Tobacco Documents Online

i I I I 1 I I I I I I I I I i I a I ASPECTS OF THE EPIDEMIOLOGY OF LUNG CANCER IN SMOKERS AND NONSMOKERS IN THE UNITED STATES Geoffrey C. Kabat Albert Einstein College of Medicine Bronx, New York, USA Abstract While it is well-established that smoking is the predominant risk factor for lung cancer, it is clear that factors other than smoking and occupational exposure must play a role in some lung cancers, and particularly adenocarcinoma. Data from a large, hospital-based case-control study are used to examine the association of smoking-related risk factors (amount smoked, filter status, mentholation, and differences in smoking habits between blacks and whites) and selected factors other than smoking (environmental tobacco smoke, previous primary cancer and radiotherapy, reproductive and endocrine factors, and body mass index) with lung cancer. Although smoking shows a dose-response relationship with all major lung cancer cell types, the strength of the relationship is weaker for adenocarcinoma, suggesting that other risk factors must play an important role for this cell type. In both blacks and whites of both sexes, odds ratios for lung cancer increased with increasing cumulative tobacco tar intake and decreased with years since quitting smoking. Use of mentholated cigarettes was not associated with lung cancer, relative to use of nonmentholated cigarettes. Exposure to environmental tobacco smoke in childhood was associated with a doubling of the odds ratio in female never smokers; however, exposure to a husband's smoking showed no increase in risk. A history of a reproductive primary and a history of radiotherapy were each associated with a four-fold increase in risk in female nonsmokers. An association of lean body mass with lung cancer was observed in current smokers, ex-smokers, and female never smokers. These results are discussed in the context of existing studies. In conclusion, variation in lung cancer rates between populations may be due to: 1) differences in effective exposure to tobacco smoke carcinogens, 2) differences in exposure to other independent risk factors for lung cancer, or 3) differences in factors which modify the effect of tobacco smoke, including differences in host susceptibility and metabolism of carcinogens. The Magnitude of the Problem Introduction Lung cancer, which in the early twentieth century was a rare tumor in the United States, is now the leading cause of cancer death in the United States and has recently become the most common tumor worldwide (1,2). The American Cancer Society estimated 170,000 new lung cancer cases in 1993, accounting for 17% of cancer incidence in males and 12% in females, and 149,000 deaths from lung cancer, or 34% of cancer deaths in males and 22% in females (1). Between 1930 and 1989 the lung cancer mortality rate increased from 4 to 73 per hundred thousand in males and from 3 to 30 in females (1). In recent years the lung cancer death rate in males has begun to level off and decline, but in females it has continued to rise at a steep rate. During this I

I period no other major tumor shows an increase in incidence remotely comparable to that of lung cancer, although the decline in stomach cancer is almost as dramatic. Differences in age-adjusted incidence rates for specific histologic types of lung cancer by sex, race, and calendar time period suggest that different histologic types may have different etiologies. In white men in the United States, rates of adenocarcinoma and small cell carcinoma increased over the time period 1969-1988, while the rate of squamous cell carcinoma decreased. In white women, all major types showed an increase (3). Squamous cell carcinoma is still the predominant histological type among males, whereas adenocarcinoma predominates among females (3). Among nonsmokers, the proportion of adenocarcinoma is greater than in smokers (3), and particularly in females, reaching 76% in one large series of lung cancer cases in nonsmoking women (4). While cigarette smoking and specific occupational exposures have been well established as important risk factors for lung cancer, there is increasing recognition among epidemiologists that smoking and occupational exposures cannot explain all of the variation in lung cancer incidence within countries and between countries, and that other factors must play a role either as independent risk factors or as modifiers of the effect of smoking (5,6). Some issues that remain to be elucidated include: 1) the high rates of lung cancer in Chinese women, who have a low prevalence of smoking; 2) the higher incidence of lung cancer in black American males in the United States; 3) etiologic factors for adenocarcinoma other than smoking; and 