cancer is the result of carcinogens associated with our environment. 1. For instance the recorded incidence of cancer of the lung has increased some thirty or forty times in the last forty years yet the 105 T200992

I 106 Tobacco and Health incidence of cancer of the mouth, of the lip, pharynx and tongue is steadily falling. Yet by far the greatest concentration of tobacco smoke is in the mouth and moreover the mouth is exposed to the smoke for longer periods than the lungs. I would remind you that the oral cavity is lined with squamous epithelium. In general squamous epithelium is more susceptible to carcinogens than columnar epithelium which, however, is more susceptible to non-specific irritants. I would again remind you that it is impossible to induce cancer of the stomach or gut with the polycyclic aromatic hydrocarbons so why should we expect them to have carcinogenic action on bronchial epithelium. The irritant action of smoke and atmospheric pollutions will, I suspect, damage columnar epithelium of the respiratory tract more than carcinogens, but will inflict little or no damage on the squamous epithelium of the mouth, etc. 2. Again it is the universal opinion that cigarettes are more suspect in lung cancer than the pipe or cigar. Yet as can be seen from the following table the pipe and cigar contains many times the amount of benzpyrene and other polycyclic hydrocarbons than are to be found in cigarette smoke. This paradox surely makes nonsense of any carcinogen theory, as we understand the action of carcinogens today. It is reasonable to expect that the greatest risk is where the carcinogen content is highest. As this is not so there must be something wrong somewhere. TABLE I .1. M. CAMPREGL AND A. J. LINDfiSY COMPARISON OF MAINSTRICAM .5"MOKE OF CIpARETTICB, CIGARS AND PIPES 1V1ICROrRAMR PER 100 GRAMS oF ToBAcco CONSUMED Cigarettes Cigars Pipes Acenaphthylene 5.0 1.6 29.1 Anthra,cene 10.9 11.9 110.0 Pyrene 12.5 17.6 75.5 3:41ienzpyrene 0.9 3.4 8.5 I would remind you that cigarettes were called "gaspers" in the beginning of the century, no doubt because of their irritating properties. On the other hand cigars are certainly non-irritating. Neoplasia: Lungs 107 Cigarette tobacco is dried in a different manner from that of cigar tobacco: it contains 20 per cent of sugars and its smoke is acid. Cigar tobacco contains no sugars, and the smoke is alkaline. I suggest that it is the irritating qualities of the cigarettes as opposed to the non-irritating properties of cigars that may explain the difference in the risks in smoking the two. 3. Another little known point is that if you smoke forty cigarettes a day for a year you expose yourself to the same amount of benzpyrene as breathing the urban air of an English town for a year, as is shown by the figures in Table II. TABLE H R. L COoPEe, A. J. LINDSEY AND R. E. WALI.EII Total Amount of 3:4 Benzpyrene Inhaled Yearly Forty cigarettes daily Urban air for for one year one year 150 micrograms 200 micrograms Now the lung cancer hazard of smoking forty cigarettes a day is a very serious one but no-one has claimed that the corresponding risk of a non-smoker living in an English urban area is in any way comparable. Therefore these figures suggest that the risk asso- ciated with smoking forty cigarettes a day and the risk of living in an English urban area for a comparable period are not related to the carcinogen content of cigarette smoke and the urban atmosphere, if the amount of benzpyrene can be accepted as a fair indicator of the total carcinogen content of the cigarette smoke and the urban air. 4. The stains of tobacco smoke on the face and hands of smokers have never given rise to cancer; nor has anyone reported the appearance of skin tumours in mice exposed for their whole lives to cigarette smoke. This shows that the carcinogen content as found in cigarette smoke is too weak even for the delicate mouse skin. It is only when it is greatly concentrated by condensation to a tar that it is effective on mouse skin. All the foregoing makes one doubt if the carcinogens in ciga- rette smoke or in urban air are strong enough to give rise to T200993

108 Tobacco and Health cancer. The report of the Medical Research Council 1955-573 in its survey of the part played by smoking in the causation of cancer of the l1ing in discussing 3:4-Benzpyrene echoes the same queries in these words:- "Even though this substance is known to be a powerful cancer producing agent there is no certainty that it is harmful in such low concentrations." I would add that there is no evidence to suggest anything different in respect of other unidentified carcinogens which may be present. Smoking increases the risk of lung cancer. The more one smokes the greater the risk but no-one has claimed that smokers get lung cancer earlier than non-smokers or that they get it earlier if they smoke more. I have studied the smoking history of 499 cases of lung cancer: the following slides show what I have found. TABLE III MALES CigrlreGtes Smokers Age 1-10 89 56.95 11-20 181 57.03 21-30 133 57.18 31-40 53 56.60 41-80 41 56.10 Average 21 cigarettes 497 average 56.96 This shows that it does not matter if one smokes one cigarette a day or 80 a day, lung cancer appears at approximately the same period in one's lifc, with a mean age of about 57. Thus in my series the date of the appearance of lung cancer in a man with a smoking history had no relation to the amount smoked. Ilowever much he smoked, it did not hasten the fatal issue. Smoking increases the risk of having lung cancer but it plays no part in determining the date when the disease manifests itself. If the smoker commenced early he developed his lung cancer at the same timc as lie who commenced late in life although the difference in the amount smoked is very great. It will be seen that those commencing at the average age of 8.8 years developed V) N Neoplasia: Lungs 109 ~ r-f ~ TABLE IV Age Commenced to Smoke Lung Cancer Cigarettes Smoked ~ 6 62 46 7 56 10 ~ 8 64 6 ~ 8 48 20 8 69 56 H 10 57 12 ~ 10 55 20-{- 10 55 18 10 68 11 10 60 10 10 60 20 Average 8.8 59.5 20.9 50.8 smoking years 26 cases 27.8 62.15 21.1 34.3 smoking years their lung cancer at 59.5 years having smoked for 51 years whereas the 26 cases commencing at 27.8 years developed their lung cancer at 62 years having smoked for 34 years only, yet both smoked the same amount daily. Those who commenced at the age of 10, smoking 20 cigarettes daily should have smoked some 220,000 more cigarettes than those who commenced at the age of 40, and yet the disease did not commence any earlier. In those cases who stopped smoking early and developed lun8 cancer later, Table V, the disease appeared approximately at the same age as in those who continued smoking right up to the onset of the disease. In these tables the numbers are insufficient for statistical analysis but it is reasonably clear that the mean age in each group of smokers is very close to 57. In this series those who developed the disease early in life, that is before the age of 35, had smoked no more cigarettes daily than those who survived to be the oldest cases which I encoun- tered. Thus whatever the approach we make the amount smoked appears to play no part in determining the time in the indi- vidual lives that the lung cancer appeared. My figures of non- smokers are not large enough to satisfy myself or, I imagine, T200994

110 Tobacco and Health ~ al , ~ Comrrcenced A Cigarettes TABLE V Age Ceased Years Since Lung N ~ ge Smoking Smoked Smoking Ceased Smoking Cancer -4 18 10 48 10 58 V ~ 16 80 47 12 59 0 18 20 60 12 72 19 45 41 13 54 20 30 42 14 56 ~ 14 6 51 15 66 41 20 61 13 20 40 f1 34 20 54 C-' 18 0 32 21 53 20 2 22 24 46 20 25 45 24 69 20 8 44 35 79 4 Average 18 22 42.2 18.3 60.5 . 23.8 smoking years TAIILE VI CarnftETTEB .,MUBED ACCORDING To AoE AT pEATH MEN MEN A ges Cigarettes A ges Cigarettes 33 20 75 30 24 8 75 35 33 25 78 10 32 17 78 15 33 31 75 30+ 32 17 75 28 78 28 75 20 Average 31 20 Average 76 25 TABLE VII NoN-SMOFCEI4S Women Average Ag e Mine 23 58.56 I)r.l)oll's 40 55.08 Tot.al 63 57.14 D/en Mine 4 46.25 Dr. 1)oll's 7 51.14 Total 11 49.36 Neoplasia: Lungs 111 anybody else but for sake of completeness I show them and I have taken the liberty of adding Dr. Doll's' cases to them, and you will see that these non-smokers developed their lung cancers certainly no later than the heavy smokers. In short, smoking may be concerned in the development of lung cancer but it has nothing at all to do with the date of onset of the disease. In some way it is the age of the individual which is the important factor. May I quote from Sir MacFarlane Burnett ° "For any form of human disease the age incidence of death from the disease is usually one of the most valuable epidemiological clues to aetiological factors." Therefore the onset of lung cancer depends on something which is closely related to the age of the individual. Carcinogens do not act in this way. For instance in bladder cancer there is a definite period of exposure to the offending chemicals before cancer appears. TABLE VIII OCCUPATIONAL BLADDER CANCER (GOLDBLkTT, 1949) Age at First Entry into Industry Number of Cases Mean Age at Death (Years) from Cancer of Bladder Less than 30 28 47.7 30-40 18 55.9 41-52 13 67.1 If a man enters the industry early in life he gets his bladder cancer in the early thirties to forties. If he enters the industry after the age of 40, the disease will appear in the 60's. This is common knowledge. Sir Percival Pott" when he first described scrotal cancer in chimney sweeps found that it sometimes appeared just after puberty in those boys who had been sent up the chimneys in person to sweep them. Earle and others12 reporting many years later when this practice had stopped recorded chimney sweeps' cancer as occurring commonly between 30 and 40 years of age. Kennaway and Kennawayl' found that T200995

112 Tobacco and Health between 1921 and 1935 the average age was 63.8. Thus in the earliest period when the exposure was intense the cancers appeared early, then when the practice of sending little boys up the chimney was forbidden, and the exposure would be lessened, the disease appeared later, and, finally, today with baths and much greater cleanliness, and therefore much smaller doses of soot, the chimney sweeps' cancer is found around seventy years of age. So the bigger the dose of carcinogen, the earlier the appearance of the tumour. Experimentally a carcinogen induces ttunours in laboratory animals in proportion to its concentration. A weak carcinogen gives rise to tumours late in the animal's life, and a strong dose early. There is no question about that. Dr. Wynder has shown quite clearly in his experiments that the amount of cigarette tar determines the date of the appearance of tumours in the skin of mice. Yet you have seen that in smokers the dose of cigarette smoke has no effect on the date of the appearance of lung cancer. I must refor briefly to nickel and chrome lung cancers and .-. o---o Smokers National Lunq Cancers (1953) 25--30-35 40 45-50-55-60-65-70-75-80-85 Chart I Neoplasia: Lungs 113 6--d 443 LUNG CANCEKS FGLLGW1iRG SMOKIiKG •-- + 135 LUKG CAKCEIIS DUE TO NICKEL. 40- 45- 50- 53- d0- 63-70. ?i= AGE Chart II point out that while they constitute a definite and serious lung cancer hazard they too do not hasten the fatal issue. Chart I shows my 495 male lung cancer cases who have a history of smoking plotted according to age beside the expected age. It will be seen that the approximation is a good one. My series, however, contain no cases over 79 years of age. Being hospital cases that is to be expected as the very aged prefer to remain at home and do not attend hospital. Chart II shows the ages of 135 cases of lung cancer in nickel workers from one factory plotted against my lung cancers in smokers. The two curves approximate closely allowing for the different numbers in each series. The mean age is the samP T200996

114 Tobacco and Health namely 57 years. In the nickel workers the age of appearance of the disease bore no relation at all to the years of exposure to the metal. In the case of chrome, Sir Ernest Kennaway noted that the workers' deaths were not notably hastened, though he had no explanation for this. I suggest that in these cases the action of nickel and chrome salts is not a specific carcinogenic action: they give rise to lung cancer by virtue of the irritative lesions, chronic bronchitis, rhinitis and catarrh which they induce in the workers. These diseases are specially prevalent in these industries. In passing I would point out that neither chrome salts nor nickel have been satisfactorily shown to be carcinogenic. You will ask if carcinogens are not the cause of lung cancer how do you explain this? The old man in me warns me not to be dogmatic but I would suggest to you that the action of the conditions which give rise to lung cancer does not depend on chemical carcinogenesis in the sense that most suppose but on the disturbance of function established by the general irritative nature of the agents to which the respiratory system is exposed. On the other hand it appears to be universally accepted that there is some unidentified potent carcinogenic factor at play in the case of the Schneeberg miners. Whether or not this is a correct explanation of the ghastly hazard of working in the mines the mean age at death from lung cancer occurs much earlier than is generally observed in this disease. I have ploti.ed the cases in Chart III against my own series and it will be observed that the curve is much to the left, as I would expect if a strong carcinogen is the cause. It is true to the accepted pattern, a strong carcinogen inducing tumours early and in high numbers. In general, lung cancer patients give a history of (a) bronchitis and catarrh, (b) smoker's cough and often later chronic bron- chitis, (c) some chronic lung trouble such as attacks of pleurisy or pneumonia, or (d) no history at all of any lung trouble. But in the last group the lack of a definite history is not surprising because we have the same lack of history in many cases of cancer .-. . 9 F . . . . • . ~ ~o 0 . . ., . -.o--O~ b 25- 30- 35- 40- 45- 50- 55- 60- 65 - T6- 75- Chart III of the stomach, colon, rectum, cervix and cancer in other organs. In passing I would remark that many of these differing forms of cancer have this in common with lung cancer, that they often have associated stagnating, altered and excessive secretions bathing the surfaces which ultimately become malignant. What evidence have we that the incidence of lung cancer may be related to chronic lung inflammations? We have excellent statistical evidence. Case and Lea° have shown that of the British veterans of the First World War who were pensioned on account of chronic bronchitis, double the number died of lung cancer than did those who had been pensioned for the loss of a limb but who had no history of chronic bronchitis. They found, too, that the bronchitics had double the expected number of lung cancer deaths when compared with the male population at large. D.D. Reid' independently showed that the post office workers who suffered from chronic bronchitis had a statistically signifi- cantly higher lung cancer death rate than the non-bronchitic T200997 Neoplasia: Lungs 115 Smakers (Average age 56-96) 00 Schneeberg 46-45) ~elzl 6

116 Tobacco and Health TABLE IX Cancer of Lung Mortality Deaths and Pleura Ratio Neoplasia: Lungs 117 highest lung cancer incidence; Norway having the lowest lung disease and lung cancer incidence. Mustard gas 547 29 207 Found 0 expected 3 14 General ulation 357 o MeVe Respirdary Cemeer and O1 . . p p Bronchitis 932 29 201 Found Lery Disceas.(1955). N Retes per i00 DD0 N General population 673.8 14.4 expected , Amputation 383 13 84 Found ' LWXG CW4LEI1 MJLES ~ LWIC C4MCEI1 iEMJ4ES ~ UIQMCNWR6 ~ r1E6M0N1A O TWiERCWLE6K ^ General population 365.7 15.5 expected rl iooao ~ aooo, r iooa _ 000a 0 TABLr X EWOLAND & WALES NORWAY 1~-I CERTIFIED CAURF.S oF DEATR OF POSTMEN INVALIDED FROM. BRONCRITIs Cause Observed Expected booo F ~ Bronchitis 51 2 'rooo E F~I C"0 Other respiratory 11 1 Cardiovascular Coronary 13 5 Other 10 3 Cancer Lung and respiratory 12 3 Other sites 6 4 ooa 2000 a Tb 1 1 C. N. S. 6 2 2 1 ti I)i iao a ioo _Inc ges ve Other 1 3 u- a- +s- so- ss- eo- 6s- ao- AGE ]s- m-s- 90- 55- 60- bs- lo- AGE Chart IV workers, Table X. Following on from these two important obser- vations it will be found by comparing the figures given by the World Health Organization° that in general those countries with the lowest deal:h rate from total chronic lung disease* present the lowest death rate from lung cancer. This is not a firm statistical statement: there are some exceptions, in particular, Japan. But it is a fair generalisation, and at the moment it cannot be other than that because there are other very important factors which vary from coimtry to catmtry and which will influence the figures from which thcse conclusions are drawn, such as smoking habits, state of industrialisation, nomenclature of diseases, differing standards of diagnosis and recording, different medical standards of efficiency and, perhaps most important of all, vastly differing climates. In Chart IV will be seen the extremes. England and Wales with the highest total lung disease incidence has the I suspect that much of the difference in the lung cancer inci- dence in different countries is the result of differences in climate. The chilly, damp, foggy autumn in Britain favours our high respiratory disease rate. At the other extreme is the lovely, mild, dry, sunny fall in the United States. That this is so is shown unmistakeably by Eastcott° in New Zealand and later confirmed by Geoffrey Dean" in the case of South Africa. Eastcott showed that immigrants into New Zealand from Great Britain had a statistically significantly higher lung cancer incidence, especially if they had lived in Britain for thirty or more years, than the New Zealanders born of British parents in New Zealand. This of course can only mean that the British immigrants came with "a something" which the native born New Zealander did not possess. That "something," I suggest, might well be an incipient (but clinically unrecognised) bron- T200998

118 Tobacco and Health I Neoplasia: Lungs 119 chitis or some other very early lung lesion conditioned by the climate and the urban atmospheric conditions, not smoking habits, because in the case of New Zealand, the smoking habits, and cigarettes are comparable with those in Great Britain. Again, it would seem that it is not the amount smoked which alone determines the high male-female sex ratio of lung cancer. I suggest here again that an important factor may well be the incidence of total chronic lung disease. These two points are worthy of further study and it is for this reason that I draw attention to them. I conclude that lung cancer is not a result of the action of car- cinogens but follows on tissue damage of a general non-specific nature. If you ask of what importance is this decision, I should answer: (1) Until we know the correct aetiology of this and of other forms of cancer we shall be at a disadvantage and we shall fail to make progress. (2) Our approach to prevention will be entirely different. We shall attempt to make cigarettes less irritating instead of trying to eliminate the carcinogens in smoke. We shall look out for and treat all forms of respiratory illness in their early stages, however trivial, with a view to preventing permanent damage to the respiratory system. It is a popular belief that smoking reduces the weight in man, and that it reduces the appetite. It has been established (R. D. Passey ) that exposure to cigarette smoke has the same effect in animals. Now it has always been supposed that the reduction in weight caused by smoking has come about as a result of the smaller consumption of food. This is often the excuse given for smoking. It has however now been shown that it is not as simple or as harmless as that. In fact experiment has shown that animals exposed to cigarette smoke use their food in a different manner from animals not so exposed. Thus, Elson, Lewis and Passey have shown that a diet contain- ing only 10% protein is quite inadequate for young growing rats, if exposed to cigarette smoke. Yet control rats, fed on the same diet, thrive and grow readily. The experiments, thongh not yet complete, indicate that expo- sure to cigarette smoke increases the excretion of urea nitrogen. 1** CD. RATS. T09b PROTEIH DIET,10 GGS. DALY 71A[ IN pWy Chart 3 SMOMM96 RATS COMTROI RATS SMOMIMG RATS COMTROI RATS 10% M(OTEIM OICT qy. pR0TER1 O/ET 20'f. HV0TE7M OIT 20% PRfiTEMi OEIT IhN., N Iwt Snaa.j MwN. N Ilttt Swx..y N.N.. .1 1111t dNn, N 40. Sn.qxt IO1 w M0 OEATMS L_ Ip OEATMS ~ 7tlfN ~. Nd1.a4(M.71; rwwN.IwN7 ~~I~(~N1 12 I 400 3 2'50 5 15 25 35 45 5 15 25 35 45 5 15 25 35 45 DAYS DN DIET Chart ii 5 15 25 35 45 T200999

t 120 Tobacco and Health EFFECT OF CIGARETTE SMOKE ON GRdNTH ANiD PROTEIN MiETABOLISM Urea excretion in rat receiving snmoke from 10 cigarcttes a day, mnaintained on 10°b protcin diet. DAYS Chart iii Therefore it would seem that, at least in part, smoking causes a reduction in the weight of the animal exposed to smoke by effect- ing some change in the metabolism of its proteins. This experiment alone would suggest that in growing boys smoking would reduce the growth rate, especially if their diet lacks proteins. The experiments also suggest that there would be other unfavourable risks. Aerosols of nicotines have almost identical effects on animals exposed for tirne-pcriods similar to the exposure to cigarette smoke, with one important exception. Whereas cigarette smoke reduces the food intake, a nicotine aerosol, in fact, increases the food eaten. REFERENCES 1. Campbell, J. M. and Lindsey, A. J.: Brit. J. Cancer, 11, 192, 1957. 2. Cooper, R. 1.., Lindsey, A. J. and Waller, R. E.: Chemi.atry and Industry, p. 1418, 1954. Neoplasia: Lungs 121 GROWTH RATE OF RATS EXPOSED TO N4COTINiE WAPOUR Averoge weights and food intake of 10 rats maintadned on a 10'l% protein diet Number of Rats Survivinp NO DEATHS Numt+ar of Rats Surviving NO DEATHS Test antmah31L nkotine I Control anima4(water water va~pour for vapour for 2 hours 2 hours) I 1 ---+ start of exposure ---+ start of exposure , ;Foodintakafw.twtJ :Foodtntmk.(vatwt.) . DAYS ON DIET Chart iv 3. Report Medical Research Council. London 1955-6-1957. Com. 180, p. 12, Her. Mag. Stat. Off. Doll, R.: Brit. J. Cancer, 7, 303, 1953. Burnet, F. M.: Canadian Cancer Conference, 3, p. 436, 1959. Case, R. A. M., and Lea, A. J.: Brit. J. Prev. and Soc. Med., 9, p. 62, 1955. Reid, D. D., and Fairburn, A. S.: Lancet, i, p. 1147, 1958. Annual Epidenl. and Vital Statistics. 1955, W. H. O.,'Geneva, 1958. Eastcott, D. F.: Lancet, l, p. 37, 1936. Dean, Geoffrey: Brit. Med. 1., 2, p. 852, 1959. T201000

122 Tobacco and Health 11. Pott, Percival; Chirurgical Observations, London, 1775. 12. Earle, H.: Med. and Surg. J.,1, 6,1832. 13. Kennaway, E. L. and Kennaway, N. 1V1.: Yale J. Biol. Med., 17, 139, 1944. 14. Goldblatt, M. W.: Brit. J. Ind. Med., 6, 65, 194'9. 10 BRONCHIAL PATHOLOGY IN MAN STANLEY P. RErMANN, M.D. lEvElty discussion of the pathological changes in any organ should begin by considering the normal anatomy, both gross and microscopic. The latter includes not only classical microscopic anatomy but also the cytology, and now the "sub microscopic" anatomy and pathology as presented for example by H. Schulz (1) with his electron microscope pictures of the lungs. If time permit, there may be added or interspersed selected presentations of the physiology, the pathologic physiology and a modicum of biochemistry. Even though the title of this presentation in the program is "Bronchial Pathology in Man," it is hardly fair to confine remarks to the bronchi alone. The bronchi lead into the lungs, and disease of the former can hardly be isolated but necessarily contributes directly or in- directly to changes in the lungs themselves. So, however brief, the lungs at least must be mentioned. But speaking of restric- tions, again it is hardly fair to confine the presentation to the bronchi and the lungs alone. For secondary changes in other parts are induced by pathological conditions in these. We need not belabor the point that many such complications are reflected in further accentuation of trouble in the bronchi and lungs themselves. Though settling the point of which is first is important in understanding the pathological interdependencies and in drawing deductions of various kinds, it is obviously im- portant, too, for prevention and therapy. Confronted by a congerie of changes within bronchi and lungs and in other situ- ations, it is important to decide upon the sequence of a thera- W T201001

124 Tobacco and Health peutic attack, that is, assuming that we know enough in any particular case to make the attack at least partially successful. Furthermore disease outside the bronchi and lungs leads to difficulties within these organs. Distortions of the vertebral column and its bones, fixation of costo-vertebral joints, stretching of ribs, funnel or pigeon breasts are potent disturbers of bronchial and lung structure and function. Upper respiratory conditions such as nasal polyps, other obstructions, infections of sinuses, tonsils, do not leave the bronchi and lungs isolated. Nor does myocardial damage and failure spare bronchi and lungs. Muscle paralyses as in post-anterior poliomyelitis, muscular dys- trophies can be nientioned. The list could be extended. Finally the reversibility of changes comes to the fore, whether return to normal can be expected if the exciting agent is removed or whether changes once established continue on their own. Answers to these as well as other questions in individual cases depend in part on the general state of health, on nutrition, mal- function of othPr organs, blood counts and supply, and that elusive though important factor, genetically controlled constitu- tion. Then there is compensation to be considered; what happens when a lobe or a whole lung is destroyed or is removed or collapsed by thoracoplasty. It also happens that the normal anatomy of the bronchi and lungs is not subject to wide variations from time to time due to the varying factors in living, such for instance, as mirrored by the differing physiologies of the breast, the uterus, the ovaries, the endocrine glands in general, even the liver. These organs and others cannot be said to have one normal anatomy, for their structure changes in tune to varying physiological activities and demands, a fact of importance in interpretation. The most that can be said for bronchi and lungs is that their anatomic structures and relationships vary somewhat from individual to individual. Unless exeessivo, such variations do not interfere remarkably with their physiological activities once they have been fully established after birth. Tl:ere are, however, changes as a result of age, which it woulcl seem are perfectly normal phenomena. These changes consist essentially of loss of tone and reduction in the Bronchial Pathology in Man 125 elastica and of some of the muscularis of the bronchi, those most important constituents of the bronchi and lungs. There is also reduction of what elasticity is provided by the ground substance to which may be added certain degrees of stiffening from arterio- sclerosis as a result of age. Ordinarily, however, the vessels supplying the lungs do not show the degrees of arteriosclerosis encountered in other vessels such as the aorta, the renal, the cerebral, and the coronary vessels. Anomalies, congenital but probably more often hereditary, vary from wide spread so-called cystic lungs to accessory lobes, small lobes with consequent enlargement of others, various anomalies in the bronchial cartilages and distribution of muscula- ture, and others, including the difficulties produced by defects in other situations as in the heart (tetralogy of f' allot ), in the diaphragm, the vertebral column and so on. The corina may be cartilaginous in 56 per cent, membranous in 33 per cent and mixed in about 11 per cent of cases. The cartilage may be only tracheal and extend but short distances into the bronchi. There may be double right or left bronchi and various other changes from the average, whatever the word "average" may mean in this or in any other case (2). As far as heredity is concerned, we believe that it is over- whelmingly due to what is in the chromosomes. Just to remind us of the difficulties of assessing the relative importance of heredity and environment in the production of any particular set of difficulties, I am reminded of the story of the grey lethal of the rat as described by Darlington and Mather (3). Para- phrased, a particular gene area determines an anomaly of car- tilage, leading to changes in the larynx and nose, thick ribs and a narrow trachea. The ribs tend to be fixed in inspiration. The thoracic vertebrae are fixed. This leads to displacement of the thoracic viscera, dilatations of the lung passages and cavities. There is slow suffocation, coma and death. If not immediate death, there is increased resistance to the pulmonary circulation, compensatory hypertrophy of the right ventricle, capillary bleed- ing into the lungs, heart failure and death. As additional diffi- culties, there are arrested development of the face with a blunt snout, faulty incisor teeth, feeding difficulties, blocked nostrils, T201002

126 Tobacco and Health inability to suckle, starvation and death. Take your pick. Among all these complications, at least the bronchi and lungs are in the middle. I dare not speculate about events in human beings, but there are such conditions as achondroplasia, etc. The relation of heredity, with or without anatomical defects, to infections, to congestions, -yes- and to cancer itself remains in the specula- tive stage. In full inspiration, about 5,000 cc. of air can be contained within the bronchi and lungs. After death, there are about 1500 cc. present in the collapsed hmgs. The greatest amount of air ex- pelled after a full inspiration averages about 3200 cc. in males and 2500 cc. in females. The amount of air exchanged per minute is about 5800 to 5900 cc. The tidal air is about 450 cc. We mention these figures as rough approximations merely to call attention to the fact that when swelling occurs of the bronchial and air sac linings, these amounts are diminished with consequent difficulties. Allergy, that attempt on the part of an organism to maintain its species specificity, is a potent cause of disturbed function and anatomy of bronchi and lungs. Swelling of the cells of the mucous membranes, edema of the entire walls, escape of cells from lymphatics and blood vessels and infiltration of them into the fluid saturated tissues, including many eosinophiles; hyper- activity of the mucous glands, producing thick tenacious secre- tions within bronchi and bronchioles even to precipitation of Charcot Leyden crystals, the formation of Curshmann's spirals which the cilia can hardly move; spasms of the muscle bundles is a partial list of consequences. Hypersensitivity occurs to many substances; some people are allergic to tobacco even when smoked by others. The pulmonary arteries follow the bronchi to capillaries in the walls of the air spaces. The blood is returned via four pulmonary veins, two for each lung. The second set of arteries, the bronchial arteries, vary in number and origin, coming from the thoracic aorta, the first or third intercostal arteries, the anterior wall of the thoracic artery with variations. Most of their blood is returned via the pulmonary veins but some of it by way of the azygos veins. We recognize the consequences within Bronchial Pathology in Man 127 trachea, bronchi and lungs of congestion and obstruction of either or both of these sets of vessels. There are superficial and deep lymphatics; the superficial sets are in the pleura. The deeper sets accompany the bronchi and their blood vessels and communicate in the pleura and at the hilum. They drain into lymph nodes at the hilum, communicate with those in the mediastinum and from there to near and distant parts. The lymphatics of the trachea and the larger bronchi consist of two anastomosing layers. There are channels around the cartilages of the trachea and larger bronchi; others are longitudinal. The lymphatics play a predominant role in many conditions, notably in infections, as exemplified by the ofttimes fatal invasion of streptococci within pulmonary lymphatics during influenza. And there is such a thing as tuberculous pulmonary lymphangitis. Besides some of the lymphatic drainage of the head and neck descends into the lungs. The tracheal and bronchial cartilages are classified as "hyaline." The intrapulmonary cartilages are "fibro elastic;" they disappear when the bronchioles are between 0.6 and 0.7 mm. in diameter. Muscle bands twine around the bronchi and bronchioles inter- mingling with those placed longitudinally. The tubular structure of the bronchioles is mostly muscular, capable of allergic and other spasms. Elastic fibers course longitudinally but cross over and surround branches of main bronchioles. They usually enclose the alveoli. Without too much detail the epithelial cells lining the trachea and bronchi consist of at least two, and sometimes more, layers. In the basal layer are the partially differentiated cells. These are the "spare parts" from which cells shed from the surface are replaced, as wear and tear removes them. Being but partially ' differentiated, they can readily divide. Under normal conditions they ascend, so to speak, from the deeper to the more superficial layers until finally they become surface cells. Normally, many become ciliated, and some become goblet cells. It is from the, partially differentiated basal cells that metaplasia occurs. Being but partially differentiated the basal cells contain more potencies than they can express, i. e., they contain two or more possibilities T201003

128 Tobacco and Health of further development. In disease one of the potencies which normally is not expressed becomes determined, to use the expres- sion in general biology, and they turn into perhaps squamous cells, thus leading to the changes classified in pathology as squamous metaplasia. This phenomenon of multiple potency is a general biological property of undifferentiated or partially undifferentiated cells. But there comes a time in the life of an undifferentiated or partially undifferentiated cell, when, of a number of possibilities, one is chosen. This is called the point of determination. Unfortunately, it is not possible to tell by ob- servation of a cell which an experienced microscopist can call undifferentiated, whether or not of the several potencies within, one has been chosen to be exerted. The only way in which this can be told is by watching the cells to see what happens or to transplant them. Obviously, we have no techniques at present for either of these methods for bronchi and lungs, so we argue backwards on the basis of what is known in general biology. We can, of course, transfer lung and bronchial cells to tissue culture media but the pervading influence of the organism as a whole is lacking. Now the factor which influences the choice of potencies in an undifferentiated cell lies, broadly speaking, in the environment, and the environmental determinor may be general or local. Analogies in pathological processes occur in many parts of the body; for example, locally, in the gall bladder in the presence of gall stones, wherein the normal cylindrical epithelium lining the gall bladder becomes replaced by squamous epithelium. As a general cause, we mention lack of vitamin A, in which meta- plasia to sqaamo=is epithelium occurs not only in the lining of the trachea and bronchi but in other situations as well; thus in the -ureter, the bladder, vagina, etc. Veritable leucoplakia appears of general and not merely local origin. The cilia in the cells of the outer layer of the tracheal and bronchial linings sweep particles from below upwards. When the movements of the cilia are disturbed or absent, particles may not be ejected, but there is some evidence that the bronchi and bronchioles have peristaltic movements. Cilia are often absent from normal cylindrical cells, and certainly metaplastic I 4 I 4 i Bronchial Pathology in Man 129 squamous cells do not develop cilia. Incidentally, the outer free ends of ciliated cells have a thickened cuticular border which bears the cilia. There is a continuity from cell to cell along this border but injury of one cell may not involve another. Further- more, ciliated cells may extend through the entire thickness of the epithelial layer, and the lower ends often reach the basement membrane. They also line the trumpet shaped ends of the ducts of mucous glands, and they extend downward into the bronchi further than goblet cells, though not in an unbroken line. There are short strips of non-ciliated cells between those with cilia. Thus in a microscopic section of a small segment of a bronchiole, the absence of cilia may not be significant. Goblet cells become less and less in number and finally dis- appear in bronchioles of about 0.4 mm. in diameter. The tracheal glands are tuboacinar. They are in the submucosa and more of them in the anterior than in the lateral walls. They often extend in and behind the muscle layers, and their ducts pass through muscle layers. The glands are mostly in clusters situated between the cartilages. Obviously, we must not interpret this picture as of invasion or even pre-cancer or carcinoma-in- situ no matter how distorted they look from infections or other injuries. Apart from hereditary and congenital defects within or without the bronchi and lungs, pathological conditions within them may be classified as degenerative, circulatory or inflammatory, with a mental note that ofttimes all three appear. The pathological changes due to disturbances of growth processes may be con- sidered as aplasias, hyperplasias, metaplasias, pre-tumor or tumor formation. This leads to the notes of the conferences on bronchial and lung pathology held over the last few years, under the auspices of the Scientific Advisory Board of the Tobacco Industry Re- search Committee (4). Briefly, twelve pathologists of profes- sorial rank agreed to examine microscopic preparations from about fifteen areas of bronchi, bronchioles and lungs obtained in routine autopsies or in surgically removed specimens as they accumulated on their services. At the five conferences held over a period of two or more years, the twelve pathologists brought T201004

130 Tobacco and Health slides of particular interest to them, with the idea of not only attempting agreement as to classification of the changes found but also of their implications. In the discussions, these ex- perienced pathologists were not unaware of the many variables in their work, and naturally, kept them in mind. More or less agreement was reached as to a classification. It was as follows: 1. Normal, 2. Hyperplasia, 3. Squamous metaplasia, 4. Atypical metaplasia, 5. Carcinoma in situ, 6. Carcinoma. As everyone knows, nomenclature is more of an irritant in many discussions of pathology (as well as in other subjects ) than are the actual lesions. Much of the argument centered about names and not conditions. Nevertheless, agreement was reached in many cases, except those on the borderlines. Here it was obvious, as it is in so many other conditions, that not enough is known at present to resolve such differences of opinion. Inter- estingly enough, they were not merely in matters of interpretation which is easier to understand, but also in classification of the actual appearances. This was in spite of the fact that microscopic pathological anatomy has the advantage of presenting definite pictures for any one to see, i. e., well prepared slides can be preserved. The differences of opinion intrigued the statisticians, and Wilson and. Burke, obtaining the opinions of the pathologists as data, have constructed several presentations from them (5). The histories of the patients from whom specimens were removed were obtained as well as possible. Such data as age, sex, occupation, place of living, changes of occupation, smoking habits and a few others were recorded. Naturally, the correlation of changes with smoking habits was of great interest. While the techniques of obtaining specimens varied somewhat, since pathologists choose blocks for section from gross appear- ances, nevertheless, a fair enough cross-section could be obtained from 15 separate areas in each lung. Perhaps no better can he done than to restate a few of the conclusions, incomplete though they may be, from the studied data of Wilson and Burke. 1. There is no difference in the prevalence of changes, called hyperplasia or metaplasia, at any age above 25 years. Less than 30 per cent of all cases studied, including both sexes 11 Bronchial Pathology in Man 131 and all ages, had 15 sections with apparently normal mucosae. 2. Males have these changes more often than females, but both metaplasia and hyperplasia do occur frequently in females. 3. Cigarette smokers in this study have fewer normal sections • than non-smokers. 4. By a rough division of occupations into "hazardous" (those which may have a high death rate from lung cancer), and those in farming and others, no consistant excess of meta- plasia or hyperplasia appeared in any of the three groups. 5. Study of the occurrence of metaplasia or hyperplasia by the size of the largest city in which the cases have ever resided revealed no differences. 6. Metaplasia was present in more cases dying from lung cancer than in those which died from other cancers or all other causes of death. 7. Carcinoma-in-situ was less and less frequently diagnosed, and then usually only when invasive carcinoma was also present. Perhaps, in this communication more attention should be given to the arguments of the pathologists as they met informally, and exploded (I) their opinions without notes being taken. They ranged over comparative anatomy, embryology, genetics, bio- chemistry, physics, public health, industry and labor, psychology, statistics, sociology, even local, national and international politics. No one here will dispute that all these and other subjects enter medicine somehow or other, and it is left to the reader to specifi- cally apply these disciplines to the "Pathology of the Bronchi." The pathologists did. Morphological classification is not a very precise process. It was to be expected that the pathologists would differ in the names they chose as they one by one examined a particular slide, especially of border line conditions. That they differed in their interpretations was no surprise. Discussion chilled some of the differences but by no means removed them completely. From the discussions, one facet of many may be picked for further expansion, viz., pre-cancer and carcinoma-in-situ. T201005

132 Tobacco and Health "Carcinoma-in-situ was less and less frequently diagnosed and then usually only when invasive carcinoma was also present." Why? Pathology is indeterminate as are all other subjects. It is a combination of science and art, now more of one and now more of the other in a given instance. The foundations of pa- thology are in the knowledge we have of general biology, from which it may be concluded that progress in general biology should be reflected in pathology and vice versa, progress in pathology is reflected in general biology. With carcinoma-in-situ goes also the expression "precancerous." These two abstractions have acquired more and more stature in the literature, in conversations, in arguments, in diagnosis, in treatment, in statistical data, and in both inductive and deductive reasoning. Originally, chiefly of pathological import, they influence many phases of medicine. They are two different things. Pre-cancer is not yet cancer, but will become so, provided the patient lives long enough: carcinoma-in-situ means cancer has been established but has not as yet broken loose of its restric- tions. If we grant that the foundations of pathology are built upon general biology, it follows that any pathological process may be and actually should be analyzed in terms of the concepts of general biology. For this purpose we may rehearse three gen- eralities. 1. The developmental possibilities of an undifferentiated or partially differentiated cell are greater than its develop- mental fate. 2. When the fate of a cell is determined, it will proceed to its chosen difTerentiation, even in an unfavorable environment. It will strive hard to fulfill its fate, and though the tissue derived from it and its descendants becomes distorted, the cells are not autonomous. 3. The time of determination of the fate of a cell varies, and the time when that fate is exerted also varies from a few minutes, to hours, to weeks or even years. Thus, a cell just born by division of its mother cell has a choice of alternatives. a. To divide. b. To remain as is, that is, undetermined and undifferentiated. Bronchial Pathology in Man 133 It remains as a reserve cell or what may be called "spare pm,t >, c. To remain undifferentiated for_a longer or shorter time, even though its developmental fate has been determined. It re- mains as a restricted spare part. d. To proceed immediately to differentiate into some type of normal cell and to fit itself in with others into an organism of the lineage from which it came. e. To use another of its several potencies e g become meta- , . ., ~ plastic. ~ f. To mutate and turn into a cell, the descendants of which are f more or less different than those normally to be expected. To determine its internal mechanism in an entirely new direction, foreign to the organism and inimical to it, i.e., into a cancer cell. Translating these generalities into the situation known as pre- cancer and/or carcinoma-in-situ, the questions to be asked are: 1. Have the division-capable cells in a precancerous lesion undergone the necessary internal change and become can- cerous, call it mutation if you like, or has the defect in differentiation leading to the precancerous appearance been in an abnormal but not cancerous potency? 2. When the expression carcinoma-in-situ is used shall we say that the division-capable cells in this lesion have in fact undergone the determination which makes them cancerous, but have not as yet exerted the new potency? That is to say, they have not as yet become invasive. But the very expression signifies that sooner or later invasiveness will appear. This argument, as can be seen, calls attention in no uncertain terms to the difficulties of prediction, for this is what is done when these two expressions are used. Unfortunately, micro- scopists see cells and tissues as of the instant when they were plunged into a fixative. What happened before the piece of tissue was removed for examination involves deductions from the accumulated experience in pathology for many years. What would have happened bad the piece of tissue not been removed when it was, but later, is a prediction. To verify the prediction g• T201006

134 Tobacco and Health we must watch the lesion to see what happens, or transplant it to try to determine its potencies. Either or both of these pro- cedures are difficult to carry out in human beings, for reasons which we need not belabor. Since we cannot have the specimen in the patient and also under the microscope, we must fall back on the experience which teaches that some of both lesions as near as we can diagnose them without the microscope, do at times become invasive. Experience also seems to teach that lesions classified as hyperplasia, sqamous metaplasia and atypical meta- plasia can and do retrogress and the area can again be clothed in normal cells. The probabilities are also that carcinoma-in-situ is capable of retrogression, but there is no retrogression when inva- sive carcinoma appears. The fact that we deal with individuals in medicine does not imply that numbers of individuals are a simple sum. In practice we adopt a dualism: remove what appears grossly to be pre- cancerous and/or carcinoma-in-situ as a form of preventive medicine, but carry the mental reservation that we are not sure the offending lesions, if allowed to remain, would in fact have turned into invasive cancer. This presentation has been, as you have noticed, much of a generalization, written deliberately in this vein, for I was sure that my confreres in this symposium would present many details along with photographs of actual lesions. It has inquired sketchily into some of the fundamentals, not only of pathology of the bronchi, but of other situations as well. It recognizes that re- peated inquiries into the fundamental beliefs and practices of any subject are of value to progress in the subject and to framing questions awaiting solution. Lest we forget, it also lists a number of variables. SUMMARY It was deemed unrealistic to confine this presentation to "Pathology of the 13ronchi" so sketches were added to include the lungs, to the influences of changes in other parts on bronchi and lungs, anci in turn to their influences on other parts. Since the normal anatomy influences pathological events and pictures, the anatomy is rehearsed. Bronchial Pathology in Man 135 00 N 111Z r--1 The principles of pathology are usually in tune with the principles of general biology, so a few generalizations from the latter subject are invoked in their relation to certain aspects of ~ the pathology of the bronchi and lungs. O ~ ~ In a cooperative study of many microscopic sections by 12 pathologists from various parts of the U.S.A. different opinions arose as to nomenclature, interpretation and implication. Questions without precise answers are explanations of why the h ~ differing opinions, often diametrically opposed, were expressed and why closer approximations cannot be given at present to some of the important questions rehearsed. I REFERENCES 1. Schulz, Heribert: The Submicroscopic Anatomy and Pathology of the Lung, Berlin, Springer, 1959. 2 Miller, Wm. S.: The Lung. Thomas, Springfield, 1937. See also Drinker, Cecil K.: The Clinical Physiology of the Lungs. Thomas, Springfield, 1954. Bloomer, W. E., Liebow, A. A., Hales, M. R.: Surgical Anatomy of the Bronchovascular Segments. Thomas, Springfield, 1960. 3. Darlington, C. D., and Mather, K.: Elements of Genetics. Allen and Unwin, London, 1950. Gruneberg, Hans: An Analysis of the Pleiotropic Effects of a New Lethal Mutation in the Rat, Proc. Royal Soc. London, Series B. 125, p. 123-144, 1938. 4. Reimann, Stanley, P.: A Cooperative Study of the Pathologic Anatomy of the Bronchial Tree and Lungs. In press, Arch. Path. to be published Nov., 1960. 5. Wilson, Edwin B., Burke, Mary H.: "Some Statistical Observations on a Co- operative Study of Human Pulmonary Pathology." Proc. Nat. Acad. Sci., 43, 1073-1078, 1957; 45, 389-393, 1959; 46, 561-566, 1960. R T201007

01 N ~ ~ ~ O I I ~ ~ CHANGES IN THE TRACHEOBRONCHIAL TREE IN CONNECTION WITH CIGARETTE SMOKING ~ N OSCAR AUERBACII, M.D., ARTIIUn PUSDY STOUT, M.D., E. CUYLER HAMMOND, Sc.D. and LAWRENCE GArtFINKEL, M.A. DURiNC the past six years we have been engaged in a study of the changes in the tracheobronchial tree brought about by inhaled irritants, especially those in cigarette smoke. The work tl-,us far done has been on white men and has been performed in three stages. The final report on this phase of our study is now being prepared. Our study was first stimulated by the fact that we had the opportunity to study at autopsy a case of bronchogenic carcinoma in a man who had been exposed to chrome, a substance generally thought to be a causative agent in lung cancer, for many years prior to his death. Multiple areas of carcinoma-in-situ, squamous metaplasia with extensive atypism of the cells and basal cell (ryperplasia w,~re found in the remaining parts of the bronchial mucosa. These findings stimulated an interest in possible epithelial changes in the tracheobronchial tree that might be related to smoking. To our knowledge at that time there had been only one report in the literature in which an attempt had been made to determine whether smoking produced definite epithelial changes. Wittekind and Striider` took many different sections from the bronchial tree and designated them according to the lobe from which they were Supported by a research grant from the American Cancer Society. 136 Changes in Tracheobronchial Tree 137 Fig. 1. Coronal section of the anterior half the tracheobronchial tree. taken. They were particularly interested in the occurrence of squamous metaplasia because many others were of the opinion that there was a relationship between metaplasia and cancer of the bronchus. They concluded that there was no relationship between tobacco smoking and metaplasia-that the metaplasia was related to a chronic bronchitis. We doubted these conclusions of Wittekind and Struder and we wanted to find out more from our own studies. It was decided to study the tracheobronchial tree of every available case of carcinoma of the bronchus and to make an attempt to compare T201008

138 Tobacco and Health these tissues with those obtained from non-lung cancer cases, of smokers and non-smokers, unless the latter presented overt bronchopulmonary disease.2 Method. The lungs used in this study were first insuftlated with 10 per cent formalin via the trachea and then suspended in formalin for at least twenty-four hours. The lungs were cut meticulously in their coronal plane in the following manner: The trachea, main and upper-branch bronchi were divided with scissors at the junction of the anterior cartilagenous and posterior membranous portions. The inferior (mesial) aspect of the main bronchi and carina were similarly subdivided at the border of the cartilage rings and posterior membranous portion. Malleable, round-tipped probes were then passed through the lateral periph- eral margin of the lungs via the branch bronchi of the upper and lower lobes. The lungs were then cut in their transverse coronal plane. with the probes serving as guides. By dividing the lung in this manner, approximately two-thirds of the circum- ference of the trachea, one-half of the circumference of the main and proxirnal-branch bronchi, and the entire right middle lobe bronchus were included in the anterior half of the specimen. Distal to the primary bifurcation of the branch bronchi, the entire circumference of the segmental and basal-branch bronchi were included in the anterior half of the specimen (Fig. I). The tracheobronchial tree was then freed from the pulmonary parenchyma by blunt dissection. The primary and secondary bifurcations of the segmental and basal-branch bronchi were readily dissected intact. The tracheobronchial specimen was next carefully subdivided into its various anatomical segments and each segment sub- divided into blocks 2.0 and 4.0 mm. wide. The tracheobronchial tree specimen generally yielded between 180 and 200 blocks. Each case was studied individually and recorded on a sche- matic outline of the tracheobronchial tree. Three types of changes were recorded, namely, basal-cell hyperplasia, stratification and squamous metaplasia. Other changes, such as carcinoma-in-situ, invasive carcinoma, and metastatic carcinoma, were not con- sidered in the compilation of our charts, but were recorded on the individual cases. Changes in Tracheobronchial Tree 139 Material. There were ninety-seven cases in which the tracheo- bronchial tree was sectioned and from which approximatel y 27,000 slides were studied. Because of extensive denudation, 56 cases were excluded and 41 remained. The 41 cases included fourteen cases of carcinoma of the bronchus and 27 cases without lung cancer. Of the latter, eight of the patients had been non-smokers, 14, light to heavy smokers, and five, heavy smokers. All cases of carcinoma of the bronchus were either moderate or heavy smokers. Changes Observed--®asal-Ce11 Hyperplasla. The basal cells lie along the tunica propria, generally as a single layer. These cells have a scant cytoplasm and oval or round nuclei. We con- sidered the process as slight basal-cell hyperplasia when there were three or four rows of basal cells; of moderate extent when there were four to six layers of such cells, and advanced, when there were seven or more layers of such cells present. Stratification. The epithelial lining here is composed of multiple layers of flattened epithelial cells lying parallel to one another. Ciliated cells are absent over this zone. Squamous Metaplasla. These layers are generally thicker than the foci of stratification. There is a similar layering of the cells, but their appearance is similar to the epithelial cells in sites where squamous epithelium normally resides, such as squamous mucosa. The three types of epthelial changes were present to a greater degree in smokers than in non-smokers. We felt then that our findings warranted a more extensive study along similar lines to determine whether these findings would hold in a larger series. The second phase9 was an attempt to test two hypotheses. Namely, (a) when a carcinogenic agent is applied to the mucosa a number of changes such as hyperplasia and metaplasia usually precede the appearance of cancer, and, (b) neoplastic changes are apt to occur at several different points. Therefore, if inhalants of one sort or another are a major factor in the causation of lung cancer, one might expect to find hyperpla~ia, metaplasia and early neoplastic changes in the remaining bronchial epithelium of persons who died of bronchogenic carcinoma. Furthermore, one would expect to find similar changes, but to a less marked T201009

140 Tobacco and Health I Changes in Tracheobronchial Tree 141 degree, in the bronchial epithelium of persons who died of some other cause, but who had been heavily exposed to potential carcinogenic inhalants. There were 117 cases used in this study of white men whose ages ranged from 22 to 88, but over two-thirds were between the ages of 50 and 70 years. Most of the patients lived in urban industrial areas of New Jersey. The histories of smoking habits, occupations and places of residence were obtained from the family by a trained social worker. The material was studied only after the smoking histories had been obtained. All patients dying of lung cancer were included. IVlethod. After fixation in formalin the tracheobronchial tree was divided into 208 portions as shown in Figure II. The pathologists received all the sections from a given case in a group although they did not know the smoking history or clinical features of the cases. The findings recorded in this study included carcinoma-in-situ in addition to basal cell hyperplasia, stratification and squamous metaplasia. We applied the same criteria here as are accepted for intraepithelial carcinoma in other sites" in the body and as those applied by Black and Ackerman' in their study of broncho- genic carcinoma. These are as follows: The basement membrane is intact; there is cellular disorganization, with loss of the usual layering, the nuclei show a great variation in size, shape and chromatin content (nuclear hyperchromatism is frequent); there is an increased number of mitoses, and they are often atypical; the nuclear-cytoplasmic ratio is increased. To test the two hypotheses the cases were divided into two groups. In one, manifest carcinoma of the bronchus was found by ordinary autopsy procedure. The other group was evaluated according to smoking history. Lung Cancer. Thirty-four patients died of bronchogenic car- cinoma, all of whom were smokers. The average age was 58. Findings in RelatioM to Smoking. These 83 patients who died without evidcnce of lung cancer by ordinary autopsy procedure were evaluated according to the same histologic criteria and were divided into three groups according to their smoking histories. Group 1. Patients who never smoked regularly. There were 16 -5 Fig. 2. Schematic diagram of the tracheobronchial tree, showing the dis- tribution of the 208 sections. patients who had never smoked regularly. Group 11. Patients who smoked less than one package of cigarettes a day. In this group there were 20 men who smoked regularly, but less than one package a day. Group 111. Patients who smoked one package or more a day. There were 47 cases in this group. Based upon the results in this histologic study it was seen that among people who died of causes other than lung cancer, basal cell hyperplasia, stratification, squamous metaplasia and car- cinoma-in-situ were least frequent in the group that never smoked regularly, with a progressive increase in the moderate and heavy smokers. The same, but more extensive changes were observed in those who died of carcinoma of the lung. Although definite carcinoma-in-situ was interpreted as present T201010 -2 -3 0

142 Tobacco and Health in all groups, with a parallel rise in proportion to increasing cigarette consumption, there was an almost similar distribution of this change in those who smoked more than one package a day (6.0 per cent ) and in the cases of bronchogenic carcinoma (6.3 per cent). These findings are fully consistent with the hypothesis that inhalants of one sort or another are important factors in the causation of bronchogenic cancer. The findings are fully consistent with the theory that cigarette smoking is an important factor in the causation of bronchogenic carcinoma. A high incidence of carcinoma-in-situ in non-smokers in the second phase of this study led us to believe that we were in- cluding cases of squamous metaplasia in this category. We then set up strict criteria for the recognition of carcinoma-in-situ and atypical cells and the exclusion of all other lesions from that category. Thits, in our final study, we did not find a single instance of a carcinoma-in-situ in the non-smokers or occasional smokers and only 0.3 per cent of the slides in less than half a pack a day cigarette smokers showed these findings. Those who smoked two packs a day or more showed 11.4 per cent of the slides with carcinoma-in-situ and those who died of lung cancer had 15.0 per cent so involved. SUMMARY AND CONCLUSIONS During a six year period a systematic histopathological study of the epithelial changes in the tracheobronchial tree in relation to lung cancer and to cigarette smoking has been carried out. Four changes in the epithelium were evaluated: basal cell hyperplasia, stratification, squamous metaplasia and carcinoma- in-situ. Our histologic studies show that among people who died of causes other than lung cancer, basal cell hyperplasia, stratifica- tion, squamous metaplasia and carcinoma-in-situ showed a pro- gressive increase as the smoking habits increased. In our final study there were no slides showing carcinoma-in-situ in the non-smokers or occasional smokers. Very few such lesions were seen in men who smoked less than one pack a day. There was a Changes in Tracheobronchial Tree 143 sharp increase of such findings in the heavy smokers. The same, but more extensive changes were observed in those who died of carcinoma of the lung. Our studies lead us to conclude that cigarette smoking is today the single most important cause of bronchogenic cancer in men. ACKNOWLEDGMENT The writers gratefully acknowledge the cooperation of the Medical Illustration Department of the Veterans Administration Hospital, East Orange, New Jersey. REFERENCES 1. Wittekind, D., and.Struder, R.: Beitrag zur Histogenese des Bronchialcarcinoms. I. Uber Epithehnetaplasie in Bronchialbaum. Frankfurt. Ztschr. f. Path., 64: 294-311, 1953. II Uber die Beziehungen zwischen Epithelmetaplasien und Carcinombildung im Bronchialbaum. Frankfurt. Ztschr. f. Path., 64: 405- 437, 1958. 2 Auerbach, 0., Petrick, T. G., Stout, A. P. Statsinger, A. L. Muehsam, G. E., Forman, J. B., and Gere, J. B.: Anatomical Approach to Study t>f Smoking and Bronchogenic Carcinoma: Preliminary Report of Forty-one Cases. Cancer, 9: 76-83, 1956. 3. Auerbach, 0., Gere, J. B., Forman, J. B., Petrick, T. G., Smolin, H. J., Muehsam, G. E. Kassouny, D. Y., and Stout, A. P.: Changes in the Bronchial Epithelium in Relation to Smoking and Cancer of the Lung. New England J. of Med., 256: 97-104, 1957. 4. Bowen, J. T.: Precancerous Dermatoses: Study of Two Cases of Chronic Atypical Epithelial Proliferation. J. Cutan. Dis., 30: 241-255, 1912. 5. Cullen, T. S.: Early Squamous-cell Carcinoma of Cervix. Surg., Gynec. & Obst., 33: 137-144, 1921. 6. Foote, F. W., and Stewart, F. W.: Anatomical distribution of intraepithelial epidermoid Carcinoma of Cervix. Cancer, 1: 431-440, 1948. 7. Black, H., and Ackerman, L. V.: Importance of Epidermoid Carcinoma in Situ in Histogenesis of Carcinoma of Lung. Ann. Surg., 136: 44-55, 1952. T201011

12 HISTOLOGICAL TYPES IN RELATION TO ETIOLOGICAL FACTORS AND TRENDS IN LUNG CANCER LRIV KREYBRRG, M.D. usuALLY lung cancer morbidity and mortality statistics refer to cases registered under nos. 162 and 163 in the International list. Already in 1937, in a discussion in the Medical Society in Oslo, I emphasized the very heterogeneous tumour material embraced by the term "lung cancer" and stressed the necessity of distin- guishing between the different histological types, when etioligical factors being studied, as presumably the different histological types also represent biologically different tumour entities. In addition, always looms the uncertainty as to the tumour in question actually being a primary lung tumour or not. Not until after the war came the opportunity to study this question systematically. At this time also the manifestation of a new lung cancer situation could be perceived, adding to the pertinency of such studies. The new situation is characterized by: a considerable and progressive increase in lung cancer cases in males in urban areas, and a much lower increase in males in rural areas and in women, Figure 1. More thai, ten years ago I started collecting as many lung cancer cases as possible in Norway. A minor part were autopsy cases, the major part surgical, including a considerable biopsy material. From the very beginning of my effort to classify the pulmonary epithelial twnours according to histological types, I met method- Histological Types, Etiological Factors and Trends LUNG CANCER MORTALITY 24 r 22 20 le 16 14 12 10 8 6 4 2 in NORWAY 29-31 34-39 39-41 45-47 49-51 54-58 57-5m Incidence per 100 000 per year adjusted to standard population cDc Pedenen. 1950 The Norwegian Cancer Registry) Fig. 1 145 ological difficulties on one hand and a good deal of skepticism from many colleagues on the other, both natural obstacles, in- timately related. The position can not be better exemplified than by the contrast between my recent papers and a paper from 1955 of my esteemed colleague professor Trauchi (1955). I therefore find it necessary to discuss my criteria in some detail. 144 1 T201012

