Tobacco Documents Online

A Scientific Perspective The Tobacco Institute 1875 1 Street, Northwest Washington, D.C. 20006

..~ . ... .•.a.. ...•h.~..._....w..1....n..........w:w .... . .... e.......w..a..-. l..i....n.... .~... Summary of Contents Overvi ew . . . . . . . . . . . . . . . . . . . . . . . . . . page 1 Cancer of the Lung . . . . . . . . . . . . . . . . . . . . . page 1 1 The assertion that cigarette smoking is the cause of lung cancer ignores basic, unresolved questions about laboratory data, smoking patterns and mortality rates, diagnostic var- iations and other confounding factors. Cancer of the Esophagus . . . . . . . . . . . . . . . . . . page 33 Interpretation of a statistical relationship between smoking and esophageal cancer must involve considerable guesswork, because the association is inconsistent on a worldwide basis and among Western nations. Cancer of the Larynx . . . . . . . . . . . . . . . . . . . page 39 Data do not warrant a conclusion that smoking causes laryngeal cancer: Population studies provide insufficient information, and disease patterns and smoking trends do not fit a causal hypothesis. Cancers of the Oral Cavity and Pharynx. . . . . . . . . . . page 43 Suspected factors in the workplace, in the environment and in the diet have been statistically related to the development of oral and pharyngeal cancers. Causation has not been determined for these diseases. Cancer of the Pancreas . . . . . . . . . . . . . . . . . . . page 53 Epidemiological research in the U.S. and abroad has suggested many possible risk factors, including smoking, but no specific environmental factor has been consistently associated with the development of cancer of the pancreas. Cancers of the Urinary Tract . . . . . . . . . . . . . . . . page 63 Inconsistencies in mortality trends in bladder and kidney can- cers, associations with diet and occupation and suggestions of other factors requiring investigation leave unanswered ques- tions about reported links with smoking. Cancer of the Lung in Nonsmokers . . . . . . . . . . . . . . page 75 Emotion must not be allowed to obscure the fact that claims linking cigarette smoke to lung cancer in nonsmokers have not been proven. Mortality Data: How Reliable? How Accurate? .......page 91 Extreme caution is warranted in evaluation of epidemiological studies based on mortality data -- such as those used to con- demn smoking -- because errors resulting from clinical misdiagnoses and recording mistakes can affect the accuracy of the mortality data and any conclusions drawn from them. 1

The Board of Editors welcomes suggestions for future editorials that succinctly summarize current work toward a clearly defined hypothesis regarding the causes or cure of cancer. Journal of the National Cancer Institute February 1981 Uncertainty about the causes of cancer is strikingly reflected by the vast amount of public and private funds spent on basic cancer research, the scientific debate on theories of causation and the frequent suggestions that common things in our daily lives might cause cancer. Yet it is not surpris- ing that cancer receives so much attention because it is one of the most serious problems in biology. Unfortunately, solu- tions have not and will not come easily. Only well-conceived and objective scientific research can lead to the resolution of the problem. Despite the recognized limitations in our scientific understanding of cancer, claims often are made that smoking causes the disease. Foremost in the litany of charges against smoking is that it causes lung cancer. In recent years, alle- gations also have included sites ranging from the pancreas and bladder to the kidney and larynx. 1

This document, while not intended to be exhaustive, is an attempt to provide a balanced perspective on some of the issues in the smoking and cancer controversy. The discussions in the following chapters rely on scientific and medical literature to point out why no firm conclusions can be drawn. Unknown Mechanism For Carcinogenesis Even after many years of investigation and millions of research dollars, scientists still are unable to describe the mechanism by which normal cells are transformed into malig- nant, or cancerous, cells. A biomedical scientist noted this fact in the Journal of the National Cancer Institutel in August 1981: A fundamental unsolved problem in the field of cancer biology is the nature of the primary event leading to the production of abnormally proliferating transformed cells. In particular, whether tobacco smoke plays any role in cancer causation is still undetermined. Even a scientist who accepts the conclusion that smoking is causally related to lung cancer wrote in 1979: "Indeed, no mechanism for human tobacco carcinogenesis has yet been successfully formulated and tested."2 2

Animal Studies One of the most common approaches to the study of cancer causation is the exposure of laboratory animals to pos- sible causative agents. Studies of animals exposed to tobacco smoke have produced inconclusive results. Despite considerable efforts scientists essentially have failed to produce in animals the kind of lung cancer most often associated with smoking. In summing up the experimental work in this area, a physician at Yale University said that "no well-designed and well-conducted experiments have shown that cigarette smoke causes lung cancer in animals."3 Epidemiological Studies Advocates of the theory that smoking causes cancer claim substantial support from epidemiological studies. Epi- demiology is a statistical science -- the study of a group of people for both the occurrence of a disease and the detection of factors that might be related to it. Epidemiology deals with statistical relationships and comparisons. It cannot determine cause. As an American Medical Association president once explained, "A statistic is a fact -- the result of a survey -- and that's all it is. Conjectures made on such a statistic are not facts. They are conjectures."4 3

Yet a common mistake made in the interpretation of epidemiological data is the identification of a statistical association as a causal connection. Statistics can never prove a cause and effect relationship; as the AMA official noted, "Statistics pose questions, they don't answer them." What the statistical associations reported in epidemio- logical stu dies can do is su ggest possible leads for further investigation. For example, statistical associations have led to suggestions that diet, place of residence and personality characteristics may be related to cancer. So while epidemiology as a science has its role, it also has serious limitations. The shortcomings of cancer epidemiology in particular were noted emphatically in a state- ment by a prominent biostatistician: "I know of no other scientific activity that's been so naive, so inappropriate, and so unscientific in its lack of attention to the basic data."5 Cancer Rate Patterns The Surgeon General's reports on smoking and health have cited various prospective and retrospective population stu dies to support the claim that a causal link exists between cigarette smoking and cancers of the lung, bladder, kidney, esophagus, larynx, pharynx, pancreas and other sites. Close 4

examination of incidence and mortality rates, however, reveals many patterns that are irreconcilable with the hypothesis that cigarette smoking causes these diseases. Under this hypothesis, disease (or death) rates would be related to smoking patterns or tobacco consumption in a specific way. For example, after tobacco smoking increased (or decreased) in a population, a measurable increase (or decrease) in disease rates would be predicted some years later. This time interval is called the lag period. This refers to the time from first exposure to an agent to the clinical appearance of the disease. Estimates of lag periods published in some scientific papers for smoking and various cancers have ranged from approximately 20 to 40 years. The causal theory cannot, for example, explain the varied patterns in bladder, oral/pharyngeal and esophageal cancer charted in Figure 1.6 Incidence rates of these diseases seem to lack any coherent pattern. They rise, fall or remain stable, depending on disease, gender and race. Even if one considers the so-called lag period, the varied trends still could not be explained by the hypothesis that cigarette smoking is the causal agent. Cancer and the Workplace The view that cancer may be relate d to environmental hazards in the workplace has received increasing attention N U1 0 ~ .~ 4h. N `0 ~ ~ 5

Figure 1 Incidence Rates* per 100,000 Population from Second National Cancer Survey (SNCS) (1947-49) and Third National Cancer Survey (TNCS) (1969-71) SNCS 24 21 -h 18 + 9 -1- 6 -t- 3-h TNCS SNCS TNCS SNCS TNCS 0-' CANCER ORAL CAVITY SITE & PHARYNX 0 0 e I BLADDER ESOPHAGUS Q- white male A - nonwhite male p- white female 0 - nonwhite female N U1 O ~-+ -P ~ N -0 *Age adjusted to the 1950 U.S. population standard v N (6) 6

recently. Research reports suggesting that certain per- centages of cancers are attributable to occupational exposures have stimulated considerable interest in the media. An edi- torial in a recent issue of the Journal of the National Cancer Institute addressed the "need for a systematic approach to environmental carcinogenesis in the workplace." The authors advocated a new epidemiological method to study occupational exposures suspected of increasing the risk of human cancer.7 In some discussions of workplace exposures and cancer, tobacco smoking, alcohol consumption, diet and other lifestyle characteristics are mentioned as possible co-factors. Some investigators have placed the major portion of the blame for the increased risk of cancer reported in some industrial workers on smoking. However, other researchers have expressed concern that focusing on a worker's smoking habit may serve to divert attention away from the hazards of the workplace.8 Mortality Data In order to study the possible relationship between factors, like smoking, and specific disease, researchers generally use information on the causes of death in population groups. Unfortunately, such mortality data often contain errors and inaccuracies. This is especially true in cancer stu dies because of mistakes in diagnosis and record-keeping. W 7

The reliability of death certificates is a major con- cern because they are the main source for mortality data. A group of scientists summarized the problem by saying that death certificates "are sometimes materially inaccurate and research based on them alone may not be secure.°9 A more detailed discussion of the reliability and accuracy of mortality data appears in a subsequent chapter. Conclusion Questions about cancer causation are complex, as demon- strated by divergent theories and the uncertain meaning of research findings. A director of the National Institutes of Health alluded to the problem in a statement to Congresslo a short time ago: In terms of the jig-saw puzzle analogy, medi- cal researchers do not know whether the blue piece they have found is part of the sky, the lake in the background, the evening dress of the girl on the porch or the body of the stationwagon in the driveway. And sometimes they manage to fit it very convincingly into what later turns out to be the wrong place. Independent scientific research, while not always pro- ductive, must continue to be supported in order to close the gaps in current knowledge. In January of 1982, the tobacco industry's research commitment in the smoking and health area 8

will surpass $100 million. Only with such continued research can the pieces of the cancer "puzzle" be put together. N cn 0 ~ ~ ~ N 0 ^.t Ln 9

References for Overview l. Straus, D., "Somatic Mutation, Cellular Differentia- tion, and Cancer Causation," J Natl Cancer Inst 67(2): 233-241, August, 1981. 2. Enstrom, J., "Rising Lung Cancer Mortality Among Non- smokers," J Natl Cancer Inst 62(4): 755-760, April, 1979. - 3. Feinstein, A., Comments in "Discussion" of presentation by P. Burch, "Smoking and Lung Cancer: The Problem of Inferring Cause (With Discussion)," J R Stat Soc A 141 (Part 4) : 437-477, 1978. 4. Appel, J., "The Numbers Game," Address, Newsmakers Luncheon, Overseas Press Club, New York City, New York, June 23, 1965. 5. Anonymous, "Is Ca Epidemiology More Than Merely Gazing Into Entrails?," Internal Med News: 2, February 15, 1981. 6. Devesa, S. and D. Silverman, "Cancer Incidence and Mortality Trends in the United States: 1935-74," J_ Natl Cancer Inst 60(3): 545-571, March, 1978. 7. Siemiatycki, J., et al., "Discovering Carcinogens in the Occupational Environment: A Novel Epidemiologic Approach," J Natl Cancer Inst 66(2): 217-225, February, 178I: - 8. Sterling, T., "Does Smoking Kill Workers or Working Kill Smokers? or The Mutual Relationship Between Smoking, Occupation, and Respiratory Disease," Int J Health Serv 8(3): 437-452, 1978. 9. Medical Services Study Group of the Royal College of Physicians of London, "Death Certification and Epidemiological Research," Brit Med J: 1063-1065, October 14, 1978. 10. Lamont-Havers, R., Statement, U.S. Congress, House, Subcommittee of the Committee on Appropriations, Departments of Labor and Health, Education and Welfare Appropriations for 1976, Hearing, 94th Cong., lst Sess., April 9, 1975 (Washington: Government Printing Office, 1975), pp. 6-7. 10

Cancer of the Lung Foremost among the charges against cigarette smoking is the assertion that it causes bronchial carcinoma, or lung cancer. Few have not heard that allegation. An American Cancer Society president even declared some time ago that "without smoking we would not have lung cancer."1 That assertion belies the facts. Lung cancer appears in nonsmokers as well as smokers. Moreover, there is an indication that lung cancer seems to be increasingly occurring in nonsmokers -- especially in males.2 In an April 1979 report on this phenomenon, a Cali- fornia epidemiologist who believes that smoking does cause lung cancer conceded that factors besides cigarette smoking must have had a significant effect on the tality rate.2 lung cancer mor- Certainly, a higher proportion of the cases in lung cancer stu dies are smokers. But such findings cannot estab- lish cause. They can only raise questions -- ones that remain unanswered in the dilemma of smoking and cancer. Animal Experiments A major question about lung cancer is why experiments in which.animals supposedly mimic human smoking repeatedly 11

have failed to produce the type of lung cancer for which smoking is most often blamed. The Surgeon General's report of 1968 admitted this deficiency. Using a variety of animals, large-scale inhala- tion studies have, it said, "essentially failed in producing squamous cell cancer of the lung."3 Today, the situation remains basically unchanged4 -- despite much-publicized efforts to promote findings reported from a beagle inhalation experiment5 as laboratory proof that cigarettes cause lung cancer. Indeed, one of the in- vestigators in that experiment announced in the press that he and his colleagues had "closed the circle in linking cigarette smoking and lung cancer."6 Shortly afterward, the methodol- ogy and findings of the dog stu dy were questioned by other scientists, inclu ding members of a special National Academy of Sciences panel.7-10 Further doubt was cast on the project with subsequent reports that the dogs might have been diseased before the experiment began.ll-12 Other animal experiments have been cited as proof that smoking causes lung cancer. These entailed daubing shaved animal skin with tobacco smoke condensate (tar), which is the particulate matter collected in smoking machines by passing smoke through a cold trap at extremely low temperatures. r•l) cn 0 ~ ~ 4N N -o J m 12

