Tobacco Documents Online

THE CIGARETTE'- LUNG CANCER ENIGMA Talk presented before small group in Boston January 4, 1960 by Edwin B. Wilson, Ph.D. Office of Naval Research and Harvard School of Public Health Boston, Mass. The subject announced is: The Cigarette - Lung Cancer Enigma. By the use of Enigma I mean that whatever relationship there may be has not been satisfactorily explained. The difference between finding statis- tical evidence of a relationship and the explanation of the relationship is great, and has been, in many human activities, observations and beliefs, medical and other. In the public health field~one need cite only malaria. If we go no further back than the Century Dictionary which ap- peared during the so-called "gay nineties,"'one finds this definition of malaria: "1. Air contaminated with some pathogenic substance from the soil; specifically, air impregnated with the poison producing intermittent and remittent fever. 2. The disease produced by the air thus poisoned... The disease is contracted by presence in the locality and not from the sick, nor do the latter seem to transplant it to the new places where they may go. The disease may apparently be introduced into the body through water that is drunk as well as through the air,"'etc. All these statements could be proved by statistics and had been jV ~ inferred from associations observed by scientists. In the past century N ~ Pettenkofer of Munich attributed the typhoid epidemics to bad air and proved ~ it to~his and others' satisfaction in a most effective way by cleaning up 0 most of the ill-smelling spots in Munich and thereby markedly reducing the (~ typhoid rate.

- 2' - When the d'efinition' in' the Century Dictionary was written the syxnptomatic differentiation' between mal.axia and typhoid was probab]1y lhaown but both had as their cause the iimpregrnation of the air with pathogenic material. When Renald Ross was trying to convince the British that mos- quitoes caused'the spread of malaria, an able statistically minded British army med'ical officer proved statistically that Rbss was al]i wrong and that the old malaria theory was right. I do not propose tonight to solve the enigma of the relation of lung cancer to cigarette smoking, or to tell you that you should or should not smoke cigarettes, but merely to try to give you some idea of why I think the problem of the explanation of the association which the statis- ticians find is still non-existent. To do this I must go back to my per- sonal professional contacts with' the sta.tistical and epidemiological work on' cancer in general. Across the summer of 1922 I transferred from M.I.T. to Harvard where a new School of Public Health was being organized. My task was to set up a Department of Vita1 Statistics for both instruction and research and for cooperation with the public health work of the State and the City of Boston. The school was set up to work in close contact with the Harvard Medical School with' many members of the two schools in common. Dr. Rosenau was professor of Preventi~ve Medicine in the Health School and professor of Epid'emiology in' the Health School. He had detailed to him by the U.S. Pub- lic Health' Service Dr. J. W. Schereschewsky, an: able and mature man, as- signed to study the rise in the cancer death rates i;n the ten original reg- istration States from 1900 to 1920 (28, 29). For technical statistical advice Dr. Rosenau sent him to me. The ana]1ysi;s of the change in time of any disease requires a good mature mind and a wide acquaintance not only

3 with the reported facts but with all sorts of possibilities for errors in the reports, for changes in the matter of diagnosis, for changes in man- ner of classification, etc. Such a mind and such an awareness Dr. Schereschewsky had; for there had'been much previous discussion of the ris- ing cancer rates in this country and abroad, and he knew the literature. On the general subject of the difficulty confronting a student of the history of the incid'ence of disease let me quote from a more recent paper by Major Greenwood'(12) as follows: "One should never believe that a disease is becoming more or less deadly until all other explanations have been excluded... Change of fashion in certification due to increasing know- ledge may affect comparability. For example, between 1861-70 and 1901-10 the rate of mortality from Acute and Chronic Nephritis increased; the in- dex number for males rose from 153 to 435. In 1911-20 it fell to 406 and in 1921-30 to 303. It would be rash to infer that the 'disease' increased and then diminished. A much more probable explanation is that after 1870 doc- tors increasingly spoke of nephritis and eschewed certifying a mere physi- cal sign -- dropsy -- which might be due to many diseases; and after 1910 they preferred to certify arteriosclerosis, thinking, no doubt rightly, that the change in the kidneys was secondary to changes in the vascular system."' Whether you believe Greenwood, who was a distinguished student of the history of medicine, or do not accept his interpretation of this parti- cular case and believe his general statement exaggerated, there remains a very great deal of truth in his position, particularly in periods when med- ical knowledge is rapidly increasing and conditions of life are changing a ~ good deal in the course of a human lifetime. Dr. Schereschewsky's study (~!1 0 came to the conclusion that something between 25% and~30;6 of the increase M on reported for cancer between 1900 and 1920 was real, but the rest was only ~ ~