146 Tobacco and Health If we study a series of average primary so-called bronchiogenic carcinomas we admittedly find a great variance in the size, the form and the differentiation of the cells, even within one and the same tumour. Some cells are small, others may be middle sized and others again large. Some cells may be round, others poly- hedral, others again more or less oval to spindle shaped and some even take the shape of giant cells. Accordingly it is possible to apply a number of purely morphological designations: "small round cell," "large round cell," "fusiform," "giganto-cellular" carcinomas. If we come to cytological differentiation, we may find some cells producing mucus, others not, and mucus of different kind if finer methods are employed for identification. By some stains, as for instance mucicarmine, we identify a smaller number of mucus-producing cells than by the use of the alcian stains. A technical warning must be sounded. Sometimes necrotic nuclei are stained blue or green, and even certain degeneration products of the cytoplasm may take the stain and mimic mucus. Next, the mucous glands of the bronchial walls may be infiltrated and embedded in the tumour tissue, more or less completely destroyed, and remnants of the mucus be taken as a product of the tumour cells proper. A strict analysis and a critical judgment is necessary, to avoid incorrect conclusions. The other important cytological differentiation is cornification with all stages up to complete keratosis with kerato-hyaline granules. Very often the keratinization is manifested by only a tendency of the cytoplasm to stain with eosine, orange G, saffron and similar stains, which, however, are not specific. Many degenerating and necrotic cells may take acid stains and resemble keratotic cells. A decision as to the true significance can some- times be very difficult or impossible. These various cytological differentiations may be used in the choice of designation of the tumour, such as: "keratinizing carcinoma," "mucus producing carcinoma," etc. If we next come to histological differentiation we may find: solid sheets, ribbons or garlands, buds and strings, we may find acinar or alveolar structures, even papilloma formation, but also a completely anaplastic picture. Histological Types, Etiological Factors and Trends 147 Again designations may be chosen from the actual histological pattern such as: "carcinoma solidum," "adrenocarcinoma," "car- cinoma anaplasticum," etc. In the different systems of classification a great many combina- tions of the characteristica mentioned may be construed. Pro- fessor Trauchi's tabulation is an example of the many possibilities offered and actually used. In my own work I have chosen another path, that of simplifica- tion. I tried to extract the characteristica which I considered essential, and tested these on biological criteria, such as age, sex, domicile, occupation and tobacco smoking, as well as on the general epidemiological pattern of lung cancer. First a few words on my morphological criteria. Epidermoid carcinoma is the designation for all tumours where any trace of cornification is found, due regard being paid to tech- • nical errors and to possible degenerative changes. The cytoplas- matic differentiation can be very marked or very slight. Usually also there is a tendency to the formation of sheets, ribbons or whorls, but very anaplastic tumours can also be found. Small cell anaplastic carcinomas can be polyhedral or oval celled, showing the cells in schools, in ribbons or garlands, some- times with a tendency to rosette formation. Personally I am convinced that these small celled tumours are only variants of the epidermoid carcinomas and transitions are seen: epidermoid carcinomas with marked ribbon-growth of small cells, resembling the carcinomas in situ of the cervix epithelium and with very slight keratinization, and small cell anaplastic carcinomas with tiny areas of keratinization. Complete absence of keratinization may be stipulated as a criterion, and this type will accordingly be more limited. As will later be seen these two histological types behave in a practically identical manner, when confronted with the biological conditions mentioned above. Adenocarcinoma is the next large and important type. Here the basic morphologic criterium is gland-structures. If such structures are found in any part of the tumour the base for the diagnosis is established, even if large parts are anaplastic,-large celled, or giant celled or with the cells growing in solid heaps as T201013

148 Tobacco and Health a "carcinoma solidtun:' Warning must be given that the epider- moid carcinomas may mimic gland structures, when the cells of the periphery invade alveoles. Cytoplasmatic differentiation, in one or anothe; part of the tumour, with mucus formation, or keratinization may guide to a safe conclusion as to type. A great variety of adenocarcinomas are found, including the tumours designated giant cell carcinomas of the lung (Nash and Stout, 1958). Combined epidermoid-adenocarcinomas are found in human material, but they are rather rare,-a few per cent or less. The bronchiolo-alveolar carcinomas have the well known characteristic pattern. Certain difficulties may be encountered when being cenfronted with ovarian lung metastases and some primary papillary adenocarcinomas of the lung. This problem is, however, of minor quantitative importance. Mucous gland tumours, mainly cylindromas, are limited to trachea and the larger bronchii, and these tumours are usually easily identified. The carcinoids or adenomas are, likewise, in most cases easily recognized. Sometimes, however, when deeply infiltrating, a differential diagnosis versus the small cell anaplastic carcinomas must be remembered. This histological typing and classification with definitions and nomenclature, was essentially adopted by a WHO Expert Com- mittee, meeting in Oslo in November 1958. From the purely morphological point of view, how many of a total of unselected lung cancers can now be identified according to the criteria mentioned? The best answer is not obtained from my own Norwegian material, because that was my first, and I did not keep a precise record of the number of cases discarded because of technical defects, and my own inability to come to a safe conclusion. A good answer is, however, given by a Finnish material with a total of 654 cases, studied and typed during April-July this year (1960). The material was supplied by a series of Finnish labora- tories through the co-operation of professor Saxen. As "lung cancer" unspecified were recognized all cases when the histologi- cal material permitted a diagnosis of carcinoma. This was done in in Histological Types, Etiological Factors and Trends 149 d' N ~ SEX DISTRIBUTION in a material of ~ O Primary Epithelial Lung Tumours Epidermoid Carcinomas 305 Small Cell Anaplastic C. 98 Males Females 6 6 Total /4denocarcinomas Bronchiolo-alveotar Cell C. Carcinoid Tumours Mucous Gland Tumours Total 67 10 35 7 31 10 25 3 Fig. 2. The total Norwegian material. order to be in agreement with the clinicians and the statisticians. Some of the cases could, however, not be used safely for typing, because of defects in quality or quantity of the material. A number of cases with good material, however, still left us in uncertainty as to type. In this ordinary material of routine preparations some ten per cent were discarded as unfit for typing, and of the remaining between 96 and 97 per cent could be typed and classified according to the criteria stipulated. I estimate that a similar situation has existed during my study of the Norwegian material and I may return to my own. In addition to the histological typing, information was also obtained of age, sex, main type of domicile, main or special type of occupation and smoking habits. If we examine the age occurrence of the histological types used in our classification, great differences are found, strengthening the probability of the existence of real different tumour entities (Kreyberg, 1959). Figure 2 shows the sex distribution. It is easily seen that the several types can be placed in two distinctly different groups: T201014

150 Tobacco and Health I ~ TUMOUR GROUPS N ~ DIFFERENT TIME PERIODS ~ PERIOD GROUP I GROUP II O ~ NUMBER RATIO NUMBER RATIO ~ M F M F M F M F Christiansen (1925 - 44 ) I 25 I I 2: I 15 13 1.2 : I E"+ Jakobsen (1937- 446) 46 8 6: I 24 22 I.1 : I Christiansen (1945- 552) II 40 4 10 : I 14 12 1.2 : 1 Kreyberg (1948- 60 ) 403 14 29 : I 119 66 1,8 : I Fig. 3 Group I with a most marked preponderance of males, and Group II with a nearly equal distribution of males and females, actually with a 2: 1 dominance of males. A few word:; about the sex ratio may be useful. Regardless of the formal social equality of women in many parts of the world, it is probable that males nevertheless often have greater facilities for medical supervision and care. Reference to regular medical examination of workers in factories, industrial plants and larger business concerns, as well as larger number of beds for males in many hospitals will suffice. In addition comes the greater tendency in women to resignation and noble sacrifice of own interests for those of the family, very much including those of the head of tile family. This not rarely leads to a progress of disease to stages too advanced to permit a precise diagnosis and effective treatment in women. The table, Figure 3 shows the number of cases personally typed and occurring in each sex, as well as the sex ratios, at different time periods in several Norwegian materials. It shall be men- tioned that for all the earlier periods the material is represented Histological Types, Etiological Factors and Trends 151 GROUP I : GROUP II NORWAY DIFFERENT TIME PERIODS MALES PERIOD Christiansen I NUMBER RATIO (1925-44) 25 : 15 1.7 : I Jakobsen (1937-46) 46 : 24 1,9:1 Christiansen II (1945 - 52) 40 : 14 2,9 : I Kreyberg (1948-60) 403 :119 3,4:I Fig. 4 by autopsy cases only, whereas the last period is mixed autopsy and surgical. A marked difference will be found. The Group II tumours show a fairly constant sex ratio in all periods. The Group I tumours on the other hand, show a systematic and very consider- able increase. If these figures are correlated to the graph, showing the general development of the increase in lung tumours in Norway in the same time, figure 1 periods, it can be concluded that the recent increase in the frequency in males is caused mainly, if not exclusively, by Group I tumours. All this considered, it may be inferred that the ratio Group I: Group II tumours in males may be more useful ratio to study than the sex ratio. T201015

152 Tobacco and Health ~ Histological Types, Etiological Factors and Trends 153 GROUP I : GROUP Ii N ~ GROUP I.: GROUP II NORWAY ~ ~ O NORWAY DOMICILE - MALES ~ ~ OCCUPATION - MALES ,CATEGOR Y NUMBER ~ N RATIO CATEGORY NUMBER RATIO Rural 63 : 38 1.7 : 1 "Open air" 104 : 48 2,2: I Smaller towns 159 ~ 44 3,6 :1 "Dusty" 184 ~ 46 4,0 : 1 Larger towns f38 : 26 5,3 : 1 "Sailors" 43 . I I 4,0 : I Sailors :43 : I I 4,0 : 1 "Clerical" 115 : 25 4,6 : 1 Fig. 5 The first of a series of such ratios to be examined concern differences in time, as just mentioned. In the admittedly some- what overlapping periods, the tendency is clear: There is an increase from 1,7 via 1,9 and 2,9 to 3,4 : 1. Figure 4. The next series concern the types of domicile, and here the figures 1,7 : 1 ior rural areas is contrasted to the Figures 3,6 : 1 for smaller towns and 5,3 : 1 for larger towns, mainly Oslo, the capital. It shall be mentioned that not one of the larger towns in Norway actually is large, according to world standards. Figure 5. The next table shows the ratios related to occupation, Figure 6. In a detailed study it was found that no single trade or occupa- tion in Norway showed an appreciably increased frequency of lung cancer, excepting a very small group of workers in a nickel plant. It was, therefore, decided to pool and analyse the material in larger occupational categories. The "open air" male workers will mainly consist of people Fig. 8 occupied with agriculture and fisheries, and accordingly be living under "rural" conditions. It is, therefore, not surprising that the ratio is low. The most interesting finding is, however, the heavy load on the "clerical" workers, that is people connected with business, administration and liberal professions. They show an even higher frequency than the people in "dusty" work, which means all kinds of work connected with dusts and fumes. The next table, Figure 7, shows the ratios correlated to smoking levels, one cigarette is counted as one gramme, and all kinds of tobacco smoking is comprised. Again a systematic pattern is found, which is still more visible from the following graph plotted on semi-logarithmic paper, Figure 8. The curve is so perfect that I cannot blame any person suspecting that the curve is an artefact. That the curve in spite of its perfection most probably is correct, "is supported by the following two curves. The first, Figure 9, of the pair represents the same figures as T201016

154 Tobacco and. Health 0® N ~ ~ GROUP I : GROUP I) ~ O NORWAY ~ SMOKERS MALES ~ ~ H GRAMMES SMOKED NUMBER RATIO PER DAY 0 4: 13 0.3: 1 -4 5:Ii 0.5:1 - 9 69 : 31 2.2 : 1 -14 147 :39 3.8:1 -19 67:12 5.6:1 - 29 75 : 9 8.3 : I 30 + 36 : 4 9•0 : 1 Fig. 7 in the previous graph plotted to indicate the risk to develop lung cancer Group 1, using the ratio Group I: Group II in non-smokers as zero line, that is, representing the risk to develop Group I tumours in non-smokers. The form must necessarily repeat the pattern of the semi- logarithmic curve, but now for a change, the non-logarithmic gives the straiglrt line with a certain area, namely that covering the amount smoked between 4 grammes and 30 grammes. With still heavier smoking the relative risk is increasing much less. The second curve, Figure 10, of the pair gives a nearly identical indication of the risk, and this curve, at least, cannot be an artefact. The ti:rnours of a British material plotted here, I typed all without any knowledge as to sex, age, smoking habits or Histological Types, Etiological Factors and Trends 155 9,0 8,3 GROUP I : GROUP II NORWAY SMOKERS 0 -4 -9 -14 -19 -29 30+ GRAMMES SMOKED PER DAY Fig. 8 previous histological diagnosis. The result of the typing was despatched as a list with only slide numbers and findings, and then my British colleagues Doll and Bradford Hill plotted the curve against the smoking levels without any further contact with myself. Not one of my diagnoses was discussed or revised. The British and the Norwegian materials finally give nearly identical figures for the risk to develop Group I tumours, related to the smoking levels. It is here not alone a question of near- identical findings in two different materials, but two different techniques have actually been employed. T201017

156 Tobacco and Health 29,7 RISK TO DEVELOP GROUP I TUMOURS CALCULATED FROM RATIO GROUP I: GROUP II / / / 0 -4 -9 -14 -19 -29 30+ GRAMMES SMOKED PER DAY Fig. 9 In 1956, for the first time I produced figures showing that the ratio can be used to estimate the risk as just elaborated. The underlaying assumption was that the Group II tumours represent a fairly constant factor. The British material was studied on the basis of the rate observed for the two groups at different smoking levels. The rate is not at all changed for Group II tumours. The curve strictly follows the zero line. The British material, therefore, not only confirms the reality of the original Norwegian findings, but also proves that the assumption of a relative constancy of the Group II tumours is correct. I ~ ~ Histological Types, Etiological Factors and Trends 157 ~ ~ SMOKING LEVELS ~ ® and Tumour . Groups ~ ~ / 25:I 10 • Group I Tumours •/ - ~ / o ~ • Group II Tumours / E / w o / c: 20:1 / vs / 0 / E ro / / w 15~1 / Z • / .. 2 / / f0 10;1 / I VI / m 5:1 ro / E / w 2:1 / I;I ~...........• .............. 1P ............. ~1.. 5 10 15 20 25 30 Av¢rage amount smoked (gramm¢s per day) Fig. 10. The British material typed by Kreyberg (Doll, Bradford Hill and Kreyberg). The final picture, Figure 11, shows the rate curves for "lung cancer" in Finnish and Norwegian males and females correlated to the ratios Group I: Group II tumours. The remarkable corre- spondeiYce again is evident. We have now two entirely different and independent materials where the typing has been done without any knowledge as to clinical facts. In the British material the ratio was related to smoking levels. In the Finnish material the typing was related T201018

158 Tobacco and Health 70 -I 0 0 0 40 0 ro ~ 20 ~ 10 "LUNG CANCER" FINLAND - NORWAY R _ RATIO F= N = ---- R=I1,7:I R= 3,4 : I - 4 R=0.6:1 ~ --------' ---Q----RW0.2:I 34-36 39-4i 44-46 49-51 Calendar years 54-56 57-58 to epidemiological observations. In both cases the findings are clear cut, unequivocal and systematic. The material so far presented is so systematic and consistent that the conclusion seems unavoidable: that the types listed and the criteria stipulated are based upon cyto-histological realities, and that the types also biologically represent oncological entities. The thesis that the ratio Group I : Group II tumours is useful in the study of etiological factors, thereby, seems well established. The second thesis, that the ratio may also be very useful in ascertaining the lung cancer incidence total in a population where a fair statistical service is not available on one hand, and a general chest disease centre with a representative clientele exists on the other hand, is supported by our findings, with the one provisio, that the other conditions are similar. This provisio may hold good for certain ethnic and cultural areas, but may be lacking in others. i ,! Histological Types, Etiological Factors and Trends 159 Here the ratio again may be a useful technical means for comparative geographical studies. REFERENCES Doll, R., Hill, A. B., and Kreyberg, L (1957): Brit. J. Cancer, 11, 43. Kreyberg, L (1937): Norsk Meg. f. laegevid., 98, Suppl. Forh, i Det Norshe Medi- cinske Selskab; 36. Kreyberg, L. (1956): Brit. J. Prev. Soc. Med., 10, 145. Kreyberg, L (1959): Unio Internett contra Cancrum ACTA, 15, 78. Nash, A. D. and Stout, A. P. (1958): Cancer, 11, 369. Tauchi, H. (1955): Nagoya Med. j., 3, 1. WHO (1960): Epidemiology of Cancer of the Lung. Techn. Report Series, no. 192, Geneve. T201019

Session III THE INTERPRETATION OF STATISTICAL AND EPIDEMIOLOGICAL DATA RELATING SMOKING TO DISEASE T201020

13 SMOKING AND CANCER: RETROSPECTIVE STUDIES AND EPIDEMIOLOGICAL EVALUATION MoaTOrr L. LEVZx, M.D., Da. P.H. T'HE effect on man of chronic exposure to tobacco smoke has long been the subject of medical investigation, specu- lation and surmise. An association between smoking and cancer of the lip, tongue and buccal cavity has been an accepted medical observation for a generation. In recent years interest in the subject has been further stimulated by the phenomenal increase in reported lung cancer incidence and mortality during the past thirty years' and by the concomitant.increase in the prevalence of cigarette smoking. As a result there has been a great deal of work on the biological, chemical, physical and pharmacologic proper- ties of tobacco smoke as well as many studies of the clinical, pathologic and epidemiologic aspects of the use of tobacco and the occurrence of disease and increased mortality. Much of this material has ' been presented or summarized during this symposium. We are now concerned with the evaluation of the accumulated data from the standpoint of the control of disease in the human population. Knowledge of the etiology of human disease has often stemmed from observations made first on diseased persons and later ampli- fied and confirmed by laboratory and experimental studies. This has been particularly true of chemical carcinogenesis. In evalu- ating the possible role of an environmental agent such as tobacco with respect to disease, the relevant information may include- 163 T201021

164 Tobacco and Health (1) the chemical, physical and biological properties of the agent, (2) observations on human groups with respect to exposure to the agent and (3) all other knowledge regarding the etiology of the disease or group of diseases under consideration, to deter- mine whether this knowledge is consistent with the hypothesis that the agent is an etiological factor. You have already heard discussions of the chemical, physical and biological properties of the combustion products of tobacco. From the standpoint of their possible role in the etiology of can- cer, certain general features of carcinogenesis are pertinent. One of the earliest established observations in chemical carcinogenesis was that combustion products of organic material may contain carcinogenic materials. Clinically it was early (1775) noted2 that exposure to imbedded coal soot increased the incidence of carcinoma of the scrotum in chimney-sweeps. The carcinogenicity of coal tar was confirmed experimentally in 1915 by Yamagiwa and Ichakawa.g In 1925, Kennaway showed that pyrolysis of a number of organic substances-petroleum, coal, skin, hair, yeast, cholesterol-produced carcinogenic tars.' Hieger and Cook and Hewitt (1933) identified one of the principal carcinogenic chemi- cals produced in this way as benzpyrene.4 Experimental production of the same type of tumor by different chemical or physical agents and production of different types of tumors by the same agent is also a rule rather than the exception in the field of carcinogenesis. Pulmonary adenomas in animals can be produced by at least 15 different chemicals. In man, occupational exposure to several inorganic chemicals-chromates, asbestos, uranium, nickel carbonyl-are known to be associated with increased incidence of malignant pulmonary tumors. Radia- tion, viruses, chemicals all produce multiple tumors experi- mentally. Another important observation in experimental carcinogenesis is that substances carcinogenic for one species of animal or for one tissue are rot necessarily carcinogenic for other species or other tissues. St)ecies and strain differences in susceptibility to the same carcinogenic agent are well known and must always be taken into account in experimental carcinogenesis. Tobacco smoke condensate is a complex substance containing Retrospective Studies and Epidemiological Evaluation 165 a large number of different chemicals, including several poly- cyclic hydrocarbons previously reported to be carcinogenic in animals. In addition, tobacco smoke condensate has been shown to be carcinogenic for at least one animal tissue, the skin of mice. This array of facts by themselves constitute presumptive evidence that tobacco combustion products belong to the class of chemical carcinogens. There remains the question whether human observations provide any additional evidence for this conclusion. If tobacco products are carcinogenic to some species but are not to man, two types of observations might be expected: first, persons who develop malignant tumors should have about the same distribution of smokers and non-smokers as other persons and second, if we identify a group of smokers and non-smokers and continue to observe them, the incidence of malignant tumors _ should be about the same in each. The first type of study is what is referred to as "retrospective" and the second as "prospective:" There have now been reported 26 retrospective studies from various parts of the western world of the prevalence of smokers among patients with various malignant tumors and in control groups.s We have previously reported two such studies, and I should now like to report a third based on the smoking histories of over 10,000 male patients seen at the Roswell Park Memorial Institute at Buffalo from 1945 to 1956. In 3016 patients with cancer of the lung, larynx, lip, buccal cavity, pharynx, bladder and esophagus, the percentage of non- smokers was significantly lower and the percentage of cigarette smokers significantly higher, than in 7146 control patients (Table 1). In this tabulation, age and residence distributions have been equalized in each group. These data are similar to the findings of many other such studies. If we consider these data as representa- tive of the relative smoking distribution of the cancer population and the general population, we can arrive at an approximate estimate of the relative risk of developing cancer which they indicate. Such relative risks are directly observed, rather than estimated, in a prospective study. For the seven forms of cancer which we found to have signifi- cantly higher proportions of smokers, the relative risks for ciga- T201022

I 166 Tobacco and Health rette smokers compared with non-smokers are estimated from the crude data and from the age adjusted and residence adjusted data (Table 2). The adjustments for age and residence do not alter the relative risks in any striking way. The highest relative risk among cigarette smokers is for lung cancer, the lowest for cancer of the esophagus. Because of the importance of lung cancer and its known association with certain occupational exposures, Table 3 shows the indicated relative prevalence of lung cancer when smokers and non-smokers are equalized with respect to occupation as well as age and residence. (The relative risks are not precisely the same in the previous tables because Table 3 relates to a smaller series of cases. ) Adjustment for residence reduces the relative risk from 6.81 to 6.03, for occupation as well, to 5.04. These data indicate that neither residence nor occupation account for the major portion of the excess lung cancer risk of smokers. RETROSPECTIVE AND PROSPECTIVE STUDIES When the results of the first retrospective studies were published, there was considerable discussion regarding the pos- sible role of bias in creating an artificial excess of smokers in the disease group as compared with the control group in such studies. In lung cancer, bias might explain this finding in a hospital study if, for example, it were postulated that lung cancer cases who smoked were more likely to enter the hospital than lung cancer cases who did not smoke. Although it has been shown that, given the proportion of lung cancer cases known to be hos- pitalized, this particular type of bias would be unlikely to produce the reported results,° the criticism of some possible bias was an important factor motivating the carrying out of prospec- tive studies. By and large, the results of the two types of study have confirmed each other. In Table 4 is presented a comparison of the retrospective data from Roswell Park Memorial Institute with prospective data recently reported by Harold F. Dorn. On the basis of both retrospective and prospective studies, it is now well established that cigarette smokers experience in- creased incidence and mortality from a number of diseases, of which lung cancer is first in terms of the excess incidence com- Retrospective Studies and Epidemiological Evaluation 167 pared to that among non-smokers. The mechanism of this in- creased incidence continues to excite controversy. Precisely which constituents of tobacco smoke are responsible for the increased cancer incidence is not known and how this effect is produced is as obscure as is true of carcinogenesis in general. In the human population, the available data, including retro- spective and prospective studies, as well as pathological studies of pulmonary cytology in smokers and non-smokers, are consis- tent with an etiological role of exposure to cigarette smoke in producing increased incidence of lung cancer particularly, as well as cancer of the larynx, lip, buccal cavity, and probably the esophagus and the urinary bladder. The known chemical con- stituents of tobacco smoke and the carcinogenic effect of cigarette smoke condensate on mouse skin provide further evidence for this conclusion. It should be noted also that, disregarding what particular diseases are increased in the smoking population of males, an increased mortality rate in this population, first reported by Pearl,' has also been noted in the three prospective studies reported in recent years.°, °, 10 It has been suggested by Yerushalmy and Palmer," and by Berkson,'$ that the findings of the retrospective and prospective studies can also be explained, in part or in whole, by assuming that smokers, and particularly cigarette smokers, happen to have other characteristics which predispose both to the acquisition of the habit of smoking and to increased risk of various diseases and to a higher than normal mortality rate. It seems reasonable to suppose that smokers are different from non-smokers in other characteristics than exposure to tobacco, either as a result of such exposure, or as a result of factors predisposing to or associated with tobacco habituation. A number of studies have described small psychological differences between smokers and non- smokers,l' for example. It is obvious that any human group classified according to a characteristic determined in part by the individual's choice is to that extent self-selected and also different from the group of individuals who did not make that choice. Some selection of this type usually can be postulated regarding human groups not selected by experimental procedures, T201023

168 Tobacco and Health and hence obtains with respect to most human population obser- vations. It should be noted that, if the presence of other characteristics distinguishing smokers from non-smokers is to provide an alternative explanation for the observed increased disease incidence or mortality, these characteristics should (1) be at least as strongly associated with the presence of disease as is smoking13 and (2) should provide at least as good a biologic explanation which is consistent with other known facts about the etiology of these diseases. No other characteristics of smokers which have been observed fulfill either of these criteria. Hence evidence is not available for accepting the hypothesis that some other characteristic than exposure to tobacco smoke offers a better explanation for the increased disease risk which has been established. One of the arguments advanced for selection as an added or alternative hypothesis for the smoking-disease association has been that association with more than one form of cancer and with several disease entities is inconsistent with a causal relation between an environmental agent and these multiple sites of can- cer and disease.l', 'a It is difficult to find a basis in logic or in biology for this reasoning. From the standpoint of logic, this argument apparently rests on the principle that the same cause should always produce the same effect. This principle is not applicable to a situation where one of the postulated causes, tobacco smoke, is not a single pure substance, but a complex mixture of over a hundred chemicals and alkaloids, and in which this complex substance is applied to a heterogeneous biological population, with varying susceptibility to disease. Biologically, it is a commor.place that carcinogenic agents even though rela- tively pure often produce multiple types of cancer and also multiple biologic effects. A single dose of ionizing radiation applied to the znouse, for example, can produce neoplasms of the pituitary, thyroid, lung, mammary gland, adrenal cortex, intestine, ovary, myeloir'. leukemia, and non-neoplastic lesions including cataracts, iris defects, life-span shortening, skin-greying general- ized arteriosclerosis, nephrosclerosis and ovarian atrophy.l* SUMMARY 1. The chemical and biological properties of tobacco smoke Retrospective Studies and Epidemiological Evaluation 169 condensate are sufficient to classify it among other chemical carcinogens. 2. Human retrospective studies_ are consistent with the hy- pothesis that exposure to this material increases the incidence of malignant tumors of the lung, larynx, mouth, bladder and esopha- gus in man. TABLE 1 CANCER AND SIS08IN0 HIBTOR7 MALES Roawell Park Memorial Institute 1945-56 Site N umber of Caaea Pe Non r Cent' -Smokere Per Cent' CiyareUe Only Lung 705 6.8 66.8 Larynx 333 7.5 61.0 Lip 469 14.4 45.5 Buccal Cavity 657 10.3 48.0 Pharynx 293 12.5 53.9 Bladder 377 17.3 52.1 Esophagus 182 16.9 48.7 Controls 7146 25.7 39.2 ' Adjusted for age and residence. TABLE 2 RHLA'rIVA RIBIC CANCER Ci®arette Smokers Compared to Non-Smokers Adjuated fo r Site of Cancer Crude Age Age an d Residence Lung 6.80 6.30 6.30 Larynx 5.24 5.61 5.38 Lip 1.57 2.02 2.07 Buccal Cavity 2.78 3.15 3.05 Pharynx 3.04 3.04 2.82 Bladder 1.77 1.94 1.99 Esophagus 1.65 1.71 1.58 T201024

170 Tobacco and Health TABLE 3 RELATIVE RISK LUNG CAxcER MALES RETR08PECTIVE STUDY Adjuated for Not Age, Age, Residence, Adjusted Residence Residence Occupation Non-smokers 1.00 1.00 1.00 1.00 Other Smokers 1.35 1.30 1.40 1.39 Cigarette Smokers 6.81 6.03 5.45 5.04 TABLE 4 RELATIVE RISK Cigarette Only vs. Non-Smokera Site Prospective Study' Retrospective Study2 Lung 9.85 6.30 Mouth, pharynx, esophagus 2.18 2.55 Bladder 1.93 1.99 s Harold I)orn, Social Statistics Section Proceedings of the Am. Statistical Assoc., 1958. 2 Roswell Park Memorial Institute, 1945-56. 3. No single mechanism of carcinogenesis is postulated by these data, as is true of other etiological observations in carcinogenesis. 4. From the clinical and public health viewpoint these data fully justify acceptance of an effective etiological relation between smoking and malignant tumors, particularly of the respiratory tract. This conclusion has been reached also by a number of official and voluntary scientific bodies and by national and state public health agencies.'$, 'e•17 As in the interpretation of any body of evidence, it is possible to suggest alternative hypotheses, and in so doing we may gain new insight into some of the basic problems relating to the philosophy and study of disease causa- Retrospeetive Studies and Epidemiological Evaluation 171 tion in man. From the pragmatic and practical standpoint, we must agree that "the sum total of scientific evidence establishes beyond reasonable doubt that cigarette smoking is a causative factor in the rapidly increasing incidence of human epidermoid carcinoma of the lung," and that this evidence is adequate for application in clinical preventive medicine and for "the initiation of public health measures.°'" REFERENCES 1. Dom, Harold F.: The Increase in Cancer of 23:6, 253-257, June, 1954. 2. Pott, Percival: Chirurgical Observations, 1775. the Lung; Ind. Med. & Surgery, 3. Yamagiwa, K, and Ichakawa, K.: Experimental Study of the Pathogenesis of Epithelial Tumours; Mitt. Med. Fak. Kaiserl. Univ. ZU Tokio,15:295-341,1915. 4. Haddow, Alexander: The Chemical and Genetic Mechanisms of Carcinogenesis, Chapt. 13, The Physio-Pathology of Cancer, 2nd edition, edited by Freddy Homburger, Hoeber-Harper, 1959. 5. Davies, D. F.: A Review of the Evidence on the Relationship between Smoking and Lung Cancer, J. Chron. Dis., 11:579,1960. 6. Levin, M. L.: Smoking and Lung Cancer: A Review; Proc. 3rd. Nat. Cancer Conf., 1957, pp. 473-478, Lippincott. 7. Pearl, Raymond: Tobacco Smoking and Longevity, Science, March 4, 1938. 8. Doll, R., and Hill, A. B.: Lung Cancer and Other Causes of Death in Relation to Smoking: A second Report on the Mortality of British Doctors, Brit. Med. 1., 2 (1956), 1071-1081. 9. F. Cuyler Hammond and Daniel Horn: Smoking and Death Rates; J. A. M. A., March 15, 1958, Vol. 166, no. 10, pp. 1159-1172, no. 11, pp. 1294-1308. 10. Dorn, H. F.: The Mortality of Smokers and Non-Smokers, Proceedings, Social Statistics Section, Dec. 27-30, 1958; American Statistical Association, Washing- ton, D. C., May 15, 1959, p. 34-75. 11. Yerushalmy, J. and Palmer, Carroll, E.: On the Methodology of Investigations of Etiologic Factors in Chronic Diseases; ]. Chron. D'u., 10:27, 1959. 12. Berkson, Joseph: Smoking and Lung Cancer; Some Observations on Two Recent Reports, J. Am. Stat. Ass., March 1958, Vol. 53, pp. 28-38. 13. Lilienfeld, A. M.: Emotional and Other Selected Characteristics of Cigarette Smokers and Non-Smokers as Related to Epidemiological Studies of Lung Cancer and other Diseases; 1. Nat. Cancer Inst., Feb., 1959, Vol. 22, No. 2, p. 259-282. 14. Furth, Jacob: A Meeting of Ways in Cancer Research; Thoughts on the Evolu- tion and Nature of Neoplasms; Cancer Research, April 1959, Vol. 19, No. 3, pp. 241-258. 15. Cigarette Smoking and Lung Cancer. Bureau of Cancer Control, New York State Department of Health, 1959. 16. Epidemiology of Cancer of the Lung. Report of a Study, Group. World Health Organization Technical Report Series, No. 192, Geneva, 1960. 17. Joint Report of the Study Group on Smoking and Health, Science, 125:1129, 1957. T201025

14 DEATH RATES AND CAUSES OF DEATH OF SMOKERS AND NONSMOKERS HAnoLn F. Doxrr, Ph.D. IFROM the moment of birth, an infant is exposed to a wide variety of environment agents and factors which, in com- bination with his genetic constitution, determine his length of life. Some hostile agonts such as bacteria and viruses are invisible and omnipresent. An individual by his own efforts cannot success- fully defend himself against the deleterious effects of such agents upon his health and longevity but must depend upon advances in medical knowledge to provide the necessary means for pro- tection. The amomnt of exposure to another group of environmental factors is determined by an individual's choice of residence and occupation. In most instances this choice is not made for health reasons, so that while in principle each person can exercise some control over the effects of occupation and residence upon health and longevity, in fact this is done only to a limited extent. A third group of factors affecting health and longevity are those arising from personal habits and modes of living. Of all the environmental agents and factors influencing health these are the most susceptihle to regulation by a person's own actions. The most important examples of this group of factors are those arising from habits of eating, drinking, and the use of non-alimentary substances such as tobacco. The relationship of the last of these to mortality rates is the subject of this paper. Warnings against the alleged harmful effects of tobacco upon health date frorn the introduction of tobacco into Europe more Rates and Causes o f Death: Smokers and Nonsmokers 173 than four centuries ago. These warnings, based largely on moral grounds and personal opinion, were in the main unheeded since the use of tobacco spread rapidly throughout the entire world and soon became one of the most universal human habits. The first attempt to determine in a statistically valid manner the effect of the use of tobacco upon longevity was made by Raymond Pearl in a study published in 1938 (1). This study, based on an analysis of the histories of 6,813 men, showed that the death rate of heavy smokers was twice that of non-smokers between 30 and 40 years of age. Thereafter the difference steadily decreased so that by age 70 the death rates for the two groups were essentially equal. Pearl interpreted this decreasing differential with age as an expression of the selective effect of mortality among heavy smokers, the biologically weaker and least resistant dying at a relatively high rate during early adult life so that the survivors at age 70 were tough and resistant individuals for whom no measurable effect of smoking could be demonstrated. Pearl did not clearly describe how his data were collected. His mortality table apparently was computed from retrospective histories and did not represent the subsequent experience of a cohort of men defined during early adult life. Some doubt therefore is cast on the validity of his mortality table. Since publication of Pearl's study in 1938 many reports on the use of tobacco by persons with various diseases have appeared. Most of the investigators used the same general method of study, namely, the collection of retrospective histories of patients with a particular disease and of a comparison group of persons without this disease. The results of these studies have been summarized by another speaker so that they will not be discussed here. Due in part to doubts concerning the validity of data collected by the retrospective method, three large scale prospective studies of the relationship of smoking to mortality were started shortly after 1950; one by Doll and Hill of physicians in the United Kingdom, a second by the American Cancer,Society of friends of volunteer workers of the Society, and a third by the National Institutes of Health of United States Government Life Insurance policyholders. Reports are now available from each of these 172 1 T201026

I Rates and Causes of Death: Smokers and Nonsmokers 175 174 Tobacco and Health 1.50 REGULAR SMOKERS 1.45 1.00 .50 NEVER DOLL USGLI ACS SMOKED HILL Fig. 1. Mortality of smokers and nonsmokers; ratio of death rate of regular smokers to that of nonsmokers; studies of Doll and Hill, American Cancer Society, and National Institutes of Health (USGLI) studies. The report by Doll and Hill is based on the experience of 24,354 male physicians at ages 35 years or over between 1 November, 1951 and 31 March, 1956; that by Hammond and Horn of the American Cancer Society on the experience of 187,783 white men between the ages of 50 and 69 from early 1952 to 1 November, 1955; and that by the National Institutes of Health on the experience of 294,155 male USGLI policyholders 30 years of ag(,- or over from 1 July, 1954, to 31 December, 1956 (2-5). Although the same general method of investigation was used in each of these studies, the amount of detailed information in the report of Doll and Hill is less than that in the other two reports so that most of the following comments will be based up- on the two studies conducted in the United States. I 30 40 50 60 AGE 70 80 85 Fig. 2. Mortality by age of smokers and nonsmokers; number of deaths per 1,000 per year, USGLI policyholders, July, ~954-Dec., 1958, T201027

176 Tobacco and Health USGLI ACS 1.58 1.5 0 Q ~ ~ I- J H a 0 ~ 1.0 0.5 0.0 SMOKED ONLY ~ LI~ NEVER PIPE 1.22 1.06 '. If'~ 100 N 80 ~ ~ 60 ~ 0 O 40 0 30 ~ D: W a 20 -r- I 1.68 ~ E:i n. ..~ 1.43 31 T- 1081.10 1.57 1.4 CIGAR CIGARETTE ONLY ONLY CIGAR CIGARETTE CIGARETTE AND PIPE AND OTHER TOTAL Fig. 3. Mortality ratios by type of smoking; ratio of observed to expected number of deaths by type of smoking USGLI policyholders and American Cancer Society Study. Total Death Rate Each of the three studies reported a higher death rate for smokers than for nonsmokers (Fig. 1). The age-adjusted death rate for regular smokers in the two studies in the United States was approximately 40 percent greater than that for nonsmokers. The smaller excess reported by Doll and Hill is due in part to a difference in classification of smoking histories and to the fact that pipe smokers apparently are a larger proportion of the total number of smokers in Great Britain than in the United States. The higher death rate of regular smokers persists at all ages above 30, the youngest age for which data were reported in any of the studics (Fig. 2). The mortality differential appears to be less after age 70. This cannot be regarded as firmly established due to the small number of USGLI policyholders at these ages. F-1 N 2 Q 10 W 0 8 W O 6 0: 5 . W m 4 ~ ~ 3 Z 2 I Rates and Causes o f Death: Smokers and Nonsmokers 30 40 50 60 70 AGE 177 80 Fig. 4. Type of Smoking and Mortality; number of deaths per 1,000 per year by type of smoking and age, USGLI policyholders, July, 1954- Dec., 1956. However, it is not unreasonable since the proportion of regular smokers more than 70 years old who use cigars or a pipe is larger than that for younger smokers. It is interesting that the relative excess mortality of smokers is as great at ages 30 to 35 as it is at older ages. Again the number of policyholders at these ages was small so that more.data are needed. In addition, it would be desirable to have mortality for smokers and nonsmokers between 20 and 30 years of age in order to see if the higher death rate of smokers exists at these ages as well. Type of Smoking Persons who have regularly smoked only cigarettes have the highest death rate of all classes of smokers (Fig. 3). Their death rate is approximately 60 percent greater than that of nonsmokers. Pipe smokers, both in Great Britain and the United States, ex- T201028

178 Tobacco and Health 2.0 1.5 1-0 0 I- 0.5 Q ~ 0 } i-J a ~_ 2-0 cr 0 ~ 1.5 1.0 0.5 0 CIGARETTES ONLY 167 ~ t 39 -96_ 1 29 20 199 1.5 -1.0 NDNE OCC <W 10'20 21'39 40' current daily numper ef cipare/4es N~M1E 0CG 1-2 3-4 5-® 9+ eurnnl dmily number of eimmre 0.5 CIGARETTE AND OTHER 172 r 123 - 107 -M-95 - 137 ryqNE OC, (10 10-20 21r39 40+ cwrrenl dally numbe• ef oimmrellt+ NONE OCG. (5 5-9 q-19 20+ current dmdy nwmtoeref pbmafule _ Fig. 5. Mortality of regular smokers by current amount smoked and type . of smoking; ratio of observed to expected number of deaths; death rate of nonsmokers = 1.00; USGLI policyholders, July, 1954-Dec., 1956. ~ CIGARETTE ONLY CURRENT AMOUNT SMOKED 40 30 20 r* 30 I L_ 40 LJ 50 __ . 60 70 80 AG E Fig. 6. Mortality of regular cigarette-only smokers by current amount smoked; number of deaths per 1,000 per year by age, USGLI policyholders, July, 1954-Dec., 1956. Rates and Causes of Death: Smokers and Nonsmokers 179 perience only a slightly increased risk of dying. Persons who have smoked only cigars or cigars and a pipe have a death rate of the same order of magnitude as pipe smokers. Men who have smoked only cigarettes experience an increased risk of dying compared to nonsmokers at all ages above 30 years (Fig. 4). The relative excess varies somewhat from one quinary age group to another but there is no convincing evidence for either an increase or a decrease in this excess with advancing age. From Figure 4 it can be seen that only men who have regularly smoked tobacco in the form of cigarettes have a greater death rate than nonsmokers after age 70. The death rate during this age range for cigar and/or pipe smokers is no greater than that for nonsmokers. Amoun+ of Smoking In the two studies conducted in the United States, men were asked to report both the current average daily number of cigar- ettes, cigars or pipefuls of tobacco smoked and the maximum number ever smoked. Reports on the maximum amount smoked were obtained from ex-smokers as well as from men who were still smoking. Both indices of the amount of tobacco smoked show a similar relationship between amount and mortality rates. For simplicity, the following comments refer only to the current amount smoked. Doll and Hill converted the number of cigars and the amount of pipe tobacco smoked into an equivalent number of cigarettes and reported mortality rates by amount for all groups of smokers. It is now known that the death rates of men who smoke cigars and pipe tobacco are significantly lower than those of cigarette smokers so that the mortality rates by amount smoked reported by Doll and Hill are not comparable to those in the reports of the American Cancer Society and the National Institutes of Health. The excess mortality of men who have smoked only cigarettes is directly related to the number smoked (Fig. 5, 7). Those who have smoked only occasionally have the same risk of dying as nonsmokers. The death rate of men smoking two or more packs per day is twice that of nonsmokers. Figure 6 shows that the effect of the number of cigarettes T201029

T i i 180 Tobacco and Health 2.5 2.0 USGLI ® ACS 0 CIGARETTE ONLY CURRENT AMOUNT SMOKED 2.23 0.5 00 1.96 N NEVER -1/2 1/2-I I-2 2+ SMOKED PACK PACK PACKS PACKS Fig. 7. Mortality ratios by current daily number of cigarettes smoked; ratio of observcd to expected number of deaths, regular cigarette-only smokers; death rate of nonsmokers = 100; USCLI policyholders and American Cancer Society study. smoked on mortality not only exists at all ages over 30 years but also that the relative excess is essentially the same throughout this age range. The same general impression of a nearly constant rela- tive excess mortality at ages above 30 years is portrayed by Figure 2 which shows death rates for all regular smokers and by Figure 4 which shows death rates for cigarette-only smokers. Furthermore, the excess is as great for men between 30 and 40 years of age as it is for those over 40. This observation raises the question of how soon after the beginning of smoking of cigarettes does the higher mortality of smokers become noticeable? The answer requires data for smokers and nonsmokers between 20 and 40 years of age. If Rates and Causes of Death: Smokers and Nonsmokers 181 cigarette smokers should have a higher death rate than nonsmok- ers immediately after they begin to smoke regularly two hy- potheses would be suggested, either the effect of cigarettes upon mortality is quickly manifest or persons who smoke cigarettes for some unknown reason have a greater risk of dying than persons who do not smoke them. Both of these hypotheses are difFicult to comprehend; neither should be given more than a passing mention until the appropriate death rates are available. Only very heavy pipe or cigar smokers-those currently smok- ing 20 or more pipefuls or nine or more cigars per day-experi- ence a greater risk of dying than nonsmokers; however, the mortality ratio is significantly greater than one only for men who smoke cigars. The number of deaths of heavy pipe smokers was too small to establish the existence of a higher death rate with certainty. Mortalify from Broad Groups of Causes Mortality rates by cause of death may be based on underlying causes only or upon both underlying and contributory causes. Mortality rates usually are computed by selecting one primary or underlying cause for each death, this cause being the disease that initiated the series of pathological changes that finally lead to the death of the person involved. For purposes of determining the effect of some habit, such as the use of tobacco, upon health it is important to know whether persons with this habit do or do not develop a specific disease. Whether or not this disease will be designated eventually as the underlying cause of death depends upon the method of selection and classification of causes of death and upon other factors not necessarily related to the conditions under which the disease originally developed. For comparability with data published by Hammond and Horn of the American Cancer Society mortality rates based on under- lying causes only will be presented for selected groups of causes. Rates for specific causes of death, including both underlying and contributory, will be presented for USGLI policyholders. The relationship between smoking and various causes of death may be considered from the point of view of the relative dif- ference in the death rate from a specific disease between non- T201030

182 Tobacco and Health smokers and smokers. For this purpose the mortality ratio will be used, that is the ratio of the observed number of deaths from a specific disease among smokers to the number of expected deaths if these smokers had experienced the same age-specific death rates as nonsmokers from this disease. Alternatively one may be interested in the relative rank of diseases that contribute to the excess mortality of smokers. For this purpose, the number of excess deaths from various diseases among smokers as compared to nonsmokers will be presented with an indication of the relative rank of each cause. When diseases and other causes of death are grouped into five main classes, (a) respiratory diseases including cancer of the lung, (b) cardiovascular diseases, (c) malignant neoplasms ex- cluding lung cancer, (d) violent and accidental causes, and (e) all other diseases, the data from each of the three prospective studies show that the greatest relative increase in the risk of death of smokers is from diseases of the respiratory system. For all regular smokers, the mortality ratio was 3.84 among USGLI policyholders and 3.74 among males in the study of the American Cancer Society. Among the respiratory diseases, by far the greatest relative increase was for lung cancer, for which the death rate among regular smokers was more than five times that of nonsmokers in both studies, being 6.6 and 5.4 respectively. This increased risk of death from diseases of the respiratory system was not experienced to the same extent by all groups of regular smokers. Men who had smoked only cigars and/or a pipe had only a slightly increased death rate from these diseases, but men who had regularly smoked only cigarettes were more than five times (5.7 and 5.3 respectively) as likely to die from respira- tory diseases as nonsmokers. An even greater difference exists between the death rate from lung cancer frr these two groups of regular smokers. Among USGLI policy'iolders the chances of dying from lung cancer were 9.35 times greater for regular cigarette-only smokers than for nonsmokcrs compared to an average increased risk of 1.60 times for regular smokers of cigars and/or a pipe. For no other group of diseases does the excess mortality among smokers closel) approach that for respiratory diseases, particularly Rates and Causes of 17eath: Smokers and Nonsmokers 183 CIGARETTE ONLY 12.45 USGLI ® ACS ~ 935 0 a a: 5.0 ~ J a 4.0 ~ a: 0 2 z.0 1.0 00 3.0 I 297 SUICIDE OTHER OTHER OTHER CORONARY RESP LUNG ACCIDENTS CANCER CVR DISEASES HEART DIS. DISEASES CANCER Fig. 8. Mortality of regular cigarette-only smokers from broad groups of causes of death; ratio of observed to expected number of deaths, death rates of nonsmokers = 1.00; included underlying causes only; USGLI policyholders and American Cancer Society Study. lung cancer (Fig. 8). The mortality ratio significantly exceeded one for each of the three other groups of diseases for which data are shown in Figure 8 but in no instance was the excess as much as two-fold. Regular cigarette-only smokers experienced higher death rates than cigar and pipe smokers from each of these groups of diseases as well as from diseases of the respiratory system. Regular smokers, irrespective of the form in which they used tobacco, were no more likely than nonsmokers to die a violent or accidental death. Mortality from Specific Diseases The death rates presented below for specific diseases are based on both underlying and contributory causes of death and repre- T201031

184 Tobacco and Health N0. Oz DE~ThS Cmnemr. iWmq 197 uu-out 1 52 1r ~ 217 Cmme mem~phmq~Wx ~ 2 16 bmccmleq cmvilyD.a<0.rau24 Cmneer, Almdder ueD cmmcmq mtoTmch tr50 Lylnphmmmx 1z00-2001 Cmmemr, Doneremm (16T) Cmncmr, Inlnllnn, lmCIWnA 1162-IW.1 Cmncmr, hldnmy I1e01 29 39 56 31 94 I6MI_00 LwMemim txon 23 Cmncmr. mthmr 66 Pnmmmrentm 145 Leo-4rn ®ronchin. 1 •Imphyxen~m, e Ic.l'aoo3xtf47 I W99 I ~13 *0 0 1.61 1 I .27, t t) I 2 3 4 N0. OF DEATNS Cmrmnmry 1420 hmm+l dlmmmem/un NmnrhmWlmmtlc mndmtmrdNlm Nv-azt 135 Hypertmnelmn rlth _,279 hemrl dimmmrml m 1.63 G.nermi 243~ 1.46 mrlarlo.clmrmdiHSmO Hy2e.rlamxbn vnllhmmt hemrl dinmmel.e.-..xt82 1 41 CermArml vm.oWlmv 261 1.33 tetimn. U3m-]3a Chrmnle Mplirltt. ISfx-SNt Chrmndm rheWlnmlle heal dlsmme0IUI0-.IN 26 04 42F.4 I 36 I 0 Pmrm1711. 0011mn.13501 9 0thmu ddmem.ms of liver, pmrlCrmm.lmmz-ur45 1 P5 Dam~.am 1260) 84 I lm L Uloen mt elm+mmoh duodenWlmlest-ean ClrrhmNx m4 Ilvmr /emnl ~2.63 61 62 _ 2.96 I 1 2 3 MORTALITY RATIO Fig. 9. Mortality of regular cigarette-only smokers from specific diseases; ratio of observed to expected number of deaths; death rates of nonsmokers and occasionally-only smokers = 1.00; includes underlying and contribu- tory causes; USGLI policyholders, July, 1954-Dec., 1956. sent the experience of USGLI policyholders, July, 1954-Decem- ber, 1956. If a person had both diabetes and coronary heart disease he was counted both in the group of deaths assigned to diabetes and in the group assigned to coronary heart disease. Consequently, the sum of the observed number of deaths assigned to specific diseases is greater than the total number of persons who died. For men who had smoked only cigarettes the death rate from the following diseases was more than twice the corresponding rate for nonsmokers-hmg cancer; cancer of the prostate; cancer of the buccal cavity and esophagus; bronchitis, emphysema, and allied respiratory diseases; cirrhosis of the liver; and ulcers of the stomach and duodenum (Fig. 9). Mortality ratios significantly greater than one also were found for lymphomas, cancer of the stomach, and cancer of the bladder. The death rates from the principal cardiovascular diseases ranged from 33 to 63 per cent greater for regular cigarette-only rse 2.75 103 Cancer of Poncreos (157) 67fy looc~ 2ao~'~ 3.00 r i'" .. 6 -10 5Lf,(J 62 185 ~ 5 4 3 2 I 0 0 1 2 3 MORTALITY RATIO MORTALITY RATIO Fig. 10. Mortality of regular cigarette-only smokers from specific diseases by current amount smoked in 1954; ratio of observed to expected number of deaths; death rates of nonsmokers and occasionally only smokers = 1.00; includes underlying and contributory causes; USGLI policyholders, July, 1954-Dec., 1956. t T201032 Rates and Causes o f Death: Smokers and Nonsmokers 100 1 t I I CIGARETTE ONLY i Bronchihs,E~ah'ysemo,efe Cancer of Lung (/62I63)I 1 ~; p e f500-527J IT x.j$€+.1II -10 Itk'z'-'~1'+L':'422o ® !lk so WUA ' ilEtl ~ -!IN Carrcer of Bucca/ CaTn/y,Esopla7~ h.M (140-148,1501 I a [- 2 50 400 r.:__.:~% r4 n., Cancer of B/adder (lB/) 100rr ~ .7® 7 uP20 143 VO 21+ 73=IA6 Cirrhosis of Liver (59/1 iabe/es(260) ~

186 N1L or Concer, IwnO msl6lh (t62) 4 Concer,proatarte 4 (177) Conce,esophoqus,boccat 2 coWty (140-14®,150) Concer,biodder (1®1) Ca~ncer, stornoch (151) Lym~phomas (200-203) Concer,pas+creos (157) Concer,intestines rectum (152-154) Concer, kidney (160) Leukemia (204) Concer,other Pneumonia (460-493) ®ronchiris(500-527) emphysema etc. 5 Tobacco and Health PIPE ONLY Coronary heart dseose (420) Nmnrheuawtic endocorditrs (421-422) II/. pF rru 142 1.t6 CD3 ® Cerebral vmsculmr 1V0 19 lesions (330-334) I hemrt diseosr(440-443) I50 Generml (450) 23 167 163 Chronic nephritis (592-594) Chronmc rheumtic artereosclerosis Hypertension without heart disemse(44P447) heart drtseose(4t0-416) Diabetes (260) 15 - l00 Ulcer of stomach daadenam(540-541) I 1 1 5 Pora9ysrs oqdans (350) Cirrhosis of 4iver (561) Other diseases of 76 I 79 I q3 ® l~ Over, poncreos 1562-567) In ® I 125 1 6 - 2.00 4 47 5 1.50 5 - 2.50 !_1 I 0 I 2 0 I 2 MORTALITY RATIO Fig. 11. Mortality of regular pipe-only smokers from specific diseases; ratio of observed to expected number of deaths; death rates of nonsmokers and occasionally-only smokers = 1.00; includes underlying and contrib- utory causes. USCLI policyholders, July, 1954-Dee., 1956. smokers than for nonsmokers. The highest mortality ratio was found for coronary heart disease, the leading cause of death among the entire group of policyholders. No increased risk of dying from chronic rheumatic heart disease, whose etiology is known, or from chronic nephritis was noticed among any group of smokers. Mortality ratios by current average daily number of cigarettes smoked for men who had always smoked only cigarettes are shown in Figm•c 10. The death rates for the following diseases increased with an increase in the number of cigarettes smoked- cancer of the lung, cancer of the buccal cavity and esophagus, cirrhosis of the liver, cancer of the bladder, cancer of the pan- creas, leukemia, and diabetes. IIowever, the number of deaths ) I Rates and Causes of Death: Smokers and Nonsmokers 187 1 t ~-, II/. /F TOTAL CIGAR p./t IE11716 Cancer, tucq 16 2) 13 IESfeS Corronary heart 373 _ 1.05 Cancer, prostmle (177) 2® disease (420) Navxhem®otic endecordills (421-422) 39 9 Concer,esopho©os,bmecoll2 cavdy (140-146,150) Hypertenaimn w11h heorf dFSeost (440-443) ®5 - I 16 Cancer, bladder (I61) 3 I Generml(450) 74 .9I Conc er, stomach 9 orleriosclerasis Hypmlensimn withoul (151) heart doemse(444-447)29 1.26 LyrMphovnms ) (200-203) 12 Cerebral vmsovtav 92 96 Cancer, pmncreas (157) I Iesions (330-334) I Chronic nephrH.s 9 100 (592-594) Concer,iatestines 54 Chronme rheum tb a c rectum (152-154) hearl disemse(410-416) 6 142 Cancer, kidney (160) ) CFrrhosrs of 22 314 Leukemia (204) II Ilver1561) Other dlseoses ot Ilver, panrneas (5®2-567)22 ~ Cmncer other 32 Diabetes , (260) 55 •90 Pneumonla 319 _i.93 Ulcer of stomoch ~ -. duodenmrn (540-541) 2.14 Duonchlt)s (500-527) 22 r 100 emphysema etc. i 0 I 2 3 0 1 MORTALITY RATIO 2 3 Fig. 12. Mortality of all regular cigar smokers from specific diseases; ratio of observed to expected number of deaths; death rates of nonsmokers and occasionally-only smokers = 1.00; includes underlying and contributory causes, USCLI policyholders, July, 1954-Dec., 1956 from each of the latter two diseases are not large enough to make this increase statistically significant. The death rate from lung cancer for men who regularly smoked more than a pack of cigarettes per day is nearly 16 times that for nonsmokers. Mortality ratios of approximately four from cancer of the buccal cavity and esophagus, cirrhosis of the liver and chronic respiratory diseases (bronchitis, emphysema, etc. ) were also found for these heavy smokers. The leading cause of death of the policyholders was coronary heart disease. The disease rate was significantly higher for light, moderate, and heavy cigarette smokers than for nonsmokers. However, the excess was much smaller than that for lung cancer. Moreover the mortality ratio was no higher for men who had smoked more than a pack than those who had smoked from one- half to one pack per day, although for both groups the rate was T201033

ir I 188 Tobacco and Health significantly greater than that for men smoking less than 10 cigarettes per day. The death rates for regular pipe and cigar smokers were less than those for regular cigarette smokers not only for all causes combined but also for most of the specific diseases shown in Figures 11 and 12. However, regular cigar smokers experienced higher death rates than regular cigarette smokers from diabetes, cancer of the prostate, and cancer of the intestinal tract and approximately equal death rates for cirrhosis of the liver, and cancer of the buccal cavity and esophagus. None of the mortality ratios for specific diseases for men who had regularly smoked only a pipe were significantly greater than one except those for bronchitis and emphysema, 1.88, and for coronary heart disease, 1.16. For each of these diseases the death rate of pipe smokers was less than that for regular cigarette-only smokers. Relative Importance of Causes of Excess Deaths The public health significance of the excess mortality from a specific disease depends not only upon the relative size of the excess but also upon the absolute number of deaths due to the disease. A disease with a death rate of 10 per 100,000 among nonsmokers and a rate of 100 per 100,000 among smokers does not account for as large a proportion of the total number of deaths as a disease with a rate of 90 per 100,000 among nonsmokers and a rate of 180 per 1010,000 among smokers; although the mortality ratio is five timcs as great for the first disease. The relative rank of a disease as a cause of death depends on the rules for selecting the underlying cause and on the way specific disease are grouped into cause-of-death categories. Some categories may include a single disease while others may include several diseases. Moreover, the relative rank of these categories varies with the age at death so that the ranking for two popula- tions even with identical age-specific-death rates will not be the same unless the percentage age distributions are the same. The data in Figure 13 show that no direct relationship exists between the mortality ratio and the relative rank of different causes of deatL. The death rate from lung cancer was more than 1 i r I i MORTALITY RATIO Coronary Heart Disease (420) 1.58 Lung 9,35 Cancer(162,163) Other Cardiovascular 1.40 Diseases Cancer Except Lung 1.30 All Other 1.29 Diseases Hypertensive Cardiovascular 1.56 Disease (440-447) Respiratory Exc. Cancer 2.76 \ \. ~\\\\\\\\~l 5.3 j5.3 2.5 4.4 16.8 11.9 15.2 13.9 12.6 Excess deaths ~ I Observed deaths 43.7 44.8 L I i I 0 10 20 30 40 PERCENT Fig. 13. Relative rank of diseases causing excess: mortality among regular cigarette-only smokers; underlying causes only; excess deaths computed from death rates of nonsmokers and occasionally-only smokers; USGLI policyholders, July, 1954-Dec., 1956. T201034