4) risk factors for lung cancer in lifetime nonsmokers; and 5) factors that account for the increase in the incidence of adenocarcinoma. In this paper, I will focus on a number of risk factors for lung cancer, including certain aspects of smoking (amount smoked, filter use, mentholation, and black-white differences in smoking habits) and the role of certain factors other than smoking (including environmental tobacco smoke (ETS), a history of radiotherapy and of previous cancers, reproductive and endocrine factors, and body mass index). In addition to presenting data from our studies, I will refer to other relevant studies and attempt to suggest directions for further research. Methods Since 1969 the American Health Foundation has been conducting an ongoing, hospital-based, case-control study of tobacco-related cancers in a number of U.S. cities. The objectives of the study were to examine in depth the role of smoking in its various parameters as well as other environmental and personal factors in the etiology of cancers of the lung, larynx, esophagus, oral cavity and pharynx, bladder, kidney, and pancreas. Cases enrolled in the study were newly-diagnosed, histologically-confirmed primary cancer of the above-mentioned sites. For each case, a hospitalized control was sought who was matched on age (±5 years), sex, race, hospital, and date of admission. Controls were patients with conditions not thought to be associated with smoking. Thus, smoking-related malignancies as well as nonmalignant smoking-related conditions (such as chronic bronchitis, emphysema and peripheral vascular disease) were excluded. However, controls and cases could have a history of nonmalignant smoking-related conditions. -2- I I I I I I I  I I I I I

I I I  I I I I I I I I I I It should also be pointed out that although a previous history of a tobacco-related cancer was inadmissible, a previous history of other, nonsmoking-related tumors was acceptable. The age range of subjects was 20-80. All subjects were interviewed in-person in the hospital by trained interviewers using a standard questionnaire containing a detailed history of tobacco use and questions on usual occupation and occupational exposures, alcohol consumption, demographics, and other factors. Within this large study, a substudy of lung cancer in lifetime nonsmokers was initiated in 1983 with a primary focus on environmental tobacco smoke. In the substudy, for each nonsmoking lung cancer case interviewed, 3 control patients who were lifetime nonsmokers and were matched on age (± 5 years), sex, race, hospital, and date of interview (within 2 months) were sought. Subjects were considered lifetime nonsmokers if they had never consumed as much as 1 cigarette per day for a year, or had smoked fewer than 365 cigarettes over their lifetime. The inclusion of detailed questions regarding the initiation of smoking early in life and amount smoked provides a basis for excluding ex-smokers who quit decades prior to diagnosis but have smoked more than this minimum amount. In the substudy, detailed questions were asked about exposure to other people's smoking in childhood in the home, in adulthood both at home and at work, and in social situations and transportation. For each smoker in the household and for each job held for at least one year, subjects were asked about the intensity and duration of the exposure. Exposures were examined individually and in the aggregate for different periods of life. Over the period 1969-1991, the questionnaire was modified a number of times (there are 4 major distinct versions in this time period). While some items remained constant (core smoking history items), others were amplified and new items were added. For this reason, in the analyses I will present, the time period will vary for different analyses. For the period 1985 to 1991, a number of questions were added to the questionnaire. These included: a brief food-frequency questionnaire designed to estimate intake of dietary fat and of vitamin A; a prior history of radiotherapy (yes/no), the site treated, the diagnosis requiring treatment; and a history of (a nonsmoking related) cancer. Results Amount Smoked Smoking-Related Factors . In an analysis of 2,085 male and 1,012 female lung cancer cases and two times as many matched  controls, we noted a dose-response relationship with level of smoking for Kreyberg I (including squamous cell, small cell, and large cell lung cancers) and Kreyberg II (including adenocarcinoma, bronchiolar, and ' alveolar cell carcinoma) in current smokers of both sexes (7). Odds ratios were consistently higher for Kreyberg I(Ki) compared to Kreyberg II (KII). The OR for smokers of 41 + cigarettes per day reached 64.1 for KI compared to 28.4 for KII in males. The corresponding ORs in females were 88.7 and 20.1 N ~ (Table 1). O O ~ -4 tp ~ -3- W O tn I