That experiments like this cannot be equated with the human smoking experience should be obvious, even to the lay- man, for at least three reasons: • Rabbit ears and the backs of mice are not the same as human lung tissue. • Inhalation differs markedly from "paint- ing " • The chemical properties of the so-called tar in such experiments may be quite dif- ferent from the smoke inhaled by smokers. Then, too, tar quantities used in many of the experi- ments were completely unrealistic. The amount has been de- scribed as the equivalent of an individual smoking 100,000 cigarettes a day.13 One researcher went so far as to say such experimental evidence was "claptrap."14 Even the 1979 Surgeon General's report rather gru dgingly conceded that "con- siderable criticism" has been directed at these studies.15 Statistical Studies Many of the reported findings used as "evidence" that cigarette smoking causes lung cancer come from epidemiologic studies. There are two primary experimental methods used in these stu dies. A retrospective study selects a group of lung 13

cancer patients, and a group without lung cancer, then back- tracks in time to determine smoking histories. A prospective study identifies a specific population group, supposedly free of disease, records their smoking habits and then observes the disease patterns that arise. Both kinds of studies arrange the raw data into various categories and run statistical tests to determine if one group (smokers) differs from another (non- smokers) in specific disease experience. Such statistical studies can provide a great deal of information, but, it is important to remember, as one eminent statistician has said, that "cancer is a biologic, not a statistical, problem."16 Basicall y, statistics gathered in these ways can prove nothing. They can indicate the likeli- hood that any observed patterns are not due to chance. But even "statistically significant" results may have no biologi- cal relevance. These epidemiological studies, moreover, can always be affected by inherited tendencies to develop certain diseases, the undiscovered effects of occupational or other environ- mental exposures and many other behavioral and biological unknowns. That is why this concession from the Surgeon General's 1979 report15 is important: 14

Most large scale stu dies on smoking and health have tended to investigate the role of smoking independently of other behavioral variables, such as alcohol consumption and other life style factors, occupational and environmental hazards an d certain psychologi- cal factors. These variables are known to be related to health status. Emphasis added] Thus, such statistical studies, no matter how large or covering how many years, "do not permit statements about causality. They can only point to differences between the observed groups."l7 Irregularities have been pointed out18-20 in the major epidemiological studies2l-26 that have been used to condemn cigarettes. Their precise meaning may be unclear, bu t they show, obviously, that the case against cigarettes is not as simple as some would suggest. A recent comment on the subject27 would appear appro- priate: Those epidemiological studies that purport to show a causal connection between cigarette smoking and various cancers, but particularly lung cancer, fail when examined critically to establish the causalclaim. Emphasis added] Mortality Rates vs. Consumption Levels One large dilemma for the cigarette causalists is the inconsistency of cigarette consumption patterns with lung 15

cancer rates around the world. Among the developed nations, the U.S. and Canada have among the highest per capita ciga- rette consumption figures in the world. But they rank 8th and 15th, respectively, in lung cancer mortality for males. Con- versely, while the United Kingdom, Finland and the Netherlands have lower per capita consumption than the U.S. and Canada, they have substantially higher lung cancer death rates. These examples are based on cigarette consumption and bronchopulmonary cancer rates in the middle 70's, the latest available for comparison internationally.28-29 Even allow- ing for a lag period, the trends in cigarette consumption going back to the 1930's are inconsistent with national mor- tality rates. Increased Mortality -- Real or Apparent? Recently, there have been cries of "epidemic" in those nations where sharp rises in lung cancer have been reported. Anti-smokers point to these reported increases and the simul- taneous rise in the prevalence of smoking as "proof" that smoking causes cancer of the lung. Such a simple explanation is no doubt appealing to them, but is it right? The biggest problem in the unquestioning acceptance of this simplism is the kind of data on which the cries of large-scale "epidemics" are based. Most of it is derived from 16

death certificates, a less than perfect source, to say the least. (For a discussion, see the chapter on mortality data.) Another development -- improved medical technology -- also presents problems for this argument. In the early part of this century, doctors had limited capabilities for diagnos- ing lung cancer in living patients. Not until after 1930 did the main clinical tools -- X ray, bronchoscopy and sputum cytology -- become available to diagnosticians.30 The "epi- demic," then, may reflect medical progress. But medical progress can become a two-edged sword: Clinicians can overuse their new tools. A Swedish medical team hinted as much in a report in 1975. The scientists, who believe that smoking is related to lung cancer, discussed the clinical and pathologic aspects of more than 700 primary lung cancer cases autopsied in a university teaching hospital over 11 years. They said it is clear that until recently lung cancer was frequently not diagnosed. They added that the increased interest in cancer in general and in bronchial car- cinoma in particular had le d to a more frequent recognition of it during the patient's life and, in some cases, even when it wasn't present.31 In other words, they said: It is obvious that the frequency figures based on clinical methods are too low in earlier series, whereas in recent years the figures may be too high because of over- 17

diagnosis. This m~ give a false impression of the real increase. Emphas si added) Changing Patterns Even if one assumes for the moment that part of the "epidemic" is real, the trends in lung cancer death rates still cannot be explained satisfactorily by smoking patterns. For example, statisticians have suggested that the rate of increase in lung cancer mortality may be declining. Others have been quick to suggest that changes in cigarettes and smoking habits have caused the decline. In 1975, a British thoracic surgeon asked whether such changes could explain the shifting lung cancer patterns. He said that this explanation seemed "likely" until one realized that the lung cancer changes actually began 50 years ago, long before the altera- tions in the smoking habit. He conclu ded that the observed patterns merely represented the "natural history" of the disease.32 Lung Cancer in Women Reports of rapidly increasing lung cancer death rates ~ for women have sparked the claim that increased smoking in o ~ .~ women is responsible. p N) ~ oa -P 18

But some scientists disagree. As one Canadian re- searcher commented in the American Journal of Public Health, "Much more significant than changes in women's smoking habits have been the changes in their employment.°33 More women who develop lung cancer now are women who work,34 supporting the belief of the longtime director of the National Cancer Institute's environmental cancer program that not enough attention has been paid to their on-the-job exposures. He has written that the adoption of smoking by women cannot explain their lung cancer patterns here or in other countries.35 There also seems to be no explanation for the statisti- cal findings of a British medical physicist who carefully examined lung cancer mortality patterns of males and females in England and Wales over the past century. He reported a "remarkable synchrony" in the recorded changes for males and females for the past 70 years. Thus, although the lung cancer rates always have been higher for males than for females, the patterns of increase over time have been almost exactly the same. Women's consumption of cigarettes, he argued, increased markedly 30 years after the striking rise in men's consump- tion. Thus, he observed, if smoking caused an increase in 19

women's mortality, there should be "kinks" in the patterns. That is, the female lung cancer mortality patterns should not parallel those of men in time. They should instead be out of phase by a 30-year period. He found no such kinks. This reasoning from recorded secular trends, perhaps difficult for the layman to follow, was presented to a meeting of fellows of the Royal Statistical Society and published in the Society's prestigious journal.36 In the paper, the researcher specifically remarked about the escalation of lung cancer rates for women: The most striking sustained rise in the in- crements of mortality for both sexes covers the period 1916-1920 to 1931-35, when ciga- rette smoking king can have made virtua~Iy no con- tr Eution to the large increase. [Emphasis ad ed ^ Other Factors The same type of information that is cited about ciga- rette smoking and lung cancer suggests that many factors may be involved in the development of lung cancer. These include occupational and environmental hazards, behavioral variables and "certain psychological factors" mentioned in the 1979 Surgeon General's report.15 20

Occupational Exposures Occupational exposure has been strongly related sta- tistically with increased lung cancer risk. • Excessive rates have been found37 in workers in petroleum and petroleum refining. • Navy personnel in construction and deck work38 have been reported to have up to three times the lung cancer incidence of other Navy enlisted men. • National Cancer Institute researchers have found excessive lung cancer mortality rates in counties where paper, chemical, petroleum an39t40ansportation industries are located. Excessive risk of lung cancer has been found in such nonindustrial occupational groups as dental technicians, decorators, electricians, bar and restaurant managers and broadcasting employees.41 One authority on occupational cancer cautioned that harmful industrial exposures could continue if public atten- tion were concentrated too strongly on cigarette smoking.35 His concern is shared by others, including an expert on en- vironmental contaminants,42 who wrote: A real question exists if cigarette smoking is not diverting attention from the effects of occupational exposure on industrial workers. ( 21

The Environment Industrial contaminants are not restricted to the worker in the plant, however. Consider NCI's identification of high lung cancer death rates in counties with certain in- dustries.39-40 Further suggestion that airborne industrial substances can affect area lung cancer death rates comes from two studies in the same issue of Archives of Environmental Health in 1980.43-44 Louisiana Gulf Coast parishes have been found to have some of the highest recorded lung cancer death rates in the country.45-46 A study of one such parish yielded a clue that residential exposure to industrial effluents may be a factor in the elevated lung cancer mortality rates in Louisiana.44 The researchers found excess risk with residential proximity to some industries (petroleum refining, lumber and metal manufacturing and canning) but not to others (soap and furniture making), which led them to comment: This observation validates the assumption that factors such as cigarette smoking are not responsible for the observed residential risks, and that in this urban area, cigarette smoking patterns are not related to distance from industry, and therefore, not confounding. Dietary Factors N vi 0 ~ -P .p N O ~ ~ The newest development in lung cancer epidemiology made headlines in 1981 -- the appearance of additional support for 22

a theory that carotene, a nutrient in dark green and dark yellow vegetables and fruit that is converted in the liver to vitamin A, may reduce the risk of lung cancer. Animal work in the 60's had suggested that a vitamin A- enriched diet reduced the incidence of cancer. Two small human studies in the mid-70's provided more support. One, a prospective study of men with known dietary habits,47 and the other, a biochemical investigation of serum A levels in patients and controls,48 had suggested that lower levels of dietary vitamin A may increase the risk of lung cancer. So did a 1979 retrospective dietary and smoking study at Roswell Park Memorial Institute in Buffalo, N.Y.49 and a preliminary report in 1980 of the investigation of serum A levels in British lung cancer cases.50 The new evidence in November 1981 was data from a large U.S. industrial health study.51 Although it was based on only 33 lung cancer deaths, the finding of an inverse rela- tionship of carotene with lung cancer incidence in a popula- tion of some 2,000 middle-age men suggests a protective quality in dietary vitamin A. The risk generally decreased with increasing levels of the nutrient -- in smokers and nonsmokers. 23

The researchers believe that cigarette smoking in- creases the risk of diseases in addition to lung cancer and were cautious of any possible interpretation of their data as showing that carotene protects smokers. One of them conceded in an interview that the number of lung cancer cases was too small to draw any conclusion other than that the more carotene one eats, the less statistical risk there is of developing lung cancer.52 The researcher said that scientists do not know how dietary vitamin A might protect the body against cancer, and another scientist, the director of Purdue University's cancer center, told a reporter: "What's really necessary is a good, controlled stu dy to prove that it's really what we think it is. I feel uncomfortable with any of these things until we know the mechanism."53 All of the researchers appear to agree that these investigations of dietary factors warrant further study. ~ Psychological Factors a F." 4h, -p N Increasingly, those who would ferret out causes of ~ 0 disease are looking to the minds of patients as well as to their genetic makeup. The journal Lancet referred recently to "a small body of research which suggests that certain emo- tional characteristics" may enter into the development of 24

cancer.54 This evidence has been accumulating since the 1960's, when a Scottish researcher compared a group of lung cancer patients who were smokers with a matched group of smokers clinically free of the disease and found the lung cancer patients significantly more likely to deny and repress their emotions. Such interpersonal difficulties, he wrote, could be important in the development of lung cancer.55 Further research in this area has indicated that psy- chological factors can be as much as twice as important as a history of smoking as a pre dictor that a patient with undiag- nosed chest disease will be found to have lung cancer,56 That finding supports the work of other investigators suggest- ing that psychological factors precede the development of lung cancer.57 Constitutional Factors William Osler, the Canadian-born physician, teacher and medical historian, once wrote: "It is much more important to know what sort of patient has a disease than what sort of a disease a patient has."58 The father of modern statistics, the late Sir Ronald .Fisher, suggested in the 1950's that constitutional factors might be far more important than smoking or any other variable in lung cancer.59 Only recently, a prominent medical

physicist reviewed much of the literature on smoking and lung cancer, studied time trends of the sexes and concluded that Fisher's hypothesis is still valid.60 Conclusion The claim that cigarette smoking causes lung cancer has not been proven scientifically. The charge ignores basic unresolved scientific questions concerning animal experiments, smoking patterns and lung cancer rates, diagnostic variations and many confounding factors. Lung cancer is an extremely complex disease. A one- sided attack on cigarette smoking as the causal agent does nothing to advance the search for its cause -- and its cure.

References for Lung Cancer 1. Pollard, H., Interview, "The John Scott Show," WOR-TV, New York, N.Y., February 5, 1975. 2. Enstrom, J., "Rising Lung Cancer Mortality Among Non- smokers," J Natl Cancer Inst 62(4): 755-760, April, 1979. 3. U.S. Public Health Service, The Health Consequences of Smokin . 1968 Supplement to the 1967 Public Health ~ Serv~.ce Revi.ew, Department of Healt i, Education and We are, PHS Publication No. 1696, 1968. 4. Furst, A., Statement, U.S. Congress, Senate, Committee on Labor and Public Welfare, Subcommittee on Health, Cigarette Smoking and Disease, 1976 Hearing, 94th Cong., 2nd Sess., Feruary 7, March 24 and May 27, 1976 (Washington: Government Printing Office, 1976), pp. 108-115. 5. Auerbach, 0., et al., "Effects of Cigarette Smoking on Dogs. II. Pulmonary Neoplasms," Arch Environ Health 21(6): 754-768, December, 1970. 6. Manber, M., "Dog Smoking Tests Seen Cancer Proof," Newark [N.J.] News, February 6, 1970. 7. Brower, L., "Smoking and Cancer," New York Times, February 15, 1970. 8. Sterling, T., "Comment on Smoking Dogs," Arch Environ Health 22: 631-632, May, 1971. 9. Committee on Biologic Effects of Atmospheric Pollution, Particulate Polycyclic Organic Matter, Division of Me ica Sciences, National Research Council (Washing- ton: National Academy of Sciences, 1972). 10. Anonymous, Environment: 21, May, 1973. 11. Hirth, R. and G. Hottendorf, "Lesions Produced by a New Lungworm in Beagle Dogs," Vet Pathol 10: 385-407, 1973. 12. Georgi, J., "Filaroides hirthi: Experimental Trans- mission Among Beagle Dogs~~FTirough Ingestion of First Stage Larvae," Science 194: 735, November 12, 1976. N Cft 13. Gifford-Jones, W., "'Cancerphobia': The Universal o Disease," The Doctor Game (Toronto: McClelland & ~ Stewar t, Ltd.,-IP-5). rv 27