190 Tobacco and Health nine times higher for men who had regularly smoked only ciga- rettes than for nonsmokers, but it caused only 6.8 per cent of the total number of deaths. Coronary heart disease with a mortality ratio of 1.58 caused nearly one half of all the deaths. Another way of assessing the importance of specific diseases is to rank causes of death by the proportion of the excess number of deaths that may be attributed to them. The excess number of deaths is the number of observed deaths attributed to any cause minus the expected number of deaths based on the death rate from that cause of nonsmokers. For example there were 187 deaths from cancer of the lung (underlying causes only ) among men who had regularly smoked only cigarettes. If this group of smokers had experienced the same age-specific death rates from lung cancer as nonsmokers, only 20 would have died from this disease. Thus there were 167 excess deaths from cancer of the lung in this group. The relative importance of coronary heart disease as a cause of death among men who had regularly smoked only cigarettes is the same by both method of ranking; it caused about 44 per cent both of the observed number of deaths and of the excess number of deaths. Thc change in relative ranking is most marked for cancer of the lung which caused 6.8 per cent of the observed number of deaths but two and one-half times (16.8 per cent ) as large a proportion of the number of excess deaths. Lung cancer caused only about one-half as many of the observed deaths as all other forms of cancer. However, it accounted for nearly twice as large a proportion of the number of excess deaths as all other forms of cancer combined. REFERENCES 1. Pearl, Raymond: Tobacco smoking and longevity, Science, 87, 216-217, 1938. 2. Doll, Richard and Hill, A. Bradford: Lung cancer and other causes of death in relation to smoking, Brit.Med. J. ii, 1071-1U81, 1956. 3. Hammond, E. Cuyler and Horn, Daniel: Smoking and death rates-report on forty-four months of the follow-up of 187,783 men, J.A.M.A., 166, 1159-1172, 1294-1308, 1158. 4. Dorn, Harold F.: Tobacco consumption and mortality from cancer and other diseases, Puhlic Health Reports 74, 581-593, 1959. 5. Dorn, Harold C•.: The mortality of smokers and nonsmokers, American Statistical Association, Proceedings of the Social Statistics Section 1958, pp. 34-57, 1959. I 1 0 15 STATISTICAL CONSIDERATIONS AND EVALUATIONS OF EPIDEMIOLOGIC EVIDENCE ANTorno Ciocco, Sc.D. , 1. INTRODUCTION t1s I have understood it, my task is to outline some of the principal aspects of the logical framework of biostatistics that are useful in the examination of data from so-called epi- demiologic studies; particularly data from studies on the relation- ship of tobacco smoking to such disease reactions as lung cancer, coronary heart disease, and peptic ulcer. Already, much has been written on the findings of the several studies reported in recent years; and many of the writers have had a closer view of the data and possess better knowledge of the disease reactions in question than I. Sharp criticisms have been leveled at some of the results and conclusions, and in turn the critics have been subject to . equally sharp retorts.i, ° These arguments reveal that the major points at issue involve the statistical concepts related to: 1. Estimation of the risk of acquiring a• specified stage of disease reaction. 2. Definition of the universe or population. 3. Causal relations. In this paper I shall examine these concepts principally from the standpoint of the knowledge that has been acquired from statistical studies on health and disease problems in human Paper prepared for Conference on Tobacco and Health, New York Academy of Medicine, September 26, 27, 1960. The author expresses his thanks to Dr D. J. Thompson for his assistance and ad- vice. 191 T201035

I 192 Tobacco and Health populations. Much of what I shall say constitutes a restatement of what has been said about the principles of research on com- munity health problems.2 Therefore, I can be brief. 2. STATISTICS IN EPIDE'MIOLOGIC STUDIES Epidemiologic studies are part of the broad class of demo- graphic studies to the extent that they are concerned witl} the temporal and spatial distribution in the population or community of persons with certain designated characteristics. In epidemio- logic studies, the concern is with the relative number of persons having specified disease reactions, and the objective is to deter- mine the significant factors among the many that antecede the onset of each of the stages of the specified disease reaction. The ultimate purpose is to identify those factors which are most susceptible to control by community action. In its simplest form, the application of statistics in epidemio- logic studies consists of (a) computing prevalence and incidence rates of the disease reactions for certain periods of time, and (b) comparing these rates among groups of the population having designated characteristics. The purpose in computing prevalence and incidence rates is to estimate the probability of the occurrence of the variable events; and statistical analytical procedures seek to determine the nature and degree of variation in these estimates. Implied in the purpose of estimating a probability is that all persons who constitute the denominator of the rate have an equal likelihood of achieving the event which would bring them into the numera- tor of the rate. To this end, as knowledge increases about the factors that modify the occurrence of a phenomenon, rates are computed for as homogeneous a group of persons as possible, so that comparison of rates between groups is less apt to be affected by differences :n heterogeneity of the groups compared. 3. INTERPRETATION OF FINDINGS All textbooks on statistics emphasize the need to assess sam- pling variation. In fact, the greater portion of most textbooks is devoted to the description of the several analytical procedures k i J d i J I Introduction to Statistical Considerations and Evaluation 193 for testing "statistical significance." Somewhat less recognition is given to the need to assess non-sampling errors. For such assessment the questions to be asked include the following: How accurate and specific is the classification of persons into sub- groups, say smokers and non-smokers? Were the classifications employed with the same degree of accuracy for all sub-groups? That rates and consequently the comparison of rates may be affected by (1) sampling variation and (2) bias in observation or recording of observations is widely appreciated; and, on the whole, failure to observe appropriate precautions is easily detected and quickly pointed out. Such is not the case with regard to two other important prin- ciples which must be observed for adequate interpretation of epidemiologic data or data from any kind of investigation. The first of these is that the observations and measurements should be consistent with, relevant to and pertinent to the objec- tive of the study. As has been said, the general objective of epidemiologic investigations is to identify the significant factors which antecede each stage of the specified disease reaction. The objective may be limited to determining the factors which antecede only one stage of the disease reaction as, for example, the first cellular change from normal to malignant, or at the other extreme, the breakdown of the human organism from malignancy to death. Whatever is the stage of interest, observa- tions and measurements on factors must antecede the stage of the disease reaction to which the factors are to be related. In statistical terms this means that the rate should express the probability of the occurrence of the stage of the disease reaction at a point in time when the factor is present and the probability of its occurrence at the same point in time when the factor is absent. Successful computation of such a rate requires certain knowledge of the disease reaction and of the factor; the sequence of stages and the time involved in this sequence, the sequence of the factor and time involved in this sequence, the possible mechanisms by which the factor operates on the disease reaction. The knowledge required may be deduced from Figure 1 wherein aspects of the temporal relationships between factors and the stages of disease reactions are described symbolically as T201036

194 Tobacco and Health A PROBABfiUSTIC VBEW OF POSSIiBLE TEAAPORAL RELATiONB OF FACTORS TO DISEASE REACTIONS om.o,. R.oa~ r..,. scw PM a.,Nc-l. DME.E STAsp W1CAl. SIAGE MAGE n,GE STAGE n,~ ~.~ I ... i ... j k w ~f=ul~l+ -'di P.n P_A ~ -P~ Pl.i ~ 4 ~ .................. :...... .... • {, o • o . d, ~ ro o. oj do _~......_..;....... _...... I ----------- 4fm F4GTOM LEVCL . ~ r-----i------~-l---------~ o.o 000 000 .o .o pOW G.nela ka.ui.~wns Pr.•oWX ~dW1 Gaxl.Mron Fiwwc+mer. @rw*mn.nt eonC07nav BIRTH AoaEscerucE an,rnHxoo Fig. r d5/i= ~f i dqho extractions from probability urns. Figure 1 is intended to show that: 1. The tirne scale for disease reactions may differ from the time scale for aging of the individual. As we know, the time scales for different diseases may vary considerably; some diseases may occur at any age, others at specified ages; some diseases may last a few days, others a lifetime. 2. Development for the individual of the factor which will affect the onset of some stage of a disease reaction may be re- garded as the extraction of a particular type of ball from an urn containing several types; and this "extraction" may take place once only or many times throughout life. Thus, the probability of acquiring the factor leading to six fingers is determined shortly after fertilization of the ovum, the probability of acquiring cer- tain malfoi7nations may be determined when the genetic consti- tution is established or during intra-uterine life, the events lead- ing to the probability of acquiring staphylococci or other micro- organisms ma}. take place anywhere along the age time scale. 3. Estimation of the conditional probability of acquiring a certain stage of disease reaction when a factor has been acquired f ; r I I Introduction to Statistical Considerations and Evaluation 195 obviously can be calculated only for the stage subsequent to that time point in which the factor becomes operative. In terms of this scheme, let us examine broadly the observa- tions on the relationship of tobacco smoking to lung cancer in man. First, somewhere along the pre-adult-adult age span persons become smokers, of different kinds and with different intensity as indicated by fo, fi, and f2 in Fig. 1. Currently, it is not clear how other characteristics develop leading to the distri- bution of persons according to this factor. Second, the observa- tions on presence or absence of lung cancer begin usually at the clinical stage (as suggested by the Figure). Currently, there is little or no knowledge regarding the duration of the pre-clinical stages. Based on current knowledge in man, if we wish to infer that the factor of smoking antecedes the pre-clinical stages we must assume that these stages occur after the smoking habit is acquired. Otherwise, estimates obtained from the prospective studies have meaning only for whatever pre-clinical stages the individuals had already achieved when they became members of the cohort. Incidentally, as can be deduced from the chart, the mortality rate computed on the basis of the number of persons who enter the cohort in the pre-clinical stage is affected by variation in the conditional probabilities for each stage prior to death. It is somewhat surprising that no data have been published on the relationship of length of survivorship after the onset of clinical manifestation to intensity of cigarette smoking. In the case of coronary heart disease and peptic ulcer, the genetic constitution seems to play a role in susceptibility to these conditions. Furthermore, Thomas'6 data would indicate that family history of heart disease is related to the acquisition of smoking habit. In our chart then, for this disease, the factor, smoking habit, would be tied to genetic constitution, and genetic constitution and smoking habit both tied to clinical manifesta- tions of the disease and death from it. In general, in the case of chronic diseases such as lung cancer, coronary heart disease and peptic ulcer, and with a trait such as cigarette smoking we are faced with a situation in which the nature and duration of the stages before the clinical ones that T201037

196 Tobacco and Health bring the person to the physician are unknown. In fact, not enough is now known quantitatively about the course of these diseases after diagnosis and until death. Furthermore, not much is now known about the development of the smoking habit, the differences that may exist between cigarette, cigar and pipe smoking, the dosage level of whatever is harmful in tobacco combustion, etc. Finally, there is still uncertainty regarding the process of carcinogenesis, gastric ulceration, or changes in the coronary arteries and how cigarette or tobacco smoking in other forms actually influence these processes. Does smoking serve to induce the process in a susceptible individual or does it make the indi- vidual susceptible? In brief, where on the time scale of an individual's life do we place the time scale of the disease reaction, or the time scale of the factor development? It is this uncertainty which leaves room for debate about inferences that can be drawn from these studies. Until the required knowledge is obtained, interpretation of the findings has to be arrived at cautiously. This is true not only for the relationship of these diseases to cigarette smoking but also for the relationship of these diseases to occupational or other factors about which no one seems to argue. Thus the existence of a relationship between exposure to chro- mate dust and lung cancer has been accepted on the basis of the comparatively high incidence of lung cancer among chromate workers. However, neither amount nor kind of exposure required for the development of cancer has been determined except that workers who develop cancer have been employed from 10 to 40 years in this industry. Since the action of the chromate substance itself in carcinogenesis is not yet known either, interpretation of the relationship is obviously limited, and by itself the existence of the relationship does not contribute to knowledge of carcino- genesis or of t he relationship of dose of chromate substance to carcinogenesis. The interpretation is limited from a practical as well. as a theoretical standpoint, since the only preventive measure possible now is the broadside attempt to reduce all dust exposure in chromate plants, without any guarantee regarding its effectiveness. For interprei ations to become sharper and more definitive, a r Introduction to Statistical Considerations and Evaluation 197 greater knowledge of the sequence of the stages and of the factors is required, or some hypotheses must be more firmly established about sequences and interactions. This point is well illustrated by Goldberger's success in elucidating the dietary factor in pellagra. Because of its seasonal character, it had been assumed that pellagra had as short an incubation period as most acute infectious diseases. Goldberger and his associates, ques- tioning this assumption, undertook year-round surveys of the pop- ulation, thus obtaining evidence that diet in fall-winter was related to the spring-summer manifestations of the disease. As further illustrations, one need only recall the progress made in understanding the natural history of such diseases as poliomyeli- tis, tuberculosis, syphilis, as each stage of the disease reaction was described and observational techniques were developed to identify it. In sum, inadequacy of knowledge about the sequence of events and time involved in these diseases and in the factor investigated prevents any definitive statement as to the stages of disease reactions the factor is supposed to antecede. Since there are no mathematical procedures to test whether or not a factor actually antecedes a specified disease reaction, the only way to determine this fact is through observation. It is only through appropriate observations that estimates can be made of the probability of the occurrence of eaeh stage of the disease reaction in terms of presence or absence of the factor. The most important principle to be considered in the inter- pretation of epidemiologic data is the definition of the universe or population to which the findings may be generaliaed. The importance of this principle looms larger and larger as we explore relationships that are dependent on time, as in the case of chronic diseases, and/or long exposure to environmental factors. As an example, one finds repeatedly that the mortality experience of active industrial workers is less than that of the general popula- tion or corresponding age, even for industries that involve toxic and other hazards. This fact, which sometjmes has been inter- preted as a demonstration of the beneficial effects of work, generally reflects both auto-selection by the worker and pre- employment and/or disability retirement policies of industrv. T201038

198 Tobacco and Health In other words, the universe of active industrial workers is not the same as that of the general population in terms of risk of dying but is modified by selective factors which are related to this risk. The need to define the universe in epidemiologic studies on chronic diseases stems from the realization that the rate of occurrence of certain events subsequent to a point in time, Ti, may be conditional to whatever series of events segre- gates the population into the groups we are interested in observ- ing and comparing. If we wish to conceive of this rate of occurrence as a conditional probability it has to be defined in terms of the pertinent condition prior to Ti. If more than one condition is involved we cannot consider the rate of occurrence subsequent to T, as a simple probability and proceed to a direct estimation of this probability. It is a compound probability, a combination of the probabilities of several events which may be interdependent, and there are no procedures based on proba- bility theory wh:ch provide a means of estimating these probabili- ties without making unrealistic assumptions. The natural history of a variety of disease reactions illustrates the several ways in which the occurrence of a disease in the population subsequent to a point T, is altered by changes in characteristics of the population prior to Ti, as a result of which the comparisons between sub-groups established at T, cannot be correctly interpreted without appropriate corrections or adjust- ments. Thus, if one population group has been subjected to an out- break of a disease, such as measles, and another group has not, comparison betwean the two groups of subsequent experience with the diseasr has limited meaning. The lack of comparability between the two groups may be regarded as due to differences in the immnnizing experiences prior to Ti. Or, if a disease such as tuberculosis occurs earlier in life and is more fatal in one group of the pot)ulation than in another, the findings in later life may not reflect the total experience of the two groups. The lack of comparability may be regarded as due to the selective effects of prior experir-ice on the resistance of the group subsequent to Ti. Or, as is the case with chronic disabling pulmonary disease associated with dust hazards, the differences in frequency of this Introduction to Statistical Considerations and Evaluation 199 condition may reflect not the extent of exposure to dust, or the dose inhaled, but the exposure to certain infections. A high inci- dence may be due to effects of accidental complications of the experience prior to Ti on subsequent experience. Or, the chances that respondents to a survey are more ill or less ill with a disease condition than non-respondents raises doubts about results ob- tained on volunteers. The lack of comparability is a result of auto-selection with bias. The many instances of erroneous interpretations of epidemio- logic studies have made us very sensitive to the need for a precise delineation of the origin of the group under study and of the dynamics through which the segregation into sub-groups has taken place. I believe that it is fair to say that most of the spurious associations cited in the literature derive from a failure to answer definitively: What is the universe from which the . group under study is derived? Because it is not always possible to establish control groups or to obtain data which permit the reconstruction of the universe when starting with individuals who already have the disease reaction, efforts are made to organ- ize follow-up studies of segments of the population who do not have the specified disease reaction. But the prospective nature of the study, when care is taken in avoiding selection bias, does not guarantee a definitive answer about relationships of the variable characteristics of the group and the occurrence of the disease reaction. Knowledge is required regarding the temporal behavior of the group with respect to all the variables of interest. Our follow-up study in 1956-57 of a sample of persons living in Donora, Pennsylvania, at the time of the well-known smog epi- sode of October, 1948, illustrates the importance of satisfying the principle of defining the universe. A one-third probability sample of residents of Donora had been surveyed immediately after the episode by the U.S. Public Health Service for the pur- pose of determining the occurrence of cardio-respiratory sickness during and prior to the time of the smog episode. Our follow-up study was primarily aimed at determining the effects that sickness during the smog episode had on subsequent morbidity and mortality. It was possible to follow-tip 99 per cent of the sample and we found that, in the eight year interval from 1948 to T201039

200 Tobacco and Health 1956, persons who in 1948 complained of having been affected by the smog acFually required medical attention, were hospital- ized for or died from cardio-vascular diseases more often than persons who had not complained of having been affected by the smog. However, persons who complained in 1948 also had a higher frequency of cardio-respiratory sickness prior to 1948. When these persons were considered separately, the difference in mortality between those who complained of illness during the smog and those who did not practically disappeared. Thus if we had not possessed information regarding the past cardio-respiratory history of the persons surveyed in 1948 and assumed that the probability of sickness and death after 1948 was independent of (a) conditions prior to 1948 and (b) the interaction of prior illness with the smog episode, we would have inferred that the smog episode was a significant factor in the subsequent mortality and morbidity from cardiovascular diseases. This experience and many others recorded in the literature indicate the importance of defining precisely in terms of the objective of the study the universe or population to which one can generalize. They reveal that assumptions based on the independence of the separate links in the chain of growth and other dynamic biological and sociall processes can easily lead to wrong interpretations. Whether or not bias due to selection has occurred in the three major follow-up studies that have been reported remains a moot point. It is hardly likely that all three have a bias in the same direction. However, the questions which have been raised about these studies cannot be completely disregarded. It is a fact that all three cohorts have a lower mortality experience during the period of observation than the general population. This could indicate some ;;election. The bearing that such selection may have on the association has been discussed at length by others, but data which give definitive support to one or the other point of view have not been given. It is of interest to note also that Hammond and Horn's' query in 1955-56 of a sample of the surviving respondents of 1952 reveals that a number of changes in smoking habits took place in the two-three ),ears interval between surveys among these per- Introduct{on to Statistical Considerations and Evaluation 201 REilSrpns OF ANrEcEDENr FacTOR TD osEasE aExraw ~ ~ ~ AEACMN U., ASSOC.,M ~ MWLE NOC SAF7Y and SUFFUEN't' SUFF1cIENr NECESSARY DIM"WRIACTM MW 0 W 0 © a 0 0 d 0 d 0 d RACTd4 LEVELS 019EASE REACrION H04 • MMIUM . YES  LOW • NO O roaNe 0 Fig. 2 sons past 50 years of age. For example, among the "never smoked," 2.6 per cent had started to smoke, among those who never smoked cigarettes regularly 17.3 per cent had begun to smoke cigarettes regularly. Conversely, 14 per cent of the heavy smokers had given up smoking. Considering the short span of time and the ages of the respondents, what can be said about the stability of the characteristic differentiating the groups compared, or of the meaning of the age when smoking began, or of the in- terpretation of findings in terms of dose? When we take these and other points into consideration, the need for caution in interpreting these studies finds a stronger argument. There is very little direct evidence to go on beyond the facts that persons who have achieved a certain age and smoke cigarettes have a subsequent incidence and mortality from lung cancer higher than persons who do not; and persons who are now heavy smokers, as defined, and who started to smoke cigar- ettes early (again comparatively speaking), and have survived to this age have a subsequent higher mortality from lung cancer than those who started late. Whether more or fewer heavy cigarette smokers, or smokers in general survived from the original cohort is a question mark. T201040

202 Tobacco and Health 4. ON CAUSAL RELATIONS The statement has been made that the observed association indicates a causal relationship of cigarette smoking to the inci- dence of lung cancer and mortality from it. The validity of the statement has been doubted because the same studies have revealed an association between cigarette smoking and a variety of causes of death. Causal relation by definition° implies that what is termed cause actually antecedes what is termed effect. In the simplest form, the expression of a causal relation found in a population may be represcnted by a fourfold table as follows: PERSONS Antecedent Condition Eff Yes I ect No Total Yes a b a-hb No C d c-hd Total a-I-c b+ d a-hb-I-c-}-d Where a-I- b-{- c-h d are the total number of persons. Following standard practice we are willing to say that if a/ ( a-{--b )> c/ ( c-{-d ) and if this inequality is general and pre- dictable within certain defined bounds there is an association between the condition and the effect. Let us examine through the use of a simple probability urn scheme some of the ways in which the development of such an association can take place. In this scheme, the result of drawing a ball of a specified shade from the factor or antecedent condi- tion urn leads to the drawing from one of two "effect" urns which differ in their composition as to proportion of balls of specified shades, representing differences in the probability of acquiring the effect. In Figure 2, three situations are depicted: i. The situation which results in an association table wherein b-O, c-O. This means that whenever the condition is present, the effect is also present; whenever the condition is not present Introduction to Statistical Considerations and Evaluation 203 RELATIONS OF AMECEDEM` FACTaR iO DISEASE RE11CTi0N II TYPVCAL OBSERVED ASSOCWTION DtsF.A9E FiEC+craN YES NO YES FQCTOR NO L-. CA.SE r. M(1L71PLE ANTECEDEWT FACTORS co~,or~s ~rIoN A6CTOR A ® `-•..0, 0 A m 0 ® O ® O O a b C d oISEAS@ REAOnaN ® J : ~ 01 10000 ' ASSOqATU)N TABLES JOINr fiYC70R A MICfqR e 0 0 b 0 o d c d o d Fig. 3 the effect is not present. In brief, the presence of the condition is both necessary and sufficient for the effect to occur. We find this situation in biological phenomena in which both the condi- tion and effect are highly specific. For example, this situation exemplifies the de fined relationship between the tuberculosis bacillus and tuberculosis infection. ii. The situation which results in an association table wherein b=O, but c>O. When two factors can produce the effect, one with a lesser probability than the other, an association of this type may arise. This means that the factor is sufficient but not necessary to produce the effect. Such an association may also be observed when one is unable to detect a lower dose of the factor which nevertheless is capable of producing the effect. iii. The situation which results in an association table wherein c-o, but b>o. This means that whenever the condition is present the effect may or may not be present, but if the condi- tion is not present the effect does not occur. The antecedent condition is necessary for the production of the effect. This is exemplified by clinical tuberculosis and other disease conditions with specified agents and for which the diagnostic criteria are rather sharply defined. It is a situation also observed in data T201041

204 To bac co and Health M ~ RELarlON5 OF PdVrEC EDE NT F15CTnR TO WEASE REPCT1oN N M ~ caSE :'. assoclATE D F 1SCT17RS (One antx ede&) ~ FAGT OR WM1lBAVAT10N5 DMEASE REACIIOtlV ~ ® ~ o©o ~...._ ~~ ~ ~` 0 • L. ~J o ooaa ~ 0 oJ ~ C ~A~1~ ASsOCiATnN TABLES FACTOR A fACTOR B a 0 0 d Fig. 4 o b c d from epidemiologic studies concerned with disease reactions for which inadequate dose of the agent or the resistance of the host prevents the clinical reaction from being manifest. Except as indicated, the most common form of association observed in demographic studies is that in which b>o, and c>o. This is the type observed for the association of cigarette smoking with a number of causes of death. There are many ways by which such an association can arise from different combina- tions of the urn scheme described here. Figures 3 and 4 illustrate this. An association of this type can arise from the fact that the factor under study by itself cannot produce the effect but acts in conjunction with another factor with stronger action. This situation is exemplified by the relationship of nutritional status, or that of exposure to unfavorable environmental conditions, to the clinical manifestations of certain infectious diseases. It can arise when the factor under study by itself cannot produce the effect but is highly correlated with the "true" ante- cedent factor. Poverty and its correlation with diet, sex and its correlation with occupation illustrate this. These urn combinations do not exhaust all the imaginable possibilities but are presented to indicate that an association table in which b>o and c>o does not permit a unique interpretation. Introduction to Statistical Considerations and Evaluation 205 Furthermore, it can also be shown that this form of association table arises when either the factor, the effect, or both are mis- classified. From these considerations, it would appear that data on the association of cigarette smoking with lung cancer and with other diseases do not indicate that smoking is a necessary, sufficient, or necessary and sufficient condition (as I have defined these terms) for the production of the effect. Which of these urn combina- tions, if any, actually depict the associations involving cigarette smoking? We do not know. However, if we assume that cigar- ette smoking is represented by Factor B in either Figure 3 or Figure 4, then it is possible that association of the form observed between cigarette smoking and a variety of disease conditions could be obtained. We have seen this type of association between sex and morbidity and mortality from a variety of diseases. We have seen it between socio-economic status and mortality also from many diseases and, in general, when the variable factor under study is associated with the effect because of the correla- tion..with a number of different factors that are sufficient, neces- sary, or necessary and sufficient to produce different effects. 5. SUMMARY AND DISCUSSION Hardly anyone questions that an association exists between cigarette smoking and mortality from a number of diseases. The real questions are: (1) What does this association mean in terms of the dynamics of these diseases? and (2) What does this associ- ation mean in terms of possible social action? I have addressed myself to the first question. In line with the principles of statistical methodology which I have singled out to discuss, the facts of the association can lead only to an interpretation that cigarette smoking is a factor in the overt clinical manifestation of certain diseases. 1. Since there is so little knowledge regarding the nature and duration of the stages which precede the overt manifestation of the disease, the relationship of the time seq4ence of some dose of cigarette smoking to a stage of a disease cannot be established. To do so requires not only more knowledge about the develop- ment of the disease, and on methods of identifying the several T201042

206 Tobacco and Health stages, but also quantitative studies on incubation or latency period and of the chronology of the several stages. 2. Since there is so little knowledge about the physical and mental characteristics of persons who acquire the smoking habit and about changes in their habit throughout life, in addition to ignorance about the dynamics of the chronic diseases, the possi- bility cannot be excluded that the follow-up studies that have been reported may have an inherent bias. Hence, the associa- tion must be regarded as real only for the time beyond the point of origin of the study. Logical dissertations cannot change this but only the acquisition of facts. In follow-up studies of the future more car eful consideration should be given to observation and recording of pertinent characteristics of the group to be followed. 3. Assuming that cigarette smoking actually precedes the onset of the variety of disease reactions that have been reported, one could very well reason that the association is of the same significance as that of poverty, socio-economic status and other non-specific characteristics and certain disease reactions. Clari- fication of the -neans by which this association develops is neces- sary for its potential contribution to an understanding of the diseases involved. When we seek to answer the question: what does this associa- tion mean in terms of possible social actionP we are not dealing with a scientific problem only but with one in which beliefs, ethics, social attitudes play an important role. There is always a tendency to want "to do something about it" when we believe that we can eliminate some unpleasant condition. If we review the history of public health measures we note that sometimes such actions taken on the basis of limited knowledge have in part or for a time produced some good. Quarantine regulations are a good example. On the other hand, when we order that cannons be shot to disFipate yellow fever, the resulting "sound and fury" signifies nothing. Let us keep in mind that lung cancer, peptic ulcer, coronary heart disease are only some of the disease reactions characterized by apparently long term genesis and complex etiology which preoccupy us because of their disabling and fatal consequences. Introduction to Statistical Considerations and Evaluation 207 Similarly, tobacco smoking exemplifies one of a number of be- havior traits and environmental factors whose physiologic and pathologic actions may vary with duration and quantity of exposure. In order to achieve sufficient knowledge to prevent these chronic conditions, or at least to reduce their impact on the community, it is necessary to develop appropriate schemes for collecting and analyzing data on population groups for the pur- pose of determining the extent of the relationships of these kinds of behavior traits and environmental factors to the disease re- actions. No matter what kind of social action emerges from these and similar discussions, let us at least not forget our obligations to advance knowledge about these conditions. REFERENCES 1. Berkson, J.: Smoking and cancer of the lung. Proc. Mayo Cl., vol. 35, pp. 367- 385, 1960. 2. Ciocco, A. and M. Ring: An evaluation of study method-a summary. A.J.P.H., vol. 41 (Aug. Supp.), pp. 7-17, 1951. 3. Cornfield, J. et al.: Smoking and lung cancer: recent evidence and a discussion of some questions. J.N.C.1., vol. 22, pp. 173-203, 1959. 4. Hammond, E. C. and D. Horn: Smoking and death rates-report on 44 months of follow-up of 187,783 men. J.A.M.A., vol. 166, pp. 1159-1172, 1294-1308, 1958. 5. Thomas, C. B.: Familial and epidemiologic aspects of coronary disease and hypertension. J. Ch. D., vol. 7, pp. 198-208, 1958. 6. Yerushalmy, J. and C. E. Palmer: On the methodology of investigations of etiologic factors in chronic disease. J. Chr. D., vol. 10, pp. 27-40, 1959. T201043

16 STATISTICAL CONSIDERATIONS AND EVALUATION OF EPIDEMIOLOGICAL EVIDENCE" ` J. YrMusxAisrst WITH the containment of many of the acute com- municable diseases during the past 30 years, much of the empha- sis in medical research has shifted to the chronic diseases. This shift has served to accentuate certain weaknesses of epidemio- logic method in the study of environmental factors as causative agents. In the search for etiologic factors in acute diseases it has been possible to supplement the epidemiologic evidence with ex- perimental data. This was true not only in the final step of identifying the speciflc causative agent, but also in the discovery of the vtode o f tranmnission of the disease. Three steps were typically involved in the procedure of investigating etiological factors in acute diseases: (1) (2) (3) Epidemiologic investigations directing attention and suspicion to factor F as the mode of transmission of disease D. Demonstration by experimental means or laboratory evi- dence that F is the mode of transmission of D. Identification of the specific agent S in F which causes disease D. The ehidemiologie method has been utilized in only the first of 1. From the Division of Biostatistics, School of Public Health, University of California, Berkcley. 2. This invesligation was supported (in part) by Public Health Service Grant No. B-2640 from the National Institute of Neurological Diseases and Blindness. 208 Statistical Considerations and Epidemiologic Evaluation 209 the three operations. The other two were accomplished by experimental methods. The major difference between the two types of investigation is that the epidemiologic method utilizes observations of unplanned events as they occur with little, if any, interference on the part of the investigator. In the experimental method the investigator manipulates and creates a situation which leads to the greatest confidence in the validity of the results. This is not to detract from the importance of the first step and the epidemiologic method. In fact it is perhaps the most impor- tgnt and creative of the three steps and without it the other two steps would often not have been attempted or thought of. Never- theless, the first step by itself was not thought to provide sufficient evidence even to implicate definitely a mode of transmission of a disease. It is often overlooked that also, for the latter, experi- mental evidence was required. As an example, epidemiologic observations by Goldberger (1) early directed attention to the diet and aroused suspicion that pellagra was a deficiency disease. These epidemiologic studies were followed by experimental dietary changes in hospital wards and among prisoners resulting in convincing proof of the dietary hypothesis (2, 3). It was not until many years later that nicotinic acid was identified as the specific agent concerned in the etiology of the disease. Similarly, epidemiologic evidence was used by Snow to impli- cate strongly polluted water as a mode of transmission of cholera (4). However, it took a fortunate circumstance simulating a planned experiment to provide convincing proof for the hy- pothesis. The fact that the same London area was served by two water companies in a non-selective fashion made it possible to investigate the occurrence of cholera among the members of the households served by these companies in almost an experimental fashion. The identification of the specific causative agent was not accomplished until many years later. It is of interest that concurrently with the above a number of excellent epidemiologic investigations were : conducted which implicated quite different factors as possible modes of transmis- sion of these diseases. Thus, in the case of pellagra, the Thomp- son-McFadden commission postulated a contagion factor on tho T201044

I I, 210 Tobacco and Health basis of very extensive and detailed epidemiologic studies (5), and "recommended most heartily the installation of sanitary systems of sewage disposal as an important means of restricting the spread of pellagra (6)." Similarly, the very beautiful epi- demiologic investigations of William Farr convinced him that the most important factor in the transmission of cholera was the elevation above sea level (7). He recommended that people who had the means should remove themselves to the higher inland districts, and that "armies would do better to march through the hills rather than to take the easy road over lowlands and marshes," and further "it would be a great advantage if soldiers in the field could sleep in raised camp beds." The reason these recommendations were not followed cannot be found in Ihe type of epidemiologic investigation but rather in the lack of supplemental experimentation. One might specu- late that had these recommendations been followed it is possible that they would have provided a setback to our understanding of the mode of transmission of these diseases. For example, if the population of London had been moved up to the hills the epidemics would have stopped for the people would not have carried along their polluted water and this would have provided "proof" that the mode of transmission of cholera is to be found in the "emanations from the soil." EPIDEMIOLOGIC FACTORS IN CHRONIC DISEASE In the search for etiologic factors and mode of transmission for certain of the chronic diseases we are handicapped in that the second stage in the chain of events outlined above is missing. It is not possible to provide experimental confirmation of the epidemiologic evidence. The reason is the relatively long period required between exposure to the environmental characteristic under suspicion and the development of the disease. When the period is short, as it is in the acute diseases, it was possible, even if it sometimes required great courage, to conduct an experi- ment by the use of volunteers who were willing to put their lives in jeopardy. In the chronic disease situation, however, no such alternative is possible simply because the period between exposure and development of diseases is measured, as we are I Statistical Considerations and Epidemiologic Evaluation 211 told, not in weeks and months but in years and decades. We are faced, therefore, with the situation in which experimental supple- mentation is not possible and,, consequently, much greater demands are put on the epidemiologic evidence. In previous years, the epidemiologic method was called upon only to ferret out from among a large number of possible factors the few with the strongest suspicion of involvement. The identifi- cation from among them of the factor which was indeed the mode of transmission of the disease was accomplished by experimental means. In the chronic disease situation, however, the second step must be established by other than experimental means, often by epidemiologic methods. It is therefore essential that the epi- demiologic evidence be viewed more critically. It is necessary to attempt to find substitutes for the experimental supplementation. The major difficulty with the observational method, which is. the main tool of epidemiologic investigations, is the problem of "self-selection:' The individuals being observed, rather than the investigator, make for themselves the critical decision as to which group they belong. Consequently the basic requirement of group comparability, accomplished in experimentation by con- scious effort through random selection, is lacking. Thus when differences in the frequency of disease between the group possessing the characteristic and the one not possessing it are observed, it is not possible to conclude that the differences are due to the specific characteristic under investigation for it may be due to other factors and characteristics in which the two groups might differ. It is for this reason that experimental data are usually required to supplement the circumstantial evidence derived from observational methods of study. In the case of chronic diseases we lack the opportunity to supplement the epidemiologic evidence by more crucial data derived from experimentation. It is therefore necessary to devise other means which will assist in the evaluation of evidence of the observational type. The initial circumstantial evidence must be supplemented by further investigation even ~f the latter is also of the observational type and circumstantial in nature. There must, however, be further extension of the investigation not a "repeat" of the same type of investigation. One such method is investiga- T201045

212 Tobacco and Health tion into what might be termed the "specificity" of the observed association (2,5, 26). SPECIFICITY OF ASSOCIATION The main doubt about the validity of the epidemiologic evi- dence stems, as was pointed out before, from the fact that the groups possessing or not possessing the characteristic under suspicion have been self-selected. Consequently, the association which has been established epidemiologically between the factor F and the disease D-1 may not reflect a causal relationship be- tween F and D-1, but be due to other factors and characteristics by which the two groups differ. To fix the ideas-the association observed between cigarette smoking and cancer of the lung may indicate a cause-effect relationship, but it may also be due to differences between smokers and non-smokers in characteristics other than smoking. It is at least possible that these other factors and characteristics, rather than smoking per se, are behind the observed association. It is therefore necessary to search for methods of investigation which would help elucidate, if not eliminate this dilemma. As a first approach it may not be unreasonable to suppose that if the observed association is not a reflection of a cause-effect relationship between F and D-1 but is due to the non-compara- bility of the grotips, then the association should not be confined to the disease D-1 but should exhibit itself also in other diseases. To use a very naive notion, suppose that the observed association between F and D-1 is due to the fact that, in general, individuals with characteristic F are "weaker" than those not possessing characteristic F. Then they should suffer a higher incidence not only of disease D-1 but also of many other diseases. Conse- quently, a first step in evaluating the meaningfulness of the association between F and D-1 is to investigate the relationship of F to the other diseases D-2, D-3, D-4, etc. If it is found that the association :.,annot be demonstrated for many of the other diseases, our confidence that the initial association between F and D-1 is meaningful is strengthened. If, however, significant asso- ciations can he shown to exist for a large variety of diseases, our Statistical Considerations and Epidemiologic Eualuation- 213 confidence that the initial association reflects a cause-effect relationship is reduced. This is not to say that the evidence of an existing association with the other diseases is proof that the original association is not one of cause-effect. Obviously, it is possible that the same factor F is indeed causally related to D-1 and to the other diseases as well, or it may be causally related to some of them and not to the others. The main point is that on this evidence alone it is not possible to distinguish between the different alternatives. It is therefore suggested that in a situation of this kind the process needs to be extended further. The extension again is aimed at establishing and evaluating the "specificity" of the associations observed between the factor and the diseases. This might be accomplished by further investigation to determine whether there is an association between the factor F and a disease or condition E which may not reasonably be expected to be causally related to F. If such a disease or condition can be found the investigation provides one additional step in the evalu- ation of the epidemiologic evidence, for if the new disease or condition E is found not to be associated with characteristic F our confidence in the "specificity" of the observed associations between F and diseases D-1, D-2, D-3, or D-4, etc. is increased. On the other hand, if a significant association is found to exist also between F and E, our confidence in the meaningfulness of the original association is reduced. Again, this should not be taken as proof that there is no causal relationship between F and D-1. It only serves to indicate that our confidence that the epidemiologic evidence can be taken as proof of a cause-effect relationship between F and D-1 is reduced. The observations in this case fit with the theory or hypothesis that the observed associations are due to factors other than F by which the groups possessing or not possessing F may differ. The result is that we are then less certain of the meaningfulness of the association and must search for supplementary types of evidence of different kinds before reaching definitive conclusions. ; We may illustrate this procedure by considering the epidemiologic evidence relating cigarette smoking to cancer of the lung. T201046

T 214 Tobacco and Health CIGARETTE SMOKING AND HEALTH A strong association has been demonstrated between cigarette smoking and cancer of the lung. Many studies, both retrospective and prospective, have confirmed the fact that a larger proportion of persons who suffer from cancer of the lung are cigarette smokers than is true of a comparable group not suffering from this disease. Similarly, cigarette smokers have a higher death rate from cancer of the lung than non-smokers. Moreover, heavy smokers suffer higher mortality from cancer of the lung than light smokers and past smokers who stopped the habit have lower rates than those who continue to smoke (8, 9, 10, 11, 12). The investigations by which these associations were established pre- sented at first some interesting and subtle sampling problems which could have biased the results. These were discussed by a number of investigators, especially Berkson (13, 14, 15, 16). However, if the association were due entirely to these sampling biases they would not have persisted for many years after the initiation of the investigation. The fact that cigarette smokers continued to have higher rates in the second and third year of observation would indicate that the association between cigarette smoking and cancer of the lung is not a resultant of these sam- pling pecnliarities alone. Consequently, the association itself is accepted as definitely established. The question turns to the interpretation of this association in terms of causation. The main difficulty in evaluating such association stems from the fact that the individuals observed have made for themselves the crucial decision whether they are smokers, non-smokers, or past smokers. Consequently, the groups lack the comparability necessary for definitive experimentation. Because of lack of opportunity for experimental supplementation, it is necessary to investigate the "specificity" of the assoeiation. The first step is to determine whether cigarette smoking is associated also with diseases other than cancer of the lung. The rationale, as indicated previously, is that if the observed association between cigarette smoking and cancer of the hing is merely a reflection of the differences between smokers and non-smokers, and not an indi- cation of a cause-effect relationship, then it would be expected that the differences between cigarette smokers and non-smokers I Statistical Considerations and Epidemiologic Evaluation 215 should exhibit themselves also in other diseases. As ie known, the three major prospective investigations (10, 11, 12) have shown that cigarette smoking is associated not only with cancer of the lung but also with a large variety of other diseases. The associa- tion is much stronger -for cancer of the lung, but there is a statistically significant association between cigarette smoking and a large variety of other diseases. It is necessary to emphasize again that this does not prove that the association observed be- tween cigarette smoking and cancer of the lung is necessarily not one of cause and effect. However, this finding weakens the argu- ment that the initial association may be taken to imply that it is one of cause and effect. The evidence provided thus far, there- fore, is inconclusive. The very strong association with cancer of the lung argues for a meaningful relationship, while the signifi- cant associations with other diseases suggests the possibility that the higher rates for cancer of the lung as well as for the other diseases amongst smokers compared to non-smokers are due to factors other than smoking. In light of the discussion presented above, it becomes necessary to probe further and in particular to attempt to find other diseases and conditions which a priori would not be expected to be causally related to cigarette smoking. In searching for such a condition, we utilized certain data from an investigation now in progress on the relationship of factors in the parents to the development of their infants. In the course of this investigation information was obtained on the smoking habits and on other characteristics of the husband and the wife during the wife's pregnancy and before the birth of the infant. It would appear that the smoking habit of the father might serve the desired purpose for it could not reasonably be expected to be related to the birth weight of the infant. It thus satisfies the requirement of a condition which could be used as a further test of the "specificity" of the association between cigarette smoking and cancer of the lung. These data were therefore utilized to investigate whether there is an association between the cigarette smoking habits of the father and the birth weight of the infant. Although the number of births available for study is not too large ( approxunately 1000 ), it was nevertheless thought worthwhile to present these data, not for whatever inherent value T201047

216 Tobacco and Health (459) 5.4 p 50) . (309) 5.3 5.5 Non- < 1 1-2 2 + Post Never Smokers Smokers Pack Packs Packs Smokers Smoked Fig. 1. Per cent of infant weighing less than 5# lbs. according to the cigarette smoking habits of their fathers. Includes smokers of unknown number of packs. they may have in demonstrating an association between smoking habits of father and birth weight of the infant, but to illustrate the methodology of evaluating the "specificity" of the association between cigarette smoking and cancer of the lung. SMOKING HABITS OF FATHER AND WEIGHT OF INFANT Information is now available on the outcome of 982 pregnan- cies for which the cigarette smoking habits of both husband and wife were avallahle." The information on smoking habits was obtained by interview with the mother. The smoking habits of the father at the initial stages of this investigation were obtained by means of a special form which the gravida took home for her husband to fill out. In the later stages the information about the 3. Only pregnancies where the outcome was either a live birth or stillbirth are included. Pregnancies terminating in abortion are excluded. i Statistical Considerations and Epidemiologic Evaluation 217 husband's smoking habits was obtained directly from the gravida. While it is not possible to evaluate the reliability of the smoking data, they are perhaps not less reliable than those ob- tained in similar studies, since the interviewers were well-trained and competent. More important is the fact that the information was obtained prospectively before the outcome of pregnancy. Consequently, such inaccuracies as may be present in the data cannot be biased by knowledge of the outcome. The measure for infant weight which will be used in this discussion is the percentage of infants weighing less than five pounds eight ounces. Other indices such as the proportion of large babies were also investigated but because of the small number of cases involved the latter will not be presented. Suffice it to state, however, the results obtained by using other indices were similar to those which will be presented. The main findings may be viewed in Figure 1, which indicates a strong association between cigarette smoking of the father and the per cent of infants weighing less than five pounds eight ounces. Thus 7.9 per cent of the infants whose fathers were cigarette smokers weighed less than five and a half pounds at birth, while only 5 per cent of the 459 infants whose fathers did not smoke cigarettes were in this low birth weight group.* The association between cigarette smoking of the father and the weight of the infant is even more striking when the amount of cig- arette smoking is considered. The more the father smoked the larger was the per cent of infants weighing less than five pounds eight ounces. These percentages range from 6.7 for infants whose fathers smoked less than one pack of cigarettes a day to 10.8 for those whose fathers smoked two or more packs a day. Even more 4. In this discussion only cigarette smoking was taken into consideration. Persons smoking only cigars or pipes are considered "non-smokers." Persons who supple- mented their cigarette smoking with pipes and cigars were classified according to the number of cigarettes smoked. The association shown between smoking and birth weight can thus be considered as understating the association between smoking and birth weight. It may also be desirable to indicate that the term "fathers" is used in this discussion for ease of expression. It is realized that strictly our knowledge related to the husbands of the gravidae and that there may be cases in which the smoking information obtained for the husband are not necessarily those relating to the natural father of the infant. T201048

Statistical Considerations and Epidemiologic Evaluation 219 Tobacco and Health 218 % % 5.9 Non- Smokers Smokers 8.0 <1 l+ Pack Packs 5.9 Past Never Smokers Smoked Fig. 2. Per cent of infants weighing less than 5% lbs. according to the cigarette smoking habits of their mothers. Includes smokers of unknown number of packs. striking is the fact that infants whose fathers gave up the habit of smoking did not differ in their birth weight from those whose fathers never smoked. The percentage of infants weighing less than five pounds eight ounces at birth was the same for past smokers as for those who never smoked. The association between cigarette smoking of the father and birth weight of infant presents therefore a picture very similar to that observed between cigarette smoking and mortality from a number of diseases. These results, taken by themselves, may be construed as weakening the argument of "specificity" of associa- tion between cigarette smoking and mortality of cancer of the lung and other diseases, for while it is possible that the asociation shown in Figure 1 indicates a causal relationship, few would consider this a:,trong probability. We are thus again presented with a situation where associations with cigarette smoking are found for a large variety of conditions, including one which is not easily acceptable as meaningful in terms of cause and effect. Z ~ Both Smokers (267) ~ ! Husband Smoker Wife Non-Smoker ~ (256) i Husband Non-Smoker Wife Smoker 109) Neither Smokes (350) I Fig. 3. Per cent of infants weighing less than 534 lbs. according to the cigarette smoking habits of the husband and wife. It is, therefore, necessary to explore further the latter association. One avenue for exploration is the smoking habits of the mother because of the possible association between the smoking habits of husbands and wives. It is perhaps not so unreasonable to expect an association between the smoking habits of the mother and the birth weight of her infant. Indeed, such an association has been recently reported (17, 18). In our material an associa- tion is found between the smoking habits of the mother and the birth weight of her infant as shown in Figure 2. It is of interest to observe first that the association is not as strong for the mother as that shown in Figure 1 for the father. The data can be studied more directly by investigating the association between birth T201049

220 Tobacco and Health Statistical Considerations and Epidemiologic Evaluation 221 weight of infant and the smoking habits of both father and mother as shom-n in Figure 3. The striking finding here is that the increase in the percentage of babies weigaing less than five pounds eight ounces is present only when the husband and wife both smoke. If at least one of them does not smoke-whether it is the husband or the wife-or if neither of them smokes, the proportion of babies under the critical birth weight is approximately the same. There is no difference in the percentages in these three groups of infants. However, the per cent weighing less than five pounds eight ounces, both of whose parents smoke, is significantly different (P<.05) from that of infants in the other three groups. Thus it is not that there is an independent father's smoking effect and an independent mother's smoking effect which, when they occur together-both smokers-is more striking. If this were the case, the percentage of babies weighing less than five pounds eight ounces would be expected to be much smaller when neither the husband nor the wife smokes. This, as may be seen from Figure 3, is not the case. This finding that an increase in the percentage of infants weighing less than five pounds eight ounces is found only when both parents smoke is more difficult to explain on a causal basis than the over-all finding of an association with father's smoking habits. It is difficult to visualize a biological mechanism which would exhibit itself in the combined effects of the smoking of both parents while no effect of smoking of only one of them alone is noted. A similar situation to this one was found by Frazier (17) who studied the proportion of babies weighing less than five pounds eight ounces according to smoking habits of the mother and her psychosomatic complaint score. If the mother was a smoker and also attained a critical psychosomatic complaint score, the percentage of infants weighing less than five pounds eight ounces was greatly increased. However, if only one of these conditions exi:aed, the percentage did not differ from that for infants whose mothers neither smoked nor attained a critical psychosomatic complaint score. These findings can perhaps more easily be explained on the basis that smoking acts as an index to differentiate smokers from 1-0 N ~ 7-4 ~ ~ ® ~ ~ ~ ~ ~ 0 ~ w z ~ 00 {{:,~ •. ' I M I u » ao T201050

222 Tobacco and Health } LL 0 ~ a x 0 z_ Y z ~ 0 M 0 a x c ~ ~ O O ~ n. u [X4 I t I. i Statistical Considerations and Epidemiologic Evaluation 223 SMOKERS NON-SMOKERS 32.5 36.2 37.5 39.4 Para I Para 2 Para 3 Para 4 Pava 5 Para 6+ Fig. 8. Average age of smoking and non-smoking husbands at the birth of the infant according to parity. non-smokers on a number of different characteristics rather than as indicating a causal relationship. The lack of "specificity" which runs through all the associations observed thus far would appear to indicate a strong possibility that smokers differ from non-smok- ers in many ways including their mode of life and perhaps, as R. A. Fisher suggests (19), they represent different genotyps,s. Certainly it provides sufficient cause to question the validity of the argument of causality as the explanation for the association observed between cigarette smoking and lung cancer and the other diseases. At least the lack of "specificity" weakens the argument for a causal relationship. On the basis of these findings, it was thought worthwhile to compare the smokers and non-smokers according to a number of other characteristics on which data are available in the present study. The results are shown in Figures 4, 5, and 6. It will be seen that a larger proportion of fathers who smoke cigarettes, drink coffee, whiskey, and beer, while relatively more of the non- smokers drink tea, milk, and wine. Further, of those who do drink, the average number of cups of coffee per day or of drinks of whiskey and beer per week is larger when the father is a smoker than when he is a non-smoker (Fig. 4). Similar findings T201051

1 224 Tobacco and Health Statfstical Co 'd tY d id ns e E ' l may be observed in Figure 5 for the mother. It is also of interest that the fathers who smoke are younger than the non-smoking fathers at the time of birth of their infants in each order of birth. Differences in the number of characteristics of smokers and non- smokers have been shown also by other investigators (20, 21). Thus Lilienfeld (20) found that smokers as a group are more neurotic than non-smokers and that they change jobs and spouses more often than non-smokers. In indicating the differences in personality between smokers and non-smokers it is not the intention to suggest that these characteristics are responsible for the increases in mortality from the various diseases or for the decrease in birth weight of the infants. They do, however, bring into sharp focus the fact that the observed differences in mortality according to smoking habits were derived from comparisons of groups that are unlike in many characteristics and consequently great caution must be exercised in interpreting the observed associations as proving causality. This is especially true in view of the diffuse association with such a large number of diseases and conditions. We are, therefore, forced either to accept all these associations as indi- cating a cause-eflect relationship, or we must find guide lines to determine which of these associations are, and which are not, meaningful. As an example, in Figure 7 the findings on birth weight of infant according to father's smoking habits are con- trasted with the findings of the association between coronary artery disease and cigarette smoking. On the left panel are data from a recent pr:per by Hammond on coronary artery disease according to amount smoked and for "past smokers (22)." The right panel presents thc percentage of babies weighing less than five pounds eight ounces according to fathers' smoking habits. The findings are made comparable by relating in both cases the rates among smokers to the rates among non-smokers. In discus- sing the figures shown in the left panel, Hammond states: Certainly, the observed association between cigarette smoking and coronary artery disease must have some rational explana- tion. The explanation must account for the findings that: (1) cigarette smokers as a group have higher death rates from coronary artery disease than do nonsmokers; (2) the coronary x ra ons an em p ao ogfc Eualtua.tson 225 0 OEM g N Ln N n I 0 ai C aR -a va N.(~, ~ 0 b ~ .~ 0 E tn T201052

226 Tobacco and Health artery disease death rate increases with degree of exposure to cigarette smoke; and (3) among persons who have smoked cigarettes for many years, the death rate is lower for those who have stopped smoking for a year or longer than for those who continue to smoke cigarettes. Personally, I find it hard to escape the conclusion that cigarette smoking increases the death rate from coronary artery *disease. If these points are valid reasons for concluding that a cause- effect relationship exists for cigarette smoking and coronary artery disease, then it would appear reasonable that similar statements can be made with respect to the panel at the right. Consequently, one may wish to conclude that there is a cause-effect relationship between the smoking habits of the father and the birth weight of his infant. Such a conclusion, however, would not be accep- table to most people on the basis of this type of evidence. DEGREE OF ASSOCIATION There is one other important consideration in evaluating epidemiologic evidence-namely, the degree or strength of the association. It may be argued that a strong association should be considered as more likely to indicate a cause-effect relationship than a weaker association. Thus because the association between cigarette smoking and cancer of the lung is stronger than with the other diseases and conditions, it is argued that the former is more likely to indicate a cause-effect relationship, while the other associations might be due to "self-selection" or to other causes. This approach, however, presents a number of difficulties. First, how is the strength, or degree, of association to be measured? The usual comparison is that of ratios or death rates among smokers relative to those among non-smokers. This approach overlooks the importance of the level of the rate. Is a tenfold increase of a vcry small rate really more significant than a twofold increase of a very high rate? Berkson (23), on the other hand, suggests that the relative contribution by the different causes to the excess total mortality of smokers as compared with non-smok- ers is a more appropriate measure. Others have pointed out that a different picture of relative risk is obtained from a comparison of survival. As Mindel Sheps succinctly puts it, "Shall we count I r , I I Statistical Considerations and Epidemiologic Evaluation 227 the living or the dead? ( 24 )" If comparison is made of relative survival, the differences become much less dramatic. All these are important questions which must be resolved, but even more important is the fact that there is no rational way to decide how large a difference there must be before we accept it as indi- cating a cause-effect relationship. For example, on the same evidence two competent investigators, each of whom considers the relationship of cigarette smoking to cancer of the lung as indicating causality, come to opposite conclusions with respect to coronary artery disease. Lilienfeld (20) states that in his opinion the relationship between smoking and coronary artery disease does not indicate an etiologic relationship but may be the result of self-selection, while Hammond, as was pointed out above, states that he cannot escape the conclusion that it does indicate an etiologic relationship. If we are to consider the "strength" of the association as decisive in indicating a cause- effect relationship, we must find a way by which to measure "strength" and a method for a decision on how large the measure must be before we will be willing to interpret the association as meaningful in terms of causality. SUMMARY AND CONCLUSION In the study of etiologic factors in acute diseases, it has usually been possible to supplement epidemiologic evidence with experi- mental data. In chronic disease studies, such supplementation is, in general, not possible. Consequently, nearly the entire weight of the evidence rests on data derived from epidemiologic investi- gations. The latter are observational in nature and suffer from "self-selection" and lack of assurance of group comparability, Lacking opportunities for experimental supplementation, it is necessary to assess the weight of the epidemiologic evidence by further probing in an attempt to evaluate the strength or weak- ness of the observed association. One of the methods by which this can be accomplished is to investigate the "specificity of association" by studying relationships of the suspected factor or characteristic with other diseases. If associations are found be- tween the characteristic and many other diseases, it may be necessary to probe further for relationships with more "remote" T201053

228 Tobacco and Health conditions, that is, with conditions which cannot easily be ac- cepted to be causally related to the factor or characteristic under suspicion. In evaluating the association between cigarette smoking and lung cancer, the first test of "specificity" indicates that smoking is related also to a variety of other diseases. Further probing with more "remote" conditions is therefore indicated. It is suggested that birth weight of infant in relation to father's smoking habit may be such a condition because it cannot reasonably be expected that such a relationship would indicate a cause-effect relation- ship. A preliminary analysis along these lines based on 1000 deliveries is attempted. It is shown that father's smoking habits are strongly corre- lated with birth weight of his infant. The per cent of small babies (less than five pounds eight ounces ) is greater for infants whose fathers are cigarette smokers than for those of non-smokers; the more the father smokes, the larger the per cent of "small babies." Moreover, the reduction in birth weight is noted only in the group in which both father and mother smoke. It is considered very improbable that these results are due to a cause- effect relationship. It is also shown that smokers differ from non- smokers in a variety of other characteristics such as drinking of coffee, tea, milk, whiskey, beer, wine. These findings are interpreted as weakening the argument that the association observed between cigarette smoking and lung cancer is that of a cause-effect relationship. It is realized that these findings da not negate the argument. In other words, these findings cannot be taken as proof that all the association may be spurious. It is possible that smoking is causally related to one or more of the conditions and not to the others. However, because of the diffuse effect, the interpretation of causality with any one of them is weakened. A reasonable conclusion is that the different studies provide a basis for a strong suggestion that smoking may be causally related to lung cancer and perhaps to other diseases and conditions. The evidence, however, is far from conclusive. Consequently, while it is imperative to intensify the chemical, pharmacological, and animal investigations of tobacco smoke with the hope of isolating i I Statistical Considerations and Epidemiologic Evaluation 229 the noxious agent and proving directly its relationship to disease, it is equally important to search just as diligently for other causative agents in cancer of the lung. A premature acceptance of the causal hypothesis as proven may retard the discovery of the real factors in the event that the association of smoking to lung cancer is eventually found to be not one of a cause-effect relationship. ACKNOWLEDGMENT The author wishes to acknowledge his gratitude to Miss Lucille Milkovich and Miss Helen E. Supplee for their assistance in the statistical analysis. REFERENCES 1. Goldberger, Joseph (1914): The Etiology of Pellagra; the Significance of Certain. Epidemiological Observations With Respect Thereto. Pub. Health Rep., 29: 1683-1686. 2. Goldberger, Joseph, C. H. Waring and D. G. Willits (1915): The Prevention of Pellagra. A Test of Diet Among Institutional Inmates. P. H. R., 30:3117-3131. 3. Goldberger, Joseph, and G. A. Wheeler (1920): The Experimental Production of Pellagra in Human Subjects by Means of Diet. Bull. Hygienic Laboratory No. 120, U. S. Pub. Health Serv., Wash., D. C. 4. Snow, John (1936): On Cholera. New York: Commonwealth Fund; and London: Oxford University Press. 5. Siler, J. F., P. E. Garrison and W. J. MacNeal (1914): A Statistical Study of the Relation of Pellagra to Use of Certain Foods and to Location of Domicile in Six Selected Industrial Communities. Arch. lnt. Med., 14:292-373. 6. Siler, J. F., P. E. Garrison and W. J. MacNeal (1917): An Experimental Test of the relation of Sewage Disposal to the Spread of Pellagra. Pellagra 111. Third Report of the Robert M. Thompson Pellagra Commission of the New York Post-Graduate Medical School and Hospital, New York. 7. Farr, William (1852): Influence of Elevation on the Fatality of Cholera. 1. Stat. Soc., 15:155-183. 8. Doll, Richard, and A. Bradford Hill (1952 ii): A Study of the Aetiology of Carcinoma of the Lung. Brit. M. I., 1952(ii):1271. 9. Haenzel, W., M. B. Shimkin, and N. Mantel (1958): A Retrospective Study of Lung Cancer in Women. 1. Nat. Cancer Institute, 21(5):825•842. 10. Doll, R. and A. B. Hill (1956): Lung Cancer and Other Causes of Death in Relation to Smoking; Second Report on Mortality of British Doctors. Brit. M. 1., 2:1071-1081. 11. Hammond, E. C. and D. Hom (1958): Smoking and Death Rates; Report of Forty-four Months of Follow-up of 187,783 Men. 11. Death Rates by Cause. J.A.M.A., 166:(11):1294-1308. T201054