Among males, the proportion of never smokers was 2% for Kreyberg I and 5 % for Kreyberg II and among females the proportion of never smokers was 6% and 19%, respectively. Filter Versus Nonfilter Cigarettes Use of filtered cigarettes was associated with reduced risk for Kreyberg I lung cancer in both sexes, with an apparent trend with increased duration of smoking filtered cigarettes (Table 2). Only the odds ratio for male switchers who smoked filter cigarettes for at least 10 years was statistically significant (7). There was little indication of a reduction in risk for Kreyberg H lung cancer. Mentholation Black males have approximately 50% higher lung cancer incidence compared to white males in the United States (8). Furthermore, blacks tend to favor mentholated cigarettes. While combusted menthol does not appear to be carcinogenic, it has been suggested that use of mentholated cigarettes may be associated with a different exposure to tobacco smoke constituents, either because these cigarettes have higher average tar-yields or, additionally, because they are smoked differently, possibly due to menthol's anesthetic properties (9). For this reason, we undertook an analysis of use of mentholated versus nonmentholated cigarettes among current smokers (10). Detailed information on specific brands of mentholated cigarettes was available in our dataset for the period 1985-1990, and therefore the analysis is limited to this period. No significant association was observed between either short-term (1-14 years) or long-term (15 + years) menthol users relative to smokers of nonmentholated cigarettes, after adjustment for other variables. For specific histologic types of lung cancer, there was no indication of an association with menthol usage (Table 3). We also failed to observe a consistent association of mentholated cigarette use with esophageal or oral-pharyngeal cancers (11,12), two other cancers which have a higher incidence in blacks compared to whites. Differences in Smoking Habits Between Blacks and Whites We compared smoking habits of blacks and whites in a total of 23,011 case and control patients enrolled in the American Health Foundation study between 1980 and 1990 (13). Blacks of both sexes among cases and controls were more likely compared to whites to be current smokers and less likely to be ex-smokers. Furthermore, blacks were approximately three times more likely to be light (<20 cigarettes per day) versus heavy smokers (> 20 cigarettes per day). This association did not differ according to cigarette preference, degree of inhalation, or quitting. The association of race and light smoking was present in both current and ex-smokers. Sociodemographic or smoking-related characteristics did not appear to explain this difference in smoking habits. Possible differences in the association of smoking with lung cancer between blacks and whites prompted us to carry out a case-control analysis utilizing all lung cancer cases in these two ethnic groups enrolled in the study between 1979 and 1990. In all four race-sex groups, a dose-response was observed with lifetime tar intake (Table 4). However, the odds ratios for blacks were somewhat higher than those of whites at a given exposure level. For example, the adjusted OR for extreme quartiles of tar intake in males were 4.3 (95% CI 3.6-5.2) in whites and 5.7 (9546 CI 3.0-10.9) in blacks. A reduced risk of lung cancer was seen among quitters compared to current smokers, but the effect of quitting was somewhat -4- I I I I I I I I I I I I I ' '

I I I I I I I I I I I I I I I I more pronounced among whites of both sexes (Table 5). The distribution of histologic types, while differing between males and females, was virtually the same in blacks and whites of the same sex. Risk Factors other than Smoking Environmental Tobacco Smoke in Lifetime Nonsmokers In the substudy of lung cancer occurring in lifetime nonsmokers, the majority of cases and controls reported a history of exposure to ETS (i.e. at least 1 year) in childhood or adulthood, either at home or at work. Table 6 gives the distribution of exposure in different settings. Only 7-15% of subjects reported no exposure to ETS (i.e. for at least 1 year) in childhood, adulthood in the home, or at work. In males, the OR was elevated for having a spouse who smoked (1.6, 95% Cl 0.7-3.8) (Table 7). Excess risk was limited to those whose spouse smoked 11+ cigarettes per day (OR = 7.5, 95 % CI 1.4-41.4), however the numbers were very small. The OR was elevated for those whose spouse smoked in the bedroom. In females, there was no indication of an association with