14. Rosenblatt, M., Statement, U.S. Congress, House, Committee on Interstate and Foreign Commerce, Ci arette Labelin and Advertising -- 1969, Hearing, 91st Cong., 1st Sess., April 15, 30 and May 1, 1969 (Washington: Government Printing Office, 1969), pp. 1255-1263. 15. U.S. Public Health Service, Smoking and Health. A Re ort of the Sur~e~on Genera ,-Dep-artment oFHeal'Eh, uca io`n an3 Welfare, DHEW Publication No. (PHS) 79-50066, 1979. 16. Berkson, J., "Smoking and Lung Cancer: Some Observa- tions on Two Recent Reports," J Am Stat Assoc 53(281): 28-38, March, 1958. ~ 17. Detering, K. and E. Hartmann, "Mortality Associated with the Pill," Lancet II: 1024, November 12, 1977. 18. U.S. Public Health Service, Smoking and Health. A Report of the Advisor Committee to the Surgeon General o-te $ubI c Hea t Service, Department o H-ealt , Education anc Wel are, PHS Publication No. 1103, 1964. 19. Sterling, T., "The Effects of Self-Selection Factors in the Study of Smoking and Lung Cancer," Presentation, The Annual Meeting of the American Statistical Associa- tion, Montreal, Quebec, August 16, 1972. 20. U.S. Public Health Service, The Health Consequences of Smokin : 1977-78, Department of H`eaTEh, Education anU Wel are, DHEW Publication No. (PHS) 79-50065, 1979. 21. Doll, R. and A. Hill, "Lung Cancer and Other Causes of Death in Relation to Smoking," Brit Med J II: 1071-1081, 1956. 22. Hammond, E. and D. Horn, "Smoking and Death Rates -- Report on 44 Months of Follow-up on 187,783 Men. Part I. Total Mortality. Part II. Death Rates by Cause," J Am Med Assoc 166: 1159-1172, 1294-1308, 1958. 23. Dorn, H., "The Mortality of Smokers and Non-Smokers," Proc Soc Stat Sect Amer Stat Assn, 34-71, 1958. 24. Doll, R. and A. Hill, "Mortality in Relation to Smoking: Ten Years' Observations of British Doctors," Brit Med J I: 1399-1410, 1964. 25. Hammond, E., "Smoking in Relation to the Death Rates of One Million Men and Women," E idemiolo ical Approaches to the ~StudV of Cancer and O er roni.c iseases, I~Cf Mono ra-h 19, ed. W. Haenszel, January, 1966, pp. 127-204. 2501442994 28 .- ..~

26. Kahn, H., "The Dorn Stu dy of Smoking and Mortality Among U.S. Veterans: Report on Eight and One-Half Years of Observation," Epidemiological Approaches to the Stud of Cancer and Other Chronic Diseases, NCI Monogr 3T, ed. W. Haensze , January, 1T5T,_pp-l-l25. 27. Burch, P., The Bio, logy of Cancer: A New Approach (Baltimore: University Parc Press, lm). 28. Lee, P., (ed.), Tobacco Consum tion in Various Countries, Tobacco Researc Counci , ResearcF-Paper, No. 6, th ed.; Edinburgh: T. and A. Constable, Ltd., 1975). 29. Segi, M., et al., Cancer Mortality and Morbidity Statistics: Japan and the World, Japanese Cancer Association, GANN Monograph on Cancer Research No. 26 (Tokyo: Japanese Scientific Societies Press, 1981). 30. Rosenblatt, M., Statement to U.S. Congress. 31. Berge, T. and N. Toremalm, "Bronchial Cancer -- A Clinical and Pathological Study: II. Frequency According to Age and Sex During a 12-Year Period," Scand J Res ir Dis 56(2): 120-126, August, 1975. 32. Belcher, J., "The Changing Pattern of Bronchial Carcinoma," Br J Dis Chest 69: 247-258, 1975. 33. Sterling, T., "Additional Comments on the Critical Assessment of the Evidence Bearing on Smoking as the Cause of Lung Cancer," Am J Public Health 66(2): 161-164, February, 1976. 34. Beamis, J., Jr., et al., "Changing Epidemiology of Lung Cancer: Increasing in Women," Medical Clinics of North America 59(2): 315-325, March, 1975. 35. Hueper, W., "Lung Cancer and Smoking in Perspective," Law ers' Medical C_ yclo ~edia....of Personal Injuries u~ries and A lpeci.a~ies, ecC.J. Frnke , Revise Vo~V, Part B (Indianapolis: The Allen Smith Co., 1972). 36. Burch, P., "Smoking and Lung Cancer: The Problem of Inferring Cause," J R Statist Soc Part 4, 141: 437-477, 1978. 37. Gottlieb, M., "Lung Cancer and the Petroleum Industry in Louisiana," J Occup Med 21(6): 384-388, June, 1980. 38. Hoiberg, A., "Cancer Among Navy Personnel: Occupa- tional Comparisons," Milit Med 146: 556-561, August, 1981. 29

39. Hoover, R. and J. Fraumeni, "Cancer Mortality in U.S. Counties with Chemical Industries," Environ Res 9: 196-207, April, 1975. 40. Blot, W. and J. Fraumeni, "Geographic Patterns of Lung Cancer: Industrial Correlations," Am J Epidemiol 103: 539-550, 1976. ~ 41. Menck, H. and B. Henderson, "Occupational Differences in Rates of Lung Cancer," J Occu Med 18 (12) : 797-801, December, 1976. 42. Sterling, T., "Does Smoking Kill Workers or Working Kill Smokers? or The Mutual Relationship Between Smoking, Occupation, and Respiratory Disease," Int J Health Serv 8(3): 437-452, 1978. 43. Ford, A. and 0. Bialik, "Air Pollution and Urban Factors in Relation to Cancer Mortality," Arch Environ Health 35(6): 350-359, November/December, 1980. 44. Shear, C., et al., "Evidence for Space-Time Clustering of Lung Cancer Deaths," Arch Environ Health 35(6): 335-343, November/December, 1980. 45. Mason, T., et al., Atlas of Cancer Mortalit for U.S. Counties: 1950-1969, Department of Hea , Eauca ion an We are, DHEW Publication No. (NIH) 75-780, 1976. 46. Mason, T., et al., Atlas of Cancer Mortality Among U.S. Non-Whites: 1950-1969, Department of Health, Education and Welfare, DHEW Pub ication No. (NIH) 76-1204, 1976. 47. Bjelke, E., "Dietary Vitamin A and Human Lung Cancer," Int J Cancer 15(4): 561-565, April, 1975. 48. Basu, T., et al., "Plasma Vitamin A in Patients with Bronchial Carcinoma," Br J Cancer 33: 119-121, January, 1976. '- 49. Mettlin, C., et al., "Vitamin A and Lung Cancer," J Natl Cancer Inst 62(6): 1435-1438, June, 1979. - 50. Wald, N., et al.,"Low Serum-Vitamin-A and Subsequent Risk of Cancer. Preliminary Results of a Prospective Study," Lancet II: 813-815, October 18, 1980. 51. Shekelle, R., et al., "Dietary Vitamin A and Risk of Cancer in the Western Electric Study," Lancet II: 1185-1190, November 28, 1981. 52. Bishop, J., "Lung Cancer Risk May Be Cut by Nutrient in Some Fruits, Vegetables, Study Finds," Wall Street Journal, December 3, 1981. 30

53. Fritschner, S., "Can a Carrot a Day Keep Cancer Away?" Washington Post, December 10, 1981. 54. Anonymous, "Mind and Cancer," Lancet I: 706-707, March 31, 1979. 55. Kissen, D., "Psychological Factors, Personality and Lung Cancer in Men Aged 55-64," Brit J Med Psychol 40 (Part I) : 29-43, March, 1967. 56. Horne, R., et al., "Psychosocial Risk Factors for Lung ' Cancer," Psychosom Med 41(7): 503-514, November, 1979. 57. Abse, D., et al., "Personality and Behavioral Charac- teristics of Lung Cancer Patients," J Psychosom Res 18: 103-113, 1974. 58. Osler, W., cited in Lancet I: 1302, June 16, 1979. 59. Fisher, R., Smokin : The Cancer Controvers , Some Attempts to ssess the Evi enc`ce-(E in urg and London: Oliver and Boyd, 1959). 60. Burch, P., "Smoking and Lung Cancer." :~-

Cancer of the Esophagus Prevention . . . will be possible only when the aetiology of the disease has been un- ravelled. Editorial The British Medical Journal July 17, 1976 Esophageal cancer studies sometimes have reported statistical associations with several factors, including alcohol, diet and smoking. The Surgeon General's reports have interpreted such studies as showing that smoking is a signifi- cant cause of the disease. Any such conclusion, however, overlooks the basic tenet that epidemiological studies, espe- cially of chronic diseases, cannot prove cause and effect. A prominent epidemiologist's warning that "this limitation is sometimes forgotten" seems relevant to the evaluation of claims that smoking is a cause of esophageal cancer.1 While smoking an d esophageal cancer have been statis- tically related in some studies, the link has not been re- ported worldwide. Investigators who studied esophageal cancer patterns in Greenland, for example, noted that "there would appear to be no association with the mere consumption of tobacco."2 r.~ In northern China, one area has had an extremely high 0 ~- .p .~ esophageal cancer rate for centuries. An Australian epidemi- ~ ~ ologist has note d that theories about the disease occurrence 33

in the Chinese "hot spot" relate mainly to diet.3 A recent report of suspect chemicals in the extracts of pickled vege- tables, a common component of the diet of the area, seems to support those theories.4 A British Medical Journal editorial, however, has argued that "gastronomy may not explain everything" since chickens in the area also seem to be particularly susceptible to esophageal cancer. Whatever the explanation, the disease, noted the editors, "does not seem to be connected in this environment with alcohol and tobacco."5 As assessment of worldwide patterns of esophageal cancer, by an International Agency for Research on Cancer researcher, provides further evidence of the inconsistent international picture. The review mentioned a statistical relationship between tobacco and esophageal cancer that had been reported in several Western countries but noted its apparent absence in others. N) a ~ .P .~ Although several factors have been linked with the .r1)o disease, the reviewer noted, these have not been associated "strongly enough or with sufficient consistency between coun- tries for their etiological role to be basic." He pointed out esophageal cancer's "propensity" for lower socioeconomic groups in many different populations and su ggested that this 34

reflected dietary deficiencies or nutritional imbalances asso- ciated with poverty.6 Some researchers refer only to reports from Western countries in their discussions of possible causes of esopha- geal cancer. In these, tobacco, and especially alcohol, are mentioned frequently. Yet the epidemiological studies do not consistently show even a statistical risk associated with smoking. British epidemiologists recently analyzed esophageal cancer mortality data in England and Wales for the period 1911-1975. Although alcohol was reported to be associated consistently with esophageal cancer, cigarette smoking was not. "It therefore seems unlikely," the authors conclu ded, "that cigarette smoking is an important factor in the aeti- ology of oesophageal cancer in England and Wales."7 Another British researcher started out to study the effect of abstinence from meat in Britain's Vegetarian Society. When he found the records insufficient for study, he turned to six contemplative religious orders whose sisters did not eat meat. He found an unexpectedly high rate of esopha- geal cancer in the nuns. Wrote the author of his findings: "There is an intriguing excess of esophagus cancer that cannot be attributed either to alcohol or tobacco."$ N cn a ~ -~ 41t. co 0 0 0

Even in the U.S., after alcohol consumption is taken into account, there is no consistent statistical association between smoking and esophageal cancer. One analysis was based on data from over 150 Veterans Administration hospitals. The author conclu ded that "alcoholic beverage consumption entails a significantly high risk of esophageal cancer, which is inde- pendent of smoked tobacco." Although he believes that smoking is involved in the disease, he did not find a significantly high risk for smokers when alcohol consumption was con- trolled. In fact, he calculated the relative risk of esopha- geal cancer for smokers to be only 1.1. Statistically, that is abou t the same as no increased risk.9 A similar conclusion was reached in a study of Washing- ton, D.C., black men, whose esophageal cancer rates are among the highest in the U.S. Epidemiologists from the National Cancer Institute concluded that their data "appeared adequate to identify alcoholic beverage consumption as the major factor."10 They found "no significant risk associated with cigarette smoking and also no consistent enhancement of risk following exposure to both alcoholic beverages and cigarette ~ 0 smoking." ~ .t~ This fin ding, perhaps surprising to the authors, prompted the comment that the use of a different control group might show an increased risk due to tobacco. But, even with a different control group, the newly calculate d risks they 36 . w 0 0 ~

thought might be due to tobacco were not significantly dif- ferent from those of the nonsmokers. Thus, the claim that tobacco smoking is causally related to esophageal cancer receives questionable support in. the literature. And the evidence of a statistical association is inconsistent on both a worldwide basis and in Western countries. Perhaps the best summary of the continuing dilemma about esophageal cancer is provided by a recognized medical text, which notes that certain nutritional and environmental factors are suspected, but "the cause or causes of cancer the esophagus are unknown.11ll of N cn 0 f-4 4~b ~ w a 0 ~ 37

References for Esophageal Cancer 1. Keys, A., et al., "The Diet and All-Causes Death Rate in the Seven Countries Stu dy," Lancet II: 58-61, July 11, 1981. 2. Nielsen, N., et al., "Oesophageal Cancer in Greenland: Selected Epidemiological and Clinical Aspects," J Cancer Res Clin Oncol 94: 69-80, 1979. 3. Armstrong, B., "The Epidemiology of Cancer in the People's Republic of China," Int J Epidemiol 9(4): 305-315, 1980. 4. Lu, S., et al., "Mutagenicity of Extracts of Pickled Vegetables Collected in Linhsien County, a High- Incidence Area for Esophageal Cancer in Northern China," J Natl Cancer Inst 66(1): 33-36, January, 1981. 5. Editorial, "Cancer of the Oesophagus," Br Med J II: 135-136, July 17, 1976. 6. Day, N., "Some Aspects of the Epidemiology of Esopha- geal Cancer," Cancer Res 35(11, part 2): 3304-3307, November, 1975. 7. Chilvers, C., et al., "Alcohol and Oesophageal Cancer: An Assessment of the Evidence from Routinely Collected Data," J Epidemiol Comm Health 33: 127-133, 1979. 8. Kinlen, L., "Mortality in Relation to Abstinence from Meat in Certain Orders of Religious Sisters in Britain," ~Ba~n~~~b~u~r_y Re ort 4. Cancer Incidence in Defined Populations, e s. J. Cairns, et al. (Cold Spring Harbor, New York: Cold Spring Harbor Labora- tory, 1980), pp. 135-143. 9. Keller, A., "The Epidemiology of Esophageal Cancer in the West," Prev Med 9: 607-612, 1980. 10. Pottern, L., et al., "Esophageal Cancer Among Black Men in Washington, D.C.: I. Alcohol, Tobacco, and Other Risk Factors," J Natl Cancer Inst 67(4): 777-783, October, 1981. 11, del Regato, J. and H. Spjut.(eds.), "Cancer of the Digestive Tract/Esophagus," Cancer: Dia nosis, Treat- ment, an d Pro nosis, 5th ed., (St. Lou s: The C.V. Mosby Company, ), pp. 446-462. N Cn 0 ~ -p .~ w 0 0 w