T 230 Tobacco and Health 12. Dorn, H. F. (1959) : Tobacco Consumption and Mortality from Cancer and Other Disease;. P. H. R., 74(7):581-593. 13. Berkson, Joseph (1955): Statistical Study of Association Between Smoking and Lung Cancer. Proc. Staff Meeting Mayo Clinic, 30:319-348. 14. Mainland, D., and L. Herrera (1956): The Risk of Biased Selection in Forward- going Surveys with Nonprofessional Interviewers. 1. Chron. Dis., 4:240-244. 15. Kortewig, R. (1956): Significance of Selection in Prospective Investigations into Association Between Smoking and Lung Cancer. Brit. J.Cancer, • 10:282-291. 16. Neyman, J. (1( 955): Statistics-Servant of All Sciences. Science, 122:401-406. 17. Frazier, Todd M. (1960): personal communication. 18. Lowe, C. R. (1;)59 ii): Effect of Mothers' Smoking Habits on Birth Weight of Their Children. Brit. M. J., 1959(ii):673-676. 19. Fisher, R. A. (1957 ii): Dangers of Cigarette-Smoking. Brit. M. J., 1957(ii): 297-298. 20. Lilienfeld, A. M. (1959): Emotional and Other Selected Characteristics of Cigarette Smokers and Non-Smokers as Related to Epidemiological Studies of Lung Canrer and Other Diseases. J. Nat. Cancer Inst., 22(2):259-282 21. Eysenck, H. J., Vlollie Tarrant, Myra Woolf, and L. England (1960 i): Smoking and Personality. Brit. M. J., 1960(i):1456-1460. 22. Hammond, E. Cuyler (1960): Smoking in Relation to Heart Disease. AJPH, 50(3, ii):2'0-26. 23. Berk.son, Joseph (1959): The Statistical Investigation of Smoking and Cancer of the Lung. Proc. Staff Meet. Mayo Clinic, 34:206-224a(April 15). 24. Sheps, M. C. () 958): Shall We Count the Living or the Dead? New England J. M., 259:1210-1214. 25. Yerushalrny, J., and Herman E. 1-lilleboe (1957): Fat in the Diet and Mortality from Heart Disesae. New York State J. M., 57(14):2343-2354. 26. Yerushalmy, J., and Carroll E. Palmer (1959): On the Methodology of Investi- gations of Etiologic Factors in Chronic Diseases. J. Chron. Dis. 10(/):27-40. I SeSs1011 IV SMOKING AND LUNG CANCER T201055

17 THE EFFECT OF SMOKING AND NICOTINE ON RESPIRATION J. H. CommoE, Jx., M.D. and JAY Nenm, M.D. I THE constituents of tobacco smoke and their chem- istry have already been discussed. We shall present here the effects of these substances on respiration. The effects may be divided into (A) their local actions, before absorption and (B) their systemic effects. A. LOCAL ACTIONS Inhaled materials might, theoretically, act on any structure in the airway from the nose and mouth down to the alveolar-capil- lary membranes which separate air from pulmonary capillary blood. Any of these structures may be affected by a large number of different chemicals in smoke. I. Effecf on Airway Resistance A number of investigatorsl'' have noted a decrease in. vital capacity associated with long-continued, heavy smoking of ciga- rettes; others have found no change.°, ° Decrease in vital capacity in heavy smokers is generally attributed to obstruction of airways, caused by constriction of bronchiolar smooth muscle, bronchial vascular congestion or edema, increase in mucus secretion or de- crease in ciliary activity. Therefore it seems preferable to measure airway resistance Supported in part by the California State Depiartment of Public Health, Tobacco Industry Research Committee, and U. S. Public Health Service Grant H-4(Y19. 233 T201056

I 00 234 Tobacco and Health 00 N \O SMOKERS NON - SMOKERS ~ (MEANS OF 21 SUBJECTS) (MEANS OF /5 SUBJECTS) ~ 41 0.30 O ~ ~ h Q 0 20 ~ . ~ E'°+ ~ CON- C/GA- /SOPRO- C/G. p CON- C/GA- /50PR0- CA'a fT - TROL RETTE TERENOL ISOPRO- TROL RETTE TEREM7L /50P1W (0.5')'.) TERENOL (a5'/.) TERENOL Fig. 1. The effect of inhalation of cigarette smoke on airway conductance in normal subjects. directly, rather than make inferences from changes in vital capa- city. Here, again, there is conflict among investigators,s, 7-'0 probably because a variety of methods has been used and none of those used measures airway resistance specifically. For this rea- son, in all of our studies, we have used the body plethysmograph method" whieh does, in fact, measure airway resistance, and that quite separately frorn any changes in pulmonary tissue resistance or pulmonary compliance. We convert the data on airway resistance into cond uctance (the reciprocal of resistance ). Then, because airway resistance varies with lung volume,Y2 we measure hmg volume simultaneously (again with the body plethysmo- graph, and express our results as airway conductance/lung volume. A decrease in this ratio indicates an increase in airway resistance and therefore a narrowing of the airway lumen by some mechanism; the actual cause of the narrowing cannot be determined by this method, but if the obstruction is relieved promptly by isopropylarterenol (a bronchodilator which also dilates arterioles), one can infer that the narrowing had been caused by smooth muscle constriction. Figure 1 shows the mean changes in airway conductance/lung volume after 10 to 20 inhalations of cigarette smoke by normal subjects. Effect of Smoking and Nicotine on Respiration 235 Airway conductance/lung volume decreased significantly in 31 of 36 healthy subjects (P<0.001). There was no significant difference between the response of-the smokers and non-smokers. The amount of decrease in conductance/lung volume varied from subject to subject; the mean decrease was 31 percent. These changes were not great enough to cause symptoms or ob- jective changes obvious by clinical examination. Puffing on a cigarette without inhalation of a significant amount of smoke resulted in no change in airway conductance. Smoking a cigar without inhalation of smoke had no significant effect on airway conductance ( P<0.1) . When an aerosol of 0.5 percent isopro- terenol, delivered from a Vaponephrin Nebulizer, was inhaled after the inhalation of cigarette smoke, the airway conductance returned toward control values. When cigarette smoke was in- haled after the administration of isoproterenol, the airway con- ductance did not decrease. Thus, isoproterenol appeared to block the effect of cigarette smoke on airway conductance. The dura- tion of the bronchoconstrictor response varied from 10 to 60 mintites. If a second cigarette was given after the airway con- ductance had returned spontaneously to control levels, there was again a decrease in airway conductance similar to the first. Thus, tachyphylaxis did not occur. The changes after inhalation of cigarette smoke could be reproduced in the same subject on different days. We also studied the effect of inhalation of cigarette smoke by 22 patients with various cardiac and pulmonary diseases who, in the control period had airway conductance which ranged from normal to markedly decreased. All of these subjects were smokers. The changes after smoking were of similar degree to those in normal subjects, and were statistically significant. In order to find the cause of the decreased airway conductance after smoking, we studied the effects of 15 inhalations of isotonic saline aerosol, of nicotine bitartrate aerosol (2 mg/cc. concentra- tion), of smoke from cigarettes with high nicotine content (2.3 mg. of nicotine in the smoke from 47 mm of cigarette length), and from cigarettes with low nicotine content (0.5 mg nicotine in the smoke from 47 mm. of cigarette length). Inhalation of aerosols of isotonic saline and of nicotine bitar- T201057

236 Tobacco and Health trate had no significant effect. Inhalation of cigarette smoke decreased the airway conductance significantly; the changes were similar regardless of the nicotine content of the cigarette. Thus, it does not appear that nicotine is the cause of decreased airway conductance after smoking. (Previous studies on isolated bron- chiolar strips indicated that nicotine had either a dilating or constricting effect on the smooth muscle,13 probably depending on the concentration of the drug and the previous state of muscle tones. ) Because other studies" have shown that the inhalation of a variety of pharmacologically inert, submicronic particles cause a similar decrease in airway conductance, we believe that the changes following smoking are caused by the inhalation of the very large number of small particles contained in cigarette smoke. The rapidity of onset of the changes after inhalation of ciga- rette smoke, and rapidity of the reversal of the effects of ciga- rette smoke by inhalation of isoproterenol aerosol suggest that the changes are due to alterations in the smooth muscle rather than to edema, congestion or increased secretion of mucus. The exact mechanism of the postulated bronchiolar constriction is not known; studies are under way to determine whether the effects are purely local or whether some reflex is involved. There is no information on the production of edema of airway tissues by smoke; inferences made from the edema-producing properties of smoke on the conjuctiva of the rabbit eye may not be valid for bronchial mucosa. The temperature of the gaseous phase of inhaled cigarette smoke has been reported'° to vary from 50°C to 70°C and this has led to speculations regarding heat injury. However quantitative studies have demonstrated that the temperature of hot inhaled gas will rapidly approach the temperature of gases in the upper respiratory tract, so it is unlikely that heat plays any significant role in changes in airway calibre during smoking.7e 2. Effect on Ciliary Activity The mucosa of the respiratory tract is covered with a blanket of mucus. Cilia play an important role in protecting the respira- E$eet o f Smoking and Nicotine on Respiration 237 tory mucosa and alveoli by moving upward this blanket of mucus to which foreign materials, including bacteria, adhere. Because of the difficulties in observing ciliary activity in intact human subjects, investigators have studied either animals or isolated specimens of human respiratory mucosa. Exposure to high con- centrations of cigarette smoke causes ciliary activity to cease in these.l', 1& The exact cause of this inhibition is not known, but nicotine solutions (above 1 per cent concentration ) have been shown to inhibit ciliary activity.l° 3. Effect on Mucus This is a difficult subject to study. For example, observers who time the movement of mucus up the respiratory mucosa'°, 21 are measuring many factors, such as the viscosity and surface tension of the mucus, the thickness of the layer and the activity of the cilia. There are few reliable studies of the quantity and quality of pure respiratory tract mucus as influenced by drugs and chemi- cals in reasonable concentrations. 4. Effect on Sensory Nerve Endings "Smoker's cough" is a commonly described and ill-defined entity. Studies have shown that cough receptors may be stimu- lated by mechanical stimuli such as powdered talc or chemical stimuli such as sulfur dioxide.$E . These stimuli also cause bron- choconstriction, and there may be a relationship between the two responses. Cigarette smoke contains both particulates and chem- ical irritants, both of which may sensitize the receptors of the cough reflex. 5. Effect on the Alveolo-Capillary Membranes It is reported that the pulmonary diffusing capacity is lower in smokers than in non-smokers of similar age, body surface area and height;2, ° however, it is not definitely known whether these changes are due to changes in the distribution of ventilation, changes in the alveolar membrane, or changes in the pulmonary vascular bed. T201058

r 238 Tobacco and Health Effect o f Smoking and Nicotine on Respiration 239 Fig. 2. Effect of inteacarotid and intravenous injections of nicotine in intact dog (top), after bilateral carotid denervation (middle) and after bilateral vagotomy (bottom). 14 kilogram dog, anesthetized with 125 mg of pentothal and 50 ml. of 1 per cent chloralose I. V. 6. EfFect on Pulmonary Vessels Nicotine is known to cause constriction of systemic vessels both by exciting the sympathetic vasoconstrictor nerves and also by a direct action. [t is conceivable that it also causes constriction of the pulmonary vascular bed, although this has not been demon- strated. B. SYSTEMIC ACTIONS Of the known chemical compounds in tobacco smoke, only nicotine is absorbed in amounts sufficient to cause physiological effects. Carbon monoxide may be absorbed and form HbCO to the extent of 5 to 10 per cent of total Hb; this produces no known physiological effect on respiration or circulation because the O2 tension of alveolar gas and arterial blood is decreased only a fraction of a mm. of Hg. Nitric oxide presumably is not absorbed in concentrations which permit its detection; it has an affinity for hemoglobin 1800 times that of CO and, once formed HbHO is unusually stable so that spectrophotometric analysis should identify it readily if any were present. Nitrogen dioxide and higher oxides of nitrogen should form methemoglobin in the blood if absorbed but no un- usual concentrations of M Hb have been found in arterial blood of smokers. Nicotine, on the other hand, is absorbed from the alveolar gas and does produce well defined physiological effects in minute concentrations.23 Its direct effects on respiration appear to be entirely reflex in origin; the receptors involved are in (1) the carotid bodies, (2) the aortic body, (3) the pulmonary and coronary circulations and (4) area postrema in the floor of the fourth ventricle. (1). The Effect of Nicotine on the Carotid Bodies" Figure 2A shows the immediate, intense but brief hyperpnea that occurs after the intracarotid injection of 0.1 microgram of nicotine per kilogram in the dog; there is little or no effect on blood pressure although peripheral vasoconstriction has been demonstrated by appropriate techniques.25 After denervation of the carotid bodies (Fig. 2E ), intracarotid injection of the same dose of nicotine produces no response proving that the hyperpnea is reflex and not central in origin. Figure 2B shows that the smallest effective intravenous dose of nicotine (in this case 1.0 microgram/kilo ) produced only a brief hyperpnea and no change in blood pressure. The fact that the same intravenous dose produces no change in respiration or blood T201059

240 Tobacco and Health pressure af ter denervation of the carotid bodies (Fig. 2F ) illu- strates that the lowest concentrations of nicotine that produce any measurable change do so reflexly through the carotid bodies. It is generally believed that the important effects of nicotine are those on sympathetic ganglion cells; nicotine does stimulate these directly but only when the concentration is five to ten times greater than that needed to activate the carotid body reflex. The mode of action of nicotine on the specialized chemo- receptors of the carotid body is not known. The main physio- logical stimulus to these chemoreceptors is anoxia (a decrease in the partial pressure of OZ in the arterial blood or in the recep- tors themseh es ). However, nicotine does not activate the mechanism that responds to anoxia, because hexamethonium, applied locally, will block the carotid body response to nicotine but not to cyanide or anoxia.2a (2). The Effect of Nicotine on the Aortic Body The aortic body is similar histologically and functionally to the carotid bodies.27 It is located between the arch of the aorta and the pulmonary trunk, receives its blood supply from the aorta and sends afferent nerve fibers up the vagodepressor trunks. Figure 2F shows that it is less sensitive to nicotine than the caro- tid bodies, because a dose of nicotine which previously stimulated respiration when given I.V. (Fig. 2B ) does not do so when the carotid bodies are denervated. However, the respiratory response is of the same type when larger doses are given (Fig. 2G and H); again the reflex nature is demonstrated by the absence of the immediate hvperpnea after bilateral vagotomy (Fig. 2, bottom record). In general, stimulation of the aortic body of the dog causes reflex hypertension as well as hypernea, and the carotid and aortic bodies together are responsible for the increase in blood pressure and respiration that occur after small doses of nicotine given intravenously. (3). The Effect of Nicotine on the Thoracic Chemoreflexes It is now well known that there are sensory receptors some- where within the pulmonary and coronary circulations which, when stimulated, caused the triad of apnea, bradycardia and hy- Effect of Smoking and Nicotine on Respiration 241 ~ _ ~ L E!~~- -~~- t~rIi- ~, i; EEE .,. t In I I_ - I ff .0 Fig. 3. Effect of nicotine solution applied locally to region of right area postrema. Same dog as Figure 2. potension.48 The receptors have not yet been identified histologi- cally, and their physiological stimulus is not known. However, a number of chemicals, including nicotine, stimulate them. Figure 2D and H illustrate the occurrence of apnea, bradycardia and hypotension when 50 micrograms of nicotine is given I.V. (Nico- tine apnea requires a much higher concentration than does nicotine hyperpnea. ) It can be shown by more precise studies that this effect is initiated by nicotine within the pulmonary2a and coronary~g circulations before it reaches the carotid and aortic bodies. Figure 2 (bottom record ) shows that the apnea no longer occurs when both vagi are cut. Many of the pharmacologic investigations of nicotine have been concerned only with the effect of relatively huge doses which cause ganglionic stimulation and marked hypertension as well; the latter per se initiates a secondary, reflex apnea origina- ting in the pressure receptors of the carotid sinuses and aortic arch. (4). The Effect of Nicotine on the Fourth Ventricle A'small region in the floor of the fourth ventricle is sensitive to nicotine and its congeners. Figure 3 illustrates the apnea that occurs within five seconds after nicotine (100 micrograms/ T201060

242 Tobacco and Health ml ) had been applied locally to a square millimeter of the floor of the fourth ventricle in the region of the area postrema.'°, " The effect is initiated locally but the pathways by which respiration is inhibited are unknown. This same area is thought by some32 to be the "emetic chemoreceptor trigger zone" that initiates nicotine emesis. Whether this area is concerned only with the respiratory component of vomiting or is of more fundamental importance in respiratory regulation remains to be determined. (5). The Effect of Toxic Doses of Nicotine on Respiration Nicotine in huge doses is lethal. It causes death by respira- tory failure. In high concentrations, nicotine paralyzes many nerve cells, including the neuromuscular junction (curare-like action). Death occurs from respiratory neuromuscular block at a time when the respiratory center is still sending nerve impulses down the phrenic nerve." SUMMARY The components of smoke can produce respiratory effects initiated locally (before absorption); these are bronchoeonstrie- tion and possibly a change in ciliary action and in mucus secre- tion. They can also cause effects occurring after systemic absorption; these are apnea, hyperpnea and secondary apnea. It is important to differentiate between what smoke and nico- tine can do and what they actually do in the local and systemic concentrations achieved in man. The main local effect is mild, temporary bronchoconstriction, due to inert, submicronic par- ticles settling in the finer airways. The main systemic effect of smoking is transient hyperpnea, initiated by nicotine acting through carotid and aortic body reflexes. REFERENCES 1. Blackburn, IL, Brozek, J. and Taylor, H. L.: Annals of Int. Med., 51:68, 1959. 2. Swan, 1•1. E., Jr. and Hatch, T. F.: Physiologist, 3:159, 1960. 3. Whitt'ield, A. G. W., Arnott, W. M. and Waterhouse, J. A. H.: Quart. J. Med., 44:141, 19551. 4. Short, J. J., johnson, H. J. and Ley, H. A.: J. Lab. and Clin. Med., 24:586, 1939. 5. McKee, K. T.: South Med. J., 51(2) :1110, 1958. Effect of Smoking and Nicotine on Respiration 243 6. Turley, F. C. and Harrison, T. R.: Am. J. Med. Sci., 183:702, 1932. 7. Higgins,l. T. T.: Brit. Med. J., 2:1198, 1957. 8. Attinger, E. 0., Goldstein, M. and Segal, M. S.: Am. Rev. of Tuberculosis Put. Dis., 77:1, 1958. and 9. Wilson, R. H., Meador, R. S., Jay, B. E. and Higgins, E.: New England J. of Med., 262:956, 1960. 10. Franklin, W.: J. Clin. Invest., 37:89'5, 1958. 11. DuBois, A. B., Botelho, S. Y. and Comroe, J. H., Jr.: J. Clin. Invest., 35:327, 1956. 12. Briscoe, W. A, and DuBois, A. B.: J. Clin. Invest., 37:1279, 1958. 13. Hawkins, D. F. and Paton, W. D. M.: 1. Physiol., 144:193, 1958. 14. DuBois, A. B. and Dautrebande, L: J. Clin. Invest., 37:1746, 1958. 15. Greene, C. R.: Science,122:514, 1955. 16. Maritz, A. R., Henriques, F. C. and McLean, R.: Am. J. Path., 21:311, 1945. 17. Dalhamn, T.: A.M.A. Arch. Otolar., 70:166, 1959. 18. Hilding, A. C.: New England J. of Med., 254:1155, 1956. 19. Weatherby, J. H.: Proc. Soc. Exp. Biol. and Med. 45:2.84, 1940. 20. Dalhamn, T.: Acta. Scand., 36 Suppl. 123, 1956. 21. Falk, H. L., Tremer, H. M. and Kotin, P.: J. Nat. Cancer Inst., 23:999, 1959. 22. Widdicombe, J. G.: J. Physiol., 123:71, 1954. 23. Cornroe, J. H., Jr.: Proc. New York Acad. Sci., in press, 1960. 24. Schmidt, C. F. and Comroe, J. H., Jr.: Physioi. Rev., 20:115-157, 1940. 25. Winder, C. V., Bernthal, T., Weeks, W. F.: Am. J. Physiol., 124:238, 1938. 26. ' Byck, R.: Brit. J. Pharmacol., in press. 27. , Comroe, J H., Jr.: Am. J. Physiol., 127:176, 1939. 28. Dawes, G. S. and Comroe, J. H., Jr.: Physiol. Rev., 34:167, 19'54. 29. West, J. W., Kobayashi, T. and Guzrnan, S. V.: Circ. Res,, 6:383, 1958. 30. Mitchell, R.: Personal communication. 31. Nicholson, H. C. and Sobin, S.: Am. J. Physiol., 123:766, 1938. 32. Laffan, R. J. and Borison, H. L.: J. Pharmacol. Exper. Therap., 121:468, 1957. 33. Gold, H. and Modell, W.: J. Pharmacol. and Exper. Therap., 57:310, 1936. T201061

I Environmental Factors and Lung Cancer 245 OCCUPATIONS WtITH A HIGH U1fV0 CANCER DEATH RATE 18 ENVIRONMENTAL FACTORS AND LUNG CANCER PnvL Koua, M.D. Aviscussiorr of the general topic of lung cancer necessarily incliides the consideration of a large number of fac- tors suspected of being significant in the pathogenesis of this neoplasm. Certainly, the subject of this symposium, "Smoking and Health," provides a hub around which I would like to discuss these environmental factors, especially in regard to the relative contribution of each. There are certain basic assumptions I would like to present at the outset. First, the association between cigarette smoking and an increased liability to development of lung cancer is docu- mented. Secondly, this association in all probability has patho- genetic significance, and third, the degree of significance can only be related to the general concept of pathogenesis rather than the specific concept of cause inasmuch as the latter denotes to me an indispensable facet or a crucial factor in the elicitation of a response subsequent to a stimulus. Cause in the traditional concept of the epidemiologist has been primarily related to deficiency diseases and infectious diseases. In these, the critical factor has been documented, and its crucial significance has been established. However confusion results when cause in this usual context is related to the pathogenesis of neoplastic disease. As I understand the epidemiologic approach, it includes studies of geographic, occupational, and general social factors with elucidation of the complex interrelationship between all as they relate to a particular disease. Expressed differently, 244 "4U16S: M[Kw. R JL. Fig. 1. Occupations with a high lung cancer death rate. it is an attempt to understand the occurrence of diseases in rela- tion to characteristics of the life processes of the persons affected. This definition has provided us with a guide for Jaboratory investigation. This first slide (Fig. 1) demonstrates that despite the over- whelming association of a disease entity with any one environ- mental factor, relationships with other such factors must be evaluated in relation to possible pathogenetic significance. Dr. Breslow of the California State Department of Health has been able, on the basis of environmental studies, to document an occupational association for lung cancer. It is the epidemiologist who characterizes distinct populations at risk to specific hazards. I think, however, that we must reserve for the laboratory the quantitative evaluation of the risk within populations that are exposed. Specifically, the epidemiol- ogists tell us, and we all accept, the more you smoke, the greater is your liability to the development of lung cancer. As a patholo- gist concerned with the pathogenesis of disease, I am concerned with the experimental and clinical findings that would tend either to enforce or intensify the meaningfulness of this association, minimize it, or more ideally put it in its proper perspective. Dr. Haenszel's paper may change my mind relative to the limited ability of the epidemiologist to determine quantitative responses within a population at risk. T201062

246 Tobacco and Health Numerous papers have been presented outlining the statistical association between smoking and lung cancer. I quote the chair- man in his "impartiality" who asked, "What would be necessary to satisfy the documentation of the cause?" I may be paraphras- ing him, hut I do not think I am misinterpreting him. Documen- tation requires epidemiological, experimental, and clinical evidence that is compatible with established concepts of experi- mental carcinogenesis as seen in the laboratory, and occupational carcinogenesis as seen clinically. I emphasize the latter since it in essence provides virtually the exclusive source of controlled data relative to the incidence of environmental cancer in man. In order to evaluate any single factor such as cigarette smoking or the one in my own particular area of interest, air pollution, it is necessary at all times to recognize that its effect is manifest in relation to a complex environmental experience. This is particu- larly germane since certainly there is no claim that all cases of pulmonary cancer have either smoking alone or air pollution alone concerned with their pathogenesis. There are certain incompatibilities that are evident on the epidemiological level when the data are interpreted on the basis of cigarette smoking representing the source of the carcinogenic agent. Both clinical and experimental findings provide discrepan- cies. First, despite the differences described in the respiratory epithelium of smokers and the increased risk to the development of lung cancer on the part of smokers, we find that the presumed maximum carcinogenic experience results in the development of cancer in a maximum of 10 to 15 per cent of the population at risk, e.g., the two-pack-per-day-or-more-smoker of more than ten years' duration. We know on the basis of experimental carcino- genic data and occupational cancer studies that the greater and more intense the exposure, in all likelihood the greater percentage of those exposed will develop cancer. The limited incidence in this group is unique when compared with the incidence of cancer in other groups exposed to environmental or occupational hazards. Yesterday a series of slides was demonstrated in which the incidence of inetaplastic changes, presumably preneoplastic, was described as being demonstrable in 90 per cent of heavy smokers. I Environmental Factors and Lung Cancer 247 This 90 per cent of heavy smokers was translatable in terms of overt neoplasms to approximately 10 per cent of the population exposed. This, I think, represents a carcinogenic paradox if one assumes that cigarettes are the source of the carcinogen and that the carcinogenic effect is mediated through a metaplastic re- '~ sponse. This is not to question that the pathogenetic role for ~ cigarette smokes appears to be documented. This is not a bit of platform esotericism because, if indeed smoking is a link in the [~ chain, it makes no difference to public health people where the chain is broken; however the crucial factor is that the chain must be broken at a point where reconstitution or relinking cannot take place. I submit that air pollution represents a unique environ- mental situation for the restoration of the chain independent of smoking and to our minds it is, therefore, a more significant environmental consideration. - Yesterday we saw a series of charts in which metaplasia was demonstrated as occurring maximally in the trachea. In terms of subsequent neoplasms that these metaplastic changes are sup- posed to represent, we note that the incidence of tracheal cancer represents less than 1 per cent of respiratory tract neoplasms. These data are an excellent confirmation of the original observa- tion of Cowdry and Suntzeff.l Legitimate doubt as to the significance of metaplasia in relation to subsequent neoplasia in the respiratory tract is warranted. It is universally recognized that the more intense the exposure to a carcinogenic stimulus, the earlier the resulting neoplasm becomes manifest. The role of cigarette smoke as the source of the carcinogen in relation to age of onset was one factor for discussion preempted by Professor Passey. We have confirmed his observation that no difference exists in the age of onset of lung cancer in studies of over 1,000 successive cases when smok- ers are compared with nonsmokers. To which a logical response may be-so few nonsmokers exist who develop pulmonary lesions of a specific histologic type ( squamous ) that the comparison may be impossible. I will have more to say about histologic type in a moment. However, if we assume the carcinogenic responses represent a dose response, we should see increments of delay in the onset of the neoplasm when the smokers of less than half a T201063

248 Tobacco and Health ruS 2 M mIIDmQ tll f1RI= fIH 0r So07f 011 tR a0l4M1D091 al urtiCK a ~.{-Lwzmm~R niaX )mlzmu Iwtn sTlmt }1v7[111 amwrbmN ~ 01 3 .1. ?1..r lml.¢lw f .1. {sw.. f.lr1tw ~ ~ 0 Hmt:e1. hw) l/r. N• (w) 2 M-l.A N t1aN ~ i M..rMa 1T 21 ... 4 C EwJ 1 ~ t0 t0 !0 - tG 100 - S 1t l0 7) - 10 100 - ] 17 10 /0 - 10 H - 4 17 t0 K - 2.5 N - ~ s :r to W - 2.3 w - a 32 10 N - :.S /{ - 7 ]f 10 H - 2.5 72 - e w t w - f f w a u - f t0 N IC 75 - / 11 1S0 F f 11 a79 I I 1 t / / 0 ... .~~ a..tt.. I 1 pql.. ee.a • ).a.0...r1r... Fig. 2. Metaplasia. pack, half to one pack, a pack and a half to two, and more than two are compared. At least in our own data we have not found this to be so. Apparently Professor Passey's data are in agree- ment. Umiker and Freneh2 in an analysis of 121 patients with lung cancer found only 1 of 38 squamous cancers in a man under 50 years of age. For squamous cancers as a whole the mean age of diagnosis for patients was 62.3 years as against 57.4 years for adenocarcinoma, 58.5 years for poorly differentiated tumors, and 57.8 years for undifferentiated tumors. The possible neoplastic significance of squamous metaplasia is apparently ;•elated to its etiology. The next slide (Fig. 2) lists specific factors which amply document that metaplasia is a nonspecific response, more significantly a response which carries with it markcrlly contrasting natural histories, from reversibility to, indeed occasionally, invasive cancer. Because the microscope provides little help in predicting the fate of metaplastic changes, I now believe that it is the cold room, the Warburg apparatus, enzyme assay, and similar biochemical studies which will provide us with this necessary information. SommersR in a report on host ( to our original observation that approximately 85 to 90 per cent of the group exposed to a presumed carcinogenic dose, and I define i this as the dose beyond which the increased exposure does not result in increased incidence, do not develop lung cancer. No data have come to my attention which would indicate that the 30 to 40 cigarettes a day smoker when contrasted with 40 to 50 cigarettes a day is at a greater risk to development of lung cancer. From the 0 to 30 cigarettes an increasing risk is noted, but the ~ plateau of incidence occurs at a level that permits the tremendous majority of those exposed to escape development of the disease. I What are the possible explanations for this? First, genetic ~ differences exist in those exposed. Almost certainly in the Men- delian sense, this does not appear to be a valid explanation. Second, local tissue resistance varies with the individual; there is little evidence for this. On the experimental level, we find carcinogens which more or less universally affect all tissues, as, for example, azo dyes in rats and aromatic polycyclic hydrocar- bons in mice. Third, host resistance exists on a systemic level Environmental Factors and Lung Cancer 249 factors in fatal human lung cancer refers to Auerbach's' prior work and then discusses his own findings and states, "Squamous metaplasia of bronchi did not occur frequently enough or in loca- tions that indicated any direct relation to cancer development, in " "Bronchial epithelial agreement with some European reports . hyperplasias accompanying lung cancer are observed significantly more often if specifically and systematically sought." Similarly Cunningham and Winstanleys noted in their study that such changes as basal cell hyperplasia and squamous metaplasia are two types of change differing in kind, and not merely in degree, which show no apparent correlation with smoking habits. I mention these reports to show that competent evaluation can, and indeed has, resulted in findings of a relatively conflicting nature rather than to suggest greater or lesser competency on the part of these investigators. There are other factors which would tend to indicate that the role of cigarette smoke in the pathogenesis of lung cancer is a wholly nonspecific one. This concept infers that the effect of cigarette smoke becomes manifest in the presence of environ- mental mental carcinogens derived from other sources. Let us return T201064

250 Tobacco and Health with immunity resulting from the presence of antibodies or other immune substances; no data are available to support this. Finally, differences in metabolic pathways exist between individuals. This would almost have to be interpreted in terms of item one on this list. However, since differences in enzymatic pathways would, of course, reflect nucleic acid differences, we would have to assume genetic differences on the cellular level, and little evidence for this exists. What are the possible alternatives that remain? First, in view of the fact that 85 per cent of those exposed to cigarette smoke do not get lung cancer, cigarettes, as a corollary, are not con- cerned with the pathogenesis of lung cancer. I think that we are unanimous in dismissing this complete exoneration. Second, cigarettes are indeed concerned, but on a level of weak carcin- ogenic activity. A possibility perhapsl An unlikely one when it is recognized that this weak carcinogen presumably is supposed to: (a) explain the difference between rural and urban incidence, (b) account for difference between one type of histopathological pattern of neoplasm and another, (c) determine the anatomic location of the neoplasm, and (d) compensate for the additional epidemioloi;ica( characteristics for which smoking provides no explanation, e.g., trne occupational exposure. This is much to ask from a weak carcinogen. A third alternative, and the one to which I subscrihe, is that cigarette smoke by virtue of its irritant properties and by virtue of the attenuating effect it has on the respiratory epithelia] defenses of the host provides an excellent local environment for carcinogens, independent of their source, to act. I wish tr: emphasize that exogenous factors can be most meaningfully discussed in terms of the endogenous factors that they modify. In other words, the quantitation of carcinogenic agents in our environment, a facet of this entire problem with which my associates and I have been greatly concerned, is mean- ingful primarily in the sense that their presence establishes a theoretical basis for a hazard. The demonstration of benzpyrene in the air does not necessarily imply biological effect or tissue specificity. Only with the determination of a biologic effect does benzpyrenc become meaningful as a carcinogenic agent to those exposed. The difference is as between a bullet resting inertly on a f Environmental Factors and Lung Cancer 251 table and a bullet coming from the muzzle of a gun-the first innocuous, the second lethal. I wish to devote the rest of my paper to the endogenous factors. , The first of the endogenous factors to be considered is that of histopathological pattern. Incidentally, I am not necessarily dis- cussing these factors in the order of importance. Great emphasis has been laid on the fact that epidermoid cancer is the environ- mental cancer while adenocarcinoma is the endogenous cancer. An excellent study, now in press,' using cytochemical techniques has shown that mucus- production occurs in an overwhelming number of the neoplasms heretofore regarded as being purely epidermoid. The significance of this is great. First, it indicates that the host factors or endogenous factors are operating to modify the histopathologic pattern of an exogenously induced neoplasm. Second, and perhaps more meaningful, is the implica- _ tion that epidermoid carcinoma of the lung has been relatively constant in its incidence while adenocarcinoma may well be the lesion that is increasing. This is so, at least insofar as the records of the Los Angeles County General Hospital indicate. I regret more than I can say the absence of Professor Kreyberg. I am sure that as a titan in this field and as a person well acquainted with the data I am now referring to he would have had something meaningful to say about these data. I wish to allude to our studies of the Los Angeles atmosphere; although I am presumably primarily discussing endogenous factors.' I refer to the confusion that results in the minds of laboratory personnel by the conflicting statements that epidemiol- ogists frequently make. We have been told, and I have read, that obviously air pollution in Los Angeles cannot be too significant in the pathogenesis of pulmonary neoplasms. Seventh Day Adventists do not demonstrate an increased incidence of pul- monary cancer. This statement conflicted with my concept of the latent period and so on; so I imposed upon my friendship with Dr. Breslow" for a comment and I quote from his response, "The California Health Survey disclosed in 1954 that 49 per cent of adult Californians had migrated to the State before 1946, 29 per cent had moved to California after that time, and only 22 per cent had been born in California." Since migration to Los T201065

252 Tobacco and Health Fig. 3. Benzpyrene concentration in selected urban areas. Angeles has been occurring at an even higher rate than to California as a whole, it is clear that one cannot consider death in Los Angeles as implying long years of residence or exposure to conditions there. I had no idea what Dr. Breslow was going to answer when I asked this of him, and certainly his fellow epi- demiologists in this room are well aware of his personal belief that cigarettes are a significant, perhaps the most significant, fac- tor in the pathogenesis of lung cancer. It reemphasized the validity of the concept of latency in the development of neo- plasms as frequently started by such men as Dr. Hueper and Professor Steiner. Now, back to our own work. What are the combinations of exogenous and endogenous factors which appear significant in elucidating the mechanism of pulmonary cancer development? Studies in this area are important because it is only through the latter that the question as to whether smoking is a factor or not can be answered qualitatively and quantitatively. It is well to recall at this time that benzpyrene has been identified in virtually all urban areas where it has been sought (Fig. 3). Its presence alone is of little significance unless its stability and physical state are also considered. The relative persistence of benzpyrene even in the presence of highly reactive Los Angeles type smog is shown in the next slide (Fig. 4). Environmental Factor s and Lung Cancer 253 ~ ~ BENZYPYRENE (BP) N IN SELECTED U CONCENTRATION RBAN AREAS ~ .wt V 100 CU. M. /6EA pt I1. ®AJ4M. SOOT LOS AN!lUrIES (U. S. A.) 3.0-3,25 ~ ® RETIUAVTK OCELAN'D) 0.23 152-255 ~ Af1l'CEN (NORWAY) 0.5-1.9 125-233 ~ OSLO (NORWAY) 0.09-1.52 143-300 F~~1 COPEItNACEN (DENMARK) 1.45 295 SMEFF1El0 (U. K.) 2.0-33 136-190 LONDON (U. K.) 2.6-14.7 BP CONCENTRATI ON AT SOURCE EXHAUST DIESEL ENGINE- INE MfDaUCTS GASOLINE ENGINE- FFICIENT OPERATION 120 IDIfNG 310 Fig. 4. Survival of polycyclic aromatic hydrocarbons in washed air and smog. i 1.4 t2 io d C.n .6 .4 .2 ~......~ a ...~+ ~ ~.~a' .0 P,ICNEA or .39 /OL ES ~2 OZL / LgR& BRONCN/ A4 - Frs S - Fig. 5. Particle size. r T201066 I SSAGE T/ME /SEc)

254 Tobacco and Health RETENTION /m .25.5.75 /. RETENT/ON /N LUNG 1-1 l~s. P.IRT/CLE S/ZE 2. 3. 4. 5vu P•4RT/CL E S/Z£ - F/6. 4 Fig. 6. Particle retention and passage time in relation to level of tracheo- bronchial tree. In the next slide (Fig. 5) one can readily see that particle size is a major factor in determining the extent of pulmonary reten- tion. I would also like to show that particle size must be con- sidered in relation to the diameter of the lumen of the tracheo- bronchial tree and the time required for passage of particles if one is to postulate host entry and effect. The next slide (Fig. 6), which is an adaptation of Dautrebande's" data to which we have added some of our own, demonstrates this. These data indicate that particles do settle out in different levels of the tracheo- bronchial tree. Particulate rt~tention is a crucial factor in postulating carcino- genic action of any environmental agent. We have found that particulate retention and suhsequent biological activity, at least in relation to particles in polluted air, fall in a size range that is biologically significant."'• " The excellent papers on the chem- istry and physics of cigarette smoke report findings that are extrapolatable to air pollution in terms of identification and quan- titation of carcinogenic compounds and the possible relation of these compour:ds to the epidemiological characteristics of lung J t Environmental Factors and Lung Cancer T~kte I tARTLIL LtSt Cf MTDlOG4A10M6 riOCSGR LN TOLWRD At! A!O CTWML flqA6 CCMmqtATE Ca.nouM YntlutN Atr..__ Clwrtttel..k. Cerctstent AtentphtA,teaue tto-tFr.m. Anthr/Cene trrwu• rWmr.ntAen. 1.2-lmwethrseme /nni-(..n.n)•tluoranthnn. Chryeen. 1,2•~.nrryr.w. 7,4-fnnn"tene AnthentArea. I ,12denqerYleae 11,12-t+nrlluenaatheae ),Adnartiuar., huoe Corm,rvene HrYtene I , 2,5,6-0lbenunthr-rene 255 Crrtan d.ta tre. Cowr tnA LloidatY. V!n Duuten. .nd Nnttan .nd Y,nder. Fig. 7. Partial list of hydrocarbons present in polluted air and cigarette smoke condensate. cancer. Except for the alkaloids, the aromatic polycyclic hydro- carbon content is similar.7z In addition, we have demonstrated the presence of aliphatic hydrocarbons and their reaction pro- ducts, epoxidic substances, etc. in polluted air. Because of the similarity in chemical findings, I have specifically deleted the details in this portion of my text. The chemical findings are partially summarized in slide 7 (Fig. 7). We have established by means of light microscopy, fluorescence micrctscopy, and isotopic studies that particulates from polluted air are retained in the respiratory tract. In addition to the particle size and stability of carcinogenic agents in the environment, the effect of the vehicle or carrier of the carcinogen is also of significance. Since we cannot reproduce the vehicles exactly as they exist in the air or in cigarette smoke, studies are necessarily limited to the parEicles themselves. We have infused carbon-labeled benzpyrene into the tracheobron- chial tree as part of a series of experiments which has as its T201067

1Y 256 Tobacco and Health Environmental Factors and Lung Cancer 257 tLMt . Comwf )bmRDft® tM C14WJC7017® tM TIM1 7,W7tWULR 1Wit0 0f lOL7IDf® A]t 1p0 Tpt71m 7pt CASC110CIDnIC p WRFCWCII¢611C tROf@itlN CArw1 .r nattal une fireteepwtrte7 Awttrutte.fr.trtq a /N..m4r Anapltl,Iw. - rt.a.Nrw. - / AwMt«r« . { n ». - 71wrr,Nwr . t,7aa.uswtAr.cu. IF I t.at•W..,.)•flwewtFwt . "r YIS,»s.YU,)sssw - td.tanerrr.« . /atAwWtmr . t,l7awi,rtrtr« - tl,tt•senrttwrrotAwr - ),w«rttw..wW,w. M i Iww.« )H,)N7,1T1,9dt wt t«a~ "M AU,100,Y17,YH..17 " » ),1-7nslryf« Nlw« •• " Fig. 8. Per cent retention of CI'-3,4-benzpyrene after injection. ultimate goal the actual inhalation of carbon-labeled carcinogens. The results of these studies demonstrate the entry of the carcino- gen into the epithelial cells and circulatory system. This is in turn reflected by passage from the circulatory system into the liver for detoxification and excretion of the metabolites into the bile. These studies reemphasize the importance of quantitative host entry rather than environmental presence in terms of a postu- lated biological effect. An experimental example of the effect of the vehicle I have just described is seen in the next slide (Fig. 8). Following the subcutaneous injection of a carcinogen in an aqueous, protein, or lipid medium, systemic elimination differs significantly witit the vehicle used.'g The implication, of course, is that the longer the residence time prior to elimination of the carcinogen or i+-s metabolites, the longer time available for a postulated hiolo,ical effect. A critical factor in the incidence of lung cancer is the recogni- tion that environmental carcinogens are not produced in a vacuum, nor do they exert an effect independent of other associ- ated environmental factors. Specifically, carcinogenic hydro- 4 , 4 0 "0" alt,)17,)f7,YM,)J0,)1{ " ^ AaWntFw« " » h^ 1N,Ut,)M,)10,)7t " • ^0^ NA,)11,Yp,)01,)/n,7N " " Mw7te.wMwl /01,1!),)/1,)!).)J " • . H~tnttaq setlr/V wa atu~.Ptert 1,IJ,bM1..uake...an..r .w~wreMw,7-wown~w««, w« t.r 1.« t.re.. au.at ),.•a..qyr.« Fig. 9. Compounds identified or characterized in the particulate matter of polluted air and tested for carcinogenic and anticarcinogenic properties, carbons are produced in relation to a host of associated poly- cyclic hydrocarbons and other compounds as well. Beginning with the pioneering work of Lacassagne, it has been frequently shown that inhibition, summation, or addition of carcinogenic effect may result from the simultaneous presence of carcinogens and related compounds. In our studies, we have found that in the presence of a constant dose of a carcinogen the tumor yields can be varied by altering the amounts of associated compounds and modifying the temporal aspects of their administration. On the basis of our laboratory data, we turned to polluted air for agents potentially capable of modifying carcinogens. As shown in the next slide ( Fig, 9), anticarcinogenic compounds exist in a common milieu. It is conceivable that the difference between the Los Angeles and New York air, the difference between British T201068

258 Tobacco and Health FLOW RA7E MM/MINUTE APPLICATION OF IRRITANT A = EFFECT AFT1tR 1' B - EFFECT AF'TER 16' C - EFFECT AFTER 46' A-B + BAsE uNE tai F-1s=--H•~o'~~, OT]A it} A'r 0 EEF 4if~ S~t 6~ 70 7'6 fD a W MINUTE6 Fig. 10. Ch;Tracteristic response of mucous flow to test agents. 11 l0 IRRITANT EFFECT OF SMOG ON INTACT RADDIT PROGRESSIVE DECREASE IN RATE OF MOVEMENT OF 9 SOOT PFRTICLES ON RESPIRATORY EPITHELIUM . ~ UDUDS OF ERPOSURE TO SYNTWETIC SBIDD 11 2 PPM OXIDIINT) Fig. 11. Irritant effect of smog on intact rabbit-progressive decrease in rate of movement of soot particles on respiratory epithelium. I Environmental Factors and Lung Cancer 259 too CTMTL tAURAT[ •of~ ~ ~`-- •- --•----•--• a 6 7 DAYS POST INJECTION Fig. 12. Area to left of "0" line = soot phase; area to right of "0" line = plasma phase. and American cigarettes, or the difference betwe'en any environ- mental exposure in terms of tumorigenic response is a reflection of carcinogenic activity as modified by associated compounds. The dose of the carciriogen as a factor has been studied. Con- trasted with a divided dose of a carcinogen a single dose of a car- cinogen results in a difference in tumor yield from 20 per cent to 70 per cent. So the method of exposure, e.g., breathing of polluted air at constant intervals and smoking at unfixed intervals, is a variable which must be regarded when the carcinogenic activity of an environmental agent is postulated. The effect of irritants on the respiratory epithelium, ciliary activity, and mucous secretion is of major significance in relation to carcinogenic effect. As shown in the next slide (Fig. 10), irritants first produce an initial stimulation of ciliary activity and mucous stream flow followed by an inhibition with ultimate recovery. In Figure 11, the data from a representative experi- ment are given. T201069

I 260 Tobacco and Health METAPLASIA Etiology II NUTRITION (Avilamiw"t A) 2) AH6XEMIA (koforel of Prortol.) 3) METABGLIC (Ur.mo) 4) IRRITATION Al .1.cfrWe[ u 01irYm0Al. Natural History q wEVewSIBOLITv 2) NECROSIS WITH SCAR TISSUE FORMATION 3) PR0GRES840H A) ATTIOWI .) MI sITV CAMC[q e) NIVASIV[ CAMCM Fig. 13. Influence of particle size of soots on the adsorption and elution of 3,4-benzpyrene from benzene and dilute protein solution It is possible to demonstrate both by light microscopy and fluorescence microscopy the deposition and retention of soot particles laden with 3,4-benzpyrene and the intracellular entry of the benzpyrene which has been eluted from the soot particles. The next slide (Fig. 12) demonstrates the host's ability to separate carcinogens from the particulates to which they have been adsorbed as a reflection of the action of plasma proteins. The latter act as elating agents. Figure 13 demonstrates the in- fluence of particle size on the ability of plasma proteins and ben- zene to elute 3,4-benzpyrene. Carcinogens are not found in environmental sources in crystalline form. Rather they occur as part of complex chemical particles with distinctive physical properties. Variations in the rate of liberation of carcinogens from the particle carriers have been demonstrated in various species. Prior to concluding, I would like to briefly discuss the possible II I Environmental Factors and Lung Cancer 261 significance of antecedent or co-existing pulmonary or systemic disease on the development of lung cancer. Professor Passey has alluded to the experimental and * clinical relationships between chronic pulmonary disase and pulmonary neoplasms. We have been concerned with the fact that an important area of research may have been overlooked by the universal concentration on the pulmonary parenchyma itself. This concept is in accord with a previously stated opinion of Stewart, Chief of the Laboratory of Pathology of the National Cancer Institute, who said that un- equivocal data are still lacking which show that action of carcino- gens in the respiratory tract is exclusively the result of direct and immediate effect on the cell. A systemic factor may be involved so that prior metabolism of the carcinogen must be considered. Subsequent to host entry and systemic metabolism of carcino- genic agents the respiratory cells may be exposed to a carcino-. gen from below, as it were, through the subepithelial blood supply. In many of the cases of lung cancer at the Los Angeles County Hospital a panorama of diseases, both pulmonary and systemic, was noted. Since the liver is primarily concerned with metabolism of carcinogenic hydrocarbons, we initiated experi- mental studies to determine the relationship of liver disease to metabolism of carcinogenic agents. The livers of rats were injured with carbon tetrachloride and alterations in the meta- bolism of 3,4-benzpyrene were studied on the assumption that a metabolite of the carcinogen, rather than the parent substance, present in the environment may be the active tumor-inducing substance. Data from these studies have been published and all I will say is that the profile of metabolites varies with the patho- logic state of the liver. I will conclude by simply stating that we feel the evidence warrants recognition and acceptance of the concept that multiple factors are concerned with the initiation and promotion of lung cancer. Quantitation and determination of the mechanism of action are problems primarily for the experimentalist. First and most crucial, of course, is the necessity for elucidating a step-wise or sequential series of events which is compatible with the role for the several environmental agents statistically associated with increased risk to this disease. Second, the links in the chain of T201070

262 Tobacco and Health events, if valid, should be capable of correlation with the clini- cal and pathological characteristics of the disease. We are aware of nothing that indicates, let alone proves, mutually exclusive roles for any of the suspected agents in any given case of lung cancer. The order of magnitude of each may vary and methods for quantitating the relative contribution of any or all suspected factors have not been developed as yet. Lung cancer is seen in rural as well as in urban residents, in smokers and nonsmokers. The public health implications obviously are those associated with the reduction in the disease coincidental with removal from the environment, if at all possible, of any initiating agent. It is our belief, on the basis of experimental and clinical data that following the diluting of the effect of cigarette smoking a reduc- tion in lung cancer would result. Although epidemiological data indicate that a great majority of lung cancers are statistically associated with a history of cigarette smoking, to anticipate a reciprocal reduction in incidence is wholly unwarranted. REFERENCES 1. Cowdry, E. V., Suntzeff, V., and Ide, G.: Comparison of histopathology of bronchial and tracheal epithelium of smokers and non-smokers (Abstr. in Program) Seventh International Cancer Congress, London, July 6-12, 1958, p. 152. 2. Umiker, W. and French, A. J.: Bronchogenic carcinoma-A second look at cell type. Cancer, 13:1053, 1960. 3. Sommers, S. C: Host factors in fatal human lung cancer. t1.M.A. Arch. Path., 65:104-111, 1958. 4. Auerbach, 0., Gere, J. B., Forman, J. B., Petrick, T. G., Smolin, H. J., Muehsam, G. E., Kassouny, D. Y., and Stout, A. P.: Changes in bronchial epithelium in relation to smoking and cancer of lung; report of progress. New England J. Med., 256:97-104, 1957. 5. Cunningham, G. J. and Winstanley, D. P.: Hyperplasia and metaplasia in the bronchial epithelium. Annals of the Rogal College of Surgeons, 24/5:323, 1959. 6. Herman, D. L. and Crittenden, M.: Distribution of primary lung carcinomas in relation to time as determined by histologic techniques. J. Nat. Cancer Inst., in press. 7. Kotin, P. and Falk, H. L: Role and action of environmental agents in patho- genesis of lung cancer; I. Air pollutants. Cancer, 12:147-163, 1959. 8. Breslow, L.: Personal communication. 9. Dautrebande, 7.: Physiological and pharmacological characteristics of liquid aerosols. Physiol. Reu., 32:214-251, 1952. I N O M ~ ~ 10. ~ O 11. ~ ~ 12. N 13. Environmental Factors and Lung Cancer 263 Tremer, H. M., Falk, H. L., and Kotin, P.: Effect of Air pollutants on ciliated mucus-secreting epithelium. J. Nat. Cancer Inst., 23:979-997, 1959. Falk, H. L., Tremer, H. M., and Kotin, P.: Effect of cigarette smoke and its constituents on ciliated mucus-secreting epithelium. J. Nat. Cancer, Inst., 23: 999-1012, 1959. Kotin, P. and Falk, H. L.: The role and action of' environmental agents in the pathogenesis of lung cancer; IL Cigarette smoke. Cancer, 13:250-262, 1960. Kotin, P., Falk, H. L., and Busser, R.: Distribution, retention, and elimination of C"-3, 4-benzpyrene after administration to mice and rats. J. Nat. Cancer Inst., 23:541-555, 1959. T201071

I Quantitative Evaluation of Etiotogic Factors in Lung Cancer 265 acteristic pattern. Not many findings have been reported which grossly contradict other data. Some remarks on individual effects will precede more general comments on the nature of the pre- 19 QUANTITATIVE EVALUATION OF THE ETIOLOGIC FACTORS IN LUNG CANCER WILLIAM HAENSZEL, M.A- Irr essaying some remarks on the "Quantitative Evaluation of the Etiologic Factors in Lung Cancer," it is well to state at the outset that virtually all the information presently at our disposal is derived from observations on the species of interest -man. While the results of animal experimentation may be use- ful to elucidate possible mechanisms of pulmonary carcinogenesis, it would be difficult to translate such findings into an assessment of the degree to which environmental and endogenous factors contribute to the variation in lung cancer risk in man. Several review articles summarizing from a variety of viewpoints the literature on lung cancer epidemiology have appeared in recent years.l` While major. interest was directed in several of these reviews to smoking and lung cancer, all make reference to other variables. It is not my intention to retrace all the ground covered in these papers, although some recapitulation is unavoid- able. A few comments will be offered on lines of further investi- gation suggested by the present evidence. The salient foatures of what is known about the magnitude of the associations displayed by the various factors have been summarized in Table 1; for ease of reference the material has been organized under subject-matter headings. No encyclopedic compilation was intended, the primary purpose being merely to indicate the relative importance of the various effects and to assign a rough ordering. The individual studies fall into a char- ~ sently available evidence. ~ The usual practice has been to compare mortality in occupation ~ groups with that for the general population, making some adjust- ~ ment for age, but not other characteristics. No studies of occupa- tional risks have systematically controlled for smoking history or place of residence, although residence would not be an important factor in studies of workers living in one locality. Perhaps the most profitable way to view the data on occupational exposures is to consider them as important evidence as to the neoplastic susceptibility of human lung tissue. The number of special situa- tions in which workers have experienced lung cancer risks 20-30 times that for the general male population would seem to permit . no other conclusion. There has been no serious challenge to the interpretation of this family of results as a reflection of the effects of carcinogens in the work environment. This does not mean that there has not been some disagreement as to their identity in certain situations. For example, it has been suggested that radi- ation emitted from radon may not have been responsible for -the high rates in Schneeberg and Joachimsthal miners and other factors such as pneumoconiosis had not been ruled out.6d A number of high- and low-risk occupations in the lower range, i.e., 2-3 or 1/3-1/2 times the prevailing rates in the general pop- ulation of the same sex can be identified in occupational mortality data published by vital statistics offices. The latter instances may differ from the more extreme examples cited in quantitative degree rather than qualitative manner with respect to exposures to carcinogenic hazards. The literature reveals no serious attempts to account for these results in terms of occupational selection, to wit, that the association is an indirect one and reflects a link between lung cancer and the personal characteristics of workers attracted into these pursuits. This might come as a surprise to persons who recall Morris' studies of coronary heart disease in London bus drivers and conductors which did reveal evidence on differences in certain physical characteristics be- tween the two occupational groups.66 Perhaps the magnitude 264 T201072

T 266 Tobacco and Health of the lung cancer risks observed for the miners and chromate workers have been so large as to discourage entertainment of selection hypotheses in this connection. The effects associated with socio-economic class are rather small and whetlter they reflect a dilution of specific occupational effects is uncertain. This is not entirely implausible since groups of occupations have been used to define social class. The social class difference may be a rather recent phenomonon; the Regis- trar-General's report for England and Wales indicated virtually no social class differential in 1930-31.21 Again, no direct control of smoking history has been incorporated in any of the socio-economic class results. However, what is known about the rather small social class differences in smoking habits°° would suggest that control for smoking history would not eliminate the social class differentials. The vital stas istics data of several countries agree that urban residents are subject to greater lung cancer risks. The gradient in risk tends to increase by size of the urban conurbation in most countries. It appears that the urban-rural differential is greater for males than for females. A few studies of urban-rural risks which took smoking history into account did not yield clear-cut results, hecluse of the limited number of observations. These will be discussed later. In the United States, an indirect adjustment for smoking history indicated that the urban-rural differential for males could not be accounted for in this manner, although it did appear that much of the female differential might be wiped out.n7 It seems unlikely that differences in rates recorded in England, Norway, the United States and other countries can be explained on diagnostic grounds. Furthermore, the experience of migrant populations woiild argue strongly against such an interpretation. The findings of Eastcott and Dean on the excess risk among British migrants in New Zealand and South Africa respectively, as compared to the experience of the native population, have been reported.=° 2° What has not been emphasized suffi- ciently is that the risk for these migrant groups is intermediate to those prevailing in England and the countries of destination. Figure 1, depicting the experience of English and Norwegian migrants to the United States, is of interest in this connection. Quantitative Evaluation of Etiologic Factors in Lung Cancer 267 L 0 0 0 m JO O W ~ Q S3-1dW 000`001 / SHld3a JO '0N T201073

268 Tobacco and Health The experience of English migrants in this country is fully consistent with that in New Zealand and South Africa. The Norwegian migrants on the other hand have a rate lower than that for U. S. native-whites, but higher than that prevailing in Norway. These results are typical of other migrants to the U. S. and lead to a generalization that the migrant group rates for lung cancer tend to be intermediate to those prevailing in "home" and "host" countries. With the exception of the dramatic occupational effects, the largest differences in lung cancer risk are produced by compari- sons based on smoking history. Moreover, studies in this area are in essential agreement that the risk rises in direct relation to amount smoked. In the early days when all the evidence was derived from controlled retrospective studies (and one could have reservations about the adequacy of the study designs and choice of controls ) there was serious question as to whether the results correctly reflected a true difference in risk between smokers and non-smokers (in the observational sense ) or were merely artifacts arising from defects in study procedures. The confirmation by three forward studies", °', 1' of the retrospective results seems to have disposed of the possibility that the early findings were statistical artifacts. A feature of note in the forward studies is that differences in risk persist past the initial years of observa- tion and so can scarcely he attributed to biases in selection and classification of the study cohorts. There is now general agree- ment on the truth of the observation that persons smoking one pack of cigarettes or more a day have a lung cancer risk about 16 times that of non-smokers in the sense that further collection of data on these and other groups would produce similar results. What is still in dispiite is the interpretation as to whether this represents cause and effect or merely reflects an indirect associa- tion with other characteristics or attributes of smokers responsible for the higher iang cancer risks. Even thougii the findings with respect to cigarette smoking history rest on observation rather than experiment, the interpre- tation of the excess risk for smokers as causal in nature has been based primarily on three points: (a) the great magnitude of the excess risk; (b) the regularity with which the risk increases with Quantitative Evaluation o f Etiologic Factors in Lung Cancer 269 amount smoked; (c) the decrease in risk among discontinued smokers. The study findings were deliberately oriented in Table 1 to measures of relative risk rather than absolute differences in risk. While it has been suggested that the absolute difference is a superior measure,68 I do not share this view. Different ques- tions call for different answers and the index measure selected should be appropriate to the question at hand. No one asserts the superiority of a specific measure for such indices of economic activity as cost-of-living or productivity under any and all conditions. The relative-risk measure has an important hier- archical property highly relevant to the present discussion. For example, if the observational association between smoking and lung cancer is accepted, but interpreted as reflecting an indirect association between smoking history and some other characteris- . tic really responsible for lung cancer, then it can be readily demonstrated that this characteristic must have a relative fre- quency among smokers vis-a-vis non-smokers at least equal in magnitude to the relative lung cancer risk for smokers in terms of a unit risk for non-smokers' The magnitude of the relative lung cancer risks by smoking history thus places an important limit on suitable candidates for the hypothesis of indirect associa- tion. What are the agents or attributes for which the relative prevalence is 16 times greater among heavy smokers or nine times greater among all regular smokers than among non- smokers? While much remains to be explored on genetic and constitu- tional effects, the work on ABO blood groups would not appear to offer great promise of uncovering striking effects in this direction. In view of the speculation on effects associated with personality traits it is worth remarking that no direct study by either retro- spective or forward techniques has been carried out to examine gradients in lung cancer risk among persons grouped by person- ality traits or other psychological attributes. Studies indicating differences in personality traits between smokers and non-smokers can be regarded as a device for screening items for further investigation; a test of their potential in accounting for the smok- ing-lung cancer association was given in the preceding paragraph. T201074