3 with the reported facts but with all sorts of possibilities for errors in the reports, for changes in the matter of diagnosis, for changes in.man- ner of classification, etc. Such a mind and such an awareness Dr. Schereschewsky had; for there had been much previous discussionof the ris- ing cancer rates in this country and abroad, and'he knew the literature. On the general subject of the difficulty confronting a student of the history of the incidence of disease let me quote fr=a more recent paper by Major Greenwood (12)~ as follows: "One should never believe that a disease is becoming more or less deadly until all other explanations have been excluded... Change of fashion in certification due to increasing know- ledge may affect comparability. For example, between;1861-70~and 1901-10 the rate of mortality from Acute and~Chronic Nephritis increased; the in- dex number for males rose from 153 to 435. In~19]11-201it fell to 406 and in 1921-30 to 303. It would be rash to infer that the 'disease' increased!and thendiminished. A much more probable explanation is that after 1870 doc- tors increasingly spoke of nephritis and eschewed certifying a mere physi- cal sign -- dropsy -- which might be due to many diseases; and'after 1910 they preferred to certify arteriosclerosis, thinking, no doubt rightly, that the change in the kidneys was secondary to changes in the vascular system." Whether you believe Greenwood, who was a distinguished student of the history of medicine, or do not accept his interpretation of this parti- cular case and~believe his general statement exaggerated, there remains a very great deal of truth in his position, particularly in periods when med- ical knowledge is rapidly increasing,and conditions of life are changing a good d'ea1 in the course of a human lifetime. Dr. Schereschewsky's study came to the conclusion that something between 25% and 30% of the increase reported'for cancer between 1900 and 1920 was real, but the rest was only

4 apparent (28y 29). How good'his estimates were I have no id'ea, but I do know that in one of those 21 years there was a marked rise in breast cancer from what it had been previously or what it was afterwards. That there was really any such discontinuity across that one year was inconceivable. In- quiry at the Bureau of Mortality Statistics in Washingtomrevealed the fact that the Bureau had sent out a special inquiry to a large number of physi- cians who had returned death certificates which seemed to the classifiers at the Bureau~likely to have been deaths from breast cancer instead of from the cause certified, and it had been from the obligimg answers returned by the doctors that the excess deaths from that cause were reported for that year. It would have happened~the next year if the same inquiries had been sent out in the same circumstances. What may happen over a time period may be seen in Massachusetts in our figures for deaths from appendicitis. They were 242 in a population of 2,800,000 in 1902, making a death rate from this cause of 82 per 100,000; in 1912 the rate was 92; in 1922 it was 122, and in 1932 it reached 14Z. All this time the operative technique was probably improving and the cure rate should~certainly not have been decreasing. Did we have really a 30 year rise in the incidence of appendicitis? The rate is now down to about 1 per 100,000. In the early days of this century cancer of the lung was so little in the eye of physicians and registrars of deaths that it was not separately tabulated. In 1932 the rate was 4 per 100,000 in Massachusetts; by 1942 it was up to 9; doubling in ten years; by 1952 it had doubled again to 18', and in 1957 it was 232 per 100,000 in this State. These are rough crude death rates; for careful work one has to use variously ad~justed rates to insure comparability, but such details are not necessary for this talk. I should,