exposure to spousal smoking, in spite of the larger sample size (OR = 1.1, 95% CI 0.6-1.9). When duration of ETS exposure (smoker-years) was examined (Table 8), females in the highest tertile of smoker-years in childhood had twice the risk of lung cancer (OR = 2.2, 95 % CI 1.1-4.5) and the linear trend across increasing tertiles was significant (p = 0.02). Previous Cancer and Radiotheranv For the period 1985-1990, we had additional questions including previous history of cancer and history of radiotherapy. In male never smokers, neither a history of a previous primary cancer nor a history of radiotherapy was significantly associated with lung cancer, however, the numbers of exposed cases were small (14). In female never smokers, after adjustment for age, education, hospital, ETS exposure, and body mass index, both a history of a reproductive primary and a history of radiotherapy were significantly associated with lung cancer (OR = 4.9; 95% CI 1.4-17.7, and OR = 4.4; 95% Cl 1.3-15.1, respectively). When female never smokers cases were limited to 32 cases of adenocarcinoma of the lung, the adjusted OR for having a previous reproductive primary was 4.0 (95% CI 0.9-17.6) and that for having a history of radiotherapy was 4.3 (95% CI 1.1-16.6) (Table 9). The OR for lung cancer among women who had both a history of radiotherapy and a previous reproductive primary, relative to those who had neither, was 18.8 (CI = 2.2-160.7). In female never smoker cases, the mean age at diagnosis of the first primary was 47 years (SD 12.9) and that of the lung primary was 62 years (SD 14.4). Histologic information was available for only four of the seven reproductive cancers (Table 10). No association of previous primary cancer or of history of radiation treatment with lung cancer was observed among ever-smokers of either sex. When attention was restricted to long-term ex-smokers (those who had quit more than 20 years prior to diagnosis), the adjusted OR for those reporting a previous reproductive primary was 1.3 (95% CI 0.3-7.0) and that for a history of radiation therapy was 3.3 (95 % CI 1.0-11.7). No interaction was observed between radiotherapy and previous cancer in ever- smokers. -5- I

Body Mass Index Several prospective studies have reported an inverse association between body mass index and lung cancer (15). Few of these studies, however, adjusted for cigarette smoking. We analyzed data on 3,607 lung cancer cases and 9,681 controls interviewed between 1981 and 1990 in order to determine whether the reported association could be due to confounding by smoking status, amount smoked, or other factors (16). Patients in the study were asked their weight five years prior to diagnosis. Separate analyses were carried out by smoking status and by sex. After adjustment for covariates, odds ratios for lung cancer by levels of body mass index, taking > 28 as the referent, showed an increasing linear trend with decreasing body mass index for current smokers and ex-smokers of both sexes and for female never smokers (Table 11). The association of leanness with lung cancer did not vary by level of other variables, including age, education, lifetime tar intake, alcohol intake, or race. In order to verify that the association was not due to overrepresentation of overweight subjects among the controls, we excluded diagnoses associated with obesity (cancers of the breast, endometrium, ovary, and gallbladder; fractures; back problems; arthritis; diabetes; and endocrine and metabolic disorders). This had no effect on the results. Comparison of the distribution of body mass index in the control group to that of the general U.S. population showed that both groups were similar. Discussion Association of Cigarette Smoking with Different Histologic Types In 1962, Kreyberg (17) classified lung cancer into two groups. Group I consisted of squamous cell, large cell, and small cell carcinomas, which he hypothesized were primarily due to exposure to external factors that had recently increased in prevalence and to which men were primarily exposed. Group II was heterogeneous, consisting of adenocarcinomas, bronchiolar (alveolar cell) carcinomas, and several other types, which Kreyberg hypothesized to be due to different etiological factors, possibly including developmental abnormalities, virus infection, and other factors of constant prevalence to which males and females were equally exposed. While Kreyberg's classification provides a valuable working hypothesis, it needs to be modified in the light of more recent epidemiological data. Studies examining the relationship of smoking to different histologic types show that smoking is associated with all 3 major types and shows a dose-response relationship (18-22). However, the