270 Tobacco and Health Examination of the findings at hand'°-" would suggest that none of the personal characteristics as defined appear to offer great promise in this regard. How well any of these items in fact discriminate for lung cancer risk remains to be determined by direct study. Many of the results enumerated are more or less uncontrolled descriptions of the effects of individual suspect factors. Not much has been done in way of simultaneous investigation of two or more variables. There are exceptions, most of them in studies dealing with smoking habits, in which the items controlled have included occupation, residence, marital status, and such other habits as coffee consumption.' The Doll-Hill prospective study was limited to physicians.°z Because of the relatively small numbers of lung cancer deaths observed in these studies, most of the adjustments for other factors were indirect in nature and provided little information on the interaction between smoking history and the other factors considered. Despite this, the collective data in Table 1 are strongly sugges- tive on two points, the presence of multiple effects and the magni- tude of the effects for smoking history and selected occupational exposures. While the nature of the responsible factors is not immediately obvious, there are also important effects for other population-urban residents, natives of certain countries and their migrants to other lands, which seem unlikely to be accounted for in terms of other known effects. Without engaging in a discussion of the semantics of "cause" it would be reasonable to adopt for the purposes of further work the view that pulmonary carcinoma is a response dependent on a variety of environmental and host factors, and that no single agent or attribute must always be present and precede the development of this disease. The problem of multiple effects comes to the fore when mea- sures of the importance of a given factor are attempted by estimating the reduction of cases which would follow from the removal of the suspect factor. Such a measure is obviously sug- gested by one of the standard epidemiological tests of the causal nature of an observed association-test by application of control measures. This approach requires information on sub-groups in which the suspect factor is absent. It is, of course, desirable that Quantitative Evaluation o f Etiologic Factors in Lung Cancer 271 such data be fortified by information on gradients in risk among groups for which there are quantitative differences in exposure to the suspect factor to provide further checks on the reasonable- ness of the results for the "zero" class. When two or more vari- ables are involved, the total effect to be achieved by removal of two factors is not necessarily the simple addition of two inde- pendent effects; the answer depends on the nature of the interaction between the variables. While the issues which arise in this connection parallel those dealt with in discussions of synergism and potentiation, they need not concern us here and only the simple question of how information on interaction can affect estimates of the reduction of cases to be achieved by the control of a suspect factor will be considered. The problem will be illustrated by referring to the gradient in lung cancer risk by smoking class among urban and rural residents. The desirability of information on the interaction of smoking history (large effects ) and urban environment (large population affected) has long been recognized as of prime importance; one difficulty has been to assemble sufficient numbers of observations to yield decisive answers. Some suggestive data have been published by Doll,27 Mil1s,R° Stocks,°° and Hammond and Horn,'Z but due in part to the extremely small numbers observed they must be regarded as inconclusive. Doll in assembling data from a controlled, retrospective study speculated "that the incidence of lung cancer in non-smokers may be the same in men and in women and in residents in areas of different density of popula- tion." Mills, from a study conducted in Cincinnati and suburbs and a rural Ohio county, indicated smaller absolute,, interarea differences to be present for male non-smokers than for male smokers, but from these results it would be difficult to hazard an opinion about the relationships of the relative differences. When Mills further classified by driving mileage he reported urban-rural differences to be larger among non-smokers who drove more than 12,000 miles per year. Stocks' two studies in the Liverpool-North Wales area yielded contradictory results; in one the larger urban-rural difference appeared to be among heavy smokers and in the other among non-smokers. The Ham- mond-Horn forward study, while indicating a decreasing gradient ~ ~~ T201075

I ~ O Quantitative Evaluation of Etiologic Factors in Lung Cancer 273 272 Tobacco and Healtla M in risk among non-smokers from cities of over 50,000 to rural areas, was based on a total numerator of nine deaths (well-estab- lished cases of bronchogenic carcinoma, exclusive of adenocar- cinoma ). This interaction may be particularly difficult to investigate by means of retrospective studies because of the overwhelming effects associated with smoking history. Deficiencies in repre- sentative coverage and response errors introduced into retro- spective studies, even those not important enough to lead to serious error in estimating large, primary effects, could lead to serious errors in estimation of other smaller effects. Forward studies based on cross-classification of study cohorts might do better in providing estimates of interaction effects, but the studies under way would obviously have to be continued several years to generate the additional data needed for this purpose. To secure a large number of cases in a limited time, a different approach based on dual sampling of deaths and the population base'a was recently undertaken. A 10 per cent sample of the lung cancer deaths recorded in the United States for 1958, including approximately 2,400 deaths for white males, were queried and information gathered from the certifying physician and family of the decedent on smoking history, residence history, diagnostic methods, and histologic type. The response rate was very good for a study of this type, 89 per cent. Corresponding information for a sample of the general population was obtained as a supplement to the Current Population Survey for May, 1958. The detailed study procedures will be described in another communication. It was recognized that the results from'a dual sampling approach depend upon the comparability of responses obtained from the two sources, and attention was given to the design of study procedures and forms and the classification of data which would improve comparability. The gradients in lung cancer risk by smoking class yielded by this study are similar to those found in the retrospective and forward studies. Figure 2 presents the standardized mortality ratios by smoking class for urban and rural white male residents and for a sub-group of "lifetime residents" in each area. The figures relate to all diagnosed lung cancer deaths; data taking histologic type into ~ ~ ~ ~ ~ 12 ~ Won-smokers 14 H N Occasional and 29 discontinued smokers 49 Regular smokers 104 5 I pock doily 137 Regular smokers 384 > I pack daily 566 5 Non-smokers 16 Occasional and 12 discontinued smokers 30 Regular smokers 66 5 I pack daily 109 Regular smokers 287 > I pock doily 537 STANDARDIZED MORTALITY RATIO 0 200 400 . 600 Fig. 2. Standardized mortality ratios (SMR = 100 for U. S. white males) for lung cancer, by smoking class and place of residence. United States, 1958. account will be published later. The virtue these data have over earlier efforts is the larger numbers of observed deaths on which the ratios are based. Noteworthy features are the trivial urban- rural differential for non-smokers and the wide gap in the rates for smokers using more than one pack of cigarettes daily. The urban-rural contrast is sharpened particularly for the "> 1 pack" group by the restriction to lifetime residents. Great emphasis should not be laid on the SMR of five for the non-smokers among lifetime rural residents, since only three deaths were observed in T201076

274 Tobacco and Health this group. The figure suggests that inclusion of persons with a varied residential history in contrasts based on current residence, particularly of former urban residents who have migrated to rural non-farm areas adjacent to large cities, dilutes the urban-rural effects. Part of the excess in the ratio for the urban "> 1 pack" group might be due to the presence of a greater number of very heavy smokers in this population, but this seems unlikely to provide the complete explanation. The 1955 survey of tobacco smoking patterns indicated that while the proportion of persons smoking more than one pack of cigarettes daily was higher in the urban than in the non-farm and farm populations, the ratio of "> 2 pack" to "1 to 2 pack" smokers was almost constant throughout all groups surveyed." One can scarcely avoid the conclusion that the lung cancer risk for urban residents in the "> 1 pack" group is greater than the sum of two separate "cigarette-smoking" and "urban" effects. Conversely, whatever the nature of the urban factor it would appear to evoke only a limited response among non-smokers. Because of the latent period between exposure and biologic response these statements reflect the influence of conditions at an earlier date. If the urban environment has been altered in important respects in recent years, the effects including the nature of the interaction may be subject to future change. From the present results for non-smokers one might argue that elimination of cigarette smoking could reduce the lung cancer risk to about 15 per cent (an average of the urban and rural non-smoker SMRs ) of the present national rate or an over-all reduction of 85 per cent. This was essentially the estimating procedure used by Doll when he concluded from his fragmentary data that one-fifth of the lung cancer deaths may be attributable to causes other than smoking. Figure 2 would seem to support some of the very early and tentative estimates offered by Doll and others. In the absence of substantive information on the nature of the interaction 1>etween residence and smoking history the conserva- tive course for estimation purposes would have been to assume a constant displacement, either relative or absolute, between the urban and rural rates over all smoking classes. This would lead f Quantitative Evaluation o f Etiologic Factors in Lung Cancer 275 to higher estimated risks for urban non-smokers than those yielded by the present set of observations. By taking the over-all difference in the urban and rural rates and adding it to a pro- jected figure for rural non-smokers (the rural female rate has often been used for this purpose) to derive an estimate for urban non-smokers, one can produce estimates that 10-15,000 or as many as one-half of the current lung cancer deaths in the United States might be prevented by elimination of cigarette smoking. The method employed by Shimkin®1 to calculate that one-half of the deaths as of 1956 were attributable to cigarette smoking in essence depended on a conservative assumption of a constant urban-rural differential in risk superimposed on the smoking- class gradient. The point to be emphasized, which has passed unnoticed by some commentators, is that different models of interaction will produce different estimates on "savings of lives." Obviously, computations on savings in lives might be refined and extended in principle by observing the experience of non- smokers in a wide variety of population sub-groups. Whether the nature of the data warrants very refined computations on savings of lives to be effected under given sets of conditions is question- able. Certainly there are more compelling reasons for studying interactions than the narrow one of estimating removal effects. Such studies would increase our knowledge as to whether the positive association between smoking history and lung cancer risk persists within all population sub-groups as the present evi- dence suggests. Proponents of the family of constitutional hy- potheses, who hold that the smoking lung cancer relationship is an indirect one mediated through an association of cigarette smoking with some other responsible characteristic, can also contribute. In undertaking controlled studies of the personality of lung cancer victims along the lines proposed by Eysenck," it will obviously be necessary to introduce control of smoking history in the analysis because of the strong effects linked to this factor. This would automatically generate additional sub-groups for investigation. By these means can counter-examples of groups be produced in which no positive gradient in lung cancer risk with amount smoked is detectable? Continued failure to identify any group in which a "dose-response" relationship with cigarette T201077

276 Tobacco and Health smoking does not prevail, leaving unshaken a presumption of the universal nature of the relationship, must he regarded as strong support for the causal hypothesis. It is interesting to note that the nature of the interaction between smoking and residence history reported here suggests a partial answer as to why England and Wales with a lower per capita cigarette consumption rate than the United States can still experience higher lung cancer mortality. Other factors such as differences in the proportions of the cigarettes consumed need to be taken into account also. A reconciliation of the inter- country difference would, of course, require more information on the degree of the interaction in the two countries than is now available. In a similar vein one may ask whether the smaller urban-rural differential in lung cancer risk for females is attribu- table to the greater number of non-smokers in that sex and whether the sex disparity in risk differential could be resolved by taking interaction into account. A preliminary, indirect calcula- tion, which can be confirmed later by reference to corresponding information on interaction among females, indicates this to be a likely possibility. Studies of interactions in the way of quantitative descriptions of the changes in slope and intercept of the curves relating lung cancer risk to rate of cigarette use over a wide variety of environ- mental conditions, may also contribute some consistency checks to theories siiggested by animal experimentation. Would the results of Fihure 2 and their amplification by study of more precisely defined population sub-groups fit in with the speculation of Stocks and Campbell that the total intake of benzpyrene through cigarette smoking and air pollution might be incrimi- nated?"°, ~2 From what is known about the relative concen- trations of this or other suspect agents in these two media, could the present cpidemiological findings on urban residence and ciga- rette smokins; he anticipated from the cumulative effect of a single agent? Or will more complicated models postulating the action of two or more agents be forced on us? Can the reports of Kotin and iiis associates that the physical and chemical proper- ties of air pollutants and cigarette smoke interfere with the normal function of the respiratory epithelium by slowing down I Quantitative Evaluation o f Etiologic Factors in Lung Cancer 277 the rate of mucous flow and facilitating particle accumulation and retention°s, ®' be incorporated into a model which can account for the present epidemiological _findings? The questions are raised not to seek answers here, but to illustrate the need for liaison between epidemiological and laboratory investigations. In closing, I should like to emphasize that no one should regard the epidemiology of lung cancer in its present state of develop- ment as a closed subject. Despite the formidable problems in the logistics of data collection which the study of interactions im- poses, it would appear a topic worth pursuing. Data arising from a more intensive study of interactions of several variables, while answering some questions, will pose others calling for fresh study approaches. Investigators have a right and responsibility, along with other interested parties, to interpret the presently available evidence and to recommend courses of action which seem indi- cated, but we can still ask that proponents of various theories on lung cancer strive to retain the frame of mind which leads to the testing of ideas by the design of new studies and the collection of new data. (Table l1 follows on pages 278-286. ) T201078

TABLE I SUMMARY OF OBSERVATIONS ON LUNG CANCER RlSK Group or Material Studied Type of Measure and/or Factors Comparison Used Controlled in Comparison Occupation Schneeberg miners, 1869-1939''' 1 a iners, Jo 1926-1938 ( pprox.),'"u Employees in 6 chromate plants, U. S., 1940-48' Deaths in nickel-producing districts of South Wales, 1948-56" Retired employees of London Gas Co., age 60 years and over, 1939-48'a Workers employed 20 years or more in dusty areas of an asbestos plant, 1922-53" Proportion of total deaths attributed to lung cancer Proportion of total deaths attributed to lung cancer Mortality rates for U. S. males Age Relative frequency of lung cancer Age deaths among nickel workers and other occupations Mortality rates for London males Age Mortality rates for England and Age Wales Magnitude of Lung Cancer Risk Lung cancer accounted for three-fourths of total deaths. Lung cancer reportedly accounted for 40-50 0 of total deaths. Some believe this estimate is too low. Rate 29 times greater among chromate workers. Risk estimated to be 5 times greater, among all nickel workers; among "process" work- ers risk was 7 times greater. Gas workers had 80% excess risk. Those en- gaged in the works had 100% excess risk. Risk was 14 times greater among asbestos workers. Risk was about 30 times greater for workers employed 15 or more years before reduction of hazards was under- taken. TIMN 0116310 ~ TABLE 1 (Continued) Male deaths at ages 20-64, Largest and smallest SMR's among Age England & Wales, 1949-5317 149 occupations. Based on 20 or more observed deaths. Males deaths at ages 20-64 U. S. 1950" Largest and smallest SMR's among Age 425 occupations.* Based on 20 or more observed deaths. •(SMR = 100 for all occupations). High SMR's 227 Metal molders 214 Elevator operators 206 Tailers and furriers 200 Longshoremen and stevedores 194 Shoemakers and repairers (except factory) Low SMR's 54 Farmers and farm managers 54 Farm laborers and foremen 59 Physicians and surgeons 63 Clergymen ~ ~ ~ p. C cu 0 ~ 0 Hi SMR's 230 Gas producer men 1Xj 0 215 Showmen fair and roundabout workers ~ ; , 200 Boiler scalers ou 0 190 Riveters, caulkers Low SMR's ~ ~ 32 Cl ergymea 0 38 Teachers 46 Farmers, farm managers h ~ 50 Registered medical practitioners, radiol- ogists r ~ ~

TABLE 1 (Continued) Social Class Cancer morbidity survey, ten Standardized incidence ratios for 5 Age, race, sex metropolitan areas, U. S., income classes based on grouping 194"t -4it•• of census tracts Male cases reported to cancer register, Copenhagen, Den- mark, 1943-47' Deaths classified by social class, England and Wales, 1950'= Standardized incidence ratios for 5 Age areas in Copenhagen defined ac- cording to house rent Standardized mortality ratios for 5 Age, sex social classes defined according to occupation of male. Deaths classified by social Standardized mortality ratios for 5 Age, sex class, England and Wales, social classes based on grouping 1930-32n of occupations. For white males risk varied inversely with social class, the risk in the lowest income group being double that in the highest income group. For white females, the low income class had a 25% higher risk com- pared with all other classes. Risk for low rental area was 75 o greater than for high rental area. The risk for males tended to vary inversely with social class; the risk of the lowest social class exceeded that of the highest by 45%. For married females, there was little variation in risk by social class. There was little variation in risk with social class for either sex, and for males, no consistent direction of association. For married females, risk tended to be posi- tively correlated with social class. TIMN 0116311 .....,t~ TABLE 1 (Continued) Urban and Rural Residence Forward study of 187,000 Death rate by size and type of Age American white males, place of residence 1952-1955" Cancer re er, Connecticut, Incidence rates for urban and rural Age 1947-195~ towns C a n c e r morbidity survey, Incidence rates for urban and rural Age Iowa, 1950" areas Deaths clasified by place of Death rates for urban and rural Age residence, U. S., 1948-49g areas Deaths classified by place of Standardized mortality ratios by Age residence, E ng 1 a n d and population size ciass Wales, 194649" Non-smokers among 5,000 hos- Estimated death rates for urban and Age, sex pitalized lung cancer and rural non-smokers control patients England, 1952" Mortality for males residing in cities of 50,000 and over was 60% higher than for rural residents; the differential was increased slightly when adenocarcinomas were e:- cluded. The small number of lung cancer deaths observed permitted no firm con- clusions as to the nature of urban-rural difference for non-smokers. Incidence 57% hfAher in urban towns for males; for females the urban excess was 24%. For males, urban incidence was 2.8 times that for rural areas; for females, the• urban ex- cess amounted to 47%. Mortality 85% higher in urban areas for males; for females the urban excess was ~ 31%. o For males mortality in London and other N , , large cities was double that for rural areas; ~ for females, the excess risk for all large h cities was almost 50%. 1 No consistent difference between London, ~ n other urban areas, and rural districts. Most Q of the lung cancer deaths were females. ~ ~ ~ r

TABLE 1 (Continued) Migrants Deaths among British mi- Mortality rates for persons born in Age, sex Mortality 50% higher among British mi- grants to New Zealand, i Iew Zealand grants; 100% higher among migrants aged 19-l9-5P 30 and over at time of immigration. Deaths among• British male Mortality rates for white males born Age Mortality 44% higher among British mi- migrants to South Africa, in South Africa grants age 45-64; no difference at ages over 19-I7-5P 65. Death among English male Mortality rates for U. S. native- Age Mortality 31% higher among English mi- migrants to U. S., 195070 born white males grants. Deaths a m o n g Norwegian Mortality rates for U. S. native- Age Mortality 30 0 lower among Norwegian mi- male migrants to the U. S, born white males grants. 1950'® o TIMN 0116312 ------~ TABLE 1 (Continued) Smoking History Forward study of 187,000 Death rates for smokers compared Age A m e r i c a n white males, with rate for men who had never 1952-55r' smoked Men who had ever smoked suffered a five- fold increase in lung cancer mortality. The increase in risk, for men with a history of cigarette smoking only, by current con- sumption, was: Under t/ pack 640%; t/2-1 pack, 740%• 1-2 packs,,690%; more than 2 packs, 1960%. Gradient in risk became sharper when adenocarcinomas were ex- cluded. Forward study of 200,000 SMR's by smoking history. (100 = Age male policyholders of U. S. ratio for men who had smoked Government life insurance, no more than occasionally) 1954-56" Forward study of 187,000 Death rates by most recent amount Age physicians in United King- smoked (all forms) compared with dom, 1951-56'= rates for non-smokers Men who had ever smoked experienced a five- fold increase in mortality. The increase in risk for men with a history of cigarette smoking only, by current consumption, was: Unerdei P~p la O%450%; t/2-1 pack, 900%; The increase in risk by smoking class was: All smokers, 1190%; discontinued smokers, under 10 years, 740~ over 10 years, 400%; light smokers, 570% moderate, 1130%; heavy, 2270%.

TABLE I (continued) Retrospective studies of male Risk by smoking class relative to lung cancer cases and con- that for non-smokers. trols Retrospective study of 105 Derived prevalence among cigarette Age female lung cancer pati- smokers relative to prevalence ents and 1304 controls, among non-smokers. New York City, 1953-55" Retrospective study of 108 fe- Derived prevalence among cigarette Age, place of male lung cancer patients smokers relative to prevalence interview and 108 controls, England among non-smokers. and Wales, 1948-19527° Retrospective study of 158 fe- Prevalence by smoking class relative Age male lung cancer patients to that among non-smokers. and 339 controls, 29 sources in U. S., 1955-57'0 The composite evidence from controlled retro- spective studies is consistent with the forward-study results on estimates of the magnitude of the differences in lung cancer risk between male smokers and non-smok- ers""" The prevalence among smokers exceeded that for non-smokers by 140%. The excess prevalence by amount smoked was: Under %2 pack, 20%; t/2-1 pack, 90 0; over 1 pack, 1040 0. The excess risk for all smokers was 105%. The increase in risk by number of ciga- rettes per day was: Under 5, 15%; 5-14, 100%; 15 or more, 290%. Epidemoid and undifferentiated carcinomas: The increase in risk for all smokers was 200%; for occasional or discontinued smok- ers, 70%; for 20 cigarettes or less per day, 130%; for over. 20 cigarettes, 610%. The gradient of risk for adenocarcinoma was considerably less. TIMN 0116313 TABLE I (continued) Special Groups Controlled retrospective study Relative risk for Seventh-Day Ad- Sex of Seventh-Day Adventists ventists in terms of unit risk for (whose religion proscribes non-Seventh-Day Adventists. smoking and drinking), 1952-5751 TABLE I (continued) Sex Ratio Mortality data from vital Death rates by sex Age statistics offices' Cancer registers and morbid- Incidence rates by sex Age ity surveys"• =• "• ~''" Series of surgical cases, Nor- Sex ratio of diagnosed cases way, 1948-1955" Survey of several controlled Estimated mortality rates for non- retrospective studi&' smokers of each sex Relative lung cancer risk (epidermoid type for Seventh-Day Adventist males was 14%. In most countries the male-female ratio is on the order of 5-7 to 1. Countries with deviant ratios include Finland (12 to 1) and Israel (2-3 to 1). E In general confirm the sex ratios observed in the mortality data. Sex ratio observed to vary by histologic type. For group I cases (squamous, large and small cell carcinomas) sex ratio was 42 to 1; for group II(adenocarcinomas, adeno- mas) the corresponding ratio was 1.2 to 1. Among non-smokers the male-female ratio is reduced to about a 40% excess among males, a sex ratio typical of that prevailing for many other causes of death.

Quantitative Evaluation of Etiologic Factors in Lung Cancer 287 286 Tobacco and Health E O O.•' d_ .~.Y 00 C~C4 w r. O'+ ~M `ayS ~ 0Ir ~ o ~ ~' 1 4 O') W ~ ~° a~ro~ M U .. .. ~ o .. C td ~~ ~~z .~ ~ ~- ° O ° O bO Vl ~ ~ 4 b IU a ~ 0 ° ~ r a O ZC x c~.., z ~ 0 0 WCO 7 '" y --Ww K m_ ~ U• 6 I I I REFERENCES 1. Doll, R.: Etiology of lung cancer. Advances Cancer Res., 3:1-50, 1955. 2. Hueper, W. C.: A quest into the environmental causes of cancer of the lung. Pub. Health Monogr. No. 36, Pub. Health Ser• Publ. No. 452, Washington, D. C., U. S. Gov't. Print. Office, 1956, 54 pp. 3. Rigdon, R. H.: Consideration of the relationship of smoking to lung cancer. With a review of the literature. South. M. J., 50:524-532, 1957. 4. Com6eld, J., Haenszel, W., Hammond, E. C., Lilienfeld, A. M., Shimkin, M. B., and Wynder, E. L.: Smoking and lung cancer: Recent evidence and discussion of some questions. J. Nat. Cancer Inst., 22:173-203, 1959. 5. Davies, D. F.: A review of the evidence on the relationship between smoking and lung cancer. J. Chron. Dis., 11:579-614, 1960. 6. Arnstein, A.: Sozialhygienische Untersuchungen iiber die Bergleute in den Schneeberger Kobaltgruben, insbesondere iiber das Vorkommen des sog. Schneeberger Lungenkrebses. Beihef t, Wchnschr. des Osterr. Sanitatsw., 25: 64-83, 1913. 7. Rosotoski, 0., Saupe, E., and Schmorl, G.: Die bergkrankheit der erzbergleute in Schneeberg in Sachsen. Zeitschr. f. Krebsforschung, 23:360-384, 1926. 8. Weber, F. A.: Die Bergkrankheit der Erzbergleute in Schneeberg. Arb. a.d. Reichsgesundheitsamt., 57:179-188, 1926. 9. Hueper, W. C.: Occupational Tumors and Allied Diseases. Thomas, Springfield, 1942, 896 pp. 10. Hueck, W.: Kurzer Bericht iiber Ergebnisse anatomischer Untersuchungen in Schneeberg. Ztschr• f. Krcbsforsch., 4'9:312-315, 1939. 11. Sikl, H.: Ober den Lungenkrebs der Bergleute in Joachimstal (Tschechoslowakei). Ztschr. f. Krebsforsch., 32:609-613, 1930. 12. Haenszel, W. M., Moriyama, L M•, and Sirken, M. G.: A proposed study for extending the scope and improving the quality of mortality data. Proc. Wash. Stat. Soc., 1956, pp. 57-64. 13. Brinton, H. P., Frasier, E. S., and Koven, A. L.: Morbidity and mortality experience among chromate workers• Pub. Health Rep., 67:835-847, 1952. 14. Doll, R.: Cancer of the lung and nose in nickel workers. Brit. J. Industr. Med., 15:217-223, 1958. 15. Doll, R.: The causes of death among gas-workers with special reference to cancer of the lung• Brit. J. Industr. Med., 9:180-185, 1952. 16. Doll, R.: Mortality from lung cancer in asbestos workers. Brit. J. Industr. Med., 12:81-86, 1955. 17. The Registrar General's Decennial Supplement. England and Wales 1951. Occupational Mortality; Part II, London, 1957. 18. Unpublished tabulations prepared by the National Office of Vital Statistics. 19. Dorn, H. F., and Cutler, S. J.: Morbidity from cancer in the United States. Pub. Health Monogr. No. 56. Washington, D. Q., U. S. Gov't. Print. Office, 1959. 20. Clemmesen, J., and Nielsen, A.: The social distribution of cancer in Copenhagen, 1943 to 1947. Brit. J. Cancer, 5:159-171, 1951. T201083

288 Tobacco and Health 21. The Registrar General's Decennial Supplement. England and Wales 1931. Occupational Mortality; Part IIa, London, 1938. 22. Hammond, E. C., and Horn, D.: The relationship between human smoking habits and death rates. J.A.M.A., 155:1316-1328, 1954. 23. Griswold, M. H., Wilder, C. S., Cutler, S. J., and Pollack, E. S.: Cancer in Connecticut 1935-1951. Hartford, Connecticut State Department of Health, 1955, 141 pp. 24. Haenszel, W., Marcus, S. C., and Zimmerer, E. G.: Cancer morbidity in urban and rural Iowa. Pub. Health Monogr. No. 37. Washington, D. C., U. S. Gov't. Print. Office, 1956. 25. Hoffman, E. F., and Gilliam, A. G.: Lung cancer mortality. Pub. Health Rep., 69:1033-1042, 1954. 26. Curwen, M. P, Kennaway, E. L., and Kennaway, N. M.: The incidence of cancer of the lung and larynx in urban and rural districts. Brit. J. Cancer, 8:181-198, 1954. 27. Doll, R.: Mortality from lung cancer among non-smokers. Brit. J. Cancer, 7: 303-312, 1953. 28. Eastcott, D. F.: The epidemiology of lung cancer in New Zealand. Lancet, 1:37-39, 1956. 29. Dean, G.: Lung cancer among white South Africans. Brit. M. J., 2:852-857, 1959. 30. Haenszel, W.: Cancer mortality among the foreign-born in the United States. J. Nat. Cancer Inst. (in press). 31. Dorn, H. F.: Tobacco consumption and mortality from cancer and other dis- ca.ses. Pub. Health Rep., 74:581-593, 1959. 32. Doll, R., and Hill, A. B.: Lung cancer and other causes of death in relation to smoking; a second report on the mortality of British doctors. Brit. M. J., 2: 1071-1081, 1956. 33. Sadowsky, D. A., Gilliam, A. G., and Cornfield, J.: The statiscal association between smoking and carcinoma of the lung. J. Nat. Cancer Inst., 13:1237- 1258, 1953. 34. Mills, C. A., and Porter, M. M.: Tobacco smoking habits and cancer of mouth and respiratory system. Cancer Res., 10:539-542, 1950. 35. Levin, M. L: Smoking and Lung Cancer: A review. In Proc. Third Nat. Cancer Conf. Philadelphia, J. B. Lippincott Co., 1956, pp. 473-478. 36. Wynder, E. L., and Graham, E. A.: Tobacco smoking as possible etiologic factor in bronchiogenic carcinoma; study of 684 proved cases. J.A.M.A., 143: 329-336, 1950. 37. Breslow, 1.., Iloaglin, L., Rasmussen, G., and Abrams, H. K.: Occupations and cigarette smoking as factors in lung cancer. Am. ]. Pub. Health, 44:171-181, 1954. 38. Doll, R., and Hill, A. B.: A study of the aetiology of carcinoma of the lung. Brit. M. ]., 2•1271-1286, 1952. 39. Wynder, E. L., Bross, I. J., Cornfield, J., and O'Donnell, W. E.: Lung cancer in women. New England J. Med., 255:1111-1221, 1956. 40. Haenszel, W., Shimkin, M. B., and Mantel, N.: A retrospective study of lung cancer in women. 1. Nat. Cancer Inst., 21:825-842, 1958. i I t I 1 A Quantitative Evaluation o f Etiologie Factors in Lung Cancer 289 41. Wynder, E. L, Lemon, F. R,, and Bross, I. J.: Cancer and coronary artery dis- ease among Seventh-Day Adventists. Cancer, 12:1016-1028, 1959. 42. Segi, M.: Cancer mortality for selected sites in 24 countries (1950-1957). De- partment of Public Health, Tohoktt University School of Medicine, Sendai, Japan, 1960, 277 pp. 43. Clemmesen, J., and Nielsen, A.: The incidence of malignant diseases in Den- mark 1943 to 1947. Acta Unio. Internat. contra cancrum., 8:140-159, 1952. 44. Pedersen, E., and Magnus, K.: Cancer registration in Norway. The incidence of cancer in Norway 1953-1954. The Cancer Registry of Norway Monogr. No. 1, Oslo, The Norwegian Cancer Society, 1959, 183 pp. 45. Sax6n, E., and Korpela, A.: Cancer incidence in Finland, 1954, Ann. Chir. et Gynaecol. Fenniae 47: Suppl. 79, Helsinki, 1958, 32 pp. 46. Department of Health, Wellington, New Zealand: Report of the medical statistician on cancer morbidity and mortality in New Zealand, Wellington, 1955, 123 pp. 47. Kreyberg, L: Occurrence and aetiology of lung cancer in Norway in the light of pathological anatomy. Brit. J. Prev. & Soc. M., 10:145-158, 1956. 48. Bnckwalter, J. A., Wohlwend, E. B., Colter, D. C., Tidrick, R. T., and Knowler, L. A.: ABO blood groups and diseases. J.A.M.A., 162:1210-1214, 1956. 49. Roberts, J. A. F.: Blood groups and susceptibility to deases. Brit. J. Prev. & Soc. Med., 11:107-125, 1957. 50. Heath, C. W.: Difference between smokers and nonsmokers. A.M.A. Arch. Int. Med., 101:377-388, 1958. 51. McArthur, C., Waldron, E., and Dickinson, J.: The psychology of smoking. J. Abnorrn. & Soc. Psychol., 56:267-275, 1958. 52. Lilienfeld, A. M.: Emotional and other selected characteristics of cigarette smokers and nonsmokers as related to epidemiological studies of lung cancer and other diseases. J. Nat. Cancer Inst., 22:259-282, 1959. 53. Eysenck, H. J., Tarrant, M., Woolf, M., and England, L: Smoking and per- sonality. Brit. M. J.,1:1456-1460, 1960. 54. Lorenz, E.: Radioactivity and lung cancer: a critical review of lung cancer in the miners of Schneeberg and Joachimsthal. J. Nat. Cancer Inst., 5:1-15, 1944. 55. Morris, J. N., Heady, J. A., and Raffle, P. A. B.: Physique of London busmen. The Lancet, Sept. 15, 1956, No. 6942, pp. 569-570. 56. Haenszel, W., Shimkin, M. B., and Miller, H. P.: Tobacco smoking patterns in the United States. Pub. Health Monogr. No. 45. Washington, D. C., U. S. Gov't. Print. Office, 1956, 111 p. 57. Haenszel, W., and Shimkin, M. B.: Smoking patterns and epidemiology of lung cancer in the United States. Are they compatible? J. Nat. Cancer Inst., 16: 1417-1440, 1956. 58. Berkson, J.: The statistical investigation of smoking and cancer of the lung. ' Proc. Staff Meet., Mayo Clinic., 34:206-224a, 1959: 59. Mills, C. A., and Porter, M. M.: Tobacco smoking, motor exhaust fumes, and general air pollution in relation to lung cancer incidence. Cancer Res., 17: 981-990, 1957. T201084

290 Tobacco and Health 60. Stocks, P.: 35th Annual Report for the year 1957 of the B.E.C.C. Cancer In- cidence in North Wales and Liverpool Region in Relation to Habits and Environment. London, 1958, 156 pp. 61. Shimkin, M. B.: On the etiology of bronchogenic carcinoma. In Tumors of the Chest. Grune & Stratton, Inc., New York, 1959, pp. 1-16. 62. Stocks, P., and Campbell, J. M.: Lung cancer death rates among non-smokers and pipe and cigarette smokers. Brit. M. 1., 2:923-929, 1955. 63. Tremer, H. M., Falk, H. L., and Kotin, P.: Effect of air pollutants on ciliated mucus-secreting epithelium. ]. Nat. Cancer Inst., 23:979-998, 1959. 64. Falk, H. L., Tremer, H. M., and Kotin, P.: Effect of cigarette smoke and its constituents on ciliated mucus-secreting epithelium. J. Nat. Cancer Inst., 23: 999-1012, 1959. 20 THE PERSONAL AND SOCIAL PSYCHOLOGY OF SMOKING Canni.Es C. Mc AnTxun, Ph.D. , THE first question any psychologist must answer before he does research on smoking behavior is: what am I doing here? Certainly he is not engaged in public controversy. Nor is he doing cancer research. If he is a good psychologist, he is simply pursuing his life-work: the attempt to unravel the tangled wherefores of the behavior of his fellowman. Smoking behavior is peculiarly multi-determined; the psycholo- gist is therefore forced to unravel more than the usual number of tangled wherefores. He benefits peculiarly much because he cannot retain his parochial loyalty to whatever theory or school of thought he has relied upon to advance his special reputation. This is always what happens to psychologists when they dare to venture outdoors. Within the confines of the laboratory, man- hours can be wasted studying "entities" that are the artificial products of one's own flair for imaginative conceptualizing. Just now, for example, learning theory is in a Ptolemaic stage of its history: full of epicycles upon epicycles upon epicycles, made necessary because the next new observation remains unex- plained by existent theory. (Of course psychologists don't draw new geometrical diagrams as the Ptolemaic astronomer did; psy- chologists invent new words. ) The same could be said of clinical psychology. Each "pretty" new case enables one to describe and name another theme or complex. Such st'udies have their uses but one somehow feels for Egon Brunswik, who used to plead in vain that psychologists should study "events with a somewhat 291 T201085

292 Tobacco and Health higher probability of occurring in human lives." Events like smoking. The indoor psychologist has great difficulty inventing clinical concepts or observing laboratory variables that have any relation to anything else that people do. One of the luckiest days he has is the day he strikes into a vein of what he calls "relational fertility." Only a handful of psychological variables can be related to any other psychological variables, much less to broad aspects of human experience. The rule may be observed that very sophisticated constructs seldom relate to anything while simple observations often do. For instance, the most relationally fertile variable in all of psychology is simply whether the person attended a public or private secondary school. But that is another story. . The second most relationally fertile datum is how many cigarettes per <lay a man smokes customarily. Many people havc begun to tap this rich lode of under- standable data. Matarazzo'' reviewed the psychological literature on smoking over a year ago, noting that long lists of variables relating to smoking and each other had turned up already in what was then still a scattered body of work. ( CofFee, alcohol, weight inerease, driving accidents, anxiety, netiroticism, hos- pitalization are some of the list.) Since that review, studies have increased in quantity but, far more importantly, in quality, as support for such work has begun to he available. Top notch work is yet to come.. One of the purposes of this talk is that II hope to show in what direction good work on the psychology of smoking can rewardingly but economically move. Our own basic findings,' now confirmed a dozen times over and in at least two countries, came from the Study of Adult Development, once known as the Grant Study. Our subjects were a panel of 250 Harvard ahimni who were selected in their sophomore years for lack of visible abnormality. They were studied hetwcen 1938 and 1942 with a wide range of medical, physiological, psychological, anthropological and sociological techniques. We have been following these men by visit and questionnaire ever sincc. They are now about 40 years old. Questions about smoking habits have been routine since the Study's inception. « 0 Personal and Social Psychology of Smoking 293 A striking number of our'subjects do not smoke. Sixty-one never have. In any year about 40 per cent of our men will be off smoking. The first question we asked was therefore, "Who smokes and who doesn't smoke?" In the absence of physiological, pharmaceutical or biochemical explanations of the rewards to the body from the smoking habit-an absence that remains striking during this conference-we felt that smoking at all seemed best explicable in terms of social learning or role behaviors, or intro- jected culture standards. We began by successfully predicting that there would be gross social class differences. Theorizing along the lines that smoking is seen in much of our society as a minor vice or self-indulgence, we were not surprised to find that nonsmoking is empirically a hallmark of upwardly mobile mem- bers of the lower-middle class. A cigarette is a faintly disrepu- table self-indulgence, proscribed on formal occasions and to the young. It is not so easily established as a habit among those people who concern themselves most to become respectable. Nor were we startled that our two most relationallv fertile variables should relate to each other: 40 per cent of our Harvard men who came from public schools are nonsmokers, 30 per cent of our graduates of less exclusive private schools but only 20 per cent of the graduates of very exclusive schools. There are two factors separable here: one is the effect of dormitory life in encouraging smoking but there is also a social status difference that remains significant when the dormitory effect is held constant. That smoking suffers from being viewed as a small vice seems to be reflected in significant relationships between non-smoking and devoutness of parents, church attendance and devoutness of the boy. In the hierarchy of venial sins, smoking seems to rate above alcohol; at Harvard and at Yale and at Amherst, drinking is regularly taken up before cigarettes! (Coffee is a correlated indulgence; whether because it is an oral vice or because it is also proscribed for the young is i.mclear. ) Nonsmokers, pious or not, are old-fashioned enough to have remained what Riesman has called "inner directed." They express a preference for nights at home rather than nights away from home, belonging to a few societies rather than belonging to many societies, dreaming of being scientists rather than of T201086

294 Tobacco and Health being sales managers. In fact,- often they become scientists. (Physical scientists, that is; psychologists are oral, uninhibited, other-directed, impious and smokers. ) Nor do nonsmokers have what Riesman likes to call "radar" with which to pick up the calls of the mass media; they respond emphatically with dislike to the prospect of being an advertiser or to the alternative of preparing the advertising for a hypothetical product. In summary, then, the nonsmoker seems to have been oriented by the mores of a particular American subculture. He is often of lower-middle class origin and upwardly mobile. He shows what Max Weber called "the worldly asceticism" that is our heritage from The Protestant Ethic. He is pious, reacts to smoking as to a small vice, is not naturally responsive to mass media but remains inner-directed, introjecting the morals of his youth, perhaps al- ways remaining something of a serious sort, frequently a physical scientist or engineer. (Pipes and cigars he will more often allow himself. They hav.e a different emotional meaning. ) Once a smoker, some men remain moderate off and on indulg- ers while others go on, often accelerating very rapidly to heavy smoking. To explain this contrast in the smoking fate of estab- lished smokers, one has to go to more strictly personality vari- ables. Within the smoking group, one can make a case for the poorer integration of the heavy smoker's personality. The heavy smokers are often given to what the psychiatrist calls "acting out." Whether in war or marriage or the bushim jitrtgle; thesw=w frequently find a place where they M livE dfMhatitalty., 81W~ seems likely that one type of pe1m *h6 beebfdK•'lr#1gidlyf addicted to smoking is the man so emotiorirtlly'°INi~~ .~,je he has great gain from taking flight into behavior ofa r~e A, active sort. Anxiety and heavy smoking are not so clearly related in our data as we ha(l expected. It seems likely that heavy smoking is best used as a tension reducer after it is already established high in one's habit hierarchy. In the course of our subjects' life- times, distress and smoking do not always correlate highly, yet there are many individual histories like that of the first subject in our file who smoked: I Personal and Social Psychology of Smoking 295 in 1947 0 to 5 cigarettes a day in 1949, on a brand new job 11 to 19 cigarettes a day in 1950 and 1951 6 to 10 cigarettes a day in 1952, unhappy and ` up to 20 cigarettes a day doctoring for his tension 6 to 10 cigarettes a day in 1953, with health fine and acute anxiety substantially di- minished The next history in the pile shows a second man unhappy with his job and smoking; then on a new, good job and quitting smok- ing; then divorced and smoking one to two packs; then happily remarried and having quit smoking all over again. Such correla- tions can occur but apparently represent a sub-type of smoking biography. Ability to stop goes still deeper into the personality. Among established heavy smokers, the less stable and those with a history of unresolved oral conflicts, in the psychoanalytic meaning of those words, were distinctly less able to quit smoking. All of these findings have cross-validation elsewhere. Matarazzo lists the older studies in his review. It is particularly gratifying that these trends appear in the data from England. We are especially grateful to Eysenckl for arranging a neat cross- validation of our original theory, using his own peculiar tech- niques, which are quite different from our own. It is always comforting in a science as inexact as ours to find a concept hold- ing up not only across samples and across an ocean, but even across a difference in the investigator's brand of psychologyl There are at least two ways these findings could be viewed. Perhaps these observations describe the kinds of persons who select themselves for smoking. On the other hand, it may be society that selects them and mediates the relation between traits and smoking by teaching each man a different smoking lesson young. Perhaps both self and social selection operate and there may, we suspect, be a large interaction effect between the two. This is the sort of problem in broad understanding to which the unwary psychologist can be brought by recording a simple datum like the number of cigarettes smoked per day! In effect, T201087

I 296 Tobacco and Health the attempt to es-plain smoking behavior forces the psychologist to evaluate the rc lative contributions of psychodynamics, of social psychology and of learning theory as well as the interactions among these usually self-sufficient concepts in the determination of one simple behavioral event. What we know about the inception of smoking is consistent with the directions already pointed by our individual difference data. Horn, ` for instance, finds that it is the students who partici- pate in neither curriculum nor extra-curriculum in the high school, just as IIollingshead'sZ lower class pupils in Elmtown were non-participants in the school culture, who more often begin early to smoke. It is also the students who are not in the college prep- aratory course. Presumably those with mobility ambitions are preparing for college. Horn finds them more often late smokers or possibly never going to smoke at all. Our own data tells the other side of this story: very high status children often smoke in secondary school. The mobile lower-middle class children start late or never. It is often so in psychology that lower and upper class personalities are alike, with the ambitious, hard-working, respectable lower-middle class people standing alone in their subcultural enclave. Is it then simply as a result of their late start that these young- sters become in adult life moderate smokers or nonsmokers? This is not always so; some late starters show extraordinary accelera- tion in their "smoking curve" once they belatedly adopt the habit. In general, we find a positive correlation but only a low one between timr, of starting to smoke and amount of smoking as an adult. When we try to predict adult smoking, interaction between time of learning to smoke and social class is a significant predic- tor. The modal upper-middle and upper class boys started to smoke before college, the modal lower-middle class boys started to smoke dm•ing college. The predictive meaning of these different times that are both "early" smoking is the same. In each social class, "early" starters share the same smoking fate. Upper class boys who smoked in prep school went on to become heavy smokers more often than not, while those who waited till their college years to start remained moderate smokers in years to Personal and Social Psychology of Smoking 297 come. Mutatis mutandis, the middle class boys who started during college more often than not became heavy smokers while those who waited to start after graduating from college are more often moderate smokers at age 40. The effects of an early start are visible in both groups but each subculture defines for itself when a start is to be deemed "early." Similar complexities arise when we ask if both time of start and smoking fate as an adult are mediated by the attitudes of the smokers' parents. Horn found early smokers more likely to have parents who smoked; so do we. Once again, however, the associations are not strong. When our subjects were boys (in the 1920's ) it was the rule that mothers did not smoke unless fathers did. Our boys therefore had either no parents who smoked or a father who smoked or two parents who smoked. The model family was one in which father smoked "often," the mother "never." Yet even families with two smoking parents produced offspring one-third of whom never smoked themselves. In this, the impact of mores seemed to outweigh the psychology of emulation: monkeys saw but did not do. It may be relevant to observe that these inner-directed young men are often more pious than their parents. It is also true that social class effects outweigh the effects of parent behavior. Peer culture may operate here; we have data on this point that we hope one day to run. It is possible that smok- ers are like what Lazarsfeld says voters get to be: surrounded'by a pool of like opinion, till they themselves dissolve into it. Or it may be that there are significant interactions between peer group behavior and larger subcultural factors. Of course, one must allow for parents who said, "Do as I say; don't do as I do!" Our parents frequently made such remarks. On the whole, they seem to have been ineffectual-if not in fact to have worked in reverse. A peculiarly effective way to ensure the son's ultimate fate as a heavy smoker was to promise him a bribe, commonly a thousand dollars or a car, if he would not smoke till he was twenty-one. Another parental attitude associated with heavy smoking in the son was "I don't want you to smoke but if you have to . . ." Moderate and nonsmoking futures awaited boys whose parents were permissive or made no T201088

298 Tobacco and Health issue, or whose sons felt their parents had made no issue. (Social class, as usual, cuts across this generalization sharply. Heavy smoking can be generated in the upper class home where "the cigarettes were out on the coffee table because they belong to the whole family." There is much more distinctiveness in the Pur- itanism of a nonsmoking upper-upper class home and such an environment is effective in influencing the son's fate, presumably for its very unttsualness. ) On the whole, then, it seems possible to conceptualize smoking behavior as the resultant of the interactions among subcultural mores, parental teachings, emulation of both the home and the adult social surround, personal value systems, psychodynamics, straight habituation, and the physiology of the smoker. (We have some mysteries about the latter in our data: smokers and non- smokers breathed, reflexed, etc., differently before they had fixed their smoking habits. ) What is more, in the kind of multivariate study we have, it is possible to set up concrete indices of each of these factors, as well as a quantitative criterion like number of cigarettes consumed in a lifetime, and actually to do an analysis of variance design to partition the relative importances of each of these very different factors as well as their various interactions. This broad-gauge psychology. The interlocking of the contributions of social, cultural, religious, dynamic, learning and individual difference variables to explaining some one human act has never been examined. Such multivariate observations are especially relevant when the study is also longitudinal. The question then becomes: which kinds of psychological theory contribute most to prophecy? The question may also be: how do we explain-and perhaps manipulate-huu,ian smoking behavior? Good studies of other sorts can be done but I am struck by the neglected fertility-and remarkable economy-of using the numerous longitudinal studies that were lacgnn in the 1930's and now contain decades of multi- variate longitndinal followup. The files of these studies are rich and untapped. What is needed is the small funds to pay clerks and the great imagination to pull meaning from many sources of less than ideal postdictive material that was not collected with this special pt:rposc in mind. This sort of program has been Personal and Social Psychology o f Smoking 299 dubbed "piggyback" research by Tagiuri. It is probably the richest source to which we can now turn for increased, deeper understanding of the complex psychology of smoking behavior. BIBLIOGRAPHY 1. Eysenck, H. H.; Tarrant, M.; Woolf, M., and England, L.: in Brit. Med. J., 5184: 1456, May 14, 19% 2. Hollingshead, A.: Elmtown's Youth. New York, Wiley, 1949. 3. Horn, Daniel: Cigarette Smoking Among High School Students. Am. J. Pub. Health, 49: 1497-1511, Nov., 1959. 4. Matarazzo, J. D., and Saslow, G.: Psychological and Related Characteristics of Smokers and Non-Smokers. Psychological Bulletin; in press. 5. McArthur, C.; Waldon, Ellen; and Dickinson, J.: Psychology of Smoking, J. of Abnorm. Soc. Psychol., 56: 267-275, March, 1958. T201089

21 SMOKING HABITS OF HIGH SCHOOL STUDENTS DANIF.L HORN, Ph.D. JUST within the last three years there has been a sudden interest in teenage smoking. Studies in this area have been started both here and abroad. Why are we interested in teenage smoking? For a number of years most states have required that a certain amount of time in the school curri(ulum be devoted to teaching about the use of tobacco. Usually this is thrown together in a teaching unit on alcohol, narcotics, and tobacco. Most of the health text books in general use still say the same things that were said 10, 15, or 20 years ago. But in the case of tobacco, something new has taken place. It is only within the last decade that the problem of lung cancer has impressed itself on most of us and that a wide variety of studies have implicated cigarette smoking as a primary cause of lung cancer. Reasoning runs something like this: if cigarette smoking canses hmg cancer, perhaps it is too late to do anything about our own smoking-perhaps it is too hard for adults to break the habit. But perhaps we can persuade children not to take it up-this may be feasible; this may be the eventual solution to the lung cancer problem. Hence the interest in teenage smoking as an attempt at cancer control through prevention. I see no point in further belaboring the question of whether or not the smoking of cigarettes causes lung cancer. I still find it difficult to see how anyone can study the evidence and arrive at any conclusion other than that cigarette smoking is the major cause of lung cancer. This is defining "cause" in a very practical sense-1) that these smokers who now have lung cancer would 300 Smoking Habits of High School Students 301 not have had it if they had never smoked and 2) elimination of smoking would result in fewer cases among those who now smoke, at any rate there is strong support for the initiation of public health control measures at this time. Certainly, to a large degree people have accepted this relation- ship between smoking and lung cancer. Review committees of scientists and physicians have studied the evidence in both the U.S. and England and have in each case concluded unanimously that the case is proved "beyond reasonable doubt." Surgeon- General Burney's recent report represents the position of the U.S. Public Health Service in accepting this relationship. Among thoracic surgeons and cancer research scientists even as of 1955 only 2 or 3 per cent denied the relationship. About 60 per cent accepted it and the remainder were still uncertain. More recently a sample of practising physicians shows that two out of three accept the relationship, one in five tends to reject it and one in seven is still on the fence. In the general public, about three out of five accept the relationship and among teenagers, two out of three accept it, with only 1 in 40 rejecting it. The results of scientific studies, reported in the public press, have helped pro- duce these results despite the pervasive influence of advertising and the counter-claims of the tobacco industry. Furthermore, people are agreed that teenage smoking should be reduced. A study of Massachusetts physicians in 1959 showed 93 per cent agreeing that teenagers should be warned about cigarette smoking. Of those dissenting, a number did so only from the feeling that the efforts would be ineffective rather than from opposition to the aims of such a program. In the general public this proposition also meets with strong majority approval. What is especially interesting is that this approval is voiced equally by both smokers and non-smokers. Smokers, in general, show more opposition to accepting the rela- tionship between smoking and lung cancer, and tend to dis- approve of educational campaigns against smoking carried out among adults. But even smokers approve an educational effort among teenagers. Perhaps this is an admission of an underlying conviction of the hazards of smoking, or perhaps it is simply a reaction expressed by a panelist last night of "go ahead and do T201090

302 Tobacco and Health what you can among the youngsters, so long as you don't interfere with me." What, then, do we know about teenage smoking? Most of the figures I shall quote about teenage smoking come from a national survey of high school students conducted for the American Can- cer Society by a private youth survey organization, although many of the relationships first were identified in our study of smoking in the Portland, Oregon, high schools. I should like to point out, also, that, these represent current teenagers and there- fore may show some differences from MacArthurs' series of Harvard students who were teenagers over 20 years ago. At the end of the last school year one high school student in three was a regular cigarette smoker. Most of these smoked every clay, but we included the few who smoke one or two days a week if they did so with regularity. More boys than girls smoke, 38 per cent against 29 per cent. There is a fairly regular increase from 21 per cent smokers at the end of the f reshman year to 44 per cent at the end of the senior year. Since previous experience indicates that one can expect about 60 per cent of these young- sters to become smokers eventually, we estimate that 10 per cent of those who are going to become smokers develop the habit with some degree of regularity before the teens-about 65 per cent develop it during their high school years- and the remain- ing 25 per cent take up smoking after high school. In Portland, about 10 per cent of those not smoking regularly become regular smokers each school year. It is clear that the junior high school and senior high school years are crucial years in the development of the smoking habit. What sort of smoking do teenagers do? At a time when approximately 50 per cent of cigarettes sold were filter tipped, at least 80 per cent of high school girls and 56 per cent of high school boys who smoked were using only filter tips, and others used them part of the time. This may reflect a high level of concern with the health hazards of smoking and an attempt to protect themseives-or it may merely be a sign of success of the tobacco industry in keying their filter-tip advertising to the desires of the reenage market. Solitary smoking is rather uncommon (about 5% of those who Smoking Habits of High School Students 303 smoke), smoking only in groups is somewhat more common (about 16% of those who smoke ). Most high school smokers engage in smoking both when alone and when in groups. By adult standards teenage smoking is light. Only about one in six of those who smoke regularly consume a pack or more a day and these are largely concentrated in the senior class. After all, it probably takes well over 20 hours and $1.75 a week to con- sume a pack of cigarettes every day, and restrictions on smoking in the school, near school grounds, or at home limit the time available for smoking. Why do teenagers smoke? To answer this question we sought for differences between the teenagers who smoke and those who do not, rather than relying on the direct question "Why do you smoke" which unfortunately leads to stereotyped responses in most cases. In our Portland study we found that first, and most important, is whether or not the parents smoke. For example, if two parents smoke-33 per cent of the children smoke, if only one parent smokes, this drops to 25 per cent and if neither parent smokes this drops to 15 per cent. Smoking by older siblings is frequently part of this pattern. Thirty-six per cent of high school students smoke if they have an older sibling who smokes. Among those who have at least one older sibling, but none who smoke, only 17 per cent are smokers. In any event, what seems to be impor- tant is that smoking is accepted by the family as a normal and expected form of behavior. As such, smoking by the younger members of the family is part of growing up. This factor accounts for from one-third to one-half of the smoking in the Portland study. Second is a syndrome of intercorrelated measures that seem to have in common the failure to achieve peer-group status or satisfactions. Smoking is high among those who have fallen behind their age-equals in school, do not participate in extra- curricular activities, and are taking the scholastically less demand- ing course of school work. This group-a minority in the school population-has not achieved satisfaction irom its peer-group relationships, at least as defined this way. It may well be that in this group smoking is a compensatory form of behavior, a symp- T201091

304 Tobacco and Ilealth tom of other problems of emotional health. This factor accounts for about one-quarter of the smoking in the Portland series. Third is the finding that there was more smoking in the Catholic schools than in the public schools of Portland. Several hypotheses to explain this finding have been suggested ranging from a rebellious reaction to discipline to the comparison of the tolerant attitude of Catholics towards smoking as compared to rather strong condemnation of tobacco by a few Protestant sects. Unfortunately we do not have the data to answer this question. However, in collecting information on parental attitudes towards smoking it becomes clear that roughly ten per cent of all high school smokers smoke despite parental prohibitions against smok- ing-these being parental attitudes as reported by the students. More girls than boys show this kind of rebellious smoking and there is somewhat more defiance against paternal prohibitions than against maternal prohibitions. In validating these findings in the nationwide survey we found that when both parents smoke, 40 per cent of the students smoke, if one parent smokes, this drops to 33 per cent, and if neither parent smokes, it drops to 23 per cent. What is most revealing is that if one or hoth parents have given up smoking, with neither parent smoking currently, the rate of student smoking drops down to about the same level as among children whose parents have never smoked. Even when one parent gives up smoking and the other continncs, the rate among the children is substantially be- low that in families where both parents continue to smoke. Smoking by an older brother or sister is equally striking. Among those who have at least one older sibling and have at least one oldcr sibling who smokes, 44 per cent smoke in high school. If none of their older siblings smokes, the rate is cut in half-22 per cent smoke. As part of our studies we developed a brief five item test of attihidcs towards smoking. This was highly correlated with smoking bchavior. Among the 13 per cent holding the most favorable attitndes towards smoking, three out of five smoked; among the 4 per cent holding the most unfavorable attitudes, only one out of nine smoked. I have alwady mentioned parental attitudes towards their Smoking Habits of High School Students 305 children's smoking. Strong disapproval, or even outright for- bidding of smoking results in much less smoking than does approval, indifference, or even mild disapproval. In Portland, only 8 per cent of the high school children smoked if they reported that both parents had forbidden them to smoke. The whole constellation of family smoking practices, family attitudes towards smoking, and student attitudes towards smok- ing contributes to the development of smoking by a high school student. In addition, satisfactions obtained in accepted peer- group relationships militate against the taking up of smoking. In fact, we can identify, two small groups of children who show all these factors in combinations in extreme form. Of 115 children who show all the factors militating against smoking-none smoke. In a group of 57 who show all the factors favoring smoking-90 per cent smoke regularly. A major study was carried out among teenagers of Portland, Oregon, and environs to study various approaches to reducing teenage smoking. A full description of how this study was conducted has already been reported.' Briefly, the basic design of the study was as follows: 1. Following an initial period of exploratory interviewing, testing, and pretesting, at the beginning of the school year an unsigned questionnaire on smoking habits and back- ground data was administered to all high school students in the 11 public high schools of Portland, and five Catholic high schools of the city, and five high schools in two counties outside the city limits but part of the urban area. 2. On the basis of this first questionnaire, 19 of the 21 schools were divided into six experimental groups ranging from two to four schools in number and matched as closely as possible for level of smoking. 3. One of the six experimental groups served as a control and no special educational activities were undertaken in those schools that would not have taken place anyhow. In the other five groups, three times during the! course of the year, 1. Horn, Daniel: Courts, F. A.: Taylor, R. M.: Solomon, E. S.: Cigarette Smoking Among High School Students, American Journ.al of Public Health, November 1959, Vol. 49, 1497-1511. T201092