5 however, point out that the cure rates differ greatly for different dis- eases, even for different sites of cancer, so that if one desires to com- pare the liability to different diseases or sites of a disease, one must use incidence rates if the cure rates differ substantially. For cancer of the lung and of some other sites the cure rates are presently so low that the death rates can be taken as incidence rates. One has, however, always to bear in mind that both death rates and incidence rates are reported rates subject to the reservations inherent in the accuracy and'comparability of the reporting,in different places and at different times. If yow follow Greenwood quite literally in his somewhat d'ogmatic statement that "one should never believe that a:d!isease is becoming more or less deadly until all other explanations have been excluded," you will be a long time getting,around to attributing causes for the changes in reported figures. I have seen in recent years a number of studies by sincere people, who must be considered competent and honest, indicating,that various amounts from relatively little to a large proportion of the reported increase in lung cancer are not real (10, 11, 16, 20, 21, 23, 24, 25, 26, 27, 30. As a result we do not know how much~of the increase we have to explain and how much we have to explain away. This I take to be one of the Enigmas in this subject. Thirty-five years ago, soon after I went to the Harvard School of Public Health, a great stir was made about the poor facilities in the State of Massachusetts for the care of 3indirgent cancer patients, for the purchase of radium, and for the establishment of a State Cancer Hospital. There were a number of bills before the Legislature at different times, and final]1y one passed which directed the Departments of Public Health and of Public Welfare to make a joint study of the situation with special attention to inoperable

6 cases and'the number of beds available for them. The study was made and reported in House Document 1200, General Court of 1926. The field work for the study was made by Dr. Herbert L. Lombard, who retired last Thursday (December 31, 1959) after 35 years of distinguished service to this State in administrati=and research on cancer and its control. He has put the story together in mimeographed form in a 170-page document: "The Massachusetts Cancer Program." Dr. Lombard had been a student in the School who had shown great interest in statistics and, when he began his field work for the study of the cancer situation in the State, was naturally in close contact with me and with Dr. Doering who was a member of my department. One of the early studies which Dr. Lombard conducted with the aid of Dr. Doering was a dis- cussion of the data collectedfrom a small sample of 217 cancer patients and 217 controls, matched for sex and age, each containing 55 men and 162 women (18). Use of tobacco was one of the questions asked, because it had been suggested that the use of tobacco was related to cancer of certain~sites, namely, lip, jaw, cheek, and tongue, all other sites being supposed to be unrelated to tobacco. In those days women did not smoke or did not admit to i,t, so the study of any relation of tobacco to cancer had to be restricted to men, and the sample of 55 was too small to show anything by site. How- ever when cancer as a whole was considered, it did~appear that "heavy smoking has some relation to cancer in general." This conclusion has been corrobo- rated by much larger studies since then. In 1927 lung cancer was not consid'ered~as important; also ciga- rettes were very little smoked. Pipe smoking was the favored form of using tobacco, with cigars next, and chewing was more frequently reported than cigarettes -- at least in this sample of 55 male cancer patients and their