magnitude of the association with amount smoked is, in most but not all studies (19), considerably stronger for squamous and small cell carcinoma than for adenocarcinoma. This tendency appears to be consistent between studies done in the United States and in China (23). The largest and most thoroughly analyzed study (19) indicates that duration of smoking is a more important variable distinguishing cell types than amount smoked. Duration of smoking showed a dose-response for all cell types, but the magnitude of the association was greater for squamous cell carcinoma than for adenocarcinoma. Part, but not necessarily all, of the explanation for the weaker association of smoking with adenocarcinoma is the fact that the proportion of never smokers among adenocarcinoma cases is larger compared to other cell types (19,24,25). Yang et al. (26) have suggested that the increase in adenocarcinoma of the lung may be explained in part by secular changes in cigarettes, specifically the increasing use of filter cigarettes starting in the -6- I I I I I I I I I I I , I I I

t rr . 1950s, which may be associated with inhalation of smaller particles and their deposition in the periphery . of the lung, where adenocarcinomas are thought to arise. I I I i I I I I The weaker association of smoking with adenocarcinoma compared to squamous and small cell carcinomas implies that other factors must play a proportionately greater role in adenocarcinoma. Black-White Differences Our data, and those from other studies, indicate that black smokers are less likely to quit smoking and that those who smoke, smoke fewer cigarettes per day compared to white smokers. Blacks had somewhat higher odds ratios for lung cancer compared to whites at given level of tar exposure and the reduction in the odds ratio associated with quitting was generally smaller in blacks compared to whites; however these differences were modest, and the design of the study precludes a direct comparison of whites and blacks. In a separate study, we found no elevation in lung cancer risk among smokers of mentholated cigarettes compared to smokers of nonmetholated cigarettes. Our results regarding black-white differences in smoking habits are compatible with a greater effect of smoking in blacks compared to whites. This could be explained by use of higher tar cigarettes by blacks. Recent work demonstrates that smokers of mentholated cigarettes have higher serum cotinine levels compared to smokers of nonmentholated cigarettes (personal communication from Dr. Pamela Clark). An alternative explanation of a greater lung cancer "response" in blacks for a given level of tobacco exposure is that the metabolism of nicotine may differ between blacks and whites, since, for a given level of smoking, blacks have been reported to have higher levels of serum cotinine levels compared to whites (27). Another possibility is that independent risk factors for lung cancer or effect modifiers of smoking account for the excess incidence in black males (28). Potential factors include: occupational exposures, diet, and altered genetic susceptibility for lung cancer (29-31). Environmental Tobacco Smoke , While our study was of modest size, we attempted to improve the quality of the exposure information by asking detailed questions about exposure in different settings throughout life. These , included: the average number of hours per day of exposure to different smokers in the household; whether the spouse smoked in the bedroom; and how the respondent rated each exposure. In addition, unlike many of the larger studies, all cases and controls were interviewed in person at the time of . diagnosis, thus eliminating the need to make use of proxy respondents. I Our data show a modest association of ETS exposure in childhood with lung cancer in women, but no suggestion of an association with a husband's smoking. The problem here is that bias could exert a greater role in recall of exposure in childhood than in recall of adulthood, and particularly, spousal exposure. ~ Since adenocarcinoma is the most common form of lung cancer in never-smokers, one would expect that if ETS is a detectable risk factor for lung cancer in nonsmokers that one would see the N _ strongest relationship for adenocarcinoma. Wynder and Goodman (32) have hypothesized that ETS may 00 , i V tb _ '] - 4f O t0 I