T 306 Tobacco and Ilealth mass educational material (pamphlets, flyers, and posters) was distributed to the students. Each of these pieces of printed matter represented one of the five following ap- proaches: a. Contemporary. Emphasis on those aspects of smoking which have immediate or current meaning to the high school student, such as cost, relation to athletic per- formance, attractiveness to opposite sex, etc. b. Remote. Emphasis on the relationship of lung cancer to cigarette smoking in terms of physical effects in later life. Approaches (a) and (b) were included to test which was most effective in modifying smoking habits. Interviews with students have shown that con- temporary effects of smoking are frequently reported as reasons for not smoking, but more deeply probing psychological tests tend to show more underlying concern among students with the remote effects. c. Both-Sided. Since some communication studies indi- cate that both-sided messages may be more effective thai3 one-sided messages for those who are not already committed to one opinion, this approach was designed to be somewhat permissive with respect to smoking, i.e., that it is socially acceptable. d. Authoritative. Since smoking is sometimes an act of rebellion against authority either in the home or the school, this type of material was included to test the appeals to authority that are common in most health education messages. e. Adult role-taking. Since smoking may serve as a symbol of being "grown-up" to many young people, this approach was designed to let the high school student reverse the usual parent-child role. The stu- den! thus assumes the role of the person who provides information on health, smoking and lung cancer to parents or other adults in the family circle. 4. Near the end of the school year a questionnaire was again administered to all students of the same high schools. The first pori ion of the questionnaire on smoking habits was Smoking Habits o f High School Students 307 identical with that used eight months earlier. Some changes were made in the second portion as a result of the findings of the earlier questionnaire. In our experiments in the Portland schools we found that it was in fact possible to reduce the rate of taking up smoking during the school year. Essentially, the theme of the most successful approach was the one emphasizing the eventual effects and may be paraphrased as follows: "You've heard a lot of arguments about smoking ciga- rettes, but there is something new to be said on the subject. Scientists have recently found out that the smoking of cigarettes can cause lung cancer. This is something that was not formerly known, now there is not much doubt. Here is some of the evi- dence . . . think about it before you decide whether or not to smoke." As such the appeal is a logical one to the intelligence of our youth, and it met with an excellent response. True, there were many who did not respond. But so far the evidence suggests that this approach was most effective among those who smoked as part of a family pattern and not for the more individual patterns of compensatory smoking or rebellious smoking. The extent of the response, namely the reduction over a period of months in the rate of recruitment of new smokers from 10 per cent to 5 per cent may seem small. Yet, carried on cumula- tively for a period of four years it would mean that about 20 per cent of our high school students that would otherwise be- come regular smokers by graduation time would not do so; in other words, the percentage of smokers could be cut in half. Health education can hardly depend on a one-shot program. The American Cancer Society has prepared materials suitable for science classes, others suitable for health education classes- filmstrips, posters, leaflets, questionnaires on smoking and atti- tudes towards smoking as teaching aids. Many health depart- ments are producing materials on this subject. These are not only useful in the schools, but for youth groups, in general, and to physicians, public health of6cials, health educators outside the schools. At the moment our only long-term solution to the lung cancer problem is to reduce cigarette smoking. T201093

308 Tobacco and Health I dislike talking about teenage smoking without adding a comment on adult smoking. Even if every teenager in the coun- try stopped smoking tomorrow-never to take it up again, it would be at least 30 years before it would begin to have any appreciable effect on the lung cancer death rate. How about the adult smokers who are dying of lung cancer this year-nearly 100 a day in the U.S.-perhaps double that figure 10 years from now? Many are pessimistic about the prospects of affecting adult smoking-I am not. If 30,000,000 people can switch to filters in 5 years, largely because of concern about health; if one fourth of physicians who were smoking cigarettes five years ago have quit; why cannot large numbers of adults be motivated to modify their smoking in ways that make it less dangerous? Especially now that there is another motive to add to self-preservation, that is, that childrens' smoking is largely dependent on parental example and parental attitudes. The point is that cigarette smoking is a health hazard-the greater the consumption, the greater the hazard. Any action that reduces smoking or exposure to the harmful ingredients in the smoke has a good chance of reducing the risk, even in an adult. SesaioM V OTHER DISEASES RELATING TO SMOKING , T201094

z ~ ~ h 22 THE EFFECT OF SMOKING ON THE PERIPHERAL CIRCULATION OF MAN: STUDIES ON THE SITE OF ACTION OF NICOTINE GxncE M. RoTx, Ph.D., Ricfrnxn M. Srucx, M.D. and RonEnT R. SECSEST, M.D. a h T'Txovcx it is generally agreed that smoking causes vasoconstriction of the peripheral blood vessels, both in normal subjects and more conspicuously in those with peripheral vascular disease, there is still controversy, particularly about the site of action of these effects. We wish, therefore, to review some of our previous work and present some new preliminary observa- tions. Because the man in good health may feel that smoking is bad only for one who is ill, it was important to determine the effect of smoking on normal individuals. One would expect to have little difficulty in determining the effect of smoking on normal individuals. However, all the methods of measuring blood flow in man are indirect, and each has its own inherent error. Furthermore, irrespective of the method used, certain factors concerned with the status of the individual bear directly on the measurement of the blood flow in the extremities. These are the environmental temperature, the position of the subject, particularly of his extremities, the taking of food, and the basal metabolic rate. The ~tandard smoking test that was devised considered these factors. Standard smoking tests were carried out on normal and other 311 T201095

I 312 Tobacco and Health subjects who were habitual smokers at the Mayo Clinic in a constant-temperature room kept at 25.5° C. ( 78° F. ) and at a relative humidity of 40 per cent. The subjects fasted for 15 hours before each test. During the test they wore light-weight short pajamas and rested in a supine position on comfortable beds. All unnecessary noise and other stimuli likely to cause vasoconstric- tion were excluded during the tests. The temperatures of the plantar surfaces of both the first and third toes and the volar sides of the distal phalanges of the first and third fingers were measured at intervals of 10 minutes with the copper-constantin thermocouples designed by Sheard. The blood pressure, pulse rates, and bas;il metabolic rates were determined during the control period. When the skin temperature, blood pressure, and pulse rate were fairly well stabilized, smoking was begun. Ciga- rettes of different brands bought on the open market were used. The subjects smoked two cigarettes in succession until two thirds of each had been smoked. They inhaled the tobacco smoke with the depth and frequency to which they were accustomed. The blood pressure, pulse rate, and skin temperature were determined simultaneously at intervals of one minute during the smoking, which generally lasted 12 to 16 minutes, and during the 30 to 60 minutes immediately after the smoking ceased.3 Four hundred twenty-five smoking tests were carried out on 100 normal subjects. From these studies the following conclu- sions could be made: (1) No tolerance develops to tobacco with respect to the vascular effects for the skin temperatures of the extremities of the habitual smokers decreased and the blood pressures and pulse rates increased. (2) The decrease of the skin temperatures of the extremities varies with the basal metabo- lic rate from clay to day but does not vary with the length of time that the sirbject had been a smoker or the number of ciga- rettes smoked a day. (3) Nicotine appears to be the most impor- tant factor in producing the vascular effects. Smoking tests after ingestion of alcohol permitted a fourth conclusion. (4) Although the oral administration of alcohol produces vasodilatation of the peripheral blood vessels, the vasconstriction from smoking can- not be prevented at any time during alcoholic vasodilatation. Effect on Peripheral Circulation o f Man: Action o f Nicotine 313 EFFECT OF SMOKING ON PATIENTS WITH PERIPHERAL VASCULAR DISEASE The effect of smoking on patients with peripheral vascular disease as determined by the smoking test using nonfiltered standard-size cigarettes was a significant fall in the skin tem- peratures of the fingers and toes and a rise in blood pressure and pulse rate. Again the variations were similar to those of normal subjects and frequently were related to the basal metabolic rate. Also the pathologic condition of the blood vessels, particularly of the lower extremities, determined whether the fall in skin temperatures occurred in the fingers or toes. The evidence that smoking plays a role in the progression of peripheral vascular disease is incontrovertible.2 Although tobacco smoke may not be the etiologic factor, it is certainly the most prominent contributing factor. Juergens and associates' found in- a study of 401 nondiabetic men with arteriosclerosis that 97.5 per cent were smokers at the onset of symptoms. The incidence is similar in patients with thromboangiitis obliterans. Various reports have been made of the relief of symptoms after cessation of smoking. In view of these reports, consideration of filters and filtered cigarettes, produced to remove nicotine and tars from cigarette smoke as a protection, is highly important. We found that the vascular effects were the same for filtered cigarettes as for non- filtered cigarettes. The Chemical Laboratory' of the American Medical Association has reported that the fraction of nicotine removed from the main stream of the smoke by the filter is small. To avoid the vascular effects, such efficient filters could be made that the culminating achievement would be a filter that would pass no smoke at all. It is exceedingly doubtful that the average smoker would take kindly to this. According to Wynder and Hoffmans the majority of present- day filtered cigarettes have a significantly lower content of smoke condensate than nonfiltered cigarettes, and the king-size ciga- rettes yield more tar and nicotine than the regular-size cigarettes. They said, too, that the smokers of king-size cigarettes smoke to the same butt length as smokers of regular-size cigarettes. Claims T201096

314 Tobacco and Health that tobacco is its own best filter, therefore, are misleading, since the latter part of the cigarette contains more condensate and nicotine than the first part of the cigarette. The smoker of king- size cigarettes then actually is exposed to more of these sub- stances in each cigarette than the smoker of regular-size ciga- rettes. Wynder and Hoffman suggested that the smoker should not smoke the cigarette to the butt but rather cease smoking at the halfway mark. Wright° did not find any filtered cigarettes that, when smoked, did not produce a response in the vascular system. In addition, he repeatedly observed relapse in thromboangiitis obliterans when patients smoked filtered cigarettes. He described a patient who first had active thromboangiitis obliterans in 1940. When he stopped smoking, this patient's symptoms remained quiescent until 1949 whon he resumed smoking. In six months gangrene of three toes developed. Once more, on abstinence from smoking the disease became quiescent. He later began to smoke again; this time he nsed filtered cigarettes. In 4 months his disease had been reactivated, and he had early pregangrenous involvement of two toes. THE SITE OF ACTION OF NICOTINE Various inve stigators have attributed the effect of smoking or the intravenous injection of nicotine to stimulation of the sympa- thetic nervous system, particularly the autonomic ganglia. Recent work by Pelikan' suggested that nicotine may prevent ganglionic transmission hy impairing the release of acetylcholine. STUDIES ON ANIMALS Isolated reports appearing from 1913" to 1948° showed that nicotine had a direct action on the blood vessels. In 1948, IIaiinovici" reported that the vasoconstrictor action of nicotine occurred even after removal of the sympathetic chains and the spinal nerves in animals. He concluded that the action was peripheral on the postganglionic fibers and perhaps directly on the l>lood vessels.. This was further confirmed by Ginzel and Kottegoda" in 1953. On the other hand, as early as 1938 Shaw" and, later, Raab'E (1943) demonstrated the presence in extracts of various organs i I Effect on Peripheral Circulation of Man: Action of Nicotine 315 170 100 Minutes Fig. 1. Effect of smoking on the fingers and toes after extensive lumbar svmpathectomy. Reversal of temperatures of the fingers and toes, with no effect of smoking in the temperatures of the toes but a mild decline in temperatures of the fingers, should be noted. [Reproduced by permission from Roth, Grace M. and Shick, R. M.: The Effect of Smoking on the Peripheral Circulation. Dis, Chest., 37:203-210, Feb., 1960.] T201097

T 316 Tobacco and Health of substances having the chemical properties of catecholamine. By 1946, von F:uler" demonstrated norepinephrine in the heart. In 1948, Schmiterlow" showed that norepinephrine was present in the wall of the peripheral arteries of animals. In 1957, Nor- denstam and Adams-Ray'' described a system of chromafi'in cells in human skin around nerves and blood vessels; these cells con- tained chromaffin granules of the same type as those in the adrenal medulla. In 1956, Holzbauer and Vogt'° observed that the alkaloid reserpine diminished the amount of epinephrine and norepi- nephrine in the adrenall medulla. Burn and Rand" extended similar studies and developed a new concept. They studied the effects of nicotine on the perfused vessels of rabbits' ears treated with reserpine and found that nicotine no longer produced vasoconstriction because of the depletion of norepinephrine and epinephrine by the reserpine. They suggested, therefore, that peripheral vascular disease could be explained by a hypersensitivity to the action of nicotine in releasing catecholamines from chromaffin granules located near the blood vessels of the skin. STUDIES IN MAN Although Haimovici° demonstrated that the vasoconstrictor action of nicotine occurred even after the removal of the sympa- thetic chains and the spinal nerves in animals, Rapaport, Frank and Massell'R reported that lumbar sympathectomy abolished the peripheral vasoconstriction, produced by smoking, in the lower extremities of 19 patients. They concluded that the vasoconstric- tion, therefore, was mediated by sympathetic vasomotor fibers. We'° also have found (Fig. 1) no vasoconstriction indicated by a decrease in the skin temperatures of the toes during smoking if lumbar sympathectomy was complete, but smoking decreased the skin temperatures of the fingers. Thus after lumbar sympathec- tomy the intact sympathetic nervous system seems to function in a more than adequate manner. As a result of the work of Burn and Rand" on animals, Strom- blad° injected nicotine into the brachial artery of healthy sub- jects. Nicotine caused vasoconstriction in the hand. Sympathi- I I Effect on Peripheral Circulation of Man: Action of Nicotine 317 25 E4 100 4 80 ° ~3° 60 2 ~'40 ~1 20 0 0 -Smokinsj- Room temperature L Finger Toe Minutes ® Blood suga.r-millisirams % / Epinephrine-like substance-miero9lQms % Fig. 2. The effect of smoking two thirds of two standard cigarettes on the cutaneous temperatures of the extremities and on the blood pressure of the same subject. Note the decline in skin temperatures and the rise in blood pressure during smoking and after cessation. Also note the insignificant change in blood sugar and plasma pressor amines before and during the same period of smoking. [Reproduced by permission from Rehder, Kai and Roth, Grace M.: Effect of Smoking on the Fasting Blood Sugar and Pressor " Amines. Circulation, 20: 224-228, Aug., 1959.] T201098

318 Tobacco and Health colytics and ganglion-blocking agents abolished the effect of nicotine. This observation appeared consistent with the assump- tion that nicotine caused release of sympathicomimetics from the h c romaffin system in the numan skin, out it can be explamed also by the existence of a peripheral nerve plexus containing ganglion ~ cells. Because of these various controversies we decided to carry out two type studies on man. Rehder and one of us (G.M.R.)21 attempted to determine whether our standard smoking test would increase the amount of epinephrine and norepinephrine in the circulating blood and in turn produce an elevation of blood sugar. Smoking tests were carried out on 24 normal subjects under basal conditions. We found that the levels of the fasting blood sugar and the epinephrinelike substances in the systemic blood did not rise appreciably during and immediately after smoking (Fig. 2). Apparently the amount of epinephrinelike substances secreted into the systemic circulation as a result of smoking is not sufficient to produce a detectable increase in the total epinephrine sub- stances in the blood as measured by the method of Weil-Malherbe and Bone,22 nor was it sufficient to increase the fasting blood sugar. E°°f I E f f ect on Peripheral Circulation o f Man: Action o f Nicotine 319 Fig. 3. Effect of smoking on blood pressure and skin temperatures of the fingers and toes of a normal reactor to the cold pressor test. There was no significant change in the blood pressure after the intravenous injection of 5 mg. of regitine. A significant rise of the pulse rate and blood pressure and a t, decrease of the skin temperature of the fingers and toes occurred, however, which seems to indicate a stimulation of the sympa- thetic nervous system. A second preliminary study was begun to determine whether, during a standard smoking test, epinephrine and norepinephrine were released and' whether phentolamine ( regitine ) hydro- chloride, a blocking agent for these substances, could abolish the effect of smoking. Standard smoking tests were carried out on habitual smokers, 10 normal persons from 28 to 42 years of age whose response to the cold pressor test was normal. In addition the tests were given to two hyperreactors to the cold pressor test and two patients with hypertension who had been treated with reserpine. The average rise of the blood pressure of the 10 normal sub- jects during the cold pressor test was 17/14 mm. of mercury, 16/16 during smoking and 18/16 during smoking after the intra- venous injection of 5 mg. of regitine. There was no significant fall in the blood pressure immediately after the injection of regi- tine (Fig. 3). In the first of the two individuals who were hyperreactors, the blood pressure rose from 130/90 to 158/110 during the cold pressor test. During smoking the blood pressure rose from 140/90 to 170/110. After the intravenous injection of 5 mg. of regitine the blood pressure returned to the basal level of 130/90 and rose again during smoking to 170/110. The blood pressure of the second hyperreactor rose from 130/90 to about 170/120 during the cold pressor test and during smoking (Fig. 4). After the intravenous injection of 5 mg. of regitine the blood pressure fell to the basal level and rose again on smoking (Fig. 5). Four days later 40 mg, of guanethidine was administered orally, and subsequently a standard smoking test was carried out. The rise in blood pressure on smoking was not significantly changed (Fig. 5). Guanethidine is a drug which acts on the terminals of the 4 T201099 I ~

320 Tobacco and Health 1 2 3 4 5 6 7 8 9 10 11 12 13 Time in Mlinutes Fig. 4. The effect of smoking on blood pressure and skin temperatures of the fingers and toes of a hyperreactor. The increase in blood pressure and the decrease in the skin temperatures of the fingers should be noted. sympathetic nerves. It presumably interferes with metabolism of norepinephrina. The question which might be raised is whether the dosage was sufficient to block the norepinephrine. Further studies are planned. CONCLUSION The question still remains as to whether smoking stimulates the sympathetic ganglia or the chromaffin tissue. The decision may rest on whether the smoking of two thirds of two cigarettes is adequate to stimulate the chromaffin tissue or whether the f l I Effect on Peripheral Circulation o f Man: Action o f Nicotine 321 p4gIfiRL tv 40 RIQf ('iWmNmAdW Smoking- -smow ~' )~ ` ' P Cod d r i Fxper Fnq~r " ' TO h'ig. 5. The effect of smoking on the blood pressure and skin temperatures of the toes in the same hyperreactor after the intravenous injection of regitine and after the oral administration of 40 mg. of guanethidine. No significant changes occurred in the blood pressure. The skin temperatures of the fingers was slightly lower after the administration of guanethidine. methods are sufficiently sensitive to measure the released pressor amines. REFERENCES 1. Roth, Grace, M.: Tobacco and the Cardiovascular System: The Effects of Smok- ing and of Nicotine on Normal Persons. Thomas, Publisher, Springfield, 1951, 66 pp. 2. Gifford, R. W., Jr., and Hines, E. A., Jr.: Complete Clinical Remission in Throm- boangiitis Obliterans During Abstinence From Tobacco: Report of Case. Proc. Staff Meet., Mayo Clin., 26:241-245, Jan. 20, 1951 3. Juergens, J. L., Barker, N. W., and Hines, E. A., Jr.: Arteriosclerosis Obliterans: Review of 520 Cases With Special Reference to Pathogenic and Prognostic Factors. Circulation, 21:188-195, Feb., 1960. 4. Chemical Laboratory,: A Study of Cigarette, Cigarette Smoke and Filters: Regu- lar Cigarettes, King-Size Cigarette, and Additional Filter-Tip Cigarettes. JA.M.A., 157:1309-1311, Apr. 9, 1955. 5. Wynder, E. L, and Hoffmann, Dietrich: Some Practical Aspects of the Smoking- Cancer Problem. New England J. Med., 262:540-545, Mar. 17, 1960. 6. Wright, L S.: Cigarettes. J.A.M.A., 155:666, June 12, 1954. T201100 Smoking ~ u I ~ I ~ Cold Pressor D t t Finger 65 d ^ 4

322 Tobacco ancl Health 7. Pelikan, E. W.: The Mechanism of Ganglionic Blockade Produced by Nicotine. Ann. New York Aca.d. Sc., 90:52-69, Sept. 27, 1960. 8. Handovsky, Hans, and Pick, E. P.: Untersuchungen nber die pharmakologische Beeinfluszbarkeit des peripheren GeFasztonus des Frosches. Arch. exper. Path. u Pharmacol., 71:89-101, Jan. 24, 1913. 9. Haimovici, H.: Postganglionic Site of Action of Nicotine With Special Refer- ence to Its Direct Action on Blood Vessels. Proc. Soo. Exper. Biol. & Med., 68:516-520, Juily-Aug., 1948. 10. Ginzel, K. H., and Kottegoda, S. R.: Nicotine-like Actions in Auricles and Blood Vessels After Denervation. Brit. 1. Pharmacol., 8:348-351, Sept., 1953. 11. Shaw, F. H.: The Estimation of Adrenaline. Biochem. 1., 32:19-25, 1938. 12. Raab, W.: The Pathogenic Significance of Adrenalin and Related Substances in the Heart Muscles. Surgery, 1:188-225, 1943. 13. von Euler, U. S.: The Presence of a Substance With Sympathin E Properties in Spleen Extracts. Acta physiol. scandinav., 11:168-186, 1946. 14. Schmiterliiw, C. G.: The Nature and Occurrence of Pressor and Depressor Sub- stances in Extracts From Blood Vessels. Acta physiol. scandinav., 16 (suppl. 56) :3-113, 1948. 15. Nordenstam, Hans, and Adams-Ray, Jack: Chromafl'in Granules and Their Cellular Location in Human Skin. Ztschr. Zellforsch., 45:435-443, Jan. 24, 1957. 16. Holzbauer, Margarethe, and Vogt, Marthe: Depression by Reserpine of the Noradrenaline Concentration in the Hypothalamus of the Cat. 1. Neurochem., 1:8-11, 195fi-1957. 17. Burn, J. H., and Rand, M. J.: Noradrenaline in Artery Walls and Its Dispersal by Reserpine. Brit. M. J. 1:903-907, Apr. 19, 1958. 18. Rapaport, S. I., Frank, H. A., and Massell, T. B.: The Effect of Smoking Upon Blood Flow in the Sympathectomized Limb. Circulation, 2:850-858, Dec., 1950. 19. Roth, Grace M., and Shick, R, M.: The Effects of Smoking on the Peripheral Circulation. Dis. Chest., 37:203-210, Feb., 1960. 20. Stromblad, C. R.: Effect of Intra-arterially Administered Nicotine on Blood Flow in the Hand. Brit. M. 1., 1:484-485, Feb. 21, 1959. 21. Rehder, Kai, and Roth, Grace M.: Effect of Smoking on the Fasting Blood Sugar and Pressor Amines. Circulation, 20:224-228, Aug„ 1959. 22. Weil-Malherbc, H., and Bone, A. D.: The Adrenergic Amines of Human Blood. Lancet, 1:974-977, May 16, 1953. I 23 CARDIOVASCULAR AND RELATED CHARACTERISTICS IN HABITUAL SMOKERS AND NON-SMOKERS HmRx W. BLACKBURN, M.D., JosEF BsozFx, HsrraY L. TAYLOR and ANCEL K~ts Tr>ERE is a higher cardiovascular mortality rate in cigarette smokers than in persons who have never smoked or in those who have stopped smoking longer than ten years, among large volunteer groups of men followed in this country and abroad.', 2 Most of the excess cardiovascular mortality of smokers is accounted for under the diagnostic label of arterio- sclerotic or coronary heart disease upon death certificates. Few have implied that smoking itself is a primary cause of the different rate of dying from heart disease, but an official body has suggested that "heavy smoking may contribute to or acceler- ate the development of coronary heart disease or its complica- tions:"' The causal relationship of the smoking-heart disease association has been propounded with less vigor than that of the lung cancer thesis because a) the statistical association is not nearly so great, b) the etiology of coronary and other heart disease appears certainly to be more of multi-factor origin, and c) no convincing theoretical explanation or experimental basis exists for a pathogenetic role of smoking in coronary or other heart disease. If, however, smoking were a cause of excess car- diovascular mortality, then the magnitude of the public health problem would be far greater than that in many rarer causes of The work described in this article was supported in part by Research Grant H 10 (C14) from the National Heart Institute, Public Health Service, Bethesda, Maryland, and by a grant from the Tobacco Industry Research Committee, New York, N. Y. 323 T201101

324 Tobacco and Health death which individually exhibit a much higher statistical asso- ciation with smoking habit. Though acute circulatory effects of cigarette smoking are well documented,' there is little available evidence concerning chronic effects. This is principally due to the difficulty of design- ing appropriate animal or human experiments. This study shares the defect common to group studies undertaken among adult smokers and non-smokers in that random allocation of smoking habit is impossible, so that the groups cannot be considered similar in all respects except that of smoking, and direct causal implications cannot be made with confidence from differences found in the group comparisons. . With awareness of these pitfalls and circumspection in drawing conclusions, it is nevertheless of interest to seek clues to the possible role of smoking in pathogenesis of cardiovascular disease, and tests of hypotheses in this regard, by comparison of circula- tory measurements and related characteristics in groups among which smokers and non-smokers are identified. Subjects and Methods A number of men are under observation by this laboratory in investigations concerning the relationship of physical character- istics to development of heart disease. The individuals from which the data to be reported herein derive are volunteers. There is no random allocation of smoking habit, and smokers and non-smokers wore not randomly assigned, after volunteering, into subgroups for investigation of smoking "effects." Actually all smokers and non-smokers available from the various study groups were employed for analysis. The groups were not originally selected with any regard to smoking habit. (The inter-relations of smoking habit and physical characteristics are now being investigated in total population samples, in which some of the sampling problems of this study are obviated. ) Group A consists of 286 actively employed business and pro- fessional men of the Twin Cities who have participated in a longitudinal study of cardiovascular aging, now in its 13th year. They were ages 45 to 54 inclusive at the outset of the study and Cardiovascular Characteristics in Smokers and Non-smokers 325 were selected from 916 acceptances to 1,000 invitations issued in cooperation with several local industrial and professional insti- tutions, then randomly picked to fill classes of different relative body weight. Variously, comparisons by smoking habit will be made in the total group and in a group rigidly screened for clini- cal cardiovascular disease, for blood pressure 160 mm. Hg systolic or 95 diastolic or more, and electrocardiographic abnor- malities. The latter group will be recognized by a total n = 233. Group B consists of 159 university students, ages 17 to 25 inclusive, chosen to fill categories of relative body weight, who accepted invitations made during the week of student health service pre-school examination. Subjects with known cardiovascular disease and hypertension were asked so to state on their acceptance reply and were elimi- nated at the outset in groups A and B. The comparisons with regard to smoking habit were made some years later, and there was no subsequent exclusion process in the study. Group C is made up of 414 railway employees, ages 25 to 67 inclusive, of lines operating in this area. They represent a majority of the local yards switchmen, and of the clerks from the local general offices of three railroads, and 85% of all train dis- patchers in a six state area. All subjects are volunteers to a general call to the area railroad industry and brotherhoods for participants in a pilot study concerning habitual physical activity of work and development of coronary heart disease. There was no other known selection with regard to health except that all subjects were currently employed. Group D consists of 234 City of Minneapolis firemen, ages 25 to 63 inclusive, and represents all of the firemen on duty at the time of our visit to randomly selected fire stations. In this pro- cess approximately 60 per cent of the active force was actually examined. Group E is a Finnish sample of men ages 20 to 59 inclusive and consists of men in "representative" rural areas in Finland.' Virtually all those invited responded, and subsequent selection eliminated subjects with history or physical findings of heart or thyroid disease or prominent physical disabilities. T201102

326 Tobacco and Health Smoking Categories Standardized smoking questionnaires were administered by the Laboratory staff. The proportion of individuals who smoke cigar or pipe alone is very small in each of these groups and in- tensity categories herein are based entirely on cigarette smoking habit. "Never" smokers are those who either have never smoked cigarettes at all, have had only remote and occasional smoking habit, or who have solely light cigar or pipe habits. All are con- sidered smokers who have been chronic habitual cigarette smok- ers, most of whom are inhalors. Only in the student group B were there smokers of less than five years duration. Cigarette smoking intensity classification is largely based upon current habit, with consideration of the level of consumption over the prior 5 to 10 year period. Never: Never smoked or occasional and remote history. Occasional: Not daily. Light: Regular smokers, 1 to 10 cigarettes per day. Moderate: Regular smokers, 11 to 20 ciga- rettes per day. Ileavy: Regular smokers, 21 or more cigarettes per day. The overall proportion of cigarette smokers and non-smokers of the groups is as follows, with habitual smokers here including the "Stopped Smoking" category, but excluding occasional smok- ers: (7 ro?sp Nevr.r Heavy All Habdtual Smokers A $7.8% 12.8% 68.$%a li B$.2% 1.`li'o 47.8% C 17.6% 88.4% 88.4% 1) 'v$.4`1", 80.7% 77.7% ... 68. 6% (without "Stopped" ) For Group E, stopped smokers were discarded for the analyses. The heterogeneous category of stopped smokers is considered only in the lung volume measurements and is here defined as all regular smokers who are not smoking currently and have not smoked for ovor one year, based on two annual questionnaires. Data on inhalation of cigarette smoke have not been analyzed for these groups. This would not appear to be a serious omission here since comparisons are made within, not between, individual Groups A to E, and between Never and Heavy smokers. In general, inhalation is almost universal (98.9%) among heavy Cardiovascular Characteristics in Smokers and Non-smokers 327 cigarette smokers in this country.° No pooling of Groups A to E is to be made. Details of individual experimental procedures employed have been published elsewhere.b, 7, ° In al groups, however, there is considerable confidence that acute circulatory effects of smoking were eliminated by the design of the studies.° Results Most comparisons are made between extreme smoking cate- gories of "Never" and "Heavy" smokers. In Group A, when results of comparisons by all smokers combined or between the clinically "screened" and total group are different from "Never" versus "Heavy," they will be given in the text. Age: In Table 1 mean ages are given for the various samples. It is possible to demonstrate small but statistically significant mean age differences between smokers.and non-smokers in Group B where there is great homogeneity in age and in Group C. In general, however, the mean ages are similar in smokers and non- smokers of the samples. No biologically significant age differ- ences according to smoking habit appear within Groups A to E which would likely produce significant differences between the several age-related variables to be examined. However, age trend analyses were performed for the variables reported in groups with wide age spread (C, D, E) and slopes compared between smokers and non-smokers. Heart Rate: In Table 2 the mean basal pulse rate for Groups A and B is significantly higher in smokers. In the clinically screened subgroup of 233 men in Group A, the difference is in the same direction but does not reach levels of statistical significnnce. The mean resting pulse rate is higher in smokers of Groups C and D, attaining statistical significance in Group D. In age-matched sub- groups of the railway workers of Group C, significant positive age trends for resting pulse rate were found. There was a signifi- cant difference in the slopes between smoking classes with a steeper rise by age in resting pulse among the heavy smokers. Thus, in each group tested there appears'to be a faster resting heart rate among smokers. In Table 3 are given the mean pulse rates and maximum pulse increments in men of several groups during the course of a 10 T201103

328 Tobacco and Health minute treadmill walk at 3.5 miles per hour and 5 per cent grade. There is no directional tendency of response according to smoking habit among these groups. In the clinically screened "healthy" subgroup (233 men ) of businessmen there is no significant differ- ence in work pulse rates, nor are statistically significant differ- ences found when all regular smokers are compared to subjects who had never smoked. In the total sample of 286 men in Group A, however, the mean difference in work pulse rate is higher in heavy smokers, which difference reaches the 0.05 level of probability by the t test. In Group C no age trend in work pulse was demonstrated, and no differences by smoking habit existed in age-matched sub- groups of these railroad men. In Table 3 work pulse increments over the standing pulse rate are available only for Groups A and B. In neither group is there a difference with regard to smoking habit. In Table 4 the mean of the maximum pulse increment during a standard cold pressor test (right hand immersed to the wrist in ice water for one minute ) and during carbon dioxide inhalation (6% COL. in air for 5 minutes ) shows the slightly higher pulse response of the smokers of Group A businessmen and the young students of Group B to be statistically insignificant. In total Group A (n= 286) the pulse difference was smaller, and no significant differences occurred upon combining. moderate and heavy smoking classes. Blood Pressure. Mean resting asucultatory blood pressures for all groups are given in Table 5. In Groups A, B, and E there is known selection bias in regard to blood pressure, but no selection based on smoking habit. Subjects with "known" blood pressure elevation were asked in their invitation not to volunteer. No such exclusion was made in the invitation to the railroad workers (C), and the firemen (D) represent the most nearly random sample of the five groups. Save for the highly select Groups (re hypertension )A and B, there is a tendency to lower mean sys- tolic and diastolic pressure levels in smokers. This is clearly a statistically significant difference among the Finnish men of Group E. However, age trend analysis in Groups C and D revealed Cardiovascular Characteristics in Smokers and Non-smokers 329 highly significant positive trends (slopes) for both systolic and diastolic blood pressure with age. Though the diastolic blood pressure slopes for age tend to be steeper in smokers, no statisti- cally significant differences were apparent in comparing the slopes and intercepts. The mean diastolic blood pressures by age and smoking habit in Group C are as follows: Age 20-89 80-39 40-49 60-G9 60-69 "Never" 73.1 76.6 88.8 81.9 88.$ "Heavy" 86.9 74.6 80.0 80.8 90.8 In the Finnish group systolic and diastolic blood pressures were lower in smokers at all ages.6 In Table 6 the mean work blood pressure levels and pressure increments during exercise are similar between smokers and non- smokers of the only two groups examined in this regard. This was corroborated by comparisons within the total Group A, and upon combination of moderate plus heavy smokers in that group. In Table 7 there are differences in blood presstire between smoking categories of Group A, with a slightly lower diastolic increment to cold and higher systolic rise to COs in smokers. The lesser diastolic response to cold stress is more significantly differ- ent between smoking classes when comparisons are made in the total sample of 286 men which includes cardiovascular patients, and the systolic increment difference persists. The differences in Group B are insignificant. Cardiac Output. Uncorrected acetylene method values for cardiac output are available in 76 of the middle-aged men of Group A. These values are systematically lower than those with the direct Fick method. In Table 8 are the mean outputs and A-V oxygen difference for 32 "never" smokers and 34 "moderate or heavy" smokers and they are similar. There appears to be a significantly greater A-V oxygen difference in the smokers with similar cardiac output between groups. Basal Oxygen Consumption. In Table 9, the mean basal oxygen consumption in Group A is higher among heavy smokers and the mean difference is close to the 5 per cent level of probability of significance. When values for oxygen consumption in all regular T201104

330 Tobacco and Health smokers of the group are compared to those of never smokers, the difference is significant between the 1 per cent and 5 per cent levels of probability. There is no difference between the smokers and non-smokers of student Group B. Relative Body Weight. In Table 10 it is seen that heavy smok- ers are relatively less obese than non-smokers in every group. The mean differences in some of the samples approach, but never reach, the 5 per cent probability of statistical significance. The mean difference between "Never" and "All" regular smokers of Group A was identical to that for "Never" versus "Heavy" groups. In Groups C and D there were no significant age trends for rela- tive weight. Mean relative weight values by age and smoking habit are given ;or Group C as follows: Age ®0-29 80-39 40-49 50-59 60-69 "Never" 106.6 109.0 104.3 108.8 107.6 "Iteavy" 104.9 102.7 101.7 105.2 97.6 Subcutaneous Skinfolds and Body Fat Composition. The fre- quency distribution of sums of skinfold thickness is seriously skewed in all groups of men studied in this laboratory and the comparison of mean values shown in Table 11 must be regarded in the light of this fact. Mean sums of two skinfolds (triceps and subscapular) com- monly measured in these studies are given for Group A and for Group E. It would appear that smokers are relatively less fat than non-smokers in these groups. Mean values of skinfolds standardized for age show this tendency according to smoking in the Finnish Group E for ages 30 to 49 years, but this is not con- sistent in ages above 59 or below 30.1' The most precise estimate of actual body composition avail- able at the time of these studies was derived from specific gravity measurements in underwater weighing. Comparison of mean values for per cent of body fat by this method reveals no signifi- cant difference according to smoking habit. Skeletal Measurements. Skeletal measurements are given in Table 12 for Group A and reveal no build differences in regard to height or linearity-laterality of frame between smokers and non-smokers. Cardiovascular Characteristics in Smokers and Non-smokers 331 Dietary Fat Consumption. Mean daily diet fat consumption was measured in some of the middle-aged men of Group A by . collection, during two, one-week periods, of replicate servings of their meals, with extraction and analysis of the total lipid content. In Table 13 the smokers are seen to have a slightly higher total ~, and "animal" fat consumption but nowhere are the differences ~ statistically significant. The "animal" fat estimate is based on dietary history taken during the period of food collection, made [~ several years ago. Blood Lipids. In Table 14 are the mean serum cholesterol levels in smokers and non-smokers of all groups, uncorrected for age. There is a tendency to higher levels in smokers among all groups, but the mean differences are consistently present and significant only in the Finnish Group E. Age analysis in Group C revealed a significant positive age trend for cholesterol. There was no significant difference in the slopes or overall comparisons between "Never" and "Heavy" smokers though the intercept difference was very close to significance at the 5 per cent level of probability. The mean cholesterol values by age and smoking habit in Group C are given as follows: Age £0-£'9 80-59 /0-19 s0-,69 60-69 "Never" $00.9 $08.1 96¢.6 218.6 261.1 «$eavy,) 196.! 886.6 834.6 247.3 2411.8 In Table 15 the cholesterol-phospholipid ratio is significantly lower in smokers of Group A, due to a higher mean phospholipid level, and no difference in Beta lipoprotein cholesterol is found. Analysis of Beta cholesterol levels by age in Group C revealed an insignificant age trend and no significant differences by age be- tween "Never" and "Heavy" smokers. In sum, only in the Finnish sample are serum cholesterol differences statistically significant between smoking categories. In Table 16 fasting blood sugar is higher in combined smoking categories of Group A. There was a progressive tendency to higher values according to smoking intensity, class as well. No significant difference in mean values for protein bound iodine is found, although the PBI in smokers of each intensity category was lower than in non-smokers. T201105

332 Tobacco and Health Mean hemoglobin levels are compared in Table 17 among smokers and non-smokers of Group A. There was a tendency toward higher hemoglobin values in every intensity category of smoker, compared to non-smokers, but nowhere are the differ- ences statistically significant. Table 18 presents lung volumes measured in Group A, and in- cludes forced vital capacity, residual volume ( RV ), total lung capacity (TLC), and RV/TLC ratio. The forced vital capacity is lower and the residual volume, total capacity ratio higher in smokers. The age spread is one decade only in this group, mean age and height are similar between smoking classes and possible acute smoking effects are well eliminated. In Table 19 from another railroad employee sample currently under study, certain respiratory function values in the men ages 40-49 are compared by intensity of smoking. There is a consistent trend (in forced vital capacity, timed expiratory capacities, and measurements of midmaximal and maximal expiratory flow rates ) according to smoking intensity. The height-corrected values show several statistically significant differences between "Never" and "Iieavy" smokers, and between "Stopped" and "Current" smokers, in the direction of lower values in smokers. Discussion Age differences between smoking and non-smoking groups compared herein are minor, but no pooling of samples has been made since the individual groups are not considered to be from the same population. There are obvious between sample differ- ences in age and in socioeconomic status of the businessmen, students, railroad employees, firemen and Finnish workers. Any conclusions drawn will be largely based upon the overall picture, considering the samples individually to be representative only of themselves. There is no known bias from selection according to smoking habit in these groups. Presence of skewness in the dis- tributions for any of the values considered would lead to high standard deviations and reduce the value of t, so that error intro- duced is in a conservative direction, tending to minimize the significance of differences found between smokers and non- smokers. No comparisons are attempted between Groups A to E. Cardiouaseular Characteristics in Smokers and Non-smokers 333 If differences by smoking habit are in the same direction in these heterogeneous groups it would suggest that the difference is one which might be generally found between smokers and non- smokers. A faster resting heart rate is found in smokers within each group (Table 2). The acute circulatory effect of smoking is apparently due to nicotine stimulation of post-ganglionic sympa- thetic fibers and stimulation of the adrenal medulla to release epinephrine.° Watts'° has demonstrated that the urinary epinephrine (but not norepinephrine ) output rises significantly as an acute effect of cigarette smoking over a two hour experi- mental period. The actual levels of circulating amines must, however, be rather low, and the total eight hour excretion of epinephrine and norepinephrine in men who were smoking was no different from values found in non-smokers. There are ap- parently other factors of greater influence than smoking in the sum total production of pressor substances. Since acute accelera- tor effects of smoking are well eliminated in the methodology of these studies, and a higher resting pulse rate occurs in smokers of all groups studied, it would appear that this phenomenon is a general one in smokers. On the basis of crude calculations in- volving several assumptions, within Group A businessmen, about one-half the pulse difference between smokers and non-smokers could be explained by the difference in metabolic rate. Pulse and blood pressure response to physical work are com- mon measures employed to evaluate circulatory "fitness," though there are other and superior parameters in this regard. It is a fact, however, that heart rate during steady state exercise is much less variable than the resting or standing pulse and better characterizes the state of "fitness" and of "training" than these, or the basal pulse rate. Consequently, if smoking leads to a deterioration of fitness this would be demonstrated with greatest validity in the actual work pulse values. There are many com- plexities in the consideration of work pulse increments in this regard. The results herein (Tables 3 and 6) suggest that in select "healthy" young and older men there is no evidence of deteriora- tion of circulatory fitness in smokers. In regard to resting blood pressure levels (Table 5) it is seen T201106

334 Tobacco and Health that no smoking differences occur in comparison of means within the two group:; (A and B) in which age is homogeneous and in which selection at the outset eliminated some hypertensive sub- jects. Standardization for age tends to eliminate the differences in blood pressnre by smoking group in Groups C and D. Stan- dardization for relative body weight further reduces the blood pressure differences between smoking categories in these groups.a An alternate explanation for higher blood pressure in smokers is sought in analysis of the frequency of smoking habit among all hypertensives at two levels in the distribution of blood pressure values (above and below cut-off values of 140 systolic and/or 90 diastolic, and above and below 160 systolic and/or 95 diastolic). In combined Groups C and D there is a tendency toward more nwn-sinokers in the elevated pressure groups, but the differences are not statistically significant on chi-square testing. In a random sample of elderly men in England" there was a higher mean blood pressure in non-smokers, with a tendency to more hypertensives (over 200 systolic or 100 dias- tolic ) in non-smoking categories. Exclusion of hypertensive patients at the outset, as in Group A, would result, if hyperten- sives actually bad less tendency to smoke, in a biased sample of non-smokers and more similar mean blood pressures in com- parison with smokers, such as is actually found in means for Groups A and B. There is a recent report from Finland" consisting of data from a virtually total sample of all men ages 40 to 59 in several geographical areas, in which the relationship of smoking and low arterial pressure is clear. So the question is unsettled. If the association were real and smoking were a cause, the mechanism might relate to a pharmacological action of nico- tine elucidatcd by Burn." The same mechanism might operate in the lower diastolic cold pressor response of smokers (Table 7); i.e, related to the experimental demonstration that nicotine administration causes depletion of ganglionic and vascu- lar tissue norepinephrine. The smoking subgroups of Tables 8 and 9 are not identical, but the finding of increased basal oxygen consumption in the middle-aged men of Group A (Table 9) is consistent with the I Cardiooascuiar Characteristics in Smokers and Non-smokers 335 limited number of comparisons (Table 8) which reveal greater mean A-V oxygen difference in smokers and a similar mean cardiac output. Higher basal oxygen consumption in habitual smokers has been reported previously'* and a considerable elevation of metabolic rate is an acute effect of smoking.16,'° The findings of greater oxygen consumption and faster heart rate are compatible with Pearl's concept" of an accelerated biological "rate of living" which Berkson1e has disinterred as a possible ex- planation of the excess mortality rate in cigarette smokers. In consideration of relative body weight differences no signifi- cant age trends are found among Groups C and D. The mean differences in Table 10 for the various groups according to smoking category do not reach levels of statistical significance, though the tendency to lower relative body weight in smokers would appear to be a finding common to smokers. As with blood . pressure, however, the existence of different smoking habits among obese and nonobese persons is an alternative explanation for the apparent higher average weights iii non-smokers. In an analysis of the presence or absence of smoking habit in all men of combined Groups A, C, and D it is found that of the total, 70 per cent are smokers. Among those above the median value for relative body weight, 64 per cent are smokers, and below the median 75 per cent (P chi square = 0.003). In analysis employ- ing a higher level in the distribution of relative weight values, and one which is often used to define over-weight (120% or more RBW ) 58.4 per cent of the obese are smokers, and 70.1 per cent of the nonobese are smokers (P chi square = 0.0003). In order to establish that individuals who are obese or hypertensive are less inclined to smoke it would be necessary to establish that the non-smoking characteristic existed before the development of the condition. The clinical impression of weight gain after stopping smoking has been documented by Brozek and Keys in a longitudinal study over a five year period."' There was a highly significant weight gain in a two year period after stopping cigarette smoking in a controlled comparison with age- and weight-matched men who continued to smoke. The serum cholesterol and blood pres- T201107

336 Tobacco and Health sure also rose within the stopped smoking group, but the differ- ence in rise betwen experimentall and control groups was not significant.20 Blood Lipids. Higher levels of blood lipids in smokers have been reported by Gofman," Bronte-Stewart," Thomas,2s and Karvonen et al.' These groups were also highly selected, and in some of the studies there were several age group comparisons in which no lipid differences were found by smoking habit. Data from the groups reported herein reveal no significant differences in homogeneous age group comparisons, except for the Finnish men among which serum cholesterol was higher in smokers of each age group between 20 and 49 years, but not at ages 50-59. In a more detailed analysis among that population, those who smoked most tended to have the highest cholesterol values.5 It is of interest that a recent report by Orma and Karvonen'Z based upon a "total" sample of men ages 40 to 59 in rural areas of Finland discloses an overall higher mean cholesterol level in smokers but marked regional differences in smoking "effect," and no tendency to higher levels in smokers in one of three principal areas. In Thomas' data on medical students" there was a signifi- cant excess of cholesterol values of 250 mg. per cent and over in smokers and of values under 250 mg. per cent in non-smokers. From these data and those from other populations, not yet available for publication, the smoking-lipid relationship is incon- sistent, and it is clear that smoking is not an ubiquitous determi- nant of elevated blood lipids. That it is not, moreover, a primary factor is indicated by the wide geographic differences which occur in average cholesterol level among groups of men in which cigarette smoking habit is not greatly disparate. However, the clearly significant association in some areas and age groups deserves more investigation. The elevated phospholipid level found in Group A is of particular interest in this regard. And in the Finnish area, for example, Orma (personal communication) is investigating his hypothesis of a relationship between smoking and higher blood thiocyanate levels, hypotensive and hypothyroid effects therefrom, and lipid metabolism, in those groups in which there are significant blood pressure and cholesterol differences according to smoking habit. Cardiovascular Characteristics in Smokers and Non-smokers 337 Bronte-StewartZZ, z' found a consistently higher proportion of cholesterol in the beta lipoprotein fraction of Cape Town smok- ers in each age, economic, and racial class. Diet differences may have been operative as analysis of questionnaires revealed greater totall consumption of fat calories in heavy smokers. Bronte- Stewart, Krut, and Perrin24 found that there are significant taste theshold differences between smokers and non-smokers, a higher threshold to bitter taste ( quinine ) occurring in smokers. In regard to food preferences, fat consumed by the smokers de- rived more from eating meat and eggs, whereas non-smokers ate a greater proportion of fat in the form of cakes, sweets, and chocolates. Others have sought to determine whether smoking has an acute effect upon lipid metabolism. Marder and associates2s reported higher plasma turbidity levels and delayed clearance of alimentary lipemia as an effect of smoking after a standard fat meal. Blackburn, Taylor, and Leuthold,26 in a rigidly controlled experiment, failed to confirm this finding. Page and associates27 found no effect of smoking two cigarettes on blood cholesterol or lipoprotein levels in three post-smoking samples taken within 30 minutes. Blackburn et al.z8 found no immediate effect of cigarette smoking on plasma free fatty acid levels. Such lipid studies are only early steps in the search for clues concerning the relationship of smoking habit to heart disease mortality rate. There are questions here which are not answered by these negative reports about smoking effect on certain lipid levels measured at certain time intervals after smoking a few cigarettes. The finding of elevated fasting blood glucose in smokers of Group A is of note. There are no known diabetic patients in the group, but three glucose values above 87 mg. per cent (100, 107, 107) one in Never, and two in smoking categories. Conflicting reports exist on the acute smoking effect upon blood glucose levels which are probably due to the varied choice of sampling intervals during and after smoking, and to, the lack of experi- mental controls. In a switchback, double-controlled experiment no blood glucose elevation was found immediately and 15 minutes after smoking two standard cigarettes.4B There are T201108

338 Tobacco and Health no reports on the smoking habits of diabetics or the prevalence of diabetes in smokers and non-smokers. Eisen and HammondZ° have suggested that the prolonged presence of several per cent circulating carboxyhemoglobin in the blood of smokers may provide a stimulus to red cell production. They have presented evidence of immediate and sustained eleva- tion of RBC count and packed cell volume from cigarette smok- ing. The hemoglobin data in Group A showed a slight tendency to higher levels with greater smoking intensity, but no statisti- cally significant differences were found. The related subject of a possible effect of smoking on blood coagulability requires more investigation, but an initial work in this field revealed no evidence of an acute smoking effect.30 There is increasing evidence for a real association between smoking and impaired respiratory function, symptoms, and disease.", ', ", "Z• "" Nevertheless, incongruities exist in the data, such as the purported "protective effect" of inhaling," and the absence of trends in certain ventilatory function values with in- creasing intensity of smoking habit.7, "° Higgins has very intelli- gently discussed the possible bias, from selection, of results in respiratory tests among smokers." Data of Table 19 are from measures in an extensive railroad employee population study to be reported upon in detail else- where, with consideration of the interrelations of physical ac- tivity, smoking habits, and respiratory function. Follow-up study in this group should yield additional information on those sub- jects who stop smoking in the interim. The function differences by smoking habit in Table 19 are notable in that 1) subjects were not selected with regard to their smoking habit; 2) there is for the first time a demonstrated parallel between respiratory function values and smoking inten- sity; 3) In the heterogeneous category of men who have given up smoking longer than one year, the lung function values are similar to those of "Never" smokers and significantly different from "Heavy" smokers; 4) the forced expiratory capacities and flow rates measure more dynamic functional characteristics than the anatomic compartments of earlier studies and of Table 18; 5) the results resemble those found in widely separate studies of T Cardioaascular Characteristics in Smokers and Non-smokers 339 different populations;1 1, 11, 112 and 6) the findings are compatible with a hypothetical pathogenetic process in which an irritant aerosol produces bronchial obstruction and.resistance to airflow. The sum is a clamant implication of causal relationship. A common or third factor, causing tendency to smoke and pro- pensity to disease, has been sought to explain the association between smoking and disease states. It is logical to explore both "constitutional"-genetic, and personality factors in this regard. Though the impression is widely extant, that smokers and non- smokers are actually different in certain fundamental characteris- tics, there is not much objective supportive evidence. But some data appearing would support that hypothesis. Eysenck and associates" have considered scores for "extraversion" and "rigidity" as personality characteristics and have found a positive relationship between amount of smoking. and extraversion score, a negative relationship with rigidity. The regressions are quite linear, the more extraverted, the heavier smoking, and the most rigid personality types being non-smokers. They found no support for the hypothesis that smokers are more neurotic than non-smokers. Eysenck" has previously pro- vided some evidence for genetic determination of the degree of extraversion. He thus concludes that his findings support the hypothesis that genotypic personality differences influence smok- ing habit. Health°' has reported on personality assessment from twenty- six trait-criteria and other psychologic tests administered by psychiatrists to a group of college men. The reliability ( repro- ducibility ) of these subjective assessments is not discussed and it would appear that evaluations were made with the interviewers having knowledge of the subjects' smoking habit. Non-smokers are said to have an excess of the following traits: well-integrated, inarticulate, bland affect, and physical science occupation, and a deficit of these traits: verbal richness, dominance of mood, lack of purpose and values, and less well-integrated. Heavy smokers showed excesses of these traits: least sound, less well-integrated, lack of purpose and values, cultural, and practical organizing, with deficiencies in: self-conscious and introspective, well-inte- grated, inarticulate shy, and bland affect. All subjects were T201109 1

340 Tobacco and Health within a wide definition of normal personality. The conclusion drawn was that non-smokers as a group possess the more stable qualities of dependability, and good direction of aims in life, though they are on the bland, colorless side, while smokers con- tain more of the men who are energetic, still searching for aims and purpose, and though less stable, perhaps more interesting. Thomas'° has found in medical students that an index (in part based upon heredo-familial characteristics) of susceptibility to cardiovascular disease is significantly related to students' smoking or non-smoking habit. The reports on monozygotic versus dizygotic twins by Fisher9D and Raaschou-Nielsen,°° reveal significantly greater similarity in smoking habit among monozygotic twins, including those pairs brought up in separate environments. In Group A of this study no difference in body build charac- teristics of skeletal height or laterality-linearity index is apparent between smoking classes. Observations are being extended to include blood group frequencies in smokers and non-smokers among the large railroad employee population currently under study in this laboratory. In conclusion, some differences in physiologic characteristics between smokers and non-smokers are found in the comparisons herein. The consideration of smoking as a cause of most of these differences is influenced by the sampling problem in such group comparisons. In addition, there are certain inconsistencies in the data as the lack of statistical trends according to smoking inten- sity and absence of tendencies in the same direction among several populations. Moreover, many of the differences are rather small and it is difficult to attribute to them biological significance. The findings in smokers are on occasion in a direction opposite to that which might relate smoking with disease, particularly in the blood pressure and obesity data. It is clear that smoking is not a primary factor affecting lipid metabolism, though particular findings in certain group comparisons suggest avenues of research interest. In reported early returns from longitudinal studies on coronary heslrt disease,'"• " smoking habit is less highly associated with incidence or attack rate than are blood choles- terol, blood pressure, and relative body weight. There is, at this Cardiovascular Characteristics in Smokers and Non-smokers 341 time, no convincing evidence that smoking itself is involved in the original causation or basic pathogenesis of coronary and most cardiovascular diseases. _ It is, of course, quite possible for an agent to affect death rates while having little to do with the cause of the underlying disease process leading to death. An epidemic of "colds" does this. Evidence exists that, in the overall, acute effects of smoking are in no wise more severe in heart patients than in normal subjects's However, individual responses are highly variable. Con- ceivably, the acutely increased cardiac work and other metabolic, cardio-accelerator, and pressor effects of smoking may precipitate a latent disease and aggravate a manifest process, as other stresses may do on occasion. The daily, hourly repetitive smoking effect in this regard is only a subject for conjecture such as that about "stress," driving in traffic, coffee drinking, etc. Though statistically significant differences between smokers and non-smokers may be found in such comparisons, causation can neither be claimed nor proven. The approach, however, may provide both worthwhile clues for further research and tests of hypotheses concerning the relation of smoking with pathological processes. TABLE 1 Atlffi kN SMOKERS AND NON-BMo1CEnl3 Group Tolal No. Men "Never„ "Eleauy„ 3 A Businessmen 233 M .¢9.8 (67) 48.4(36) -0.8 SD 3.1 2.9 B Students 159 M 20.i (83) 20.8'(53)* +0.7 SD 1.9 1.8 C Railroad Workers 414 M IB.8 (73) 43.4(138) -3.4 SD 12.1 10.3 D Firemen 234 M 39.806) 39.9 (60) +0.1 SD 9.5 6.4 E Finnish Men 525 M All Non- Smokers 38.6 (165) All Smokers 88.7 (380) +1.2 SD 1.4 0.8 * "Moderate plus Heavy" smokers. T201110

I 342 Tobacco and Health N Cardiovascular Characteristics in Smokers and Non-smokers 343 ~ 00 wn TABLB 2 ~ ~ BASAL OR RESTING PULSL RATE IN SMOKERS AND NON-SMOKERS C 0 (in Beats Per Minute) ~i m + I~ z Total Probability ~.°, Group No. Men "Neverit I 'Heavy" d of t p,y 00 A Businessmen 286 M 64.6 (82) 69.0 (48) +4.5 <0.01 SD 8.4 9.9 B Studente 159 M 6$7 (83) 667 (53)* .. . +30 <005 . SD 1 8.8 7 ~" m . M 0 C Railroad 4 Workers 414 M 68.1 (73) 69.E (138) +1.2 >0.05 a n' SD 11.9 12.3 9 D Firemen 234 M 7£.3 (56) 79.8 (60) +7.5 <0.001 ...,n SD 8.3 12.0 ~ o 0 an * "Moderate plus Heavy" smokers. TABLE 3 MEAN WORK PULSE RATE IN SMOKERS AND NON-SMOKERS E-4 ~03 ~ - M . (in Beats Per Minute) 9 a~ Total Probability 3 ~" Group No. l4fen "Never" "Heavy" a of t z A Businessmen 233 114.8 (52) 120.2 (28) +5.4 >0.05 q .. A Businessmen 286 114.8 (61) 121,7 (35) 9 +6 <0 05 H B Students 159 118.6 (83) 115.7 (53)* . -2.9 . >0.05 z C Railroad Workers 414 115.7 (73) 116.6 (138) +0.9 >0.05 U z H Mean Work Pulse Increment fy1 N P M v A Businessmen 233 53.2 (50) 56.4 (28) +3.2 >0.05 a ~q B Students 159 39.4 (83) 40.0 (53)* +0.6 >0.05 * "Moderate" plus "Heavy" smokers. fA .. ~- T201111

344 Tobacco and Health M TABLE 7 TABLE 5 MAXIESAL BLOOD PRE66UBT4 INCREMENT DURING MEAN RESTING BLOOD PRESSURE IN SMOKERS AND NON-SMORERS COLD PRES60R TEST AND 6% CO= INHALATION (in mm. of Hg) Yd IN SMOItERS AND_NON-BMOSERB ~ (in mm. of Hg) Total Probability Group No. Men "Never" "Heavy" o of t 0 Group A (N = 288) Probability A Businessmen 233 S 123.0 (60) 124.0 (33) +1.0 >0.05 ~ Cold Pressor "Never" "Heavy" A of t D 81.1 83.1 +2.0 > 0.05 Systolic BP Increment M 19.8 (67) 19.7 (35) -0.1 >0.05 A Businessmen 286 S 129 0 (74) 129 7 (46) +0 7 05 >0 ~ Diastolic BP Increment M 16.1 (67) 18.0 (35) -3.1 <0.05 . . . D 84.1 86.2 +2.1 . > 0.05 H ~ COz Inhalation B Students 159 S 121 0 3 (83) 120 3 (53)* -1 >0 05 . . , . Systolic BP Increment M 16.8 (67) 20.0 (36) +4.7 <0.05 D 68.1 69.4 -{-1, 3 > 0.05 Diastolic BP Increment M 9.7(67) 11.5 (36) +1.8 >0.05 C Railroad Workers 414 S 133,8 (73) 129.3 (138) -4.5 >0.05 Group B D 0 0 "Mod r t 80.6 77.5 -3.1 > . 5 e a e D Firemen 23-4 S 135.1 (56) 132,4 (60) -2.7 >0.05 Cold Preasor "Never" and Heavy" D 80.2 76.5 -3.7 >0.05 Systolic BP Increment M 1/.8 (83) 18.2 (53) -1.4 >0.05 1 (53) BP I t M 18 4(83) 14 Di t li +0 7 05 >0 . ncremen c as o . . . "Non- "All sncokers" Srnokera" P! Fi i h C0s Inhalation Systolic BP Increment M 11.4 (49) 18,1 (19) 0.7 0.05 , nn s Workers 525 S 137.6 (165) 132.3 (360) -5.3 <0.001 Diastolic BP Increment M 12.0 (49) 9.7 (19) -2.3 >0.05 0 84 D 87 8 -2 2 <0 05 . , . . * " " " " TABLE 8 Moderate Ileuvy plus smokers. RESTING CARDIAC OUTPUT (ACETYLENID METHOD) SMOKERS VERSUS NON-SMOICERS TABLE 6 GROUP A "M d r t Probability MEAN BI,OOn PRESSURES AND MAXIMAL I3LOOn PRESSURE INCREMENTS DURINO EY S N "Never" (n = 32) o e a e and Heavy" (n - 34) A of I ER(:IFE IN MOKERS AND ON-SMOKERS 0 1 0 05 (in mm. ilg) Cardiac Output (L./min.) M SD 4.8 0.6 4,7 0.6 . - . > Group Moderate Probability d- V Oxygen Difference M 65.3 69.5 +4.2 <0.05 A Bnsine4emrrn (23:3) "Never" and Heavy" o of t (cc./Liter) SD 8.0 7.2 lnli W rk S P M 1 89 49 0 12 51 5 0 5 0 05 o ys c res,snre . . ( ) ) + 9. ( . Work I)iast' nlic 1'ressure Af 77 7 77 0 -0 7 > . >0 05 TABLE 9 . . . . , BASAL OXYGEN CONSUMPTION (CC./MIN./M=) A 73usinexnicn (233) "Never" "Hcavy" IN SMOKERS AND N0N-SMOIRrERS Total Probability Systolic BP Inrrrment At 5.4(49) 5.3 (26) -0, l Di t li 131' I 11 >0.05 Group No. lllen "Never" "Heavy" 0 of t as o c ncrement f -1.3 0.4 +1.6 >0.05 A Businessmen 233 M 113.3 (67) 117.9 (36) +4. 6 >0.05 "Moderate SD 11.6 11.4 B Students (159) "Never" and Heavy" "Never" "All Smokers" S t li l' A Businessmen 233 M 118.3 (67) 117,7 (114) +4. 4 <0.05 ys o c B inrrement Af 18. R(83) 17.6 (53) -1-1 .3 ' >0.05 SD - 11.6 12.0 Diastolic 131 Increment A9 -18.8 -17.0 -1.8 >0.05 B Students 159 M 128.4 (83) 128.2 (53) -0. 2 >0.05 T201112