Cancer of the Larynx Laryngeal cancer is a relatively rare disease -- representing less than 1.5 percent of the total annual cancer incidence rate in the United States.1 Thus, researchers observe few laryngeal cancer cases, even in large epidemio- logical studies. For example, Hammond and Horn, in their study of 188,000 males, recorded only 24 laryngeal cancer deaths.2 And Kahn's stu dy of approximately a quarter of a million U.S. veterans found only 54 laryngeal cancer deaths.3 As a consequence of such small numbers, the conclusions drawn from population studies may be suspect. Nonetheless, the Surgeon General's reports have used such studies as a basis for concluding that smoking is causally related to laryngeal cancer. Yet an English specialist in diseases of the larynx who reviewed several major epidemiological studies considered them inadequate to establish such a causal relationship.4 About the Hammond and Horn study, he commented that the conclusions that can be drawn are "not clear." He described Kahn's mor- tality ratios for smokers as "slightly less impressive," once they are "viewed in the light of the number of cases -- 54 in all, and usually in single figures in each smoking group." And he contended that Doll and Hill's British physicians study "produced no firm conclusion about laryngeal carcinoma." Only

16 deaths due to this disease were observed during a 10-year follow-up.5 He also discusse d the contrast between cigarette smok- ing trends and the "remarkably constant" incidence rates of laryngeal cancer. In other words, laryngeal cancer rates have been essentially the same year after year even though tobacco consumption has increased dramatically. Consequently, he warned that any data showing an association between cigarette smoking and laryngeal carcinoma "must be interpreted with caution." Overall, therefore, he found "no irrefutable evidence" that smoking causes laryngeal cancer. The contrasting smoking and disease time trends gener- ate d a comparable reaction in another British scientist. He suggested that the striking dissimilarity "would seem to be incompatible" with the hypothesis that tobacco is causally related to the disease.6 N cn 0 ~ A similar divergence in trends was noted in a report ~ ~ w 7 0 from Scotland. Although the percentage of smokers in 0 on Scotland is high and an increasing amount of tobacco is being consumed, the authors did not find the higher death rates in smokers that would be predicted under the causal hypothesis. Nor could they account for this same divergence in an area'`of 40

western Scotland where the quality of data was high. Even though the authors stated there is scientific evidence su g- gesting a causal relationship of smoking to laryngeal cancer, they are forced to recognize that their own data presented an "anomaly." The causal hypothesis also is not supported by the work of an Argentinian who examined the smoking habits of cancer patients in three cities.8 He found that smoking did not have a statistically significant relationship to laryngeal cancer. This finding was interpreted as "not indicating a dependency between the presence of the studied cancer and the smoking habit." He concluded that "we can therefore hypo- thetically assume that other factors, besides the significance of the smoking habit, must logically affect the etiology of these pathologies." Thus, the data do not warrant a conclusion that smoking causes laryngeal cancer. Not only have the population stu dies provided insufficient information, but disease and smoking trends are inconsistent with the causal hypothesis. Two British scientists, although recognizing the considerable re- search interest in smoking and laryngeal cancer, nevertheless concluded that "a positive causal relationship has not been substantiated statistically."9 41

References for Laryngeal Cancer 1. U.S. Department of Health and Human Services, Surveil- lance, Epidemiology, and End Results: Incidence an3- Mortality Data, 1973-77; NCI Mono ra h 57, National Institutes of Health, NIH Pub i.cation No. 81-2330, June, 1981. 2. Hammond, E. and D. Horn, "Smoking and Death Rates -- Report on Forty-Four Months of Follow-Up of 187, 783 Men. II. Death Rates by Cause," J Am Med Assoc 166(11): 1294-1308, March 15, 1958.' 3. Kahn, H., "The Dorn Stu dy of Smoking and Mortality Among U.S. Veterans: Report on Eight and One-Half Years of Observation," Epidemiological Approaches to the Study of Cancer and Other Chronic Diseases; NCI Mono rg aph 19, ed. W. Haenszel, January, 1966, pp. 1-125. 4. Stell, P., "Smoking and Laryngeal Cancer," Lancet I: 617-618, March 18, 1972. 5. Doll, R. and A. Hill, "Lung Cancer and Other Causes of Death in Relation to Smoking: A Second Report on the Mortality of British Doctors," Brit Med J II: 1071-1081, November 10, 1956. Doll, R. and A. Hill, "Mortality in Relation to Smoking: Ten Years' Observations of British Doctors," Brit Med J I: 1399-1410, May 30, 1964. 6. Burch, P., "Are 90% of Cancers Preventable?," IRCS Med Sci 6: 353-356, 1978. 7. Boyle, P., et al., "Cancer of the Larynx in Scotland," Br J Cancer 41: 196-197, 1980. 8. Ferrara, F., "Ecological Analysis of Lung Cancer in the City of La Plata," Proceedings of 2nd International Clean Air Congress,-19'TS, pp. 244-24'T. 9. Stell, P. and T. McGill, "Exposure to Asbestos and Laryngeal Carcinoma," J Laryng 89(5): 513-517, 1975. 42

Cancers of the Oral Cavity and Pharynx An observation from a review of recent medical litera- ture on oral and pharyngeal cancers characterizes strikingly the need for basic work on the diseases. Said the author: "Much of the data available on oral and pharyngeal cancer is limited to case histories that characterize the clinical disease aspects and consist of few numbers." According to the author, an epidemiologist from the University of Washington, the situation is compounded further by the failure fully to analyze available data and by the existence of serious weak- nesses in current methodology.1 Discussing these weaknesses, the researcher pointed to another problem. "Even the most obvious confounding factors such as age, sex, race or SES [socioeconomic status] fre- quently were not controlled when other variables were ana- lyzed," she said. The result of such shortcomings is that "the literature often lacks comparability and continuity," she said, and this "ill-defined research" allows for limited con- clusions. In her view, these limitations certainly apply to con- clusions about alcohol and tobacco. Even though she believes N they are probably major risk factors for the two cancers, she o ~ 4, concede d: $ 0 0 00 43

To date, no one type or amount of alcoholic or smoking product has been implicated con- clusively. Also complicating the stu dy of oral cancer has been the inability to determine the natural history of the disease -- in particular, whether a condition in the oral cavity called leukoplakia is a precursor of cancer. Characterized by thickened, white patches on the tongue, gum or cheek, leuko- plakia sometimes has been associated with the use of tobacco. However, a study of leukoplakia in patients from Denmark and Hungary found no statistically significant differences between smokers and nonsmokers.2 Findings from Sweden and Denmark indicate that the question of whether leukoplakia really is a precursor of oral cancer is unresolved. The Swedish study followed nearly 800 patients with leukoplakia for up to 44 years and found that only 4 percent of the group had developed oral cancer after 20 years.3 Abou t the same incidence was observed in a group of 331 Danish patients who were followed for periods of more than a year.4 Interestingly, in both studies, nonsmokers were much more likely to develop oral cancer than smokers. N tTi O ~ Current epidemiolo ~ gical findings are of limited value. ~, w Contributing to the dilemma of oral and pharyngeal cancers are o occupational risk factors, possible involvement of viruses, unusual patterns in mortality rates worldwide an d negative t 44

animal work -- that is, failed attempts to induce cancers with cigarette smoke. Experimental research on oral cancer has been hampered by the failure to find an appropriate animal model. As the 1979 Surgeon General's report conceded, "An ideal animal model in which to study oral carcinogenesis has not been found." Of particular relevance is this comment from the same report: "Cigarette smoke and cigarette-smoke condensates generally fail to produce malignancies when applied to the oral cavity of mice, rabbits, or hamsters."5 Other investigators have examined the possible role of occupational factors. A recent study at Roswell Park Memorial Institute found that male patients who had been employed in the leather industry had significantly increased risks for these cancers, which the authors said, "could not be explaine d by differences in smoking habits."6 Another study investigated lip cancer among commercial fishermen in Newfoundland. The authors found that the fisher- men had a 65 percent higher probability of developing lip cancer than men of comparable age and that cigarette smoking was not important as a risk factor. What they called "unex- pected" was the finding that those fishermen who used the mou th as a third han d in hauling in the tar-coated nets had a N cn 0 ~ ~ ~ w a ~ 0

lower incidence of lip cancer than those who used other metho ds.7 Viruses also have been examined for a possible role in these cancers. Although the research is not conclusive, scientists have postulated that the herpes simplex virus may be related to lip cancers.8 The Epstein-Barr virus has also been investigated for its possible etiologic role in naso- pharyngeal cancer (NPC).9-11 One group of researchers described this virus as "the only factor thus far linked to NPC in patients throughout the world," but it has not been detected in all cases.12 Claims that smoking causes oral and pharyngeal cancers are not supported by disease rate and cigarette consumption patterns. Logically, if both of the cancers are caused by cigarette smoking, their rate patterns should be fairly simi- lar. Yet they do not appear to be, according to data from two large-scale national cancer surveys by the U.S. govern- ment.13 The surveys show that over a 24-year period, pharyngeal cancer incidence rose by 15 percent while cancer of the oral cavity dropped nearly 40 percent. Oral and pharyngeal cancer incidence rates by sex and race, as reported in these surveys, present another puzzle, illustrated in the accompanying chart. as N Cn 0 ~ .p .Q w 0 ~ ~ 46

Figure 1 Incidence Rates* Per 100,000 Population From Second National Cancer Survey (SNCS, 1947-49) and Third National Cancer Survey (TNCS, 1969-71) Oral Cancer Male 18.1 WHITES Female NONWHITES Male Female 8.0 SNCS TNCS -38% Change SNCS TNCS -39% Change SNCS TNCS + 25% Change SNCS TNCS -22% Change Pharyngeal Cancer Male Female NONWHITES ~ Male 5.0 SNCS TNCS +67% Change Female SNCS TNCS 0% Change N r-n 0 ~ ~ population standard * Age adjusted to the 1950 U S ~ . . (13) w 0 ~ -WHITES 47

Changes in cancer rates might be explained by improve- ments in the ability to diagnose diseases and the increased access of minority groups, especially blacks, to medical care facilities. But these factors cannot explain the decreased oral cancer rates accompanying stable pharyngeal cancer rates among nonwhite women. Even when lag period is considere d, the observed disease trends are inconsistent with the causal theory, which predicts that disease rates will rise with increased cigarette consumption. This has not occurred in oral and pharyngeal cancer incidences, which have decreased by almost 30 percent during a period in which cigarette smoking was increasing. Differences in international cancer mortality rates also challenge the validity of the causal theory. In 1975, for instance, male oral and pharyngeal cancer mortality rates in Hong Kong and Singapore were among the highest in the world. But the rates were much lower in the U.S., Denmark Japan.14 Even though per capita cigarette consumption in Hong Kong and Singapore was much lower than in the U.S.,15 and mortality rates there were about four times higher than in the U.S. Conversely, Japan, which had a reported per capita cigarette consumption rate similar to that of the U.S. and the United Kingdom, reported oral and pharyngeal cancer mortality rates roughly half those of the U.S., England and Wales and Scotland.

All of these findings suggest that vigorous scientific investigation is needed. Claims of a causal role for smoking in the development of oral and pharyngeal cancers convincing scientific foundation. are wi thou t

References for Oral and Pharyngeal Cancers 1. Smith, E., "Epidemiology of Oral and Pharyngeal Cancers in the United States: Review of Recent Literature," J_ Nati Cancer Inst 63(5): 1189-1198, November, 1979. 2. Roed-Petersen, B., et al., "Smoking Habits and Histo- logical Characteristics of Oral Leukoplakias in Denmark and Hungary," Br J Cancer 28: 575-579, 1973. 3. Einhorn, J. and J. Wersall, "Incidence of Oral Car- cinoma in Patients with Leukoplakia of the Oral Mucosa," Cancer 20(12): 2189-2193, December, 1967. 4. Roed-Petersen, B., "Cancer Development in Oral Leuko- plakia Follow-Up of 331 Patients," J Dent Res 50(3): 711, May/June, 1971. [Abstract] 5. U.S. Public Health Service, "Chapter 5. Cancer," Smoking and Health. A Re o~rt of the Su.r~ geon General, Department oTHealth, E ud catlon and Welfare, DHEW Publication No. (PHS) 79-50066, 1979, pp. 39-42. 6. Decoufle, P., "Cancer Risks Associated with Employment in the Leather and Leather Products Industry," Arch Environ Health 34(1): 33-37, January/February, 1979. 7. Spitzer, W., et al., "The Occupation of Fishing as a Risk Factor in Cancer of the Lip," N Engl J Med 293 (9) : 419-424, August 28, 1975. 8. Hollinshead, A. and G. Tarro, "Soluble Membrane Antigens of Lip and Cervical Carcinomas: Reactivity with Antibody for Herpesvirus Nonvirion Antigens," Science 179(4074): 698-700, February, 1973. 9. Henle, W., et al., "Antibodies to Epstein-Barr Virus in Nasopharyngeal Carcinoma, Other Head and Neck Neo- plasms, and Control Groups," J Natl Cancer Inst 44(1): 225-231, January, 1970. 10. Huang, D., et al., "Demonstration of Epstein-Barr Virus-Associated Nuclear Antigen in Nasopharyngeal Carcinoma Cells from Fresh Biopsies," Int J Cancer 14: 580-588, 1974. 11. Glaser, R., et al., "Human Nasopharyngeal Carcinomas Positive for Epstein-Barr Virus DNA in North America," J Natl Cancer Inst 64(6): 1317-1319, June, 1980. 50