have a greater effect on adenocarcinoma than on other cell types because the volatile components in sidestream smoke may be able to penetrate to the periphery of the lung. Several areas of inconsistency should be noted in studies of ETS and lung cancer. First, studies carried out in the United States have yielded contrasting results regarding an association with spousal smoking. The largest study, by Fontham et al. (4), reported an OR of 1.29 (95% CI 1.04-1.60) for having a smoking husband, with an increasing trend with increasing pack-years of exposure, reaching an OR of 1.79 (95 % CI 0.99-3.25) for nonsmoking women who had > 80 pack-years of exposure. Stockwell et al. (33) also noted a significantly increased OR (1.6, 95% CI 0.8-3.0) for exposure to a husband's smoking and an OR of 2.4 (95 % CI 1.1-5.3) for > 40 smoker-years in adulthood. Brownson et al. (34) reported a borderline elevated odds ratio for the highest level of cumulative exposure only (OR for > 40 pack-years of exposure from all household members = 1.3, 95 % CI 1.0-1.8). In contrast, other studies, including those by Janerich et al. (35), Wu et al. (36), and the present study provide no evidence for an association with spousal smoking. Second, there is inconsistency among studies with regard to the relationship of ETS to specific cell types. Fontham eta l. (4), whose large series of nonsmoking female lung cancer cases was predominantly adenocarcinoma (76%), showed a dose-response relationship for this type. Several studies, however, noted an association of ETS with squamous or small cell carcinoma but not with adenocarcinoma (37,38); two studies showed stronger associations with squamous cell carcinoma (or all types other than adenocarcinoma) than with adenocarcinoma (39,33); and one study reported an association only for "other/mixed" cell types, of which the numbers were small (34). Third, several studies indicate an association of ETS exposure in childhood with lung cancer (33,35), whereas others offer no support for an association (4,34). While an association of ETS exposure with lung cancer in never smokers has compelling biological plausibility and potentially important health implications (40,41), these inconsistencies as well as those from studies in other countries, point up the difficulties in quantifying ETS exposure throughout life and in detecting what on average among those passively exposed may be a small excess increment in risk. Radiation Treatment. Previous Renroductive Pritnarv, and Hormonal Factors Radiation treatment and a history of a previous reproductive cancer were so highly correlated among the female cases in our data that it was not possible to examine the effect of one factor independent of the other. These results are based on small numbers of never smokers and need to be confirmed in larger studies. Their interest lies in suggesting that endocrine-related tumors, or their risk factors or treatment, may increase the risk of lung cancer in nonsmoking women. Radiation is an established lung carcinogen (42-44), and radiation therapy, particularly for a previous breast cancer could have contributed to subsequent lung cancer (45). However, only three of the seven never-smoked lung cancer cases with a history of radiotherapy had a first primary of the breast. And the association of radiation treatment with lung cancer was comparable for those who reported radiation directed at the abdomen and neck. -8- I I I I I I I I I I I I I 1 I I I

I r I I I ~ I I I I I I I I I I I I An alternative explanation compatible with the results reported here is that endocrine factors may play a role in the development of lung cancer. The fmding of an association of a previous reproductive primary with lung cancer in female nonsmokers is consistent with reports, based on tumor registry data and one prospective study, indicating an increased incidence of second primaries of the lung in women who had a first primary of the breast, endometrium, or other reproductive sites (46-48). Only one of those studies took smoking history into account. Annegers and Malkasian (48) noted that six cases of lung cancer (compared with 1.2 expected) occurred among 526 patients with endometrial cancer who were followed for at least 10 years. The mean interval between diagnosis of endometrial cancer and lung cancer was 14.8 years, comparable to that observed in our data (15 years; see Table 10). All five of the six lung cancer cases who had adenocarcinoma were nonsmokers. The possibility that endocrine factors may play a role in the development of lung cancer is raised by: (1) the greater proportion of nonsmokers and of adenocarcinomas among female compared to male lung cancer cases (24,25); (2) the presence of steroid receptors in some lung tumors (49-52); (3) the greater than expected incidence of lung cancer among female survivors of a primary of the reproductive organs (46-48); (4) an association of short menstrual cycle and late age at menopause with lung cancer (53); (5) an association of estrogen replacement therapy with lung cancer (54,55); and a more frequent than expected family