I' TABLE 10 d' B.ELaTI4E BODY WEIGHT IN SMOKERS AND NON-SMO%ERS d' (in Per Cent) M ~ Total Probability rl Group No. Men "Never" "Heavy" a of t ~ A Businessmen 233 M 100.3 (67) 98.9 (36) -1.4 >0.05 O SD 14.8 14.6 B Students 159 M 10'8.5 (83) 104.1 (53)' -4.4 >0.05 ~ SD 14.8 14.3 ~ C Railroad F ~1 Workers 414 M 105.9 (73) 103.1 (138) -2.8 >0.05 ~ SD 11.5 12.5 D Firemen 234 M 112,0 (50) 108.6 (60) -3.4 >0.05 SD 10.8 10.4 * "Moderate" plus "Heavy" emokers. TABLE 11 OBESITY IN SMOKERS AND NON-SMOKERS Sum of 2 Skinfolds (in mm.) Neuer Heavy a Probability of t Group A 233 M 37.6 (65) 33.4 (35) -4,2 >0.05 M Never 37,6 (65) Moderate and Heavy 34.4 (86) -3.2 >0.05 Sum of 2 Skinfolds (in mm.) Group E 525 M Never 21,9 (165) All Smokers 18.3 (360) -3.6 <0.05 Per Cent of Body Tat by Specific Gravity Measurement Group A 233 M Never 24,6 (60) Heavy 24.2 (32) -0.4 >0.05 M Never 24.6 (60) Moderate and Heavy 24.0 (74) -0.6 >0.05 TABLE 12 SXELETAL MEASUREMENTS IN SMOKERS AND NON-SMORERS Probability Never Heavy A of t Group A M 176.0 (65) 176.3 (33) +0.3 >0.05 Height (cm.) Sl) 6.3 6.0 Never All Smokers Bicristal plus M 39.4 (64) 39.7 (91) +0.3 >0.05 biacromial diamet.er/ standing height (c:m.) SD 1.33 1.50 Cardiovaseular Characteristics in Smokers and Non-smokers 347 TABLE 13 DIETARY FAT CONSUMPTION IN SMOgPSRB AND NON-SMO8ER8 GROUP A All Never Smokers A Probability of t (n ! 33) (n = 44) Total Fat (Gm./day) M 114.7 116.6 +1.9 >0.05 % Fat Calories in Diet M 10.3 41.8 +1.0 >0.05 "Animal" Fat (Gm./Day) M 84.0 86.0 +2.0 >0.05 % "Animal" Fat Calories in Diet M 29.8 30.3 +0.5 >0.05 TABLE 14 MEAN SERUM CHOLEB'TEROL LEVELS IN SMOI[ERS AND NON-BMOSERB Never Moderate and Heavy A Group A 234.8 (65) 242.7 (73) +7.9 Never Moderate and Heavy Group B 173.7 (83) 179.0 (53) +5.3 Never Heavy Group C 224,8 (73) 232,2 (138) +7.4 Never Heavy Group D 241.7 (54) 251.1 (59) +9.4 Never Current Smokers Group E 222,6 (165) 248.1 (360) TABLE 15 +25.5 CHOLEBTIDROL/PHOBPHOLIPID RATIO AND BETA LIPOPROTEIN CHOLE6'TEnOL IN SMOKERS AND NON-SMOKERS GROUP A "Never'f "Heavy" a Probability of t C/P M 1.558 (59) 1,460 (30) -0,098 <0.01 SD 0.124 0.154 C/P M "Never" 1.558 (59) "All Current Smokers" 1.493 (107) -0.065• <0.01 SD 1.124 0.149, Phospholipid (mg. %) M 179.6 (59) 191.6 (96) +12.0 0.01 SD 26.0 31.0 Beta Cholesterol M 81.0 (62) 80.6 (95) -0.4 >0.05 (% of total cholesterol) SD 5.6 6.4 T201113

, 348 Tobacco and Health ~ Cardiovascular Characteristics in Smokers and Non-smokers 349 TABLE 16 C'ry ~ TABLE 19 MEAN FASTING BLooD SUGAR AND PROTEIN BOUND IODINE ~~..~ IN SMOKERS VERSUS NON-SMOKERS SMOKING AND REBPIRATORP FUNCTION IN A SAMPLE OF RAILROAD EMPLOTE'Ea AoE6 40-49 ~ (n - 432) GRouP A "All Current Probability ~ Never Stopped "Never" Smokers" A oJ t Ciparettes/day Smoked <10 10-20 21-40 41> >i Year (n = 65) (n = 109) ~ FBS (mg. %) M 69.6 74.8 +4.7 >0.01 No. of Men 105 12 136 105 7 66 SD 4 ~ 14 9 10 . . PBI (mg. °jo) M 5.77 6.fi£ -0.25 >0.05 Height (cm.) 174.5 174.1 174.0 175.2 176.2 176.9 SI) 1.40 1.15 Forced Vital Capacity (FVC)(L.) 4.532 4.105 4.416 4.334 4.189 4.156 TABLE 17 FEC 0.75 (L.) 3.185 3.108 2.955 2.838 2.554 3.276 HEMOGLOBIN LEVEI. IN SMOKERS AND NON-SMOKERS MMEF (L./aec.) M L 2.2 2.3 2.0 1.8 1.5 2.4 GROUP A EFIt ( ./aec.) FVC (cc.)/Ht. cm. 3.9 25.9 4.0 25.5 3.8 25.3 3.5 24.7 3.3 23.7 3.9 26.8 "Moderate Probability FEC 0.75 cc./Ht. cm. 18.1 17.8 16.9 16.2 14.5 18.5 "Never" and Heavy" g oj t (n = 65) (n = 73) MMEF (cc./sec.)/ Ht. cm. 13.6 13.2 11.4 10.5 8 6 13 7 Hemoglobin (gm. %) M 16.h5 15.69 +0.26 >0.05 MEFR (cc./sec.)/ . . SI) 0.93 0.97 Ht. cm. 22.2 23.0 21.6 20.4 18.9 22.1 TABLE 18 LnNO VOLUMES IN SMOKERS AND NON-aMONERS 4. Wynder, E. L (editor) (1955): The Biologic Effects of Tobacco. Boston, Little, Geour A (n - 221) pp. 35-91. "All Current Probability 5. Karvonen, M., A, Keys, E. Orma, F. Fidanza, and J. Brozek (1959): Cigarette "Never" Smokers" i; of t smoking, serum cholesterol, blood pressure, and body fatness. Observations (n - 65) (n - 125) in Finland. Lancet, 1: 492. Vital Capacity (ca) M 4320 4109 -211 <0.05 6. Hammond, E. C. (1959): Inhalation in relation to type and amount of smoking. SI) 678 644 Journal of the American Statistical Association, 54: 35. Residual Volume (cc.) M 1890 2040 +150 >0.05 7. Blackburn, H., J. Brozek, and H. L. Taylor (1959): Lung volume in smokers SD 376 608 and non-smokers. Ann. Int. Med., 51: 68. Total Lung Capacity (cc.) M 82$8 6170 -58 >0.05 8. Blackburn, H., J. Brozek, and H. L. Taylor (1960): Common circulatory mea- SD 897 905 surements in smokers and non-smokers. Circulation, in press. ItV/TLC (~o) M 30.6 85.2 +2.6 <0.01 9. VanSlyke, C. B., and P. S. Larson (1950): Observations on the role of the SD 4.1 6.9 adrenal medulla in the blood pressure response to nicotine. J. Pharmacol. and Exper. Therap., 98: 400. REFERENCES 1. Doll, R., and A. B. Hill (1958): Lung cancer and other causes of death in rela- tion to smoking: a second report on the mortality of British doctors. Brit. M. J., 2: 1071. 2. Hammond, li. C., and D. Horn (1958): Smoking and death rates-report on forty-four months of follow-up of 187,783 men. J.A.M.A., 166: 1294. 3. Cigarette Smoking and Cardiovascular Disease (1960): Joint report of Ad Hoc Committee on Smoking and Cardiovascular Disease. Circulation, 22: 160. T201114

350 Tobacco and Health 10. Watts, D. J., and A. D. Bragg (1956): Effect of smoking on the urinary output of epinephrine and norepinephrine in man. J. App. Physiol., 9: 275. 11. Brown, R. G.. T. McKeown, and A. G. W. Whitfield (1957): A note on the association between smoking and disease in men in the seventh decade. Brit. J. Prev. Soc. Med., 11: 162. 12. Orma, E., and M. J. Karvonen (1960): The relationship of smoking to blood pressure, serum cholesterol, and respiratory function in men of totallyy sampled areas in Finland. Report at the 5th International Congress of Gerontology, August, 1960, San Francisco. 13. Burn, J. H. (1960): The effect of nicotine on the peripheral circulation. Ann. New York Acad. Sci., in press. 14. Schlumm, F. (1930): Nikotinabusus und Grundumsatz. Verhandl. d. deutsch. Geselisch. F. Inn. Med. Kong., 42: 151. 15. Roth, G. M. (1951): Tobacco and the Cardiovascular System. Thomas, Spring- field. pp. 26-38. 16. Hiestund, W. A., H. J. Ramsey, and D. M. Hale (1940): The effect of cigarette smoking on metabulic rate, heart rate, oxygen pulse and breathing rate. J. Lab. and Clin. Nted., 25: 1013. 17. Pearl, R. (1933): Tobacco smoking and longevity. Science, 87: 216. 18. Berkson, J. (1959): The statistical investigation of smoking and cancer of the lung. Proo. Staff Meet., Mayo Clinic, 34: 206. 19. Brozek, J., antl A. Keys (1957): Changes of body weight in normal men who stop smoking cigarettes. Science, 125: 1203. 20. Blackburn, H., J. Brozek, H. L Taylor, and A. Keys (1960): Comparison of cardiovascular and related characteristics in habitual smokers and non-smokers. Ann. New York Aead. Sci., in press. 21. Gofman, J. W., F. T. Lindgren, B. Strisower, O. de Lalla, F. Glazier, and A. Tamplin (1955): Cigarette smoking, serum lipoproteins, and coronary heart disease. Gcriatrics, 10: 349. 22. Bronte-Stewart, B. (1956): Smoking and the cardiovascular system. Brit. M. J., 2: 659. 23. Thomas, C. B. (1958): Familial and epidemiologic aspects of coronary disease and hypertension. J. Chronic Dis., 7: 198. 24. Bronte-Stewart, B., I.. E-I. Krut, and M. J. Perrin (1960): The relationship of smoking to ischemic heart disease. South African Med. 1. 34: 511, June 11. 25. Marder, L., G. I-I. Becker, B. Maizel, and H. Necheles (1952): Fat absorption and chylomicronemia. Gastroenterol., 70: 43. 26. Blackburn, If, I•f. L., Taylor, and E. Leuthold (1960): Unpublished observa- tions. Cigarctle smoking and alimentary lipcmia. 27. Page, I. H., 1. A. Lewis, M. Moinuddin (1959): Effect of cigarette smoking on serum cholesterol and lipoprotein concentrations. J.A.M.4., 171: 1500. 28. Blackburn, H, H. L. Taylor, E. Leuthold, and W. Carlson (1960): Unpub• lished obseivations. The effect of cigarette smoking on plasma free fatty acid levels. 29. ~ ~ 30. O 31. ILI ~ 32. ~ 33. I 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. , Cardiotaascular Characteristics in Smokers and Non-smokers 351 Eisen, M. E., and E. C. Hammond (1956): The effect of smoking on packed cell volume, red blood cell counts, hemoglobin, and platelet counts. Canadian M.A.J., 75: 520. Blackburn, H., E. Orma, G. Hartel, and_S. Punsar (1959): Tobacco smoking and blood coagulation. Am. J. Med. Sc., 238: 448. Higgins, L T. T. (1957): Respiratory. symptoms, bronchitis, and ventilatory ca- pacity in random sample of an agricultural population. Brit. M. J., 2: 1198. Franklin, W. (1958): The effect of smoking on pulmonary function in a work- ing adult population. Abstract. J. Clin. Invest., 37: 895. Higgins, L T. T. (1959): Tobacco smoking, respiratory symptoms, and ventila- tory capacity. Brit. M. J., 1: 325. Fisher, R. A. (19'59): Smoking. The Cancer Edinburgh and London. Controversy. Oliver and Boyd. England (1960): Smoking and Eysenck, H. J., M. Tarrant, M. Woolf, and L. personality. Brit. M. J., 1: 1456. Eysenck, H. J. (1957): The Dynamics of Anxiety and Hysteria. Routledge and Kegan Paul, London. Heath, C. W. (1958): Differences between smokers and non-smokers. A.M.A. Arch, Int. Med., 101: 377. Thomas, C. B., and E. A. Murphy (1959): Observations on some possible pre- cursors of essential hypertension and coronary artery disease. Ann. Int. Med., 50: 970. Fisher, R. A. (1958): Cancer and smoking. Nature, 182: 596. Raaschou-Nielson, E. (1960): Smoking habits in twins. Dan. Med. Bull., 7: 82. Dawber, T. S., F. E. Moore, and G. V. Mann (1959): Some factors associated with the development of coronary. heart disease: Six years follow-up experience in the Framingham study. A.J.P.H., 49: 1349. Doyle, J. T., A. S. Heslin, H. E. Hilleboe, and P. F. Formel (1959): Early diag- nosis of ischemic heart disease. New England 1. Med.,'261: 1096. Levy, R. L. (1947): Effects of smoking cigarettes on the heart in normal persons and in cardiac patients. J.A.M.A., 135: 417. e T201115

24 TOBACCO ALLERGY AND VASCULAR RESPONSES WALTER REDISCH, M.D. To smoke or not to smoke is still, at least within the realm of cardiovascular disease and its prevention, a question asked by many a patient and difficult for the honest physician to answer. However, from the Conference of the New York Academy of Sciences in March of this year and from the pro- ceedings of this conference, which so far have been most interest- ing, a few facts have emerged concerning the effects of tobacco use on the cardiovascular apparatus of man. From the work of Bing, Iiedell Thomias, Grace Roth, Travell, Rinzler, Montgomery, Rottenstein, Edwin Wood and others the following seems to be impressed on one's mind: 1. work performance of the heart may be altered 2. heart rate and arterial pressure may be altered 3. skin perfusion may be altered 4. perfusion of the skeletal muscles may be altered 5. all these e[t'ccts may be due to the pharmacologic action of the absorbc d nicotine and consequently the mode and the way of administration of the tobacco, that is whether it is smoked in cigarettes, cigars or pipes, or is chewed or snuffed, is of minor importance here 6. there seems to exist a critical threshold of nicotine sensitivity 7. responses lo tobacco encountered in patients with cardio- vascular disease are not necessarily different from those observed in healthy subjects, and 1 ! r Tobacco Allergy and Vascular Responses 353 8. some of these changes occur in a certain percentage of test subjects while failing to occur in the rest. If there are any direct roads from these basic leads to their clinical application, they are likely to be crowded with open and hidden pitfalls. It seems so much better then, to travel on the devious but much safer avenues of rationally justified investiga- tory approach, which every such lead opens up. Let us see where travel along one of these many avenues has carried us so far: the fact that some circulatory changes can be observed only in a certain percentage of healthy test subjects as well as of patients with vascular disease has suggested the possibility of an immunologic mechanism at play. Lights on this avenue of approach have so far been few and far between indeed. In 1932, Harkavy and his co-workers first explored the question of tobacco allergy as a possible causative factor in vascular disease. Their investigations were followed by those of Sulz- berger and his group along similar lines. While these two groups of workers did find suggestive evidence for a tobacco sensitivity mechanism active in the production of some vascular pathology, such evidence was denied by Trasoff and his group, by Chobot, and by Westcott and Wright. It was quite natural and logical that most of these early investigations centered around so-called thromboangitis obliterans according to the concept of the late Dr. Buerger. This concept was accepted without opposition in those days, but has been subsequently questioned. Many workers in the field of vascular disease feel that most cases of so-called Buerger's disease are cases of what we call obliterative athero- sclerosis and the rest are due to various forms of endangitis or angitis. The attempts at clinical correlation in this early work consequently call for more elucidation. There remains the impor- tant fact that in continued efforts, Harkavy as well as Sulzberger in spite of their disagreement in important details have both been able to demonstrate immunologic specificity of the skin reaction to tobacco extracts, if not to nicotine. Moreover, there remains the fact that some cases of obliterative arterial disease are obviously aggravated by smoking, while in others such a deleterious influence of tobacco is not demonstrable. In 1947, Boyle, Wegria and their co-workers investigated the effects of 352 1 T201116

354 Tobacco and Health intravenous injection of nicotine on the circulation of healthy persons and of patients with cardiovascular disease. They found considerable variation in responses and this variation was as great in healthy persons as in patients with cardiovascular dis- orders. The authors concluded that the wide variation in re- sponses to tobacco depends upon individual susceptibility rather than upon presence of disease. Ten years later, Silvette, Larson and Haag reviewed the Immunologic Aspects of Tobacco and Smoking. They concluded that "much of scientific interest and value can hardly fail to result from a careful, and critical, observation of natural and acquired idiosyncrasy to tobacco in man," and pointedly sum- marized the state of affairs in the words: "A well designed and properly controlled program of clinical investigation of tobacco sensitivity in relation to disease is greatly to be desired." Meanwhile Harkavy and co-workers had started their studies to demonstrate that among smokers, individuals who were skin- sensitive to tobacco extracts were significantly more prone to develop early coronary disease and its sequelae than the non- allergic controls. In most interesting survival studies, Harkavy and I'erlman have shown that among 106 survivors of myocardial infarction, 5,6 or about 53 per cent had positive skin reactions to tobacco extracts with both direct and passive transfer. Sulzberger, Fontana and their co-workers found a significant correlation between skin sensitivity to tobacco extracts and other allergic manifestations. TABLE 1 CORRICLATION BLTWE1CN "TOBACCO ALLIBROY" AND OTIIER ALL%ROIC MANrlr]ESTATIONS iN SMOKER.4 Total Percentage Total Season.al Bronchial of Other Number ,Skin Test Ir'hin.itis Aethma Urticaria Allergies 1'osi4ivo 57 77o lR%n 26% 51 377 Net;a+ive 320 4rjo 4% 5% 13 Tobacco Allergy and Vascular Responses 355 They likewise found a similar correlation to what they called "peripheral vascular symptoms," to wit, coldness, numbness and tingling of the extremities in response to smoking. TABLE 2 CORRELATION BETWEEN "TOBACCO A.LLEROY" AND "PERIPHERAL VASCULAR S7CddPTONB" IN SIIOHBRB Total Cold Numbness of Tingling of Number Skin Test Extremities Extremities Extremities Positive 57 30% 28% 30% 377 Negative 320 13% 14% 15% As interesting as these findings are, the evaluation of the subjective peripheral complaints in response to smoking and their identification as being truly vascular in nature is difficult. It has been known for many years, that about one of four persons tested shows a decrease in surface temperature in re- sponse to smoking a cigarette. Later, it was reported that plethys- mographically measured blood flow to the fingers and toes was diminished in about the same percentage of test subjects after smoking. Testing of skin sensitivity to tobacco extracts yielded 20 to 30 per cent positive results. Recognition of the similarity in percentage figures obtained was responsible for an investigation of possible relationships be- tween skin sensitivity to tobacco extracts and vascular responses to the smoking of tobacco. As the result of an initial study of 80 healthy smokers, one significant correlation became apparent: almost 90 per cent of those who had negative skin tests, failed to show a decrease in total blood flow to the lower extremity. However, since blood flow responses to smoking in the lower extremity seemed erratic, in that some test subjects showed unequivocal increase in total flow, it soon became evident that it was necessary to employ some acceptable means of separately estimating skin flow and muscle flow. A method of differential plethysmography has been developed in our laboratory on the T201117 i

356 Tobacco and Health basis of the rather constant difference between skin-mass/muscle mass ratio in the foot and in the leg. The formula shown on the next slide permits a rough estimation of skin and muscle flow by correcting for the most obvious errors. VL3 V F VL VL. WFs I Fc"~4'"`l FL + FF t =FcG~.~M,~.) VL +VF V FLS Fig. 1 P=f (CC/lOOCt/M1N.) Ps=PF +CF 0 CF•0.20 P-=PL-CLA CL•0.26 ~~ Fk-PL 1-CF'CL F=FLOW RATE V = VOLUME P=PERFUSION RATE L=LECz F=FOOT S=SKIN M = MUSCLE C= CORRECTION FACTOR Of course, all experiments are done in the Constant Tempera- ture Laborato.y; the ones reported here were done at 25°C and 55% humidity. In order to be reasonably sure that differences be meaningful, we considered as significant only changes of more than 1 1/2 mI. per 100 ml. of tissire/min. At the present time 34 subjects have been tested with this method. There were 25 males and nine females. Their ages ranged from 24 to 87 years with an average of 56.3. Seven of the 34 subjects exhibited evidence of obliterative arteriosclerosis, while the rest had no demonstrable vascular disease. Nine of the group were significantly sensitive to one or more specific I I Tobacco Allergy and Vascular Responses 357 tobacco extracts. Skin and muscle flow responses of these 9 subjects were tested for the smoking of tobacco types to which their skin was sensitive. TABLE 3 Skin Flow Muscle Flow Decrease No Change Increase Decrease No Change Increase Total Allergic 9 8 1 0 2 3 4 No OAS 8 7 1 0 2 2 4 OAS 1 1 0 0 0 1 0 In response to smoking, in eight of these nine tobacco-sensitive persons a significant decrease in skin blood flow was observed. One had a decrease not significant according to our criteria and is therefore listed as showing "no change"; none had an increase in skin flow. TABLE 4 ; Skin Flow Muscle Flow Decrease No Change Increase Decrease No Change Increase Total Non- Allergic 25 2 15 8 2 12 ii No OAS 19 1 10 8 2 9 8 OAS 6 1 5 0 0 3 3 In contrast, of the 25 subjects who were skin negative, eight showed an increase in skin flow, 15 had no change and in only two a significant decrease in skin flow was observed. There was no conclusive trend in muscle $ow changes in either group. The differences in skin flow responses, however, are highly significant. T201118

358 Tobacco and Health Everyone interested in studying vascular responses to physio- logic and pharmacologic stimuli must soon realize that there is no uniform re-sponse of the peripheral blood vessels. The admittedly convenient terms vasoconstrictor and vasodilator ac- tion can be applied solely to the effects on specific vascular beds; and even there only under certain given conditions. It has been customary to take one parameter, such as systemic arterial pres- sure, and to assume that if this increases, peripheral resistance has increased, and that this indicates "overall" vasoconstriction; and that the reverse situation implies "overall" vasodilation. It has become more and more debatable whether this sort of con- cept is justified as a working hypothesis. There are 2 major reasons for this doubt: First, evidence has been accumulating suggesting that specific vascular beds have specific functions within the complex of body economy. Second, evidence is begin- ning to appear indicating that the purely physical mechanical concept of blood circulation may no longer be tenable. That is, that changes in the hio-chemical milieu of the organism as well as of the vessel wall may be capable of changing vascular responses. To cite just one example, directly concerned with the problem of tobacco consumption: Burn and his co-workers have demon- strated that cardiovascular responses to nicotine, occurring with predictable quantitative accuracy, can no longer be elicited after depletion of catechol-amine stores by reserpine; but can be elicited after restoration of the catechol-amine milieu by infusion. Thus it seems hardly permissible to simply use results of measurements of once or the other vascular bed and its responses in man for generalized conclusions. Much further evidence is needed demonstrating that reproducible responses of specific vascular beds to tobacco can be elicited in a significantly greater percentage of people with positive skin tests to tobacco extracts, than in those with negative skin tests. If this proves so, we may be on the way to a practical screening procedure for the detec- tion of those individuals prone to react with their vasculature to the use of tobacco. Obviously, many more questions will have to be answered before we may consider ourselves on terra firma: extensive further immunologic investigation is needed. The question of nicotine idiosyncrasy and sensitivity in man has to be 1 Tobacco Allergy and Vascular Responses 359 investigated; for little is known about this problem beyond the fact that nicotine may have some haptenic activity and that it does produce occasionally both qualitative and quantitative sensitivity phenomena. The difficulties in attacking this problem are of course appreciated. Lowell has emphasized the necessity of clear-cut differentiation between the pharmacologic action of nicotine and a possible antigen-antibody reaction. What is the antigenic "something" in tobacco extract? Is it nicotine con- jugated to some protein? What about the nicotine occurring so widely in the botanical world apart from the tobacco leaf as we know from the work of Dawson, Solt, and Christman? The ques- tion whether skin sensitivity to tobacco extracts can be considered identical with skin sensitivity to, shall we say, tobacco smoke con- densates has been a delicate point with many people in the field; however, we may expect this question to be answered before too long. Little is known of the interaction of nicotine and histamine, nor of nicotine and the catecholamines. These ques- tions of purely investigative nature will have to be dealt with before the clinical problems of specific disease-relationships may be considered. Some of us question the "specificity" of some of these "disease entities." In any case, we who are concerned with vascular responses are a great distance from the answer to the essential question=`to smoke or not to smoke." However, we feel strongly that the answer will not be a simple yes or no; that it will be qualified and, we hope, that such qualification will be based on sound and reliable information. If we are humble and will accept Pascal's religious philosophy, as well as his teachings in science, we may find comfort in the words of his God: "Console yourself, you would not seek me if you had not found me." T201119

I 25 H ko M ~ ~ O Z ~ CORONARY HEART DISEASE INCIDENCE AND MORTALITY AMONG SMOKERS AND NON-SMOKERS WILLIAM J. ZUKEL, M.D., TIIOMAS R. DAWBER, M.D., WILLIAM B. KANNEL, M.D., HAROLD A. KAHN, M.A. and HARoLn F. DORN, Ph.D. OBSERVATIONS that tobacco smoking can induce symptoms in individuals with angina pectoris have been reported in the medical literature for nearly one hundred years 1-e This ischemic effect has been confirmed by the production of electro- cardiographic or ballistocardiographic abnormalities during smoking in individuals with coronary heart disease.'-° A number of studies have now observed frequency of heavy cigarette smoking among individuals with coronary heart disease and an increased mortality rate from coronary heart disease among cigarette smokers as compared with non-smokers.°''° The difference in prevalence or incidence of coronary heart disease in some of thesestudics has ranged from none to a fourfold excess among cigarette smokers compared with non-smokers (Table 1). Note that the `rcatest difference in the ratio of coronary heart disease among cigarette smokers compared to non-smokers appears in the vounger age groups. While this degree of increase may appear minor in comparison with the more than ninefold range found for cancer of the lung, the fact that nearly one-half of all excess deaths among regular cigarette smokers is attributed to coronary heart disease empha- sizes the need io determine the significance of this association. Coronary Heart Disease Incidence and Mortality 361 This report will present findings from three studies carried out by the Public Health Service in which this question of the rela- tionship between cigarette smoking and coronary heart disease is considered in further detail. These three studies were carried out by different investigators in different population groups on a prospective basis. Information on smoking habits of the indi- viduals in these studies was determined at the beginning of each study and the populations have been followed to determine the incidence of coronary heart disease in relation to the character- istics of these individuals at the time of entry into the study. Data analyzed for this report cover a period of from one to eight years of follow-up of these populations. Data from the first of these studies are an amplification of the findings from the follow-up of 294,155 Governmental Life In- surance policy holders in the study of Dr. Harold F. Dorn, reported ori earlier in this conference. During the two and one-half year period of July, 1954 through December, 1956, there were 5,775 deaths in which coronary heart disease (420) was an underlying cause. The diagnosis was Cawv, lunq 1162-1631 (460- 463) P-" 6rmralwfiy x(1 0r aE~TH6 197 23 0 145 CI6+16TTE 0NL7 1 19 1,09 1 61 - 327 Caawry Mal diumd4201 Nmnp-mlw .ndoemrdYlM(421-4221 HypFlHn~d+ w.M hq1 di<I®M (440'447) 6.nMrol N+.auOM.mn M601 HypalAns.me .n6,pu1 Nqrl diwaHrW44•447) Grob.d rouub. 1-6 (330•3341 Ch- "low41. (592 •494) Nor1 d~~'N4roi •4e6) ybmdraN n6l~mna ro601 NO oi OEATH3 1420 135 276 243 62 261 26 42 & qArr d4Ha1R1 d I~wr, ymwea. YS62•5671 43 0.o@yNS 1260) 64 UIeH ml Nd-N, dwdmwn 1540-6411 ~~ l~a~ 51 62 2, 63 _2.95 mdqrHnm,N0(500-527) 147 0 I 2 3 4 0 1 2 3 MORTALITY RATIO Figure 1. Mortality of regular smokers of cigarettes only from specific diseeises; ratio of observed to expected number of' deaths; death rate of nonsmokers and persons who have smoked only occasionally equals 1.010. (Data include contributory causes in addition to underlying causes of death.) 360 i T201120

362 Tobacco and Health confirmed by autopsy in 17 per cent, by electrocardiogram or other laboratory tests in 35 per cent, by history or clinical find- ings alone in 23 per cent and the remaining 25 per cent were coroner's cases or the basis for diagnosis was not given. Figure 1 presents the relative excess mortality among regular cigarette smokers for each of the diagnostic categories of cardio- vascular disease. The 63 per cent excess deaths from coronary heart disease for persons who smoked only cigarettes is higher than for any of the other diagnostic categories of cardiovascular disease. No excess risk is apparent for chronic rheumatic heart disease. Table 2 presents the mortality ratios for coronary heart disease by age and by type of lifetime smoking history among these policy holders Using the age-adjusted mortality in the group of individuals who never smoked or only occasionally smoked as a reference figure of one, you will note that very little increase is present for cigar and/or pipe smokers but that a 52 per cent in- crease is present for all persons who smoked cigarettes and a further increase to 63 per cent for those who smoked cigarettes only. The highest ratio of excess risk appears in the 30 to 54 year age group with a persistent but lesser effect with increasing age. Table 3 presents the mortality ratios for coronary heart disease by the amount of cigarette smoking. Again, using the mortality experience of persons in the population who never or only occasionally smoked as a reference figure of one, there is a clear increase of age-adjusted risk for those who smoke less than 10 cigarettes per day and a further increase for those smoking 10 to 20 cigarettes per day. A further increase of risk occurs with heavy smoking in the age group 30 to 54, but is less apparent in the older age groups. Table 4 relates the experience of cigarette smokers to non- smokers in rm•al areas. The age adjusted mortality ratio for those who snioked one pack or less is 64 per cent greater than for rural non-smokers. A further increase to a 96 per cent excess occurs among those who smoked more than one pack of cigarettes per day. Table 5 coutrasts the mortality experience from coronary heart disease for cigarette smokers and non-smokers in areas with a Coronary Heart Disease Incidence and Mortality 363 population of 50,000 or more. Again the age-adjusted increased mortality ratio for *cigarette smokers persists and increases with the amount smoked. This experience was also found to exist for policy holders who lived in intermediate sized communities in this study. It appears from these data that a significant excess risk of dying from coronary heart disease occurred among regular cigarette smokers when compared with persons who never smoked or smoked only occasionally. This excess risk is present regardless of the size of the community in which these persons lived. The second study providing information on the relationship of smoking to coronary heart disease is the Framingham Heart Epidemiology Study which was established in 1948 by the National Heart Institute in Framingham, Massachusetts. This long-term study of a random sample of the adult popula- tion between the ages of 29 and 62 was established to observe the natural development of heart disease and to assess the role of constitutional or environmental factors of possible importance. Sixty-nine per cent of the individuals within the desired sample responded and were given a thorough cardiovascular examin- ation. Table 6 indicates that 4,469 persons responded and were examined at the initiation of this study. An additional group of volunteers has participated in this study giving a total population of 5,127 adults who were initially free of coronary heart disease. The incidence of coronary heart disease in this group of volun- teers is essentially the same as that of the selected respondents and for this reason the two groups have been combined. Biennial examinations have been carried out on this group to identify the individuals who developed coronary heart disease after entry into the study. An attempt is now being made to determine which factors are associated with an increased risk of developing coronary heart disease. In a previous report based on six years' experience, smoking was found to be associated with an inczeased incidence of non-fatal myocardial infarction and of death from coronary heart disease among men age 45 to 62. It did not appear to be associated with an increased incidence of angina pectoris. T201121

364 Tobacco and Health The following preliminary data are now presented on the basis of an analysis of eight years of follow-up which has just been completed. Table 7 presents the experience of the total study group in which 114 cases of myocardial infarction occurred among the males and 25 cases among the females who were initially free of coronary heart disease at entry into the study. This total ex- perience for the males is given an incidence ratio of 100 for reference purposes. You willl note the lower than expected incidence of myocardial infarction among the "never-smoked" group (53). All non- cigarette smokers considered as a group have an incidence ratio of 60. Non-cigarette smokers are defined as those who never smoked, had stopped smoking, or who smoked cigars or pipes onlv. In contrast to these non-cigarette smokers, the cigarette smokers have an incidence ratio of 125, double the rate among the previous group. This is a significant level of difference. Some gradation of increased risk appears in relation to the amount smoked with a ratio of 155 being found for men who smoked over 20 cigarettes per day, a threefold increase above the experience of those who never smoked. The females show the same trend of increased risk of myocar- dial infarctior, with a ratio of 159 for cigarette smokers and 75 for non-cigarette smokers. Considering cases of angina pectoris among males the m.unbers are insufficient to establish a significant trend; however, the data to this point suggest an increased risk among the men who smoked over 20 cigarettes per day, as indicated by the ratio of 158 in this group. To this point the data do not indicate any relationship between cigarette smoking and an increased incidence of angina pectoris among women in the Framingham study. Table 8 prrsents data on the incidence of myocardial infarc- tion by age and smoking status of the males in this study. At age 30-39 the non-cigarette smokers have an eight year incidence of mvocardial infarction of 5.8 per 1000, in contrast to a rate of 21.3 for those smoking less than 20 cigarettes per day, and a further increase to 38.8 per 1000 for those smoking over 20 Coronary Heart Disease Incidence and Mortality 365 cigarettes per day. Note the higher incidence among cigarette smokers at each age level. To adjust for the possible effects of serum cholesterol level, blood pressure level and weight, the average of rates for the eight high-low combinations of cholesterol, blood pressure and weight has been calculated in developing this adjusted incidence ratio (Table 9). In doing this type of analysis, the effect of cigarette smoking can be separated from the effects of these three factors which are known to influence the risk of developing coronary heart disease. When the effects of these three factors are corrected for, the excess risk of both myocardial infarction and total coronary heart disease among male cigarette smokers persists. You will note that the excess risk of myocardial infarction for cigarette smokers is nearly sixfold at age 30 to 39, 7.5-fold at age 40 to 49 and drops to 1.2-fold at age 50 to 59. The increased risk also persists with regard to total coronary heart disease for males. A four-fold excess risk appears for female cigarette smokers in rela- tion to myocardial infarction, but is not apparent when total coronary heart disease is considered in this manner as may be expected since angina pectoris cases are included in the grouping of total coronary heart disease. Table 10 shows an analysis of the mortality experience within the Framingham population. Again the excess risk among ciga- rette smokers is reflected among males who smoked 20 cigarettes per day or less and a more marked excess (154 vs. 62) for those smoking over 20 cigarettes per day. A parallel type of experience is seen for mortality, from coro- nary heart disease. Non-smokers have a mortality ratio of 60, in contrast to the ratio of 121 for male cigarette smokers. In Table 11, the eight-year mortality rates from all causes are adjusted for high-low differences in cholesterol, blood pressure and weight between smokers and non-smokers. The excess mor- tality for cigarette smokers is not significantly affected by this adjustment. As in the case of morbidity data the highest relative risks are in the younger age groups. The third prospective study from which data on the relation- ship of smoking to coronary heart disease are available is the T201122

366 Tobacco and Health North Dakota Coronary Disease Study. This was carried out by the Heart Disease Control Program of the Public Health Service in cooperation with the North Dakota State Department of Health and the physicians of two District Medical Societies in six counties of North Dakota during 1957. A special health survey was made by the Bureau of the Census to enumerate a 10 per cent sample of all occupied dwellings in the six county area. From this enumeration 1,886 males age 35 and over were identified and information was obtained on their smoking habits as well as other characteristics. From the estimated population of 20,330 males age 35 and over in this six-county area, 228 cases of coronary heart disease were diagnosed by the practicing physicians during the year 1957. Of these, 160 cases were considered by a review committee of internists to be new cases without previous manifestations of coronary disease. The number of new cases within the age range of the Framing- ham Study w,~:s small; however, the data parallel the Framingham findings rather closely (Table 12). Using the total actual experience of incidence of new cases of definite or probable myocardial infarction as a reference ratio of 100, the "never-smoked" males had a lower incidence of myo- cardial infarct ion (61) than did the group as a whole ( Framing- ham figure was 53). The cigarette smokers had a greater than expected incidence of myocardial infarction (112) with the greatest incidence observed among those smoking 21 cigarettes per day or over. This ratio of 125 among heavy cigarette smokers is greater than the ratio for those smoking less than 21 cigarettes per day (108) and double that of the non-smokers in this popu- lation. IIeavy smokers in Framingham had a three-fold excess risk you recall (155). When all diagnostic categories are con- sidered together, the same unfavorable trend for smokers persists. Previous analysis of data from this study indicated that farmers had only hal: the incidence of coronary heart disease as non- farmers during the year of study (Table 13). The numbers are small but the actual incidence ratio of cases of coronary heart disease is higiier (113) among farmers who smoked than among those who never smoked cigarettes (73). An unfavorable relationship among non-farmers who smoked Coronary Heart Disease Incidence and Mortality 367 appears in relation to the amount smoked. The ratios for all coronary heart disease are 65 for those who never or occasionally smoked, 99 for those smoking less than 21 cigarettes per day, and 124 for those smoking 21 or more cigarettes per day. When the ratio of incidence rates of either myocardial infarc- tion or of total coronary heart disease is considered for cigarette smokers in comparison with non-smokers in different age groups (Table 14), we again find the highest unfavorable risk occurring at the younger ages. The incidence of myocardial infarction is 4.4 times as high among cigarette smokers at age 35 to 49 as among non-smokers; the differential lowers at the older ages. DISCUSSION Data provided by these three studies further reinforce the association between cigarette smoking and an increased mortality and incidence of coronary heart disease in males. This relation- ship appears to be more marked among the younger group of coronary-susceptible males than among older males. The unfavorable experience of cigarette smokers in increased mortality and incidence from coronary heart disease occurs in both rural and urban populations. The increased incidence of myocardial infarction among male cigarette smokers in the Framingham Study appears to be in addition to any effects re- lated to other recognized and measured factors such as serum cholesterol, blood pressure and weight. More detailed exploration of the relationship between these factors or other factors of pos- sible importance in the development of coronary heart disease will require the accumulation of additional cases before exten- sive analyses are possible. It is apparent that such a correlation with multiple factors will be slow in coming unless arrangements for pooling of data from current epidemiological studies can be accomplished. Such a step is now underway between the Fram- ingham Study and the Albany Cardiovascular Health Center. It would be extremely helpful to extend this pooling of basic data to other prospective studies where classification of cases is made by similar criteria. IN SUMMARY Data relating the mortality and incidence of coronary heart T201123

368 Tobacco and Health disease among smokers and non-smokers have been analyzed from three prospective studies-one a population of 294,000 Govern- ment National Service Life Insurance policy holders whose mor- tality experience has been followed for two and one-half years, the second from a general population of 5,127 adults in Framing- ham, Massachusetts, followed with biennial examinations for eight years and the third from a rural general population of 20,000 males in North Dakota where the incidence of coronary heart disease was recorded for a one year period. In these studies male cigarette smokers experienced a signifi- cantly higher mortality from coronary heart disease and a signifi- cantly higher incidence of myocardial infarction than did males who never smoked or only occasionally smoked cigarettes. This unfavorable relationship was more marked in the younger age groups and occnrred among males living both in rural and urban communities. The greater incidence of myocardial infarction among males who smoked cigarettes appears to he separate from the effects of elevated serum cholesterol, elevated blood pressure or excessive weight. ACKNOWLEDGMENT The authors acknowledge the important assistance provided by the staffs of the Biometrics Research Branch, the Framingham Heart Epidemiology Study and the Heart Disease Control Pro- gram of the Puhlic Health Service in preparation of data for this report. Coronary Heart Disease Incidence and b4ortality 369 '1'ABLh; 1 5'rUDIFS OF CIOARETTE ShfOKINO AND COROtdARY 1)ISEASE Ratio of CHD Age in Smokern In.uesligalor 1'opulalion Size Group Measure Non-Smokers ~ Fnglish, Willius, 416 40-49 Prevalence 4.8 and I3erkson 776 50-59 of coronary 2.4 1940 808 fi0-h disease in 0.8 consecutive admissions P 1)oll and Hill 73,000 person yrs 35-54 CHD Mortality 4.2 1956 20,000 '` 55-64 " 1.0 12,0011 " " 65-74 " 1.3 7,000 " " 75+ " 1.6 ~ Hammond and 112,000 " " 50-54 CHf) Mortality 2.5 Horn 1958 89,000 " " 55-59 " 2.5 59,000 " " 60-64 " 1.9 37,000 " " 65-69 " 1.6 ~ Dorn 1959 479,000 " " 30-80 CHDae prirnary 1.8 or associated cause of death ~(.onserlltlve records of 1,000 males 40 years of age and older were selected on t he basis of tobacco smoking. The prevalence of coronary disease was then determined in comparison with the prevalence of coronary disease in 1,000 matched non-smokers. .1. n. M. A., 15: 1328, 1940. s Itat.io relates smokers of one pack and more per day to non-smokers among British physicians. Pipe and cigar smokers included according to grams of tobacco smoked. Cirrulalion, 17: 1085, 1958. a ltatio relates smokers of one pack or more per day to non-smokers. Data from white males obtained by volunteer workers of American Cancer Society. Circulalion, 17: 1085, 1958. ' Male U. S. Covermment Life Insurance Policyholders, smokers of over one park per day. Pvblir. Hralfh Relrorls, 7.4: 581, 1050. T201124 l

370 Tobacco and Health '1'ABLI: 2 Coronary Heart Disease Incidenee and Mortality 371 TABLE 5 MORTALITY RATIOF FOR CORONARY HEART DISEASE BY AGE AND TYPE OF SMOKING, ~ USGLI POLiCYIIOLDERS LIFETIME HISTORY ~ MORTALITY RATIOS FOR CORONARY HEART DISEASE AMONG REGULAR CIGARETTE SMOKERS IN AREAS OF 50,000 OR MORE POPULATION Total Age 0 USGLI POLICYHOLDERS-LIFETIME HISTORY-MAXI1rSUM AMOUNT SMOKED Total Age Type of Smoking A ge-A djucted Rati.o 9Y0-5J 55-5fJ 60-64 65-69 70-1- ~ I -1 Never Smoked 1. 00 1.00 1.011 1.00 1.00 1.00 ~ Type and Amount Smoked Age Adujalcd Ratio 60-54 66-64 66-I- Cigar and/or Pipe 1,08 1.23 1.15 1.08 1.07 .97 Neoer Smoked 1.00 1.00 1.00 1.00 Total Cigarettes 1.52 1.88 1.59 1.50 1.47 1.23 Cigarette Only Smokerk Cigarettes only 1.63 1.77 1.66 1.67 1.59 1.24 One pack or less 1.47 1.11) 1.47 1.50 ~ More than 1 pack 1.59 1.22 1.68 1.37 'rASl.r 3 r MORTALITY IIATIO4 FOR CORONARY I-IF.ART l)ISEASE AMl/NG CIGARETTE SMOKF.RS TABLE 6 t BY (CURRENT) AMOIINT SMOKEn j1SGUI POIdCYHOLDERR CONSTRUCTION OF FRAMINGHAM STUDY GROUP Total Age Number of persons Type and Currrnl ~-. _- ~- - Amoun.t SmoF.ed age-Adjustrd Ralin .40-6/ 55-5h 60-64 65-{- Total Men Women Nelrer Smoked 1. 00 1. 00 1.00 1.00 1.00 Random Sample 6510 3090 3420 Cigarette Only Respondent-total 4469 2024 2445 Less than 10 per day 1.31 1.39 1,26 1.40 .89 Respondents-free of CHD 4393 1976 2417 10 2'0 1.77 1.45 1,93 1.74 1,17 Vohmteers-free of CHI) 734 307 427 21 or more 1.72 2.25 1.07 1.95 I .95 Total-free of CHI) (Population at Risk) 5127 2283 2844 '1'ABL1i, 4 MORTAI,I'rY 1LATIOt1 FOR CORONARY HEART I)ISSEASE AMONti REGULAR CIGARICTTE SMOKERS IN RURAI, AREAS 1.1.5'GI,I POLI(:YNOLnERS-I,iFETIME 1'1ISTORY-MAYIMUM AMOUNT SMOREn Total Age Type and Amovnl Slnnkrd Agr-/1 d,luRted Ra110 60-6/r 5fi-64 6.5+ Rural Never Smoked 1./10 1.00 1.00 1.00 Rural ('i,garelle Only One pack or less I.64 ... 1.56 1.63 More than I pack I.06 ... 2.11 1.16 T201125

Coronary Heart Disease Incidence and Mortality 373 Tobacco and Health 372 '1'ARIX 7 '1'ABLh. 1) 1' RAMINGIIAM--hRELIMINARY fl-YI9AR SMOKING I)ATA ~ I'RAhiINGHAtd-hRELIMINARY 8 YEAR.S'MOKING DATA ~ Cases of /)isrr.er Orrrn-rinrl :I mnng 7'hn.%c il ge :Il)-B2 and Frer, of Disea.se on lielation.vhip of Smoking and Occurrence of Disease Among Those Age 80-59 ~ and Free of Disease on Entry Into the Study Entry IrtAo lhe Study I~-1 Males Females 1)rf nily dl yorarrlial Dcfinitc Angina Pectori.s ~ lnfarrlion Only Category 80-89 40-49 60-69 I0-/r9 50-59 " fllalr h'rmalc Afale Female C /3asic 1)ata- ('igarrttr ,Smokrrx ' Cases of 1)efinitl: M. 1. 16 '3ts 3:1 •1 i : fin krr Ratio Iratio Ratio . Cases of DuL CHI) 19 42 50 4 13 A'o. nf of JVo. of o/' No. of of No. of of I'opulatiott Sizc 563 488 3'l!1 :157 201 ('atrs Ohr, to Cases Ohs. ln ('aReR Ohs. to Cases UhR. to Ob- I:xp. Ob- Exp. Oh- IE,rp. Oh- F'xpr. Basic 1)ata-- Non-G'ig. Smokrrs (excl'1ulCs lapsns) Smoking ('atrgur,rl vrrrMrl ('asrs served Cases srrved ('ascs serUed Cases Cases of 1)cfinitc M. 1. 1 1 13 'L 11 Cases of uef. CHI) 1 6 24 13 36 1'otal 114 1(141 25 10(1 57 100 56 100 Population Size 171 189 177 510 531 Never timokcd R 5:i 1:3 7:3 /l 42 109 l,alrshd R lifi 1 * :3 * 4 * (;rtitde 8-Year 1 ncid. Itates/1000 Cigar-Pipr Only 11 (12 .. .. 8 * Tot al Nml-Ciaruet l r• 27 60 14 75 'L(1 87 46 114 M. L-Smokers 28.5 57.4 100.3 11.2 24.9 Total (.•iKarr't-io 85 125 10 159 :17 110 10 6(i -Non-Smokers 5.8 5.3 73.4 3,!) 16.9 10 pvr dny Itl 116 3 3 * :3 * 1(1 • 10 per tlny I:i 117 a * 6 * 1 * Total (7HI)-,5mokers :3a.8 86. 1 152,0 11,2 6•1.7 90 prr day 30 Ill!1 I * 12 8R rr * ,Non-timokur+a 5.9 31.7 135 , 6 25.5 67.8 Over 20 per tlnv 32 155 I li 159 I * Unknown 2 * 1 * * w Adjacsted^ 8-Year 1nr.irl. 1Zatas/1000 -M. L--Smokers 28,0 55.4 82.1 14,4 30,5 * Not romllulrll if hlu; (hun 10 rnsrs ohcorvotl, -Non-Smokers 3,5 7.4 67.2 3,6 8,1 '1'ARLh: tt Total ('H1)--,Smokers 34.5 81.41 131.2 14.4 70.0 I''!t,\MIYiiII.\M PR@;LIMINARY };-N'I:AR tiMOKiNG I)A'1'A -Non-Smokers 3,5 30,8 11,37 23.1 44.4 8-1'H:An INrlur•,Wr; nr• ti1YnrnuulAl, INrARI°rln.N AMrING A'tALI0N AGx. 3t/- 51) ANu IeRM,1. OM' I)Itil,Atib, ON EN'rnv IN'r0'rnM S'rnnl Ratio of 1 ncid. Ila.tes- I?ntr/1(X)0 Sntokers/Non-Snwkers -M. I-Crude •1.1) 10.8 1.4 3.9 1,5 .I,rfr rd ISntrrtnrr lsxrnu.ina'iar .i0-,il1 40 41) Ill-.5:1 _-Adjttsterl" 5.11 7.5 1.2 4.0 :3.8 Non-ril,~nrrttoti»tokrrs :i.8 5.3 73.4 Total CHl)-Crude 5.8 2.7 1.1 0.4 1.0 Less than'30/rla)• 21.a (i4.4 78.4 -Adjuste+ 5.9 2.7 1.2 0.6 1.6 20 tutrl ovor/dnv 39.9 55 2 155.41 * Polnllalion flrv of ('111) al-initial exalruination. ^ Average of rates for the 8 high-low comhinations of cholesterol, hlood pressure alnl weight. T201126

374 Tobacco and Hcahh I Coronary Heart Disease Incidence and Mortality 375 ABLE 10 C:*J kl) M ~ TABLE 11 ' FRAAiUIORAM-PRELIMINARY 8-YEAR SMOICINO DATA ~ I' RAMIN(IIIAhi-PRELIAIINARY 8-YEAR SMOKINq DATA ~ Ilforlalilp Annonp Those Afle 30-62 and Free of Disease on Entry into the Study 0 Relationship of Smoking and Mortality from All Causes Among Those Age 60-69 and Free of Disease on Entry Into the Study Dealhs Frone All C¢Iases CHD-Dealhs DI a te f ern alr nra le Fem ~ ale H ~ m.nkinp Ca.le,rlrrry No. of tYISCs ob- scrverl Ratio nfob- xrrr~rrl to rx- pcclcd cases No. of ca.ecs ob- served Ratio I of ob- eenled to ex- pecled cases No. of 4/IRC5 ob- served ! Ratio of ob- served to ex- pected cases No. of cases ob- served Ratio of ob- served to ex- pected cases Total 120 10U 72 100 46 100 10 100 Never Smoked l 77 39 86 4 6 Lapsed Cigar-Pipe Only i 12 * 61 1 * 2 5 * * 0 * Tot.al Non-Cigarette :31 62 4(1 84 11 60 6 * Total Cigaretlo dll 122 29 118 33 121 3 a i (1 Pnr Day IO 105 I) * .1. * 1 # L(1-t9 1'or 1)n.v 13 107 a .1 . 2 20 Per 1)ay aa I()(1 11 Over 20 I'rr 1)ny :15 154 .I I ti 190 Unknown I , .1 2 * I ` Not comlruted if Icxs I,hnn 10 ca.ses ohscrved, Males Females 80-39 40-49 60-69 40-49 60-69 Basic Data-Cigarette Smokers Deaths 15 26 38 10 8 Population Size 562 489 329 357 201 I3aAlc n2tta-Non-CIQ',ar(:ttd'- Smokers (exrlndm lapxce) Deaths 1 2 17 7 22 Population Size 171 1811 177 51(1 531 Crude 8-Year Mortality Itates/10110 Cigarette Smokers `16,7 53.3 115.5 28.0 39.8 Non-Cig. Smokers 5.8 10.6 96.0 13.7 41,4 AdjuetedR 8-Year Mortality Rates/1000 Ciaarette Smokers 24,2 53.4 101.5 28,8 33,2 Non-Cig. Smokers 3,5 10.6 86.6 11.9 37.4 Rat.io of Mortality Rates Cig. 8mokem/Non-Smnkere Crude 4,6 5.0 1,2 2,0 1.0 Adjusted 6.9 5.0 1.2 3.4 0.9 a Average of rates for the 8 high-low combinations of cholesterol, blood pressure and weight. T201127

376 Tobacco an.cl Health ~ Coronary Heart Disease I ncidence and Mortality 377 ln TARLf': 12 ~ l TABLE 13 Nrxrrn I)AKO9.1 CARONARY 1)ISxnsr:STnnY (1057) r ~ NORTH DAKOTA CORONARY DISEASE STUDY C:ASER (lY' I)I:+EARE AMr1N0 hZALES ACE 35-64 AND I'REN OF ~ CASES OF DISEASE AIriONO THOSE AOE 35-64 AND FREE OF l)rSP•.ASF. ON IiNTRY INTr) THT. STUDY DISEASE ON ENTRY INTO THE STUDY .lt yorardial Infarrlion or /-1 ncclirsion* All Coronary Hcart DiseaRe** ~ Farm,ers Ratio Ratio H Ar(/Exp ~ Act/F x Myocardial Infarction or . p Slnokinq l'atrqm•y .IrGral Erprrled X 100 Arbr.al Expected X 100 Occlusion All Coronary Heart Disease Total 5:i 53. () 1(10 R3 83.0 100 Ratio Ratio Nevrr srnokr•d 12 19.6 (31 21 31.3 67 Act/Exp. Act/Exp. 1 apwod 5 6.9 7.1 4 10.5 7 (3 Smoking Category A ctual Expected X 100 Aclual Expected X 100 I'otal GiQnxetlc 131) 2.5 .9 112 4(3 40.4 114 Total 21 21.0 100 37 37.0 100 <21 per day 31 19. •1 108 :3O 30. !'i JR Never Smoked 9 10.1 89 13 17.9 73 <21 and nver 9 6.4 125 1(1) 9.11 162 Lapsed 2 2.9 ... 4 5.0 ... Ifiknow•n 7 11.4 ... R 0.7 ._. Total Cigarette 7 8.3 84 16 14.1 113 * Inr.ludrs drfinitr and prohahlo rnses. <21 per day 7 11 11.9 92 ** Inrludev all ditegno.stie catrgorieg of definite nnd probable dinr•arr: Angina 21 and over 0 ... ... 5 2.2 227 ltiectoris, coronary inqnlTirionr,y, myocirrdial infnrrl.ion nnd sudden rlrath dne to Unknown 3 0.0 ... 4 0.0 ... coronary honrt dirtrnso. Non-Farmers Total 32 32.0 100 46 46,0 100 Never Smoked 3 8.5 35 8 12.4 65 Lapsed 3 3.8 79 4 5,4 74 Total Cigarette 22 19.5 113 30 28.0 107 <21 per day 14 13.2 106 19 19.1 99 21 and over 8 6.3 127 11 8.9 124 Unknown 4 0.4 ... 4 0.6 T201128