12. Easton, J., et al., "Nasopharyngeal Carcinoma in the United States. A Pathologic Study of 177 U.S. and 30 Foreign Cases," Arch Otolaryngol 106(2): 88-91, February, 1980. 13. Devesa, S. and D. Silverman, "Cancer Incidence and Mortality Trends in the United States: 1935-74," J Natl Cancer Inst 60(3): 545-571, March, 1978. 14. Segi, M., et al., (eds.), Cancer Mortality and Morbidit Statistics: Ja an and the World, Japanese Cancer ssocia ion, GANN-RonograpF onZ~ancer Research No. 26 (Tokyo: Japanese Scientific Societies Press, 1981). 15. Lee, P. (ed.), Tobacco Consum~t~ion in Various Coun- tries, Tobacco Research Council, Research aper No. 6 (Mi-ed.; Edinburgh: T. and A. Constable Ltd., 1975). l 51

Cancer of the Pancreas Coffee Is Linked to Cancer of Pancreas, But Stu dy Cau tions Tie Is Only Statistical Headline The Wall Street Journal March 12, 1981 Just about every newspaper and newscast that day carried word of what one medical editor referred to as "the latest entry in the carcinogen-of-the-week series."1 Re- searchers from Harvard had just reported finding an associa- tion between America's favorite beverage, coffee, and pan- creatic cancer.2 But most of the headlines and news bulletins contained only half of the information in the Wall Street Journal's headline: They omitted the point that the Harvard study was purely statistical. As the Journal put it, the stu dy "didn't show whether coffee caused pancreatic cancer, but said only that there is a statistical link between coffee consumption and the ailment."3 Fortunately for America's coffee drinkers, The New York Times and Medical World News later carried detailed criticisms from Yale, Columbia, Mt. Sinai Medical Center and even White River Junction, Vt., all of which suggested that the research was flawed.4-5 53

Meanwhile, the editor of a medical magazine use d the study and the publicity surrounding its publication to express his concernl abou t the way in which research dealing with cancer and suspected cancer-causing agents is reported to the public: Too often, both scientists and the press present the latest findings about a suppected carcinogen in incomplete and misleading fashion. Certainly, the Harvard study reported finding "a strong association" with coffee consumption. But the authors also emphasized "the need to determine whether the association exists in other data and to evaluate its causal or noncausal nature."2 According to the researchers, their stu dy was planne d to reevaluate the relationship between smoking and pancreatic cancer. But they found only a "weak positive association" between pancreatic cancer and smoking. On the other hand, they were surprised at the "unexpected association" with coffee drinking. A critical review of this article commended the investigators for pursuing other suspects when "the original analyses showed nothing substantial to incriminate" tobacco.6 54

In addition, the Harvard researchers noted that their observed association with coffee consumption seemed to be consistent with other findings. Specificially, they commented on the apparent increase in pancreatic cancer in recent decades in the U.S. an d on the low rates observed in Mormons and Seventh-day Adventists, who advocate abstaining from tobacco, alcohol, tea and coffee. Although they speculated that this pattern could be "compatible" with a causal role for either coffee consumption or cigarette smoking, they conceded that "the relatively small excess of men with the disease in proportion to women would seem to be more suggestive of a role for coffee rather than for cigarettes." Of course, this study neither proves nor disproves that pancreatic cancer is caused by coffee, tea, alcohol or smok- ing -- nor can any epidemiological stu dy. And the lead Harvard researcher appeared to address this point later that summer when he was quoted in a news story about research sug- gesting that green coffee beans may contain a powerful anti- cancer factor. He stated: "We're not saying the association existed. We're just saying it existed in our data" and "there should be further research."7 Unlike the Harvard study, two reports published in 1981 from the International Meeting on Pancreatic Cancer made no headlines.8-9 Perhaps that was because both of these 55

National Cancer Institu te stu dies stressed a much less drama- tic point -- that little is known about the cause or causes of cancer. One of these stu dies, which was au thore d by three NCI researchers, discussed several problems in analyzing data on pancreatic cancer. For example, the researchers noted that although pancreatic cancer mortality rates have increased in most countries, "it is difficult to determine how much of any increase" is actually due to a more frequent occurrence of the disease and how much is due to other factors. These could inclu de increased recognition of the disease because of im- provements in diagnostic procedures or increased accessibility to medical care.8 The NCI researchers also cited a worldwide 1976 study by the International Agency for Research on Cancer10 which, they said, demonstrated that "one of the persistent problems with descriptive epidemiology of pancreatic cancer is the fact that histologic confirmation of pancreatic cancer compared with other cancers is consistently low in most countries of the world."8 In other words, very few patients are examined with such thoroughness that a portion of the suspected cancer is removed for microscopic examination. 2501443020 In their own analysis of thousands of pancreatic cancer cases and deaths in the U.S. dating from the mid-1930°s, the 56

NCI researchers conceded that smoking and diabetes, two risk factors commonly associated with pancreatic cancer, explained "only a small proportion of the disease." They concluded that "much epidemiologic work remains to be done."8 .The authors of the second study also emphasized the continuing need for obtaining accurate information about the incidence of this disease.9 They believed that their research, in contrast to other studies, achieved a "rather complete ascertainment" of cancer incidence in Los Angeles County over a six-year period. By using hospital and clinic pathology records as well as death certificates, they were able to analyze 3,614 pancreatic cancer cases and determine that this cancer "still preferentially afflicts the old, the black and men, although the differences in risk with factors other than age are modest." Further, they found "no compel- ling evidence," they said, to support any specific environ- mental cause (such as occupational exposure or smoking). And they did find "substantial evidence which is inconsistent" with environmental hypotheses. Although they noted that a causal interpretation of the association between smoking and pancreatic cancer described in some stu dies has "credibility," they also stated that "a major etiologic role for either smoking or drinking is inconsistent" with certain disease rates. In other words, groups thought to be at high risk because of their smoking and drinking habits 57

had disease rates approximating those of so-called low risk groups. And they concluded: Despite this most detailed description of the pattern of pancreas cancer occurrence in a large, diverse, contemporary population, there is no very credible biologic explana- tion for the pattern. In an attempt to explain the reported patterns of disease rates, other recent stu dies have looked at a variety of factors possibly related to pancreatic cancer. For example, an overrepresentation of metal workers among pan- creatic cancer deaths in one Minneapolis county was considere d in dicative of the nee d to examine occupational exposures. The stu dy, which was co-authored by a researcher on assignment from the U.S. Center for Disease Control, identified and re- viewed records of all confirmed pancreatic cancer deaths in that county from 1935 to 1974. Although the researchers cited stu dies which have suggested that smoking is a risk factor for pancreatic cancer, they cou ld not examine this issue because they did not have smoking data available.l1 A research team from the University of Maryland School of Medicine, publishing in the Journal of the American Medical Association, discussed previous studies and presented original data that linke d pancreatic cancer with at least a dozen factors. These inclu ded chronic alcoholism, pancreatitis, allergies or dermatitis, gallstones, gallbladder removal, 25o1443p22 58

dietary factors, decaffeinated coffee, occupational exposures, steroids and certain gynecologic conditions and procedures. Although the researchers mentioned cigarette smoking among the risk factors listed in other stu dies, their own data showed "absence" of a relationship in males and an ambiguity in females. They did not report finding a statistically sig- nificant relationship between smoking and pancreatic cancer, but they did observe an increased risk among smoking women who drank decaffeinated coffee or had certain gynecologic problems. The authors had no explanation for this increased risk and called for further research.12 The confusing and even conflicting findings in this area are illustrated in another article written by a group from the University of South Carolina and Duke Medical Center. Unlike the investigators from the University of Maryland, they were unable to find any association between smoking and pancreatic cancer in either males or females. But, also unlike the Maryland researchers, they did find a statistically significant relationship between diabetes and pancreatic cancer in females.13 As these contradictory findings suggest, any considera- tion of the possible causes of pancreatic cancer should take 59

into account an observation in the 1973 edition of a medical text14 that is still relevant: The cause of pancreatic carcinoma remains unknown, and indeed there is little circum- stantial evidence from either experimental animals or epidemiologic studies on which to base even a tentative hypothesis. As the research indicates, little is known about the cause or causes of pancreatic cancer, which represents abou t 3 percent of all cancers diagnosed yearly in the U.S.15 Many possible risk factors have been suggested, but no specific environmental factor has been consistently associated with the development of pancreatic cancer. The conclusions that have been reached, and even the data from which these conclusions have been drawn, are open to question. Certainly, much research remains to be done. 60

References for Pancreatic Cancer 1. Editorial, "Putting Suspected Carcinogens in Perspec- tive," Medical World News 22 (9) : 7, April 27, 1981. 2. MacMahon, B., et al., "Coffee and Cancer of the Pancreas," N Engl J Med 304(11): 630-633, March 12, 1981. 3. Bishop, J., "Coffee Is Linked to Cancer of Pancreas, But Stu dy Cautions Tie Is Only Statistical," Wall Street Journal, March 12, 1981, p. 16. 4. Schmeck, H., "Critics Say Coffee Study Was Flawed," New York Times, June 30, 1981. 5. Anonymous, "Storm Brews Over Study Tying Coffee to Pancreatic Cancer," Medical World News 22(8): 10-11, April 13, 1981. 6. Feinstein, A., et al., "Coffee and Pancreatic Cancer. The Problems of Etiologic Science and Epidemiologic Case-Control Research," J Am Med Assoc 246 (9) : 957-961, August 28, 1981. 7. Cohn, V., "Flap Percolates Over Effects of Coffee Beans," Washington Post, August 29, 1981. 8. Levin, D., et al., "Demographic Characteristics of Cancer of the Pancreas: Mortality, Incidence, and Survival," Cancer 47 (6) : 1456-1468, 1981. 9. Mack, T. and A. Paganini-Hill, "Epidemiology of Pan- creas Cancer in Los Angeles," Cancer 47(6): 1474-1483, 1981. 10. Waterhouse, J., et al. (eds.), Cancer Incidence in Five Continents, Vol. 111-1976, IARC Scientific Publica- tions, No. 15 International Agency for Research on Cancer, 1976). 11. Maruchi, N., et al., "Cancer of the Pancreas in Olmsted County, Minnesota, 1935-1974," Mayo Clin Proc 54: 245-249, 1979. 12. Lin, R. and I. Kessler, "A Mulitfactorial Model for Pancreatic Cancer in Man: Epidemiologic Evidence," J Am Med Assoc 245(2): 147-152, January 9, 1981. 13. Shingleton, W., et al., "A Case-Control Study of Cancer of the Pancreas," Am J Epidemiol 110(3): 357-358, 1979. [Abstract]

14. Moertel, C., "Exocrine Pancreas," Cancer Medicine (Philadelphia: Lea & Febiger, 197 )~ pp. 1559-1570. 15. U.S. Department of Health and Human Services, Surveil- lance, Epidemiology, and End Results: Incidence and Mortality Data, 1973-T7; W!I Monograph 57, NaMonaIT Institutes ot Health, NIH Publication No. 81-2330, Bethesda, Md., June, 1981. 62

Cancers of the Urinary Tract Cancers of the kidney and bladder, which are by far the most common urinary tract cancers, account respectively for about 2 percent and 4-5 percent of all cancer cases diagnosed in the U.S. yearly.1 Both of these cancers have been asso- ciated with smoking in some epidemiological studies.2 How- ever, as with the other cancers discussed in this document, these urinary tract cancers have been related to numerous other factors as well.3 Exposures to various substances in the workplace, the residential area and even the home have been suggested as possible suspects. Dietary factors also have received atten- tion, and it has even been theorized that high blood pressure or psychological depression may place some people at increased risk. Not only must any attempt to study these diseases consider such factors, it also must account for the curious anomalies in the incidence data of bladder and kidney can- cers. An analysis of the most recent U.S. data from the National Cancer Institutel brings to light some enigmatic patterns: r") cn 0 ~ 4h, ~ • Overall, the incidence of bladder cancer ~ is twice that of kidney cancer. r°v ~ 63

• In white males bladder cancer incidence is almost three times that of kidney cancer. But in white females it is only half again as high as kidney cancer. • Bladder cancer occurs twice as frequently in white men as in black men, but their kidney cancer rates are basically the same. • White female bladder cancer incidence is almost a third higher than black women's, but kidney cancer is equal in women of both races. Given these varying rates in men and women of both races, these two cancers hardly can be blamed on a single agent, such as cigarette smoking. Cancer of the Kidne An international review of cancer incidence publishe d in 1976 reported sharply contrasting rates of kidney cancer worldwide. High rates occurred in Scandinavia and central Europe, while the lowest rates were found in southern Europe, most of Asia an d several parts of South America. Intermediate rates were reported for the U.S. and Canada.4 However, these rates and information on cigarette consumption do not fit the patterns predicted by the causal theory when an appropriate lag period is considered. For N t11 example, Sweden and Norway, which rank very high in kidney ° .~ cancer incidence, have shown consistently low per capita a N W 64

cigarette consumption over the last 50 years. In contrast, Japan and the United Kingdom, with relatively low kidney can- cer rates, have exhibited markedly higher cigarette consump- tion over this same period.5 Most epidemiological studies of kidney cancer have examined the possible significance of a variety of factors in addition to smoking, with diet and occupation receiving the most attention. For example, epidemiologists at the American Health Foundation reviewed international data on kidney cancer and presented findings from their own study of 202 kidney cancer patients. Although they found "a significant but moderate" association with cigarette smoking in their patient population, they concluded that "no correlation worldwide basis."b exists on a On the other hand, they did find significant associa- tions with consumption of sugar, meat, milk, fats and oils. On the basis of these and other findings, they suggested "a working hypothesis that a dietary factor, most likely fat- cholesterol intake, contributes directly or indirectly" to the development of kidney cancer. In a lengthy discussion of other suspected factors, the same authors pointed out that kidney tumors had been produced in animals by administering high estrogen doses, as well as by infection with viruses. Other animal work has suggested the 65

possible influences of the analgesic phenacetin,7 the flame retardant Tris8 and the synthetic estrogen DES.9 Workplace exposures also have been investigated in connection with kidney cancer. Some retrospective studies have failed to establish occupational associations with kidney cancer.6 Other research, however, has found elevated risks among workers in a number of industries, inclu ding petrochemi- cal and petroleum refiners,10 laundry and dry cleaners,11 newspaper press workers12 and coke oven workers.13 Another study of plant workers, the Western Electric Health Study, proposed that high blood pressure and psycho- logical depression might be involved in kidney cancer.l4-15 The stu dy population of more than 2,000 plant workers has been followed since the late 1950's. The investigators decided that a relationship between high blood pressure and cancer mortality observed in preliminary data merited further study. A follow-up analysis, which controlled for a number of other factors, inclu ding smoking, found that elevated blood pressure "appeared to be related most strongly" to kidney cancer mor- tality,14 In a separate article, the team reported that psycho- logical depression was associated with a number of different cancers. Although the cancer deaths were too few for a mean- ingful analysis by specific type, 60 percent of the kidney 66