7 55 matched controls. In respect to the form in which tobacco is used there has been a great change over the years, and if the form in which i;t is used is of great importance in' the incidence of any type of disease we cannot accept Greenwood's generalization. Despite the smallness of the sample it is interesting to go back to the old tables of Lombard and Doeri~ng to note what are the statistical results by type of tobacco use. Pi;pe smoking was significantly assoeiated with cancer; cigarette smoking was not, though by any formula used to measure the amount of association it would show almost as much associationy the lack of significance being due to the small number of users of cigarettes; cigar smoking was negligibly associated with cancer, as it seems to be in recent studies; but chewing had a significant associa- tion of about the same magnitude as pipe smoking. Recent studies omit chew- ing, and they do not show the association with' pipe smoking which Drs. Lombard and Doering found (2,3,4,5,6,7,8,13,14-). Statistics, even when sig- nificant by the accepted tests, can be tricky, and conditions can change. It is not easy to be sure that what appears to be a control is in fact a good one. In another paper of about the same old time, more than 30 years ago (L7), Drs. Lombard and Doering showed that there was a great dif- ference in the reported rates for cancer among the native born of native parents (for whom the rate was 85), the native born of foreign parents (for whom' the rate at 160 was nearly twice as much)i, the native born of mixed parents (l1.l)', and the foreign born (137,)'. It can therefore be seen thaty both in samples and controls, ma.tching with respect to these variables would be desirable. Nobody ever can be sure that there is no unnoticed variable which has a large enough effect to make an appreciable bias be- tween the sample and its supposed control.

8 This difficulty I will not call an enigma of cigarettes and~lung cancer, or of smoking and cancer in general, but of statistics in general. And now how did the poor little cigarette get into this messy situation in which we find it? In 1933 Dr. Evarts Graham of St. Louis was operating on a fellow physician for lung cancer, expecting to have to take out only a part of a lung, but he found the condition so bad~that it was necessary to remove the whole lung -- and the patient lived:. That was the first successful removal in one stage of an entire lung. It was very suc- cessful, for the patient was still living and practicing his profession, as an obstetrician, twenty years later when he was good~enough~to come to Boston to help open our April cancer drive; he is still alive and active, at last reports. On March~ 4, 1957, DDr. Graham died' of lung cancer; he could not be saved because both lungs were involved. But in the intervening years he inferred from his clinical observations that persons with lung cancer had usually been unusually heavy cigarette smokers, and thus pointed the finger of suspicion at the cigarette as a possible causative agent in most cases. I say "in most cases" because there are cases of lung cancer where the pa- tient has never smoked cigarettes or anything else and there is no known agent of any kind on which one can lay the blame. Also there are old per- sons who have smoked cigarettes furiously for many years and seem to be in ~ •perfect]1y good health for their age. You can note the same of alcohol, or C underweight or overweight, and a great many other things that seem to be QQ ~ bad for some individuals. It has long been remarked in the proverb: One ~ man's meat is another man's poison. C Probably many medical generalizations and many generalizations in other fields have been made by observant practitioners of the arts and crafts and professions. One hears a nauseous amount of talk about how much more

9 important it is to do basic research than applied; if there had been no applied research there might have been no tools, or even ideas, with which to do basic research. But one thing seems certain; the basic research folks distrust the generalizations made by the practitioners quite as much as the practitioners distrust the wisd=of the theoreticians in practical matters. And so when Dr. Graham and his collaborators had pointed a finger at the cigarette as responsible for many of the cases of lung cancer, two studies were set up systemmatically to check the generalizationy one by Doll and Hill in Englandy one by Hammond and Horn working for the American Cancer Society. Dr. Doll and Professor Hill asked the physicians of Britain to let them know what and how much they smoked and~have been following them ever since, and I hope will long continue to follow them, for lengthening follow- up rapidly increases the total yield of this sort of study (2-7). Hammond and~Horn asked a large number of the volunteer workers of the Cancer Society to find out what and how much ten of their male acquaintances between 50 and 69 years of age smoked and further to keep them informed as to their being still alive or to have died (13, 14). The follow-up was good on nearly 200,000 men for about four years, when it was stopped; for one cannot expect to keep the interest of volunteers indefinitely. These two studies have shown that the rates of death from lung cancer among cigarette smokers have been much in excess of those of non- smokers or cigar or pipe smokers and that the rates of death have increased with the rate of cigarette consumption. They have done more than this; for they have shown that the rate of death from all causes is greater for ciga- rette smokers and~that the same is true for a considerable number of indi- vidual causes of deathy though not for all. There is a further study of