4 378 Tobacco and Health n ~ TABLr 14 M ~ NORTIi DAICOTA CORONARY DISEASE STUDY RELATIONSHIP OF,5'MOKING ANI) OCCURRENCE OF CORONARY HEART DISEASE* ~ AHIONG MALES A(;E 35-64 AND FREF. OF 1)ISEASE ON TNTRY INTO STUDY ~ Category All Ages 6.5-49 60-h9 60-69 70-1- ~ Basic Data-Cigarette Snwkers Cases M. 1. 40 12 17 8 3 ~ Cases CHD 61 20 24 3 4 Population size 9190 5360 2300 1170 360 ~ Basic Data-Non-Cip. Snwkers (excludes lapses) Cases M. 1. 30 1 6 7 16 Cases CHI) 58 3 10 19 26 Population size 7950 2050 1750 2240 1910 Crude I ncidence Tlates/1,000 M. I. Among Smokers 4.4 2.2 7.4 6.8 8.3 M. I. Among Non-Smokers 3.8 0.5 3.4 3.1 8.4 Total CH D Among Smokers 6.6 3.7 10.4 11.1 11.1 Total CII I) Among Non-Smokers 7.3 1.5 5.7 8.5 13.6 Ratio of In.cidence Itatas- Srrlokers/Non-Suwkers M.I.-Crude 1.2 4.4 2.2 2.2 1.0 Total Ci-Ill-Cru(le 0.9 2.5 1.8 1.3 0.8 * Includes definite and probable cases. REFERENCES l. Beau, M. t1n memnire sur la fuml<e de tabac considerce comme cause de 1'angine de poitrine. Arch. GM. de M(fdecine, 20:122-123, 1862. (cited by Roth, ref. 2) 2. Roth, G. M. and Shick, R. M.: Effect of smoking on the cardiovascular system of man. Circ., 17:443-459, March, 1958. 3. Huchard: Traitc clinique des maladies du coeur et de i'aorte, ed. 3, Paris, Gaston Doin, 1899. (cited by P. D. White in ref. 4) 4. White, P. D., and Sharber, T.: Tobacco, alcohol and M. A., 102: 655-657, March 3, 1934. angina pectoris. J. A. 5. Graybiel, A., Starr, R. S., and White, P. D.: Electrocardiographic changes follow- ing the inhalation of tobacco smoke. Am. Heart J., 15:89-99, January, 1938. 6. Davis, F. W., Jr., Scarborough, W. R., Mason, R. E., Singewold, M. L., and Baker, B. M., Jr.: The effect of exercise and smoking on the electrocardiograms and ballistocardiograms of normal subjects and patients with coronary artery disease. Am. Heart J., 46:529-542, October, 1953. 7. Glendy, R. E,. Levine, S. A. and White, P. D.: Coronary disease in youth. Comparison of 100 patients under 40 with 300 persons past 80. J. A. M. A., 109:1775-1781, Nov. 27, 1937. , Coronary Heart Diserrse Incidence and Mottality 379 8. English, J. P., Willius, F. A., and Berkson, J.: Tobacco and coronary disease. J. A. M. A., 115:1327-1329, October 19, 1940. 9. Yater, W. M., Traum, A. Ii., Brown, W. G., Fitzgerald, R. P., Geisler, M. A., and Wilcox, B. B.: Coronary disease in men eighteen to thirty-nine years of age. Report on eight hundred sixty-six cases, four hundred fifty with necropsy examinations. Am. Heart J., 36:334-372, Sept, 1948. 10. Mills, C. A.: Tobacco smoking: Some hints of its biologic hazards. Ohio State M. J., 46:1165-1170, Dec., 1950. 11. Dolgoff, S, Schrek, R., Ballard, G. P., and Baker, L. A. Tobacco smoking as an etiologic factor in disease. 2. Coronary disease and hypertension. Angiology, 3:323-333, August, 1952. 12. Sigler, L. H.: Tobacco as a contributing cause of degenerative coronary disease. New York J. Med., 55:3107-3113, November 1, 1955. 13. Hammond, E. C., and Horn, D.: Smoking and death rates-Report on forty- four months of follow-up of 187,783 men. II. Death rates by cause. J. A. M. A., 166:1294-1308, March 15, 1958. 14. Hammond, E. C.: Smoking in relation to heart disease. Am. J. Pub. Health, 50: Pt. 2, 20-26, April, 1960. 15. Doll, R., and Hill, A. B.: Lung cancer and other causes of death in relation to smoking: A second report on the mortality of British doctors. Brit. M. J., 2:1071-1081, Nov. 10, 1956. 16. Breslow, L., and Buechley, R.: Factors in coronary artery disease; cigarette smok- ing and exercise. Calif. Med., 89:175-178, Sept., 1958. 17. Buechley, R. W., Drake, R. M., and Breslow, L: Relationship of amount of ciga- rette smoking to coronary heart disease mortality in men. Circ., 18:1085-1095, Dec., 1958. 18. Doin, H. R: Tobacco consumption and mortality from cancer and other diseases. Pub. Health Rep., 74:581-593, July, 1959. 19. Wynder, E. L, and Lemon, F. R.: Cancer, coronary artery disease and smoking: A preliminary report on differences in incidence between Seventh Day Ad- ventists and others. Calif. Med., 89:267-272, Oct., 1958. 20. Rigdon, R. H., and Kirchoff, H.: Smoking and disease; a study based upon 12,050 individuals. Texas Rep, on Biol. and Med., 16:116-132, Spring 1958. 21. Doyle, J. T., Heslin, A. S., Hilleboe, H. E., and Formel, P. F.: Early diagnosis of ischemic heart disease. New England 1. Med., 261:1096-1101, Nov. 26, 1959. 22. Dawber, T. It, Kannel, W. B., Revotskie, N., Stokes, J., III, Kagan, A., and Gordon, T.: Some factors associated with the development of coronary heart disease. Am. J. Pub. Health, 49:1349-1356, Oct. 1959. 23. Zukel, W. J., Lewis, R. H., Enterline, P. E., Painter, R. C., Ralston, L. S., Fawcett, R. M., Meredith, A. P., and Peterson, B.: A short-term community study of the epidemiology of coronary heart disease. Am. 1. Pub. Health, 40:1630-1639, Dec., 1959. T201129 I

26 CHRONIC BRONCHITIS, SMOKING AND AIR POLLUTION C. M. FLr•.Mirn, M.D., F.R.C.P. IN speaking of chronic bronchitis to an American audience it is essential to state what we mean by this term in the United Kingdom, for much of the evidence that smoking and air pollution are concerned in the causation of this disease has been obtained in my country. I have described the disease in greater detail elsewhere (Fletcher, 1959) and will now only emphasize that it may be regarded as having three main components. First, there is a hypersecretion of bronchial mucus, manifest as chronic cough productive of mucoid sputum, which may be described as simple bronchifis. It appears that this alteration in mucus secre- tion increases the liability to hronchial infection, since we find that suhjects with the greater volume of sputum have an in- creased liability to recurrent episodes of bronchial infection dur- ing which the spianm becomes purulent; while others may have persistent infection with continuously purulent sputi.im. These cases may he described as having chronic bronchitis with recur- rent or persistent i>>(ection. Lastly, a proportion of these cases have impaired ventilutory capacity and may be described as having chronic hronchitis with generali.sed airways obstruction which, whcn it is severe, causes dyspnoea. Such cases are often diagnosed as having emphysema, although significant anatomical emphysema is not always found at autopsy. Death may occur trom respirilt(1Ty or cardiac failure, usually during an exacerbation of infection. The wide range of clinical conditions covered by the term chronic broncl-iitis must lbe borne in mind when we consider 380 Chronic Bronchitis, Smoking and Air Pollution 381 evidence of the role of smoking and air pollution in its aetiology. The clinical diagnosis of chronic bronchitis is usually applied in the United Kingdom to cases with a combination of all the three parts of the syndrome. In the U.S.A. many of these cases would be labelled asthma or emphysema without mention of the associ- ated bronchitis. SMOKING AND CHRONIC BRONCHITIS Mortality Statistics We can obtain no direct evidence from mortality statistics about the relationship between smoking and deaths attributed to bronchitis, for information about smoking habits is not recorded in national censuses. Comparisons between mortality rates from bronchitis in different countries are liable to serious fallacies be- cause of differences in diagnostic customs from country to coun- try. Mork (1960), however, combining mortality in middle-aged men from bronchitis and pneumonia has found that, while there is only a slight relationship between recent mortality rates and cigarette consumption in countries for which information is avail- able (Fig. 1a), the relationship is closer, with a significant correlation, if cigarette consumption 20 years ago is considered (Fig. lb ). The extreme divergence in these figures between the United Kingdom, with a gross excess of mortality, and the U.S.A. where, in relation to cigarette consumption, mortality is very low, is noteworthy. While differences in diagnostic custom may account for some of this divergence, there must be other impor- tant factors involved here. More direct evidence has been obtained on the relationship be- tween mortality from bronchitis and cigarette smoking by prospective studies of selected population groups both in the United Kingdom and the U.S.A. Among British doctors, Doll and Hill (1956) found a steady gradient of bronchitis mortality with increasingly heavy smoking, those smoking more than twenty- five cigarettes a day having a bronchitis mortality six times greater than non-smokers. In the U.S.A., Dorn (1959) found that mortality from bronchitis and emphysema was 3.27 times as great in regular cigarette smokers as in non-smokers. Pipe and cigar smokers had a mortality no greater than non-smokers. T201130

382 Tobacco and Health CIGARETTE CONSUMPTION IN 1930-35 IN VARIOUS COUNTRIES IN RELATION TO AGE STANDARDISED MORTALITY PER 100,000 IN MEN AGED 40-64 FROM BRONCHITIS AND PNEUMONIA 1955 120 IIO ^ -- o60 < z 0 ~ W 50 z a C Z < Ln 40 r= 0•44 • ITALY .NEW ZEALAND • NETHERLAMDS • FRANCE • CANADA SWEDEN• • SWITZERLAND •NONWAY *DENMARK • BELGIUM • AUSTRIA U.K.* • FINLAND • U.S.A. I • ITALY • BELGIUM •FINUND • NEW ZEALAND • NETHERLANDS • AUSTRIA • FRANCE • CANADA •SWEDEN •SWITZERLAND •NORWAY •DENHARK 1 I I ~ I , 1 I I I II I i I( 0 200 400 600 800 1000 1200 1400 ANNUAL CIGARETTE CONSUMPTION PER ADULT 1930-34 Fig. la Fig. 1. Standm-dised mortality rates per 100,000 from Bronchitis and pneumonia in men aged 40-64 in various countries related to (a) Average cigarette consumption per adult 1950-1954. (b) Average cigarette con- sumption per adult 1930-1934. Morbidity Studies of "Chronic Bronchitis" Leese (1956), comparing the smoking habits of 100 patients admitted with chronic bronchitis to a hospital in West London with those of agc-and-sex-matched control patients without Chronic Bronchitis, Smoking and Air Pollution 383 CIGARETTE CONSUMPTION IN 1950-54 IN VARIOUS COUNTRIES IN RELATION TO AGE STANDARDISED MORTALITY PER 100,000 IN MEN AGED 40-64 FROM BRONCHITIS AND PNEUMONIA 1955 a ---- ~ - - - - - 0 60 r: 0•21 ~ W Z °' S0 G l_ U.S.A.0 (wMile) I 1 II I I 1 I 1 1 1 1 1 1 I 0 500 1000 1500 2000 2500 3000 3500 ANNUAL CIGARETTE CONSUMPTION PER ADULT 1950-54 Fig. lb bronchitis, found that nine per cent of the bronchitics compared with 22 per cent of the controls were non-smokers whereas 14 per cent compared with five per cent were heavy smokers. Ed- wards (1957) calculated from these data that those who smoked 20 or more cigarettes a day were four to five times more likely to be admitted to hospital with chronic bronchitis than non-smokers. In population studies "chronic bronchitis' defined in various T201131 . U.K.

384 Tobacco and Health PERCENTAGE OF MEN AND WOMEN AGED 40-64 DIAGNOSED BRONCHITIC BY G.P.S ACCORDING TO S M 0 K I N G H A B ITS (COLLEGE OF GENERAL PRACTITIONERS SURVEY 1957-58) 30 25 20 LU 0 a ~ z 15 W U 0:: W °- 10 5 MON 1-14 15-24 25+ SMOKERS G/DAY G/DAY G/OAY SM0KERS EX SMOKERS Fig. 2. Percentage of men and women aged 40-64 in random samples drawn from the practices of 92 General Practitioners in the United King- clom who were diai;nosccl as having chronic bronchitis by the practitioners after completing a symhtomatic qi.icstionnaire, in relation to their present tobacco consnmption (College of General Practitioners; 1960). ways, has been fomid to be more frequent in smokers than non- smokers by Oswald and Medvei (1955), Stnart-Harris and Hanley (1957), Edwards et al. (1959), Fletcher et al. (1959), and the College of General Practitioners (1960). Fignre 2 is derived from Chronic Bronchitis, Smoking and Air Pollution 385 the latter survey and shows that in comparable smoking groups the prevalence of chronic bronchitis in men and women is similar except in those smoking less than 15G/day, where the prevalence is much less in the women. In this group, however, there are many more very light smokers than among the men. Similar findings have been reported by Oswald and Medvei (1955) and Ogilvie and Newell (1957). It appears that the sex difference in prevalence of chronic bronchitis may be largely due to differences in smoking habits between the sexes. Fry (1954) appears to be the only investigator to have found no difference between the smoking habits of bronchitics and controls, but he did not separ- ate non-smokers from ex-smokers. To elucidate the association between bronchitis and smoking more clearly we must consider the relationship between the various components of the bronchitis syndrome and smoking. Smoking and Simple Bronchitis It is a matter of common experience, expressed in the phrase "a smoker's cough," that smoking causes cough and expectoration. The validity of this simple observation has been confirmed by a large number of investigations both in population samples and in hospital patients. These have shown that cough and expectora- tion are more prevalent in heavier than in lighter smokers (Green and Birkowitz, 1954; Palmer, 1954; Philips et al., 1956; Ogilvie and Newell, 1957; Fletcher et al., 1959; Higgins, 1959; Liebeschultz, 1959; Flick and Paton, 1959; Olsen and Gilson, 1960; College of General Practitioners, 1960). A varying propor- tion, up to 50%, even of the heavier smokers, however, are re- ported to deny any cough or expectoration. Common experience also shows that most, but not all cases of simple chronic bronchi- tis who discontinue smoking cigarettes lose their productive cough, and the population surveys all agree in showing that the prevalence of simple bronchitis in ex-smokers lies between that of light smokers and non-smokers (Fig. 3). Pipe and cigar smok- ers are found to be much less affected than cigarette smokers (Philips et al., 1956; Higgins, 1959; Olsen and Gilson, 1960). In a large proportion of people, therefore, cigarette smoke appears to act as a bronclual irritant causing cough and hyper- T201132

386 Tobacco and Health cf' ~ RELATIONSHIP BETWEEN PRESENT SMOKING HABITS ~ ~ AND PREVALENCE OF PRODUCTIVE COUGH,DYSPNOEA AND RECURRENT CHEST ILLNESSES IN MEN AGED 40-64 ~ O Nr.OF MEN ( COLLEGE OF GEBERAL PRACTITIONERS SU'RYEY 1957-SR ) NON-SMOKERS 1-14 15-24 54 G/DAY G/DAY S H 0 K E 307 189 25+ G/DAY R S 96 EX-SMOKERS 141 and the Fig. 3. Relationship between present tobacco consumption prevalence of productive cough, dyspnoea on walking on the level and recurrent chest illnesses in random samples of men aged 40-64 on the practice lists of 92 General Practitioners in the United Kingdom. (College of General Practitioners; 19'60). i Chronic Bronchitis, Smoking and Air Pollution 387 secretion of bronchial mucous, but it seems that the irritant effect is reversible in many cases. The acute effects of exposure •to cigarette smoke have been studied in experimental animals by Hilding (1957), Dahlmann (1958) and Falk et al., (1959). There is an out-pouring of mucus with a slowing of the ciliary move- ment, changes which, in man, would be expected to result in productive cough. The chronic effects in man consist of hyper- trophy of mucus secreting glands and bronchial metaplasia (Auerbach et al., 1957; Chang, 1957; Sanderud, 1958; Ide et al., 1959). The severity of the changes is greater in the heavier than in the lighter smokers, but a proportion of even the heavier smokers retain a normal bronchial mucosa. The increase of mucous glands is what would be expected in subjects with chronic productive cough and has been described as the essential feature of the pathology of chronic bronchitis by Reid (1960). Smoking and Recurrent Chest Illnesses and Dyspnoea Higgins (1959), in three samples of elderly men found that chest illnesses of a bronchitic character and dyspnoea on slight exertion were much more frequent in smokers than in non-smok- ers, but there was little difference between lighter and heavier smokers. Similar findings have been reported by Fletcher et al., (1959); Fletcher and Tinker (1960); College of General Prac- titioners (1960), (Fig. 3). Smoking and Impairment of Lung Function Disability from dyspnoea in chronic bronchitis is due chiefly to obstruction to air flow in the lungs. Although this is commonly attributed to emphysema, clear evidence of emphysema is often lacking so that the condition is best diagnosed in purely descrip- tive terms as generalized obstructive lung disease (Ciba Foun- dation, 1959). In the United Kingdom, apart from questions about dyspnoea, the tests used in epidemiological studies to measure this dis- turbance of function, have been forced expiratory volume ( F.E.V. ) in 0.75 or 1.0 seconds and peak expiratory flow rate, measured by the Wright peak flow meter (Wright and McKer- row, 1959). Population studies have uniformly shown an in- T201133

r 388 Tobacco and Health creased prevalence of dyspnoea and reduction of mean ventilatory capacity in cigarette smokers compared with non-smokers, with ex-smokers and pipe smokers having intermediate values (Higgins, 1959; Higgins et al., 1960; Fletcher et al., 1959; Fletcher and Tinker, 1960; College of General Practitioners, 1960). In sonie surveys, the heavier smokers have been found to have more dyspnoea and a lower ventilatory capacity than lighter smokers, but in general the average level in those two groups has not been found to be significantly different (Figs. 3 and 4). In a sample of Danish men in Bornholm, Olsen and Gil- son (1960) found a significant impairment of ventilatory capacity in cigarette smokers, while pipe and cigar smokers had, if any- thing, higher mean values than non-smokers. They considered that the difference in mean ventilatory capacity between Danish men and men in British rural areas might be explained largely by the greater cigarette consumption by the latter. Impairment of ventilatory capacity in smokers has also been observed in several American studies. Flick and Paton (1959), in a study of hospital patients, showed impainnent of mean peak expiratory flow rates in elderly smokers compared with non- smokers and also a significant decreasing trend of ventilatory capacity with increasing smoking among the smokers. Blackburn et al., (1959) measured lung volumes in 221 men aged 47 to 57 and fonnd a significant decrease of vital capacity and increase of residual volume (consistent with airways obstruction) in smokers who had respiratory symptoms, but not in those who had no symptoms. No clill'erence was found between lighter and heavier smokers. Wilson et al. (1960), in a detailed study of 14 male smokers of more than 20 cigarettes a day who had no respiratory symptoms, found, even in these subjects, significant impairment of ventilatory capacity and also of diffusing capacity in compari- son with 14 matchcd non-smokers. Previous to the observation reported by Dr. Nadel this morning, cigarette smoking appcared to have little immediate effect on ventilatory capacity according to observations by Bickerrnan and Barach (1J56); l'lich et al. (1957) and Attinger et al. (1958). Only a few subjects, preclominantly those with obstructive lung I I , Chronic Bronchitis, Smoking and Air Pollution 389 disease, showed some deterioration of ventilatory capacity after smoking one or two cigarettes. All the components of the bronchitis syndrome are more fre- quent and more fatal in smokers than in non-smokers. It is reasonable to assume, on a priori grounds that the relationship between cigarette smoking and simple bronchitis is one of cause and effect. We do not know what components of tobacco smoke are responsible for this effect, nor do we know why cigar and pipe smoke causes so much less irritation. There are chemical differences between these different kinds of smoke, but the chief difference may well lie in the simple fact that few pipe or cigar smokers inhale the smoke deeply. The evidence incriminating smoking as a cause of the disabling consequence of chronic bronchitis is perhaps rather less definite chiefly because the quan- titative relationship between disability and tobacco consumption, which would be expected if cigarette smoking itself increased liability to recurrent infection and obstructive lung disease, is lacking in many studies. It is possible, however, that some of those who are most seriously affected tend to reduce their ciga- rette smoking so that in surveys where current smoking habits are considered (which has been the general practice ) these peo- ple are classified in the lighter smoking group rather than that to which they had previously belonged. The heavier smokers may thus be a self-selected group of relatively resistant indi- viduals. Higgins (1960) has attempted to overcome this objec- tion by a prospective study in which be related changes in ven- tilatory capacity to cigarette consumption during a period of five years. In a sample of non-mining men, but not in a sample of miners, he found a greater rate of decline of ventilatory capacity in heavier than in lighter smokers although no difference was found between the mean ventilatory capacity of the two groups at the beginning of the survey (Fig. 4). AIR POLLUTION AND CHRONIC BRONCHITIS Vital Statistics , The gradient of bronchitis mortality from rural to urban areas T201134

390 Tobacco and Health RELATION BETWEEN VENTILATORY CAPACITY ~ ~ (INDIRECT MVV) AND SMOKING HABITS IN ~ NON-MINING MALES AGED 20-69 SURVEYED ~ IN 1954 AND 1959 (HIG6INS 1960) ~ O MMEAN IMMVV T 110 `MEAN OF WOTH SURVEYS) (L/MIN.) MEAN ANNUAL FALL IN INDINECT MVV SETWEEN SURVEYS (I./MIN.) Na OF MEW I0 45 26 EX SMOKERS Ifl Fig. 4. Relationship between ventilatory capacity (indirect maximum voluntary ventilation) and smoking habits in a random sample of non- mining men aged 20-69 in the Rhonddl Valley, South Wales, who were surveyed in 1954 and 1959. Above: mean of measurements at both surveys. 13elow: D1ean change between surveys. (Higgins; 1960). , . Chronic Bronchitis, Smoking and Air Pollution 391 in England and Wales (Table 1), which is approximately the same in both sexes, is often adduced as evidence of an effect of air pollution. But this gradient only shows a two-to-one ratio of mortality between the largest towns and the rural areas, although the contrast in pollution is far greater. Mere size of a town, how- ever, is a poor index of pollution and many attempts have been made to get more accurate indices. The most complete analysis of mortality data in relation to more detailed indices of pollution is that carried out by Daly (1959). He derived indices of air pollution by domestic and by industrial smoke in residential areas from figures of industrial and domestic fuel consumption, and from measurements of the built-up areas in 83 large towns in Great Britain. He expressed pollution as the amount of smoke produced per annum per acre of built-up area in each town. The domestic pollution indices in the study ranged from 17 to 80 units, and the bronchitis mortality in middle-aged men from 44 to 200 per 100,000 per annum so that there was a wide range of both pollution and mortality. The bronchitis death rates showed a highly significant correlation with both domestic and industrial smoke pollution. Bronchitis mortality was also significantly correlated with four indices of social status, but the correlation with pollution remained highly significant after allowance for these social differences. Daly did not find it possible in this analysis to distinguish between pollu- tion by smoke and by sulphur dioxide, but he made a separate analysis of pollution from power stations (which emit sulphur products but little smoke) and found no relationship between such pollution and bronchitis mortality. Pemberton and Gold- berg (1954), Stocks (1959) and Fairbairn and Reid (1958) have also shown highly significant correlations of bronchitis mortality with various other indices of air pollution in various areas of Great Britain. In the U.S.A. (Manos, 1957) deaths from emphysema show a rural-urban gradient similar to that for bronchitis in England and Wales. The chief difficulty about accepting these regional corre- lafions between bronchitis mortality and pollution as evidence of cause and effect is that it is impossible to be sure that the many immeasurable occupational, social, climatic and other variables T201135

392 Tobacco and Health that are associated to various degrees with exposure to pollution may not be largely responsible for them. It was this that led Reid (1958) to consider rates of sickness absence due to bronchitis among postmen and civil service clerks who form stable groups of men with uniform selection who do the same jobs for the same remuneration in various parts of the country. Reid derived an index of pollution from the frequency of fog in 37 towns of widely varying smokiness and lie divided these towns into four groups providing populations of equal size arranged in increasing severity of the fog index. He found that sickness absence rates from bronchitis, in contrast to any other causes, showed a uniquely steep gradient with increasing pollution, being nearly three times as great in the most polluted as in the least polluted towns. Further, when sickness absence rates among postmen, who are exposed in their work to the air of the towns in which they reside, was related to those among clerks, who work in- doors protected to some extent from pollution, the postmen/ clerk ratio for bro-.ichitis was found to increase steeply on passing from the least to the most polluted towns, while the ratio for other diseases was similar in all the areas. Bronchitis was the one illness whose incidence was differently increased among the postmen in the more polluted areas. While this does not prove that air pollution is the cause of the increased bronchitis mor- bidity in the postmen, it is hard to think of any other explanation. This study provides most convincing evidence that air pollution is an important cause of the high bronchitis morbidity and mor- tality in our polluted towns. Epidemiological Surveys In all studies of mortality and morbidity the disease index is a crude one derived from death or sickness certificates provided by doctors who may diffcr considerably in their diagnostic customs. This might impair or exaggerate contrasts between different areas if doctors in the less polluted areas were more or less liable to diagnose bronchitis than those in the more polluted areas. Nor do these crude ilidices reveal what part of the bronchitis syn- drome air pollution exerts its effects up'on. Some preliminary light is thrown on this by recent epidemio- A t , Chronic Bronchitis, Smoking and Air Pollution 393 PERCENTAGE OF MEN & WOMEN AGED 40-64 DIAGNOSED BRO'NCHITiC BY GA AND PERCENTAGE WITH MORNING SPUTUM, CHEST COLDS & DYSPN'OEA. (COLLECE OF GENERAL tRACnriON'ENS SURVEY 1'9S7-S0) SYNDROME G.P$ DIAGNOSIS tUNAL ANEAS 0•t7 SMALL TOWN'S 0.41 LAN'GE TOWNS 0•SS WOMEN 6 RURAL ANEAS 0•22 O.Es DIAGNOSIS fNONCNITN: SYNDROME SMALL TOWN'S 0•20 ~IE IANGE TOWNS 069 Fig. 5. Percentage of men and women aged 40-64 in random samples of the practice lists of 92 General Practitioners in various areas in the Unitea Kingdom who:-(a) were diagnosed as having bronchitis by the practitiondr after completing a standardised symptomatic questionnaire, and (b) admit- ted to regular morning sputum at least in the winter, recurrent chest colds and dyspnoea on hurrying, in their replies to the questionnaire. logical studies made in Great Britain using standardist'd ques- tionnaires about respiratory symptoms (Table 2). There are no important differences in the proportion of men who have chronic productive cough in the different areas. This symptom is closely related, as we have seen, to smoking habits and there are only small differences between town and country in tobacco consump- tion. Recurrent chest illnesses, however, are considerably more frequent in urban than in rural men. Significant dyspnoea was not found to be more prevalent in urban than in rural areas by Higgins and his associates, but a striking gradient from rural areas T201136

394 Tobacco and Health I Chronic Bronchitis, Smoking and Air Pollution 395 to large towns was found by the College of General Practitioners (1960). The higher prevalence of symptoms reported by the general practitioners may be due to differences in the tech- nique of asking questions about symptoms despite the use of standardised questions in all these surveys. There was evidence in the practitioners' survey that doctors in rural areas diagnose bronchitis in less severely affected cases than do their urban colleagues. Urban-rural comparisons based on general practi- tioners' diagnoses may thus underestimate the true differences (Fig. 5). Acute Effects of Pollution Contrast between urban and rural areas in respect of recur- rent chest illnesses rather than in respect of persistent cough and sputum suggests an episodic rather than a persistent chronic effect of air pollution. It is, of course, well known that the short periods of intense pollution which occur during periods of atmos- pheric inversion in our cities are associated with sharp increases of morbidity and mortality chiefly attributed to bronchitis. Recent episodes of this kind in London have been described by Logan (1953), Gore and Shaddick (1958), Bradley and Logan (1958) and Martin and Bradley (1960). Waller and Lawther (1956) have shown that men with chronic bronchitis tend to become ill during periods of increased pollution and that the number and severity of these illnesses are related to changes of pollution rather than to changes of temperature or humidity. We have no contemporaneous evidence that these peaks of pollu- tion initiate disabling bronchitis in subjects previously unaffected, hut this possibility is suggested by the experience of chest physi- cians in London who in recent years have seen many cases of advanced bronchitis who date their disability from the great smog of London in December, 1952. Leese (1956) found that 33 per cent of lDronchitic patients had become ill during this period* compared with only 2 per cent of the controls. DlSCUSSION AND CONCLUSIONS The association of morbidity and mortality from chronic brony chitis with smoking (in particular cigarette smoking) which I 1 ~f~~r: ~ vKi&b~~Vfi:tlx;ai due4nqAyvt**:hiSbar_ suh"trs: The demonstration of impaired lung function in smok- ers without symptoms by Wilson et al. (1960) does, however, suggest that smoking may damage the lung independently of the complications of simple bronchitis. The contrast between the U.S.A. and Great Britain in respect of bronchitis mortality is perhaps relevant in this connection. There is some evidence (Pemberton, 1956; Flick and Paton, 1959) that simple bronchitis may be not much less prevalent in American than in British smokers, and yet the experience of clinicians who have worked on both sides of the Atlantic sug- gest that there is a very real difference in the prevalence of bron- chitis with recurrent infection and obstructive lung disease. Clearly something else than smoking is at work. Berkson (1958) has suggested that the lower mortality of non-smokers from various causes compared with smokers may be due to their being "biologically disposed to self-protective habits." This might well be relevant to the complications of simple bron- chitis: for neglected respiratory infections might be more likely to produce chronic disability. This may be a factor in the steep social class gradient of bronchitis mortality in the U.K. (Fletcher, 1958). I doubt whether American smokers are much more self- protective in their habits than English smokers and I think we must look to environmental contrasts to explain the difference. Air pollution is the obvious factor to incriminate: for few of your cities suffer from the levels of pollution found in our smoky towns. It is a reasonable hypothesis that, during periods of high pollution, cases of simple bronchitis are more liable than normal subjects to some acute irritative effect on the bronchi which, by altering the bronchial secretions in some way, may encourage the onset of acute infective bronchitis and that such illnesses may both exacerbate the chronic simple bronchitis and produce a permanent increase of resistance to bronchial air-flow. The T201137

V 396 Tobacco and Health validity of this hypothesis might be tested by prospective studies of the times of onset, nature and consequences of such recurrent infections in our city dwellers. Such studies are now being planned. But if this hypothesis were correct it would be difficult to reconcile with the impression one has that even in cities in the United States, such as Pittsburgh, which were heavily polluted until relatively recently, there does not appear to have been a prevalence of chronic disabling bronchitis in them comparable to that found in our cities. It has recently been reported that in Sheffield, 40 per cent of cases of heart failure admitted to hospital have chronic pulmonary heart disease secondary to chronic bron- chitis (Stuart-Harris et al., 1959). It seems to me unlikely that such a high prevalence could fail to have been noticed in your cities which have been heavily polluted sufficiently recently for the effect to be stilll noticeable. It is most desirable that we should obtain more direct evidence of the true contrasts in the prevalence of simple chronic bron- chitis and its complications between our two countries, in both urban and rural areas. IIypotheses based on impressions await the test of accurate observation. Much might be learnt from studies of the types of heart failure entering all the hospitals serving areas of contrasting experience of pollution, and from epidemiological surveys of chronic respiratory disease in repre- sentative samples of the general population in such areas, which, if the same standardised technique were used, could be directly compared with the results of surveys recently carried out in Great Britain, Scandinavia and Finland. Such studies might provide many of the facts we need to advance our understanding of a disease whose importance as a cause of chronic disability is outstanding in my country and may be more important over here than current vital statistics suggest. Smoking and air pollution are not the only agents concerned in the aetioingy of chronic bronchitis and its complications. Combining the results of Iliggins (1957), Higgins and Cochran (1958) and the College of General Practitioners (1960 ), an esti- mate of the prevalence of chronic bronchitis with recurrent in- fection in rural non-smoking males aged 40 to 64 can be obtained of approximately 5 per cent. Detailed clinical study would be Chronic Bronchitis, Smoking and Air Pollution 397 TABLE 1 ~ STANDARDISED MORTALITY RATIO OF MORTALITY FROM BRONCffiTI8 IN ~-I ENOLAND AND WALEB 1951-53 p"I ACCORDINO TO RESIDENTIAL AREA O ~ ~ Standardised Mortality Ratio Area Male Female Conurbations 131 126 Greater London 123 117 Urban areas over 100,000 107 101 Urban areas 50,000-100,000 90 96 Urban areas less than 50,000 85 84 Rural districts 62 72 TABLE 2 PREVALENCFJ OF RESPIRATORY SYMPTOMS IN MEN AGED 55-64 IN vARI0U8 POPULATION SAMPLES IN GREAT BRITAIN Percentage with Ile1ui.l Two or Number Cough Breathless- More Loca- of and ness Clrest "Chronic Author tion Man, Sputum Grade 3-{-* Illneasest Bronchitis•'$ Higgins (1957) Higgins and Rural 173 22.5 8.7 4,6 6.4 Cochran (1958) IIiQRins et al. ^ Urban 165 24.8 9.1 12.7 '. 12.7 (1956, 1959) ~~ Rural 66 34.9 10.62 7.6 9.1 Small 1M1 " 33.3 21.5 8.3 15.3 College of reneral towns I'ract.itioners (1960) Large 108 34.3 37.9 16.7 23.2 towns * Men who have to walk slower than others on the level owing to breathlessness. t Illnesses with an increase of phlegm during previous three years. $ Usual cough and sputumandatleastonechest illness in previous three vPA--m T201138

398 Tobacco and Health required to elucidate the nature of the disorder in these cases, but smoking and air pollution are not responsible for them. Nevertheless, the evidence incriminating smoking and air pollution as major contributory causes of chronic bron- chitis is sufficiently clear to justify action. Smoke control mea- sures have been widely applied over here and are beginning to be effective in a small way in my country. But we still await some real attempt by public health authorities in any country to attempt to control the modern pandemic of cigarette smoking. In the field of chronic respiratory disease this is perhaps the greatest challenge facing preventive medicine today. REFERENCES 1. Auerbach, 0., Gere, J.B., Forman, J.B., Petrick, T.G., Smolin, H.J., Muehsam, G.E., Kassouny, D.Y. and Stout, A.P. (1957): Changes in the bronchial epithelium in relation to smoking and cancer of the lung. New England J.Med., 256, 97. 2. Attinger, E.O., Goldstein, M.M. and Segal, M.S. (1957): The effects of smok- ing on the mechanics of breathing. (1) In normal subjects. (2) In patients with cardiopulmonary disease. Amer.Rev.resp.Dis., 77, 1 and 10. 3. Berkson, J. (19,i8): Smoking and lung cancer. Some observations on two re- cent reports. 1. Amer.statist.A., 52, 28-38 4. Bickerman, H.A. and Barach, A.L. (1954): The effect of cigarette smoking on ventilatory function in patients with bronchial asthma and obstructive pul- monary emphysema. J.Lab.Clin.Med., 43, 455. 5. Blackburn, I-I., Brozek, J. and Taylor, H.L. (1959): Lung volume in smokers and non-smokers. Ann.intern.Med., 51, 68. 6. Bradley, N.H. and Logan, W.P.D. (1958): The London fog of December 2nd-5th, 1957. Mon.Bull.Min.Hlth., 17, 156-166. 7. Chang, S.C. (1957): Microscopic properties of whole amounts and sections of human bronchial epithelium of smokers and non-smokers. Cancer, 10, 1246. 8. Ciba Foundation (1959): Terminology, definition and classification of chronic pulmonary emphysema and related conditions. Thorax, 14, 286-299. 9. College of General Practitioners. (1960): Chronic Bronchitis in Great Britain: A national survey carried out by the respiratory disease study group of the College of General Practitioners. In preparation. 10. Dahlman, T. (1956): The effect of cigarette smoke on ciliary activity in the upper respiratory, tract. A.M.A.Arch.Otolar fng,, 70, 166-168. 11. Daly, C. (1959): Air pollution and causes of death. Brit.J.prev.soc.Med., 13, 14. 12. Doll, R. and Hill, A.B. (1956): Lung cancer and other causes of death in rela- tion to smoking: second report on the mortality of British doctors. Brit.med,J., 2, 1071-1081. Chronic Bronchitis, Smoking and Air Pollution 399 O 13. Dorn, H.F- (1959): Tobacco consumption and mortality from cancer and other ~ diseases. U.S.Publ.Hlth.Rep., 74, 581-593. ~ 14. Edwards, J.H. (1957): Contribution of cigarette smoking to respiratory disease. ~ Brit.J.prev.soc.Med., 11, 10. ~ 15. Edwards, F., McKeown, T. and Whitfield, A.W.W. (1959): Association between O smoking and disease in men over sixty. Lancet, 1, 196-200. ~ 16. Eich, R., Gilbert R. and Auchincloss, J.H. (1956): Effects of smoking on respira- tory mechanics in chronic pulmonary emphysema. Clin.Res.Proc., 4: 151. ~ 17. Fairbairn, A.S. and Reid, D.D. (1958): Air pollution and other local factors in ~ 18. respiratory disease. Brit.J.prev.soc.Med., 12, 94-103. Falk, H.L., Tremer, H.M. and Kotin, P. (1959): Effect of smoke and its con- I 19. stituents on ciliated mucus secreting epithelium. J.Nat.Cancer Inst., 23, 999. Fletcher, C.M. (1958): Disability and mortality from chronic bronchitis in rela- 20. tion to dust exposure. A.M.AArch.indtutr.Hlth., 18, 308. Fletcher, C.M. (1959): Chronic bronchitis: its prevalence, nature and patho- 21. genesis. Amer.Rev.resp.D'u., 80, 483-493. Fletcher, C.M., Elmes, P.C., Fairbairn, A.S. and Wood, C.H. (1959): The sig- 22. nificance of respiratory symptoms and the diagnosis of chronic bronchitis in a working population. Brit.med.J., 2, 257-266. Fletcher, C.M. and Tinker, C.M. (1960): Chronic bronchitis: a further study 23. of simple diagnostic methods in a working population. In preparation. Flick, A,L. and Paton, R.R. (1959): Obstructive emphysema in cigarette smokers. 24. A.M.A.Arch,intern.Med., 104, 518. Fry, J. (1954): Chronic bronchitis in general practice. Brit.med.J., 1, 190-194. 25. Gore, A.T. and Shaddick, C.W. (1958): Atmospheric pollution and mortality 26. in the county of London. Brit.J.prev.soc.,Med., 12, 104-113. Green, B.A. and Berkowitz, S. (1954): Tobacco bronchitis, an anesthesiologic 27. study. Ann.intern.Med., 40, 729. Higgins, I.T.T. (1957): Respiratory symptoms, bronchitis and ventilatory capac- 28. ity in a random sample of an agricultural population. Brit.med.J., 2, 1198. Higgins, I.T.T. (1959): Tobacco smoking, respiratory symptoms and ventila- tory capacity: studies in random samples of the population. Brit.med.J., 1, 325-329. 29. Higgins, I.T.T. (1960): Ventilatory capacity in miners: a five-year follow-up study. In preparation. 30. Higgins, I.T.T. and Cochran, J.B. (1958): Respiratory symptoms, bronchitis and disability in a random sample of an agricultural community in Dumfries- shire. Tubercle, 39, 296-301. 31. Higgins, I.T.T., Cochrane, A,L., Gilson, J.C. and Wood, C.H. (1959): Population studies of chronic respiratory disease. A comparison of miners, foundry work- ers and others in Staveley, Derbyshire. Brit,J.industr.Med., 16, 255. 32. Higgins, I.T.T., Oldham, P.D., Cochrane, A.L. and Gilson, J.C. (1956): Respira- tory symptoms and pulmonary disability in an industrial town. Brit.med.J., 2,904. T201139

400 Tobacco and Health 33. Hilding, A.C. (19.56): On cigarette smoking, bronchial carcinoma and ciliary action. II. Experimental study on the filtering action of cows' lungs, the depo- sition of tar in the bronchial tree and removal by ciliary action. New England J.Med., 254, 1155-1159. 34. Ide, G., SuntzelF, V. and Cardly, E.V. (1959): Comparison of the histopathology of tracheal and bronchial epithelium oF smokers and non-smokers. Cancer, 12, 473. 35. Leese, W.L.B. (1956): An investigation into bronchitis. Lancet, 2, 762. 36. Liebeschutz, H.J. (1959). Respiratory signs and symptoms in young soldiers and their relationship to smoking. J.R.Army M.Corps., 105, 76. 37. Logan, W.P.D. (1953): Mortality in the London fog incident, 1952. Lancet, 1, 336-338. 38. Manos, N.E. (1957): Comparative mortality among metropolitan area of the United States, 1949-51. 102 causes of death. U.S.Publ.Hlth. Service Publ., No. 562. 39. Martin, A.E. and Bradley, W.II. (1960). Mortality, fog and atmospheric pollu- tion. Mon.Bull.Min.Hlth., 19, 56-72. 40. Mork, T. (1960): Ph.D. Thesis. University of London (unpublished). 41. Ogilvie, A.G. and Newell, D.J. (1957): Chronic Bronchitis in Newcastle-upon- Thyne. E. & S. l.iving.stone, Edinburgh. 12. Olsen, I-1.C. and Gilson, J.C. (1960) Respiratory symptoms, bronchitis and ven- tilatory capacity in men: an Anglo-Danish comparison with special reference to smoking habits. Brit.med.J., 1, 450-456. 43. Oswald, N.C. and Medvei, V.C. (1955): Chronic bronchitis: the effect of ciga- rette smoking. Lanoet, 2, 843-8•14. 44. Palmer, K.N.V. (1954): The role of smoking in bronchitis. Brit.med.J., 1, 1473. 45. Pemberton, J. (19 v6): Chronic bronchitis, emphysema and bronchial spasm in bituminous coal workers. A.M.A.Arch.industr.Hlth., 13, 529. 46. Pemberton, J. and Goldberg, C. (1954): Air pollution and bronchitis. Brit.mad.J., 2,567. 47. Phillips, A.M., Phillips, R.W. and Thompson, J.L, (1956): Chronic cough: an analysis of etiologic factors in a survey of 1274 men. Ann.intern.Med., 45, 216. 48. Reid, D.D. (1958) Environmental factors in respiratory, disease. Lancet, i, 1289. 49. Reid L. (1960): Measurement of the bronchial mucous gland layer: a diagnostic yardstick in chronic bronchitis. 'Thorax. 15, 132-141. 50. Sanderad, K. (1958): Squamous metaplasia of the respiratory tract epithelium. Acta Path.ct Mirrobud.Scand,, 43, 47. 51. Stocks, P. (1959) Cancer and bronchitis mortality in relation to atmospheric deposit and smoke. Brit.mrd.J., 1, 74-79. 52. Stuart-Harris, C.I•I. and llanley, T. (1957): Chronic Bronchitis, Ernphysema and Cor Pulmonale. John Wright and Sons Ltd. Bristol. Chronic Bronchitis, Smoking and Air Pollution 401 ~ ~ 53. Stuart-Harris, C.H., Twidle, R.S.H. and Clifton, M. (1959): A hospital study M of congestive heart failure with special reference to cor pulmonale. Brit.med.J, ~ 2,201. ~ 54. Waller, REZ and Lawther, P.J. (1957): Further observations on London fog. ~ Brit.med.J., 2, 1473-1475. ~ 55. Wilson, RH., Meador, R.S., Jay, B.E. and Higgins, E. (1960): The pulmonary ~ pathologic physiology of persons who smoke cigarettes. New England J. Med., 262, 956. C.B. (1959): Maximum expiratory flow rate as a M. and McKerrow Wright B 56 F 1 , . , . ~ measure of ventilatory capacity. Brit.med.J., 2, 1041-1045. T201140

IF, 27 INTERRELATIONSHIP OF SMOKING AND PEPTIC ULCER PHILIP CooPLR, M.D. and HAxoLn K. Tsuji, M.D. EVALUATION of this problem is very difficult be- cause of the paucity of reported experimental and clinical investi- gation, and the difficulty of evaluating experimental or clinical investigation because of the many factors and variables involved in the genesis, treatment, and healing of the peptic ulcer. To add to its complexity, clinical impressions, not subjected to critical or scientific analysis, have been accepted, by many, as established facts. Review of the available literature on the effect of smoking on gastric physiology revealed some interesting and provocative reports. Most were based on limited investigation with little or no scientific valid data or suitable and adequate controls. A number of investigators as I. Gray,' Rosenblum,Z Ehrenfeld and Sturtevant,' Steigman et al.,4 Piper and Raine,° reported an increase in gastric acidity after smoking in patients with peptic ulcer. Schnedorf and Ivy,^ Hodges and Gilmour,° and others reported no increase of gastric acid after smoking. Toon, Cross and Wangenste.en,' in 1951 reported a study on dogs with tracheostomies. They found that histamine in beeswax, or cigarette smo'.ce alone produced no peptic ulcers. Both agents combined, however, did produce ulcers. They also noted that an isolated gastric pouch showed no change in the volume and free acid of gastric secretions following smoking. The limited and controversial nature of the investigation stimulated our interest, and as a result, in 1956, we reported on a 402 Interrelationship of Smoking and Peptic Ulcer 403 controlled study of cigarette smoking on patients with duodenal ulcer.° There were 60 "smokers," and 60 "controls." Gastric secre- tions during the basal, smoking and control periods, and the post smoking or post control periods were studied. Comparison be- tween the "smokers" and the "controls" revealed no significant differences in the volume, pH, free acid, peptic concentration or peptic output of gastric secretions. Of interest was a further limited study on some pre- and post-operative duodenal ulcer patients. This included 27 patients, 11 had had a subtotal gas- trectomy, and 16 patients were being prepared for such surgery. The pH of the gastric juice was determined for the basal, smok- ing, broth stimulation and insulin stimulation periods. It was found in both the pre- and post-operative groups that the pH was elevated during the smoking period. We subsequently accomplished further studies to determine the possible effect of smoking on dissolved gastric mucins and the viscosity of gastric juice.l° The study group was composed of 30 subjects, 20 with uncomplicated duodenal ulcer, and 10 without gastro-intestinal disease. They were equally divided into two groups, "smokers" and "controls: " Under the conditions of the study smoking did not significantly alter the concentration of the dissolved gastric mucins, or the relative viscosity of gastric juice. Further studies were accomplished on the effect of smoking on intragastric balloon pressure and intragastric temperature of patients with duodenal ulcer.ll Previous works on the effect of smoking on gastric peristalsis had not reported consistent findings. It was felt that alterations in intragastric temperature may reflect changes in blood flow. Twenty patients were in each phase of the investigation. They were evenly divided between "smokers" and "controls:' Comparison of the two groups revealed that ciga- rette smoking caused no significant difference in the number and duration of intragastric balloon pressure waves, and resulted in no significant differences in intragastric temperature. Twenty patients without evidence of gastro-intestinal disease were later found to have a statistically significant decrease in the concentration of plasma pepsinogen during a smoking period that followed a period of abstinence.'a Very little is found in the literature in regard to smoking as an T201141

404 Tobacco and Health etiological factor in peptic ulcer. H. T. Gray" implied that smok- ing was a predisposing factor, whereas Barnett" expressed the opposite view. The literature is more extensive on the effect of smoking on the healing or recurrences of peptic ulcer. Dworken et al.,ls Brown et al.,'° state that there is no effect. Many authors as Bat- terman and Ehrenfeld,'r Doll et al.,'R Jones,79 and Friedrick,' report that smoking does have an effect. Mitty, Rousselot and Delany=° in a report on 171 post-gastrectomy patients stated that there was no relationship between smoking and weight gain or dumping, but noted that of the eight patients who had late complications all were smokers. Jordan, Bolton and DeBakey,Z' in a similar study on post-gastrectomy patients found that the proportion of patients who smoked was no greater among the group with poor results than among those with good results. Marginal ulcers which developed however, were in smokers. To further study the findings and to evaluate the impressions of others a qnestionnaire, was sent to 368 physicians primarily interested in gastro-intestinal diseases. Ninety-two questionnaires were completed and retnrned, and 16 replies by letter were received, making a total of 108 replies. Thirty-nine questionnaires were not completed because of the death or retirement of physi- cians, or the lack of forwarding addresses. Some form of response was thcrefore obtained from 147 of 368. We recognized the marked limitations and value, in general, of a questionnaire. It was, however, prepared to allow one to report scientific data and impressions based on material available to each physician. It was of interest that most of the information was based on impressions. This fact was readily acknowledged by most of those completing the questionnaire. There was little information based on or reporting scimntific ohjectivc data. The replies to the question on the number of patients who smoked at or -,vithin a short period before the onset of ulcer were too few for evaluation. The remainder of the information obtained from the questiomnaires is presented in the following tables. Because of incomplete answers or omissions of answers to certain questions, figures in the tables are presented as absolute rather than as percentage figures. Interrelationship of Smoking and Peptic Ulcer 405 TABLE I SMOKING AND PEPTIC ULCER No. Replies to Questionnaire-92 Increased Decreased No Effect No Opinion Gastric secretion and acidity 54 1 14 9 Gastric Motility 42 11 12 13 TABLE II SMOKING AND PEPTIC ULCER No. of Replies to Questionnaire-92 Increased Decreased No Effect No Opinion Appetite 0 77 4 4 Nervousness 26 20 24 5 Weight 1 67 10 3 TABLE III SMOKING AND PEPTIC ULCER No. Replies to Questionnaire-92 Recommendations on Smoking: Continue Limit Stop Limit or Stop Limit and Stop Continue and Limit TABLE IV SMOKING AND PEPTIC ULCER No. of Replies to Questionnaire-92 A. Etiological Factor in Development of Ulcer Yes-21 No-60 B. Symptom.n of Ulcer In general Occasionally Rarely Never Aggravates 50 25 14 3 Relieves 1 7 33 0 C. Interferes or Delays Healing of Ulcer In general 42 Oacssionally 21 Rarely 17 Never 5 T201142

407 Interrelationship of Smoking and Peptic Ulcer 406 Tobacco and Health As expected comments sent by letter varied. Many physicians felt that there was an association between smoking and the aggravation of ulcer symptoms or the disease itself. Others did not concur. Very few however, felt that smoking was a factor in the etiology or the production of a peptic ulcer. In general most of the comments in letters stated or implied that patients with peptic ulcers should stop smoking. Some stated that smoking was a manifestation of nervousness which many patients with ulcer display, and that the stoppage of smoking may not have its anticipated effect. Those answering by letter, in general, emphasized that their views and recommendations were also based on personal impres- sions. Where reference to specific studies were made it did appear that the relationship between smoking and ulcer became more controversial. One example of comments relative to investigative work, re- ported in the replies is presented. B. Crolu,2stated that years ago he practiced fractional test meals on control cases and on cases of peptic ulcer, duodenal and gastric, with the idea of establishing the influence if any of cigarette smoking. Neither on the control nor on the ulcer cases was the secretory gastric curve influenced by smoking. Rarely did he feel that smoking influenced the course of prognosis of ulcer cases, then only in excessive or chain cigarette smokers. On only 3-4 occasions has abrupt cessation of cigarette smoking led to sub- sidence of symptoms. To evaluate the problem from a patient's standpoint, 242 patients with peptic ulcer, being treated on a medical regimen, were requested to fill out a questionnaire. At the time, 190 patients were smoking, and 52 were not. The answers were re- viewed in a very critical manner. It is of interest however, to note the tabulation of two categories of answers. The first on reasons for smoking and the. second on the effect of smoking. It is noted that, among the "Reasons for Smoking," pleasure or relaxation was the most evident, and habit was the next most prominent reason. On "The Effect of Smoking," it was found that most patients who continued smoking had no evident effects, and that they as a group found that there was a lessening of nervous- ness while smoking. ~ M TABLE V ~ .5'MOSINO AND PEPTIC ULCER ~ Non-Smokere 0 (at Present) Smokere ~ Eeasons for Smoking Gives pleasure 69 Patients 19 190 Patients 104 ~ Relaxes 16 142 Stimulates 2 9 ML Decreases appetite 11 34 r Keeps one busy 2 23 I Is a habit 2 79 Pleasant aroma 10 43 Some patients reported multiple reasons TABLE VI SMOKING AND PffiPTIC ULCER Non-Smokers Effect of Smoking No effect Causes nausea Causes stomach pain Increases nervousness Lessens nervousness (at Present) Smokers 52 Patients 190 Patients 10 79 9 23 7 14 10 15 9 93 Some patients reported multiple effects The various approaches taken to evaluate this problem have led to a realization that very little valid adequate scientific inves- tigation has been accomplished to the present to justify final statements on the effects of smoking on gastric physiology. Certainly we cannot assume that smoking increases gastric acidity or mortality, nor is it a factor in the etiology of peptic ulcer. Why non-smokers, or smokers who refrain from smoking, appear to heal their ulcers more rapidly cannot be explained on present available information. An indirect rather than a direct effect on the stomach is possible. The patient who smokes heavily may disregard other advice relative to treatment and therefore may not respond as well. The smoker, may represent a different constitutional type. Smoking by reducing the appetite may reduce the amount of food intake available for buffering the gastric acid. Smoking without the diminution of food intake, or at T201143

408 Tobacco and Health time of or following meals may conceivably have little or no effect on ulcer symptomatology. There is need for considerable scientific investigation in this field. It will be of great interest, and at the same time, may resolve many other problems involved in the etiology, treatment and healing of the peptic ulcer. REFERENCES 1. Gray I.: Ann. Int. Med., 3:267, 1929-30 2. Quoted by Kraemer M. and Lehman D.: Gastroenterology, 8:202, 1947 3. Ehren[eld L and Sturtevant M.: Am. J. Med. Sci., 201:81, 1941 4. Steigman F., Dolehide R.H., Kanusiski L: Am. J. Gastro., 22:399, 1954 5. Piper D.W., Raine J.M.: I,a.ncet, 1:5, 1959 6. Schnedor[ V.G., Ivy A.C., JA.M.A., 112:898, 1939. 7. Quoted by: Wynder E.L. The Biologic Effect of Tobacco, Little, Brown & Co., Boston, 1955, Chapt. V 8. Toon R.W., Cross F.S., Wangensteen O.H.: Proceedings of the Society for Experi- mental Biology and Medicine, 77:866, 1951 9. Cooper P., Knight, J.: New England J. Med., 255:17, 1956 10. Cooper P., Saltz M., Harrower H., Burke D.: Gastroenterology, 33:959, 1957 11. Cooper P., Harrower H., Stein I-L, Moore G.: Gastroenterology, 35:176, 1958 12. Cooper P., Stein H., Moore G., Harrower H.: Rhode Island Med. J. XL:215, 1957 13. Gray, H.T.: Brit. Med. J., 1:1040, 1924 14. Barnett C.W.: Boston Med. and Surg. 1, 197:451, 1927-28 15. Dworkcn H.J., Roth H.P., Duber H.C., Berger D.G.: Gastroenterology, 33:880, 1957 16. Editorial by W.G.: Canad. Med. A., 78:959, 1958 17. Batterman R.C., Ehrenfeld L: Gastroenterology, 12:575, 1949 18. Doll R., Jones F.A., Pygott.: The Lancet, 1:657, 1958 19. Jones A.: Brit. Med. J., 2:238, 1958 20. Mitty F., Rousselot L.M., Delany G.: Ann. Surg. 150:76, 1959 21. Jordan C.L., Bolton B.F., Debakey M.E.: J.A.I'VI.A., 161:1605, 1956 22. Crohn, B.B., Personal Communication. T201144

Recommended To Increase Your Professional Ef fectiveness George E. Burch and Nicholas P. DePasquale - HOT CLIMATES, MAN AND HIS HEART. E. Grey Dimond - THE EXER-_ C I S E ELECTROCARDIOGRAM IN OFFICE PRACTICE. 180 pp. (11 x 8 1/2), 338 il., $ 10.00 Hans Kraus and Wilhelm Raab - HYPOKINETIC DISEASE: DIS- EAS_ ES PRODUCED BY LACK_ OF EXERCISE. 204 pp., 146 il., $ 7.50 It Irvine H. Page - HYPERTEN- SION: A Manual for Patients with High Blood Pressure. 2nd Ed., 128 pp., 5 il., $ 3.00 Travis Winsor - PERIPHERAL VASCULAR DISEASES: An Ob- jective Approach._ 864 pg.-r 435 large drawings, $ 16.50 Grace M. R o t h - T O B A C O AND THE CARDIOVASCULAR SYSTEM. 72 pp., 12 il., (Amer. Lec. Circulation), $ 2.25 John S. Chapman - THE ANONY- MOUS MYCOBACTERIA IN HUMAN DISEASE. 184 pp., 18 il., $ 7.50 Seymour Iv[. Farber - CYTO- LOGIC DIAGNOSIS OF LUNG CANCER. 90 pp. (8 1/2 x 11), 60 il. in full color, $ 7.50 Seymour M. F a r b e r-' N G CANCER. 157 pp., 29 ii. (6 in full color), (Amer. Lec. Chest Diseases), $4.75 Ted F. Leigh and H. Stephen W eens - T H E MEDIASTINUM. _ 256, pp. (7 x 10), 440 il., $ 11.50 Richard H. Meade - A HISTORY OF THORACIC SURGERY. 960 pp., 117 il., $ 27.50 William W. Mushin, et al. - AU- TOMATIC VENTILATION O F THE LUNGS. 366 pp., 240 il., $10.50 R u d o l f N i s s e n and R. H. L. Wilson - PAGES IN THE HIS- TORY OF CHEST SURGERY. 176 pp., 90 il., $7.50 Franz Bergel - CHEMISTRY OF ENZ YMES IN CANCER. 136 pp., 76 il., (Amer. Lec. Living Chem- istry), $ 5.50 R. Lee Clark, Jr. - CANCER CHEMOTHERAPY. 268 pp., 14 il., (Amer. Lec: Living Chem- istry), $10.50 Gilbert H. Fletcher and William S. MacComb - RADIATION THERAPY IN THE MANAGE- MENT OF CANCERS OF THE ORAL CAVITY AND ORO- P H A R Y N X. 408 pp., 346 il., (Amer. Lec. RadiationTherapy), $ 16.50 M. Lederman - CANCER OF THE NASOPHARYNX: Its Natural History and Treatment. 128 pp., 109 ii., (Amer. Lec. Radiation Therapy), $ 6.75 CHARLES C THOMAS • PUBLISHER • SPRINGFIELD • ILLINOIS J TIMN 0116376

. To Increase Your ta.....v..~....,... . . Professional Ef fectiveness George E. Burch and Nicholas P. DePasquale - HOT CLIMATES, MAN AND HIS HEART. E. Grey Dimond - THE EXER- c' T S E ELECTROCARDIOGRAM IN OFFICE PRACTICE. 180 pp. (11 x 8 1/2), 338 il., $10.00 Hans Kraus and Wilhelm Raab - HYPOKINETIC DISEASE • DIS- EASES PRODUCED BY LACK OF EXERCISE. 204 pp., 146 i1., $ 7.50 Irvine H. Page - _HYPERTEN- SIO1`T: A Manual forPatientswith High Blood Pressure. 2nd Ed., 128 pp., .5 iL, $ 3.00 Travis Winsor - PERIPHERAL Y,ASCULAR. DISEASES: An Ob- Lective Approach. 864 pp.T 435 ,'' large drawings, $16.50 , ; . . ,r . Gxace M. Roth - TOBACCO_ "',' AdD THE CARDIOVASCULAR SYSTEM. 72 pp., 12 il., (Amer. Lec. Circulation), $ 2.25 John S. Chapman - THE ANONY- MOUS MYCOBACTERIA IN HUMAN DISEASE. 184 pp., 18 'il.,$7:50 Seymour M. Farber - CYTO- LOGIC DIAGNOSIS OF LUNG CANCER. 90 pp. (8 1/2 x 11), 60 i1. in full color, $7.50 Seymo M. Farbe r-' NG CANCER": r 157 pp., 29 il. (6 in full color)'., (Amer. Lec. Chest Diseases), $ 4.75 Ted F. Leigh and H. Stephen W eens - T H E MEDIASTINUM 256 pp. (7 x 10), 440 il., $11.50 Richard H. Meade - A HISTOic,l OF THORACIC SURGERY. 96.0 pp., 117 ii., $ 27.50 William W. Mushin, et al. - AU- TOMATIC VENTILATION OF THE LUNGS. 366 pp., 240 il., $10.50 Rudolf Nissen andR. H. L. Wilson - PAGES IN THE HIS- TORY OF CHEST SURGERY. 176 pp., 90 i1., $ 7.50 Franz Bergel- CHEMISTRY OF ENZYMES IN CANCER. 136 pp., 76 il., (Amer. Lec. Living Chem- istry), $ 5.50 R. Lee Clark, Jr. - C A N C E R CHEMOTHERAPY. 268 pp., 14 il., (Amer. Lec: Living Chem- istry), $10.50 Gilbert H. Fletcher and W illiam S. MacComb - R A D I A T I O N 1~ THERAPY IN THE MANAGE- MENT OF CANCERS OF THE ORAL CAVITY AND ORO- P H A R Y_N X. 408 pp., 346 ii., (Amer. Lec. Radiation Therapy), $16.50 M. Lederman - CANCER OF THE NASOPHARYNX: Its Natural History and Treatment. 128 pp., 109 il., (Amer. Lec. Radiation Therapy), $ 6.75 CHARLES C THOMAS • PUBLISHER • SPRINGFIELD • ILUNnic T201146 613.8 J