cancer deaths occurred in employees with psychological depres- sion at the start of the study. The authors concluded that "psychological depression in middle-aged men is associated with increased risk of death from a variety of cancers inde- pendently of age, cigarette smoking, alcohol consumption, family history of cancer and occupational status.1115 Thus, it is clear there are no easy answers for the questions regarding the causes of kidney cancer. Cigarette smoking has been associated with kidney cancer in some epi- demiological studies, but an analysis of worldwide data did not show a statistically significant association. Incon- sistencies in mortality trends, associations with diet and occupation, suggestions of possible influences by such factors as high blood pressure and depression--all require further investigation. So long as the issue remains clou ded by these scien- tific uncertainties, no answers can be expected. Cancer of the Bladder As with kidney cancer, comparisons of international bladder cancer rates with cigarette consumption figures cast N ch 0 dou bt on the claimed causal link between the two. Ireland, F-& -~ Finland and Japan all rank low or intermediate in bladder o P.- &7

cancer mortality but high in cigarette consumption throughou t the relevant time period. Conversely, Denmark, the Nether- lands and Belgium rank quite low in cigarette consumption but have high bladder cancer rates.5,16 Analyses of U.S. bladder cancer rates on a regional basis also show departures from causal-theory predictions. Two government epidemiologists found elevated mortality from bladder cancer in clusters of counties throughout the U.S., especially in the Northeast. While accepting cigarette smoking as a bladder cancer risk factor, they conclu ded that it "is not likely to be responsible for the elevated mortality in the Northeast, inasmuch as the available national surveys show only small regional differences in smoking practices."l7 These same researchers proposed that occupational ex- posures might explain the high rates in certain areas. Their statistical analysis, they said, "in dicated that bladder can- cer rates among males throughou t the country were signifi- cantly higher in counties with chemical manufacturing plants," with the highest rate recorded in Salem County, N.J., where nearly one-fourth of the working population is employed in chemical production. Similarly, the three parishes in Louisiana with the highest bla dder cancer rates "had 'high' involvement" in the chemical industry. The particular chemicals associated with

0 high risk were: dyes, dye interme diates and organic pigments, pharmaceutical preparations and perfumes, cosmetics and other toilet preparations. Indeed, the contribution of numerous industries to the risk of bladder cancer has been stu died since 1895, when German investigators reported a significant increase among aniline dye workers. Many subsequent studies have also examined the possible influence of cigarette smoking. One large British study, which carefully assessed workers' histories, found excess bladder cancer risks in several occu pations. These inclu ded textile and clothing workers, tailors, electricians and electrical engineers, as well as workers in the chemical dye industry. The authors found, however, "no clear association" between cigarette smoking and the occurrence of bladder tumors.18 In another stu dy of the chemical industry, researchers noted a striking increase in risk for workers manufacturing benzidine but found no correlation between smoking and bladder cancer.19 A recent study of leather workers identified "signifi- cantly high risks of bladder cancer." There were so few female cases that their smoking information was not analyzed. However, among males, the author said, "the relative risk N cn 0 ~ -~ ~ w 0 w w 69

increased with duration of employment and remained elevated after adjustment for smoking habits."20 Researchers investigating workers at a rubber manufac- turing plant in Akron, Ohio, pointed out that "cancer of the urinary bladder has long been recognized as a disease of rubber workers."21 Their stu dy of 29,087 men and women employed in the plant between 1940 and 1978 verified this association.22 Additional evidence that the presence of industry may account for regional differences in bladder from National Cancer Institute scientists. on pet dogs diagnosed for cancer the U.S. and Canada and found nificant association" between of manufacturing industries. cancer rate comes They studied data at 13 veterinary centers in strong and statistically sig- "a bladder cancer and the presence They speculated that this asso- ciation was related "to industrial carcinogens that have es- caped into the general environment."23 Regional differences in bladder cancer risk may also be influenced by local water sources, according to some scien- tists. A recent NCI study associated certain chemical sub- stances called trihalomethanes (THMs) in some water supplies N C31 with increased risk of bladder cancer. Said the researchers: O ~ .p .~ co 0 w 70

Reasonably strong associations between bladder cancer and THM levels in drinkinq_ water appear in both sexes after control for differences in social class, ethnic group, urban versus rural residence and overall county industrialization. Further analyses indicated that the presence of specific high-risk industries for bladder cancer did not confound this association.24 A researcher with the U.S. Environmental Protection Agency, who reviewed this and other epidemiological studies, concluded that "there is potentially an increased risk of bla dder, colon and rectum cancer from drinking waters con- taining trihalomethanes."25 The controversy over bladder cancer is not limited to smoking. The debate surrounding saccharin and bladder cancer has received much attention in the headlines of major news- papers and needs no repetition here. It need only be noted that saccharin is one of many suspects in the search for the causes of bladder cancer. The tentative nature of previous conclusions is illus- trated by the reported remark of an NCI official in a discus- sion of saccharin and cancer: "Bladder cancer has, in fact, increased. We've generally a ttributed this to smoking, bu t perhaps we were wrong."26 ~ cn 0 ~ ~ .P w 0 cn 71

References for Urinary Tract Cancers 1. U.S. Department of Health and Human Services, Surveil- lance, Epidemiology, and End Results: incidence anc- Mortality Data, 1973-77; NCI Mono raph 57, National Ins it tutes of Hea th, NIH Pub i.cati.on No. 81-2330, June, 1981. 2. U.S. Public Health Service, The Health Conse uences of Smoking. The Changing CigareFte~Repor. o e~ Surgeon General, Department of Health and Human Ser- vices, DHHS Pu5lication No. (PHS) 81-50156, 1981. 3. National Cancer Institute, National Institute of Environmental Health Sciences, National Institute for Occupational Safety and Health, Estimates of the Fraction of Cancer in the United States Relatec~to Occu at onal Factors, U.S. Department of Heath, ~ E ucation and We are, September 15, 1978. 4. Waterhouse, J., et al., (eds.), Cancer Incidence in I~R Five Continents, Vol. 111-1976, Pub-Iications No. 15 (Lyon: International Agency for Research on Cancer, 1976). 5. Lee, P. (ed.), Tobacco Consumption in Various Coun- tries, Tobacco Research Counci , ResearcePaper, No. 6 TTi ed.; Edinburgh: T. and A. Constable Ltd., 1975). 6. Wynder, E., et al., "Epidemiology of Adenocarcinoma of the Kidney," J Natl Cancer Inst 53(6): 1619-1634, December, 1974. 7. Isaka, H., et al., "Tumors of Sprague-Dawley Rats Induced by Long-Term Feeding of Phenacetin," Gann 70 (1) : 29-36, February, 1979. 8. Reznik, G., et al., "Renal Carcinogenic and Nephrotoxic Effects of the Flame Retardant Tris (2,3-dibromopropyl) Phosphate in F344 Rats and (C57BL/6N X C3H/HeN)F1 Mice," J Natl Cancer Inst 63(1): 205-212, July, 1979. 9. Reznik-Schuller, H., "Carcinogenic Effects of Diethylstilbestrol in Male Syrian Golden Hamsters and European Hamsters," J Natl Cancer Inst 62(4): 1083-1088, April, 1979. 10. Thomas, T., et al., "Mortality Among Workers Em~ployed in Petroleum Refining and Petrochemical Plants,' J Occup Med 22(2): 97-103, February, 1980. 72

11. Katz, R. and D. Jowett, "Female Laundry and Dry Cleaning Workers in Wisconsin: A Mortality Analysis," Am J Public Health 71(3): 305-307, March, 1981. 12. Paganini-Hill, A., et al., "Cause-Specific Mortality Among Newspaper Web Pressmen," J Occup Med 22(8): 542-544, August, 1980. 13. Redmond, C., et al., "Long-Term Mortality Study of Steelworkers. VI -- Mortality from Malignant Neoplasms Among Coke Oven Workers," J Occup Med 41(8): 621-629, August, 1972. 14. Raynor, W., et al., "High Blood Pressure and 17-Year Cancer Mortality in the Western Electric Health Study," Am J Epidemiol 113(4): 371-377, April, 1981. 15. Shekelle, R., et al., "Psychological Depression and 17-Year Risk of Death From Cancer," Psychosom Med 43(2): 117-125, April, 1981. 16. Segi, M., et al., (eds.), Cancer Mortalit and Morbidit Statistics: Ja an and t e Wor ,-U-apanese ancer Associati.on, GANN MonograpYon Cancer Research No. 26 (Tokyo: Japanese Scientific Societies Press, 1981). 17. Blot, W. and J. Fraumeni, "Geographic Patterns of Bladder Cancer in the United States," J Natl Cancer Inst 61(4) : 1017-1023, October, 1978. 18. Anthony, H. and G. Thomas, "Tumors of the Urinary Bladder: An Analysis of the Occupations of 1,030 Patients in Leeds, England," J Natl Cancer Inst 45(5): 879-895, November, 1970. 19. Horton, A. and E. Bingham, "Risk of Bladder Tumors Among Benzidine Workers and Their Serum Properdin Levels," J Natl Cancer Inst 58(5): 1225-1228, May, 1977. 20. Decoufle, P., "Cancer Risks Associated with Employment in the Leather and Leather Products Industry," Arch Environ Health 34(1): 33-37, January/February, 1979. 21. Monson, R. and L. Fine, "Cancer Mortality and Morbidity Among Rubber Workers," J Natl Cancer Inst 61(4): 1047-1053, October, 197$. 22. Delzell, E. and R. Monson, "Mortality Among Rubber Workers. III. Cause-Specific Mortality, 1940-1978," J Occu Med 23(10): 677-684, October, 1981. 73

23. Hayes, H., et al., "Bladder Cancer in Pet Dogs: A Sentinel for Environmental Cancer?," Am J Epidemiol 114 (2) : 229-233, August, 1981. -' - 24. Cantor, K., et al., "Associations of Cancer Mortality with Halomethanes in Drinking Water," J Natl Cancer Inst 61(4): 979-985, October, 1978. 25. Williamson, S., "Epidemiological Studies on Cancer and Organic Compounds in U.S. Drinking Waters," Sci Total Environ 18: 187-203, 1981. 26. Gillette, R., "A Look at Tests, The Ban on Saccarin: How? Why?," Los Angeles Times, March 20, 1977. N 6n 0 ~ ~ ~ w 0 ca oo 74

Cancer of the Lung in Nonsmokers Among the more recent charges against smoking is the claim that exposure to other people's cigarette smoke -- some- times referred to as "passive" or "involuntary" smoking -- can cause lung cancer in nonsmokers. These assertions rely heavily on studies in Japanl and Greece,2 which claimed nonsmokers are at higher risk of developing lung cancer because of their exposure to cigarette smoke. Both studies have received extensive criticism in the scientific literature since their publication. In addition, findings from a recent American Cancer Society study3 con- tradict the Japanese and Greek research. Thus, these claims of increased cancer risks are highly suspect and not proven. The Japanese study, by Dr. Takeshi Hirayama, chief epidemiologist of Japan's National Cancer Centre Research Institute, reportedly found nonsmoking wives of heavy smokers to have a much greater risk of developing lung cancer nonsmoking wives of nonsmokers.l Shortly after the appearance of Hirayama's study, than questions were raised by other scientists about its design and the validity of its conclusions.4 Subsequently, a number of 75

highly critical letters appeared in the British Medical Journal emphasizing several of these deficiencies. For example, an American cancer epidemiologist questioned the make-up of Hirayama's study population, explaining that "certain biases" could have been introduced if the population were not selected carefully.5 The health districts included in Hirayama's study contained most of the heavy industries in Japan, she noted, suggesting that occupational or environ- mental exposures could have affected the disease rates he reported. In another letter a mathematician called attention to "some apparent inconsistencies" in Hirayama's data.6 Recal- culating from the published data, he determined that the crude lung cancer death rate for unmarried nonsmoking women actually was higher than the rates for nonsmoking wives of both non- smokers and smokers. He concluded it is "problematic to relate the higher risk for lung cancer in nonsmoking women married to smokers to passive smoking." In an editorial, the director of the Central Institute for Industrial Medicine in Hamburg, West Germany, reviewed these and other "problems" in the Hirayama study.7 Because of the high percentage of agricultural workers in the study population, he concluded that the sample was not representa- tive of the Japanese population as a whole. He summarized: 76

It is not possible to show scientifically convincing proof that the risk of lung cancer is increased by so-called passive smoking. women8 or to account for their exposure to indoor air pollu- tion from household heating and cooking equipment.9 One Other reviewers noted the study's failure to record the histological types of the lung cancers found in the Japanese group of German scientistslo concluded: In view of so many open questions Dr. Hirayama's conclusions do not appear very well founded. Finally, the cancer epidemiologist who had written earlier to the journal about Hirayama's study population wrote back to comment on his published reply.11 She stressed that Hirayama had not addressed "most of the questions raised." She noted that Hirayama's extrapolations from the husbands' smoking habits did not quantify accurately the amount of smoke to which the wives actually had been exposed. Based on her review of the literature on the issue, she concluded that "on the strictly toxicological level there is no hazard for non- smokers." In the study of Greek women in three Athens hospitals, Dimitrios Trichopoulos and his colleagues concluded that a nonsmoking woman whose husband smokes has twice the risk of N ch 0 ~ ~ ~ co 0 ~ ~ G

developing lung cancer as a nonsmoking woman married to a non- smoker.2 Yet even the authors of the Greek study conceded its "obvious" limitations. Trichopoulos and his colleagues ad- mitted that "the number of cases are small," as the study consisted of only 51 women with lung cancer, 40 of them non- smokers. A group of German researchers focused on this de- ficiency, stating that the "small number of inappropriately selected cases appears unable to yield convincing results.1110 In addition, the authors offered little information about the population from which the sample was selected, nor did they indicate that they considered factors such as diet, family history or occupational exposure in reaching their conclusions. After reviewing problems such as these in both the Trichopoulos and Hirayama studies, a scientist from the German Heart Center, Munich, concluded that "passive smoking may mean an annoyance for healthy adults, but in all proba- bility it is not connected with damage to health."12 Casting further doubt upon the conclusions reached by the two studies is a recently published report by Lawrence Garfinkel, American Cancer Society vice president for epidemiology and statistics.3 In a follow-up study of almost 180,000 American women, Garfinkel compared lung cancer mortality rates of nonsmoking wives reportedly exposed to N cn 0 ~ ~ .p w 0 -p ~ 78

different levels of tobacco smoke. He found that "none of these differences were statistically significant" and concluded: Compared to nonsmoking women married to nonsmoking husbands, nonsmokers married to smoking husbands showed very little, if any, increased risk of lung cancer. In an interview, Garfinkel emphasized his inability to detect any increased risk in nonsmoking wives of heav smokers.13 He said, "We could find no increased risk, no dose-response relationship." And he added: Passive smoking may be a political matter but it is not a main issue in terms of health policy. Following publication of the ACS study and criticisms of the Greek and Japanese research, the managing director of the American Lung Association stated late in 1981: "Whether or not passive smoking causes lung cancer is a subject for more research. 1114 79