history of reproductive cancer among female lung cancer cases compared to controls (56). Gao et al. (53) noted that women with late menopause were at increased risk of adenocarcinoma of the lung and that the OR increased with decreasing length of the menstrual cycle, with a 3-fold excess among women who had shorter cycles. In a study of adenocarcinoma of the lung in women, Wu et al. (57) reported a decreased OR for lung cancer in women who had taken oral contraceptives compared to those who had never used oral contraceptives (OR for <2 yrs = 0.9, 95 % CI 0.5-1.6; OR for > 2 yr = 0.4 (95 % CI 0.2-0.8). Cases and controls did not differ significantly in their use of estrogen replacement therapy. Taioli and Wynder (55) examined reproductive history and endocrine factors in relation to adenocarcinoma of the lung. Of the reproductive variables, an early age at menopause, < age 40 years, was associated with decreased risk of adenocarcinoma (OR = 0.3; 95% CI 0.1-0.8). Use of estrogen replacement therapy (ERT) was associated with adenocarcinoma (OR = 1.7, 95 % CI 1.0-2.8 for ever users relative to nonusers). There was a significant interaction between smoking and ERT. Relative to those who neither took ERT nor smoked, the OR for women who smoked and used ETR was 32.4 (95 % CI 15.9-665.3) and among women who only smoked, the OR was 13.1(95 R6 CI 6.8-25.2). Women who took ERT but never smoked had an OR for adenocarcinoma of 1.0 (95 % CI 0.3-3.8). At present, the existing evidence for a role of hormones in adenocarcinoma of the lung is limited and circumstantial. Some of the studies (i.e. Adami et al•) did not adjust for smoking or distinguish between histologic types. Further studies are needed that would obtain in-depth information on reproductive and endocrine factors as well as differences in serum and urinary estrogen levels between lung cancer cases and controls. N 0 00 ~ co W 0 -9- _' ~ I

Body Mass Index While increased body mass index is a risk factor for certain cancers, e.g. post-menopausal breast cancer and endometrial cancer, several studies indicate an inverse association with other tumors (including those of the lung, larynx, esophagus, bladder, and stomach) (58). Since these cancers are associated with smoking and since smokers tend to be leaner than nonsmokers, it is crucial to adequately adjust for smoking. The large numbers of cases and the detailed information on smoking history permitted us to stratify by smoking status and to adjust finely for cumulative tar intake in current and ex-smokers. Such adjustment did not reduce the magnitude of the association, but, rather, slightly increased it. Furthermore, the strongest association of leanness was observed in women who never smoked (Table 11). Self-reported weight 5 years prior to diagnosis was used in computing body mass index in order to minimize any effect of weight loss due to disease in the period immediately preceding diagnosis. However, when body mass index based on weight 1 year prior to diagnosis was used, the results were unchanged.. Further studies are needed to confirm whether there is in fact an association of lean body mass with lung cancer risk independent of smoking status and amount smoked and whether such an association is with leanness which is not a consequence of disease. If this association is not due to confounding or to weight loss due to disease, possible explanations include: 1) an association of leanness with decreased levels of nutrients that may be protective or with increased levels of dietary risk factors; 2) an association of leanness with increased metabolic rate and with accelerated cell turnover in the lung. Conclusions 1. Differences in smoking habits and tobacco products may contribute to observed differences in lung cancer incidence between populations. It is important to quantify lifetime tobacco intake (taking into account the tar-yield of cigarettes smoked throughout life, as well as amount smoked at different periods of life and duration of smoking, inhalation, etc.). It is also necessary to examine different histologic types separately, since the association with smoking differs by histologic type. 2. In addition to differences in tobacco smoke exposure, differences in lung cancer rates could be due to: 1) differences in host susceptibility, including the metabolism of carcinogens; 2) differences in exposure to independent risk factors; or 3) interactions of smoking with other risk factors or protective factors. 3. Since there is consistent evidence that cigarette smoking is not as strong a risk factor for adenocarcinoma as it is for squamous and small cell carcinoma, other factors must play a major role in the etiology of adenocarcinoma. -10- ' I I I I I I I I I I I I I