Tobacco Smoke Components Some investigators base their claims of increased health hazards to nonsmokers on reported measurements of certain tobacco smoke components in the atmosphere. However it must be noted that the methodologies for measuring these substances in public places are still evolving. Moreover, simply because a component can be detected by highly sensitive instruments, its presence does not necessarily imply any sig- nificance to human health. One widely studied component of environmental tobacco smoke has been carbon monoxide (CO). Research has shown that the main sources of CO in the environment are motor vehicles and industrial processes.15 Indoor levels of CO also are affected by activities such as cooking and heating, even the number of persons present, because CO is generated by body metabolism. Yet environmental tobacco smoke often is blamed for contributing significant amounts of carbon monoxide to the environment. CO from tobacco smoke has been measured under realistic as well as unrealistic conditions. Under realistic condi- tions, measurements indicate that CO levels indoors where smoking is allowed will rarely exceed 10 parts per million ~ a (ppm).16 In public places with normal ventilation, the CO rN .A w concentration would be expected to be in the range of 5 0 44 -p 80

ppm.l7 Both figures are well below the limit of 50 ppm recommended by various health agencies here and abroad for workers exposed over an eight-hour period. A New Jersey pharmacologist concluded18 after review- ing the literature that: Environmental studies suggest that tobacco smoke has little impact on the CO content of room air except under highly artificial conditions. However, dependence upon measurements of CO to deter- mine ambient smoke levels has been criticized repeatedly be- cause carbon monoxide is produced by many sources other than tobacco smoke. In contrast, nicotine is considered a much more reliable indicator of the amount of tobacco smoke in the environment since it is produced almost exclusively by burning tobacco. Studies using nicotine as an indicator suggest the con- tribution of tobacco smoke to the atmosphere is minimal. Drs. William Hinds and Melvin First of the Harvard School of Public Health, for example, found only very small amounts of nicotine in the atmosphere of bars, bus and airline terminals, res- taurants and cocktail and student lounges.19 In a 1980 publication,20 Hinds further described their findings: 81

Nevertheless, the typical average airborne concentration of tobacco smoke in public rooms containing smokers provides the non- smoker with an amount of inhaled tobacco smoke equivalent to less than 0.01 cigarettes per hour. Commenting on their findings, a physician at the Harvard Medical School said that under the most severe concen- trations reported in the study, the nonsmoker would be exposed to "an amount of tobacco so small that the risk of development of any adverse health effect would be nonexistent, on the basis of any available data in the literature today."21 European researchers, using different methods to measure the presence of tobacco smoke in the atmosphere, found slightly higher concentrations of nicotine than Hinds and First.22 Yet basing their conclusion on their measurements, they, too, emphasized that "smoking does not represent a risk to nonsmokers." Studies of other tobacco smoke constituents indicate that their atmospheric levels, too, are minimal.23 One such study described the amount of volatile organic compounds added to the atmosphere by cigarette smoking as "insignificant."24 The claim has been made, based largely on reports by American Health Foundation scientists, that nitrosamines from 2501443046 82

tobacco smoke represent a cancer risk to nonsmokers.25 it is important to note, however, that even these researchers have emphasized that there are at present "no epidemiological data linking human respiratory cancers to volatile nitro- samines."26 One of the same researchers27 recently stated: There are no epidemiologic studies suggesting a measurable increased risk for tobacco- related diseases among nonsmokers and for people working in smoke-polluted environments. A study by James Repace and Alfred Lowrey, who attempted to measure the concentration Washington, D.C., area establishments, support the claim that cigarette smoke of particulates in often is cited to may be harmful to nonsmokers.28 They reported that in buildings is permitted, levels of particulate matter are than in places where smoking is not permitted. 19 where smoking much greater Serious questions were raised about the study's scien- tific methodology when Repace presented additional data to a recent meeting of the American Society of Heating, Refrigerat- ing and Air-Conditioning Engineers.29 A consulting engineer asked Repace whether he and Lowrey had measured the quantity of outdoor air introduced by rv ventilation systems.30 The answer was, "No." cn 0 ~ 83

In addition, criticisms of the research were submitted for inclusion in the printed record of the meeting.31 One commentator noted such deficiencies as their failure to measure the particulate levels before smoking began, their failure to present information on actual ventilation rates and their failure to determine the specific contribution of tobacco smoke to particulate levels. In fact, his analysis of their earlier data suggested that particulate levels were far more strongly related to the number of e~ ople present in the testing situation than to the percentage of those people who were smoking. He also expressed serious concern about the reliability of the testing instrument. He suggested that "the claim that smoking is responsible for indoor air pollution is an oversimplification of a complex, multi-source problem." Other research has not supported claims of health effects associated with cigarette smoke components in the am- bient air. For example, Danish scientists exposed nonsmoking subjects to atmospheric tobacco smoke and measured levels of smoke components in the test room.32 Their measurements of particulates indicated, they said, "the passive smoker will never inhale more than what equals 1/2-1 cigarette per day." They concluded that no data exist to support the contention that ambient tobacco smoke would "have a lasting adverse health effect in otherwise healthy individuals." N

 In a similar study, Canadian researchers exposed healthy nonsmokers to extremely high levels of tobacco smoke in a small, unventilated area under conditions which some sub- jects described as the worst they had ever experienced.33 Even under these extreme experimental conditions, only "mini- mal" physiological responses were reported. In evaluating these results, the researchers concluded that "the main argument for smoke-free air seems symptomatic rather than physiological." Statements made by more than a dozen scientists and physicians in oral testimony and written statements submitted to a subcommittee of the U.S. House of Representatives in 1978 are still valid today.34 Dr. Edwin Fisher, a professor of pathology, testified that "there is a lack of scientific in- formation incriminating atmospheric tobacco smoke as a health hazard."35 Simiarly, Dr. Hiram Langston, past president of the American Association for Thoracic Surgery, concluded:36 The weight of evidence as it exists in the world literature does not support a claim of adverse health effects for those exposed to 'passive smoking'. Even critics of tobacco have acknowledged that smoking has not been established as a cause of disease in nonsmokers. Dr. Ernst Wynder, president of the American Health Foundation, admitted that while he thought environmental tobacco smoke 85

could be annoying, "it has no influence on the health."37 And Dr. Nicholas Wald, a well-known British researcher with anti-smoking views, recently questioned the emphasis placed on the health aspects of involuntary smoking: "The most im- portant issues regarding exposure to other people's smoke in the present state of knowledge are aesthetic and social rather than medical."38 Conclusion Obviously, claims of increased health risk for non- smokers exposed to other people's cigarette smoke are not proven. But the considerable publicity received by studies purporting to show such risk have resulted in insufficient attention to the scientists who caution that emotion and fear must not be allowed to obscure scientific facts as they currently exist. The president of the German Society of Industrial Medicine in Berlin, in a recent review7 of studies on the public smoking issue, cautioned: One should remain conscious of the fact that, especially in the case of an emotionally treated subject matter such as passive smok- ing, unsubstantiated statements and premature requirements for regulative measures or a reversal of burden of proof do not, in the final analysis, serve the cause of scientific credibility nor facilitate the search for truth. 86

References for Lung Cancer in Nonsmokers l. Hirayama, T., "Non-Smoking Wives of Heavy Smokers Have a Higher Risk of Lung Cancer: A Study from Japan," Brit Med J I: 183-185, January 17, 1981. 2. Trichopoulos, D., et al., "Lung Cancer and Passive Smoking," Int J Cancer 27(1): 1-4, 1981. 3. Garfinckel, L., "Time Trends in Lung Cancer Mortality Among Nonsmokers and a Note on Passive Smoking," J Natl Cancer Inst 66(6): 1061-1066, 1981. 4. Anonymous, "Japan Study Linking Passive Smoking with Strong Ca Risk Startles Experts," Med World News, February 16, 1981. 5. Macdonald, E., "Non-Smoking Wives of Heavy Smokers Have a Higher Risk of Lung Cancer," Brit Med J_ II: 915-916, October 3, 1981. 6. Rutsch, M., "Non-Smoking Wives of Heavy Smokers Have a Higher Risk of Lung Cancer," Brit Med J_ I: 985, March 21, 1981. 7. Lehnert, G., "Krank durch Passivrauchen? [Ill by Pas- sive Smoking?]," Munch med Wschr 123(40): 1485-1488, 1981. Translation. 8. Schmahl, D., Quoted in "Lungenkrebs durch Passivrauchen? Studie aus Japan bestatigt Verdacht/ Widerspruch aus Deutschland [Lung Cancer from Passive Smoking? Study from Japan Confirms Suspicion/ Contradictory Opinion from Germany]," Suddeutsche Zeitung, March 18, 1981. Translation. 9. Sterling, T., "Non-Smoking Wives of Heavy Smokers Have a Higher Risk of Lung Cancer," Brit Med J I: 1156, April 4, 1981. - 10. Grundemann, E., et al., "Non-Smoking Wives of Heavy Smokers Have a Higher Risk of Lung Cancer," Brit Med I: 1156, April 4, 1981. J 11. Macdonald, E., "Non-Smoking Wives of Heavy Smokers Have a Higher Risk of Lung Cancer," Brit Med J II: 1465, 12. November 28, 1981. - N cn 0 Schievelbein, H., "Lungenkarzinom bei Passivrauchern ~ [Lung Carcinoma in Passive Smokers]," Munch med Wschr ~ ~ 123(17): 668-669 1981 Translation w , . . 0 cn I" 87

13. Garfinkel, L., Interview, "Nicht vom eigentlichen Problem ablenken, [Let's Concentrate on the Main Issue]," Munch med Wschr 123(40): 1483-1484, 1981. Translation. 14. Swomley, J., Letter to Carol M. Thomas, Secretary, Federal Trade Commission, October 19, 1981. 15. Stewart, R., "The Effects of Low Concentrations of Carbon Monoxide in Man," Environmental Tobacco Smoke Effects on the Nonsmoker: ~Re po_rtrom a Worksho , ed., R. Ryl an~er, (Denmar s: P. ~.~cFim33~, Vo3ens, 4) , 56-62. 16. Bridge, D. and M. Corn, "Contribution to the Assessment of Exposure of Nonsmokers to Air Pollution From Ciga- rette and Cigar Smoke in Occupied Spaces," Environ Res 5(2): 192-209, 1972. 17. Anderson, G. and T. Dalhamn, "Halsoriskerna vid passiv rokning [The Risks to Health of Passive Smoking]," Lakartidningen 70: 2833-2836, 1973. Translation. 18. Hutcheon, D., Statement, State of New Jersey Public Health Council, Public Hearing on Regulation of Smoking in Certain Public Places, Trenton, New Jersey, December 12, 1978. 19. Hinds, W., and M. First, "Concentrations of Nicotine and Tobacco Smoke in Public Places," N Engl J Med 292(16): 844-845, 1975. 20. Hinds, W., "The Lung and the Environment," Seminars in Respiratory Medicine 1(3): 197-210, January, 1980. 21. Huber, G., "Smoke and Heat," N Engl J Med 293: 47-49, 1975. 22. Badre, R., et al., "Pollution Atmospherique par la Fumee de Tabac [Atmospheric Pollution by Smoking]," Ann Phar Pr 36(9-10): 443-452, 1978. Translation. 23. Klosterkotter, W. and E. Gono, "Zum Problem des Passivrauchens [On the Problem of Passive Smoking]," Zbl Bakt Ha, I Abt Orig 162: 51-69, 1976. Translation. 24. Holzer, G., et al., "Gas Chromatographic-Mass Spectrometric Evaluation of Exhaled Tobacco Smoke," Chromatography 126: 771-785, 1976. J ` 88

25. Brunnemann, K., et al., "Assessment of Carcinogenic Volatile N-Nitrosamines in Tobacco and in Mainstream and Sidestream Smoke from Cigarettes," Cancer Res 37: 3218-3222, 1977. 26. Brunnemann, K. and D. Hoffmann, "Chemical Studies on Tobacco Smoke. LIX. Analysis of Volatile Nitrosamines in Tobacco Smoke and Polluted Indoor Environments," Environmental Aspects of N-Nitroso Compounds, eds. E. A. Walker, et al., (Lyon: IARC Scientific Publica- tions, 1978), pp. 343-355. 27. Hoffmann, D., et al., "The Role of Volatile and Non- volatile N-Nitrosamines in Tobacco Carcinogenesis," Banbury Report: A Safe Cigarette?, eds. G. Gori, et al., (Cold Springs Harbor: Cold Springs Harbor Laboratory, 1980), pp. 113-127. 28. Repace, J. and A. Lowrey, "Indoor Air Pollution, Tobacco Smoke, and Public Health," Science 208: 464-472, 1980. 29. Repace, J. and A. Lowrey, "Tobacco Smoke, Ventilation, and Indoor Air Quality," Presentation before The Sym- posium on Ventilation and Indoor Air Quality, Semi- annual Meeting of ASHRAE, Houston, January 25, 1982. 30. Spielvogel, L., Comments on J. Repace and A. Lowrey Presentation, "Tobacco Smoke, Ventilation, and Indoor Air Quality," The Symposium on Ventilation and Indoor Air Quality, Semiannual Meeting of ASHRAE, Houston, January 25, 1982. 31. Sterling, T., Comments on J. Repace and A. Lowrey Presentation, "Tobacco Smoke, Ventilation, and Indoor Air Quality," The Symposium on Ventilation and Indoor Air Quality, Semiannual Meeting of ASHRAE, Houston, January 25, 1982. 32. Hugod, C., et al., "Exposure of Passive Smokers to Tobacco Smoke Constituents," Int Arch Occup Environ Health 42: 21-29, 1978. 33. Pimm, P., et al., "Physiological Effects of Acute Passive Exposure to Cigarette Smoke," Arch Environ Health 33(4): 201-213, 1978. 34. U.S. Congress, House, Committee on Agriculture, Sub- committee on Tobacco, Effect of Smoking on Nonsmokers, Hearing, 95th Cong., 2ess., September 7, 1978 (Washington: Government Printing Office, 1978). 89 N cn 0 1-"

35. Fisher, E., Statement, U.S. Congress, House, Committee on Agriculture, Subcommittee on Tobacco, Effect of Smoking on Nonsmokers, Hearing, 95th Cong., 2nd -9ess., September 7, 1978 (Washington: Government Printing Office, 1978), pp. 2-20. 36. Langston, H., Statement, U.S. Congress, House, Commit- tee on Agriculture, Subcommittee on Tobacco, Effect of Smoking on Nonsmokers, Hearing, 95th Cong., 2nd Sess., September 7, 1978 (Washington: Government Printing Office, 1978), pp. 158-184. 37. Wynder, E., Quoted in "Unter Vier Augen: Wenn Schon-dann lieber harmios raucher [Face to Face)," Schweizer Illustrierte, October 26, 1976. Translation. 38. Wald, N., "Breathing in Their Smoke," Oxford Times, March 6, 1981.

Mortality Data: How Reliable? How Accurate? For many years, the U.S. and other countries have compiled "vital statistics" to monitor and stu dy the health status of their populations. Included in these vital statis- tics are numbers of deaths, usually classified by cause. Such mortality data can be useful in identifying and evaluating trends in specific diseases, in setting health research priorities, and, by way of epidemiological studies, in sug- gesting factors that may be associated with certain diseases. It is obvious that for valid conclusions to be drawn from these types of statistical studies, mortality data must have a high degree of reliability. Sources of Mortality Data: Death Certificates Mortality data generally come from death certificates. As a result, scientists and statisticians have been concerned with the quality of this source of information. Some comments from the medical and scientific literaturel-4 indicate the depth of concern: It has become obvious that in many cases the diagnosis on a death certificate may have actually no bearing whatsoever or any rela- tion to the cause of death of a given patient. Disquiet exists about the accuracy of the diagnostic information contained in the medical certificate of cause of death even in cases when a full post-mortem examination is carried out. 91

Comparison of death certificates with clini- cal and autopsy records has led to the con- clusion that mortality statistics on cause of death are of dubious accuracy. In 1973, the frequency of lung cancer in each county in the United States was reported by the Epidemiology Branch of the National Cancer Institute. These data are now used for our vital statistics, but I wonder how reliable such data can be if based on the kinds of death certificates that I have found. At first it might seem surprising that the cause of death listed on a death certificate may not be accurate. How- ever, a closer look at how death certificate information is obtained reveals that errors can occur, chiefly from mis diag- nosis and recording mistakes. The physician's diagnosis of the patient's fatal disease, usually found in the clinical or hospital record, often serves as the basis for the cause of death listed on the death certificate.* Thus, the information entered on the death certificate will be inaccurate if the physician has failed to diagnose the patient's disease accurately and the error is not corrected on the basis of laboratory or autopsy results. * The internationally accepted death certificate form requires that a physician indicate (a) the immediate cause of death, (b) the con dition which led to the immediate cause of death and (c) the underlying disease or condition which led to the conditions listed in (a) and (b). It is the underlying cause of death (c) that is then coded and compiTed anff- eventually reflected in national and international mortality statistics. r. Cn 0 ~ ~ ~ w 0 cn m 92

A number of recent stu dies have used autopsy findings to evaluate the accuracy of clinical diagnoses. A 1980 study in Finland, for example, compared clinical and autopsy diag- noses in 377 cancer patients and found substantial disagree- ment. The investigators conclu ded: "Our study indicates that accurate clinical diagnosis remains a significant problem despite modern medical resources."5 Even if the correct clinical diagnosis is made, it may not be recorded properly on the death certificate. In 1979, U.S, researchers compared the hospital diagnosis for 9,724 deaths with the underlying cause of death recorded on the death certificate and found agreement in only 72 percent of the cases.3 Various reasons have been suggested for recording errors. One pathologist has commented that many physicians "regard the death certificate as a document which simply declares that the death was due to natural causes and does not have medico-legal significance." He noted also that "doctors certifying deaths often fail to realise that the information they record is utilised by the statistician for compiling data of epidemiological significance."6 Most recently, the whole question of death certificate accuracy and its effect on mortality statistics was addressed 93

by a research team led tists.7 They compared death certificate with estimation of the true by National Cancer Institu te scien- the cause of death listed on the what they considered to be the "best cause of death" -- the cause "assigned by a physician based on au topsy findings in combination with pertinent clinical data." Disagreement was found in 42 percent of their 257 cases. The scientists also considered the "cancellation ef- fect" -- that losses and gains for a specific cause of death, caused by misdiagnoses or recording errors, might cancel each other out, resulting in valid overall reporting of mortality statistics. This possibility was rejected by the researchers, who concluded that their stu dy "affirms the need to assess the reliability of U.S. mortality statistics."7 Cancer Mortality Data When researchers compare medical records and autopsy reports for cancer of specific sites, they often find high rates of clinical "misdiagnosis." In this context misdiag- nosis means underdiagnosis (cancer not detected by the attending physician but found at autopsy), overdiagnosis (cancer clinically diagnosed but not found at autopsy) or a 94

combination of the two (cancer clinically diagnosed but as- signed to the wrong primary site). The view of one researcher that "there are many pitfalls in the radiologic and biopsy diagnosis of cancer, particularly, with regard to site of origin"8 is well supported. A study of 1,000 consecutive autopsies at a New York City hospital concluded that "a considerable number of neoplasms were found at autopsy in which the site of origin had not been established during life."9 For pancreatic cancer, 48 percent of the cases found at autopsy had not been clinically diagnosed. In a study at Boston City Hospital, "incorrect clinical diagnoses of cancer appeared in 40 percent (1,094) of the 2,734 cancer patients."10 These errors involved either undiagnosed cancer or diagnosis of the wrong primary site. Reports of cancer misdiagnosis have not been limited to the U.S. Finnish researchers found in an au topsy series of cancer patients that the clinically assigned site of the primary tumor was incorrect in 20 percent of the cases.5 Results of a Scottish stu dy indicated that for the more common sites of cancer, there was agreement between autopsy findings and clinical diagnosis in less than half of the 327 cases.11 95

Investigators in Israel conducted a similar study and found that 23 percent of autopsied cancer cases were either clinically undiagnosed or assigned to the wrong primary site.12 In addition, 18 percent of those cases with a clinical diagnosis of cancer were unconfirmed at autopsy. It also was noted that misdiagnosis rates did not occur with equal frequency within subgroups of the population but varie d with the patient's ethnic origin, age and days hospitalized. These considerations led the authors to caution: Our findings are of particular importance when related to the interpretation of cancer incidence studies . . . . the fact that misdiagnosis does not occur randomly in the population has to be taken into account whenever inter-ethnic comparisons are made. Overreporting of certain types of cancer also occurs, as evidenced by a 1981 letter to the editor of the New England Journal of Medicine.13 The authors investigated the "allegedly high mortality rates for cancer of the pancreas" in male residents of two counties in Kansas, which had been identified in a 1973 government cancer mortality atlas.l4 All death certificates coded to pancreatic cancer for these counties were obtained for the appropriate time period, and all available hospital and medical records were examined for the in dividuals concerned. The researchers found that, ~ Cn 0 "according to the most liberal criteria," only 45 of the 63 ~ ~ ~ w 0 0 96

men certified as dying of pancreatic cancer "could possibly have had the disease."13 All mortality statistics, including those for cancer of specific sites, may be affected by errors made in the codin of the underlying cause of death listed on the death certifi- cate.* For example, when more than one cancer site is men- tioned on the death certificate, coding of the underlying cause has always been difficult.15 It is also confusing when the word "metastatic" appears, because this word is used ambiguously, sometimes to designate a primary site from which the cancer has spread, and sometimes to designate a site to which cancer has spread.l5 Lung Cancer Mortality Data Of the cancer sites, the lung frequently has received special attention from investigators interested in questions of diagnostic accuracy. Their research findings, while dif- ferent in some respects, have been consistent on one point: The rate of misdiagnosis of lung cancer has been high. Yet * Death certificates are collected by offices of vital statistics in each sta te and country and the information is "translated" into an internationally established code number denoting the "underlying cause of death." The code numbers and corresponding diseases are listed in the International Classification of Diseases. State, national and international mortality data for specific diseases are compiled using the coded death certificates. 97

there seem to be no consistent patterns or obvious explana- tions for the discrepancies. In one stu dy of 493 cancer cases in a New York City hospital, only 44 percent of the clinical diagnoses of lung cancer were confirmed at autopsy.9 Also, the authors found that "carcinoma of the lung was the only major neoplasm in which no cases were found at au topsy in addition to those diagnosed clinically." [Emphasis added] It was suggested further by the researchers that the overdiagnosis problem observed in their stu dy might be due to the similarity in appearance between primary cancer of the lung and other can- cers that had metastasized or spread to the lung. Conversely, a Boston hospital stu dy found that lung cancer cases were missed clinically. Of lung cancers con- firmed at autopsy, 27 percent had not been diagnosed in the hospital.10 The Boston researchers did not address the question of clinical overdiagnosis. Other studies, from Israel,12 Finland,5 ~1) Scotlandll and New York,16 have found varying rates of cn 0 ~ ~ overdiagnosis and underdiagnosis of lung cancer. .~ w 0 m N Researchers from Yale Medical School have offered a possible explanation for misdiagnoses of lung cancer, based on their patient data that included smoking information.17 In 98

a study of 654 cases in which primary lung cancer was found at autopsy, the investigators reported that 16 percent of these cases had not been clinically diagnosed. They hypothesized that physicians may use more diagnostic tests for lung cancer when a patient is a smoker than when the patient is a non- smoker. They called this tendency "detection bias." This could account for the failure to diagnose lung cancer in some nonsmokers. Indeed, data from the Yale study "suggested that the more patients smoked, the more likely they were to have their lung cancer detected during life." Further, cytological examination of the sputum (a diagnostic procedure for lung cancer) had been ordered for a significantly higher percentage of smokers than it had been for nonsmokers. In fact, it appeare d that the rate of ordering such a test was related to the amount that the patient smoked. The authors conclu ded that "the rates of diagnosis may be affected by bias in the way that doctors order and deploy the available diagnostic technology." Conclusion Mortality statistics are often relied on by investiga- tors with little recognition that such data may contain inac- curacies. Death certificates, the main source for mortality N) ~ 0 ~ ~ ~ w 0 m w r. 99

data, have been analyzed in numerous studies and shown to have errors resulting from clinical misdiagnoses and recording mistakes. The reliability of cancer mortality statistics in particular has been questioned by researchers for the same reasons. Thus, extreme caution is warranted in the evaluation of epidemiological studies using mortality data. 100

References for Mortality Data 1. Briggs, R., "Quality of Death Certificate Diagnosis as Compared to Autopsy Findings," Ariz Med 32(8): 617-619, August, 1975. 2. Busuttil, A., et al., "The Accuracy of Medical Certifi- cates of Cause of Death," Health Bull 39(3): 146-152, 1981. 3. Gittelsohn, A. and J. Senning, "Stu dies on the Reli- ability of Vital and Health Records: I. Comparison of Cause of Death and Hospital Record Diagnoses," Am J Public Health 69(7): 680-689, July, 1979. 4. Rigdon, R., "Reliability of Data from Death Certifi- cates," N Engl J Med 303(24): 1422, December 11, 1980. 5. Stenback, F. and H. Paivarinta, "Relation Between Clinical and Autopsy Diagnoses, Especially as Regards Cancer," Scand J Soc Med 8(2): 67-72, 1980. 6. Gwynne, J., "Death Certification in Dunedin Hospitals," NZ Med J 86(592): 77-81, July 27, 1977. 7. Engel, L., et al., "Accuracy of Death Certification in an Autopsied Population with Specific Attention to Malignant Neoplasms and Vascular Diseases," Am J Epidemiol 111(1): 99-112, 1980. ~ 8. Rosenblatt, M., et al., "Diagnostic Accuracy in Cancer as Determined by Post Mortem Examination," Prog Clin Cancer 5: 71-80, 1973. 9. Rosenblatt, M., et al., "Causes of Death in 1,000 Consecutive Autopsies," NY State J Med 71(18): 2189-2193, September 15, 19TE. 10. Bauer, F. and S. Robbins, "An Autopsy Stu dy of Cancer Patients. I. Accuracy of the Clinical Diagnoses (1955 to 1965) Boston City Hospital," J Am Med Assoc 221(13 ) : 1471-1474, September 2T,1972 . 11. Cameron, H. and E. McGoogan, "A Prospective Study of 1152 Hospital Autopsies: II. Analysis of Inaccuracies in Clinical Diagnoses and Their Significance," J Pathol 133(4): 285-300, 1981. - 12. Ehrlich, D., et al., "Some Factors Affecting the Accuracy of Cancer Diagnosis," J Chronic Dis 28(7/8): 359-364, August, 1975. - 101

13. Holmes, F., et al., "More on Reliability of Death Certificates," N Engl J Med 304(12): 737, March 19, 1981. 14. Mason, T. and F. McKay, U.S. Cancer Mortalit b County 1950-69, National Institu`~es o eal , u ic~ealth Service, e, U.S. Department of Health, Education and Welfare, DHEW Publication No. (NIH) 74-615 (Washing- ton: Government Printing Office, 1974). 15. Percy, C. and A. Dolman, "Comparison of the Coding of Death Certificates Related to Cancer in Seven Coun- tries," Public Health Rep 93(4): 335-350, July/August, 1978. 16. Jimenez, F., et al., "Cancer of the Lung in Males," Bull NY Acad Med 51(3): 432-438, March, 1975. 17. Feinstein, A. and C. Wells, "Cigarette Smoking and Lung Cancer: The Problems of 'Detection Bias' in Epidemio- logic Rates of Disease," Trans Assoc Am Physicians 87: 180-185, 1974. 102

This document is published by the Tobacco Institute in the belief that the controversy about smoking and health must be resolved by scientific research and in the belief that full, free and informed discussion of the controversy is in the public interest. ~