Tobacco Documents Online

e C C ` ~' ~, a~ ~ "V 7 DR. KABAT: Thank you. P9 ~ 0 s!,I w', ~ Tc r / 9 9 0, /~. B 7- The problem of passive smoking and lung,cancer has provoked a good deal of debate bothh on a scientific and on a public policy level. Do the studies that purport' to ~ show an association of' exposure to environmental tobacco smoke (ETS) and lung cancer occurruig;in lifetime nonsmokers provide adequate evidence to resolve the issue? As Nancy Haley has just shown, she and her, colleagues are very good I at' measuring recent exposure to ETS using cotiniae measured in saliva, serum, and urine. Unfortunately, these biomarkers are not helpful for assessing exposure over the severall dt.cades relevant to the induction of lung cancer. Given the lack of a biomarker for long- term exposure to ~ ETS, epidemiologic studies have had to rely on self-reports or proxy-reports of ETS exposure. I propose to raise what' I consider to be some of'tlie key aspects of the roughly 15 epidemiologic studies of'the issue of' ETS and lung cancer and to point' out certain areas that require further study; I will briefly refer to our own study which is still in progress at the American Health Foundation. Finally,, I! will suggest a possible direction for further study of this issue. EPIDEMIOLAG'IIC Table 1 lists studies exam' g the lung cancer risk of'non-smoking wives of smoking husbands compar',d to the non-smoking wives of'non-smoking husbands: : One notes that the, greatest magnitude of the overall relative risk (RR) is 2.1. After the Tnchopoulos and Cocrea studies, the highest RR is 1.65' (Lam et al.). The national Research Council's committee on passive smoking carried our a mcta-analysis of' the existing studies in 1986 and came up with an overall RAt of 1.34 (95!% aonfidence interval: 1.18-1-53) (1). In four out of the fifteen studies listed, the overall RR is statistically significant. When one examines the data by level of exposure, i,e., number of cigarettes per day smoked' by the husband stratifie& into two or more ltvels, 8 of'the 15 studies show evidence of a dose-response relationship.. HISTOLOGY When~ we look at the effect of ETS' exposure by histologic type, we see an interesting discrepancy (Table 2). Dalager et al. (2) and Pershagen et aL (3) show roughly comparably elevated odds ratios (OR) for squamous cell and small cell carcinomas combines, but not for adenocarcinoma. In contrast, Lam et all (4) obtained a significant effect for adenocarinoma~ but not for squamous cell carcinoma. The results of Hirayama's study (5) presumably agree on this point with those of Lam et al., since the majority of'his lung cancer cases were apparently adenocarcinoma. Trichopoulos et al. results (6) presumably weigh im on the side of Dalager et al. and Pershagen et al., since Trichopoulos excluded adenocarinoma and terminal' bronchial carcinoma from their series. Since adenocarinoma occurs more commonly in never smokers than in~smokers and~ generally more commonly in women than in men (7), one would expect that if' ETS exposure is an appreciable risk factor for lung; cancer, it is associated with adenocarcinoma, as well as possibly with other types. Tlie inconsistency in the results to date regarding histology indicates that this is one area that merits further study. ' 187 ' /'r'7

ASSESSMENT OF DISEASE STATUS' Misclassification on disease status occurs when diagnoses other than primary carcinoma of the lung are included in ! the case series or when, a primary cancer of the lung, is included among the controls due to its having gone undetected: Garfinkel et al. reported that of 283 women.listed: as having lung cancer in hospital rerords but with no mention of'their having smoked, 36 (12.7%'0) turned' out to, have diagnoses other than lung cancer when the histology was reviewed by one of the authors (8). In.istudies in which histologic verification of lung cancer is a criterion for inelusion in the study, n+isclassiFcation should'be minimaL However,,some of'tlte studies listed in Table 1 were lacking this for all cases. It should also be mentioned that even when lung cancer is histologically verified, it is possible that some :cases jttdged'to be primary cancer of the lung are actually secondary to a, cancer of another site that has gone undetected. ASSESS14f I+IT OF EXPOSURE STATUS'. This is a, greater problem than assessment of disease status, and for some investigators it is the key problem of epidemiologic studies of ETS' and lung cancer (9,10). Misclassifieation of'exposure status can occur in a number of'ways. First, subjects who have smoked for some period! of'their life can be erroneously included in a study of never smokers. Second; subjects may under-report (minimize) or over-report (inflate) their ETS exposure, or this may be done by proxies: A third type of misclassification can occur wheni some indirect measure (such as whetber the subject is married to a smoker or how much the spouse smokes) is used as an indicator of ETS exposure. The effect of miselassification ~ on the estimate of the RR depends on whether the misclhssifeation is random or differential (t.hat is systematic). Random misclassification will bias the estimate of the RR toward the null, thus making an effect, if there is one, more diffcult to detect. If misclassification on ex asure differs between cases and controls, the estimate of the RR ' can be biased either upwards or downwards depending on the direction of the bias (11). Mirsclassification of active smokers as never smokers. Garfinkel and co-workers found that among lung cancer cases identified as 'nonsmokers^ or lhcking, any mention 1 of smoking in the hospital record, 40% were revealed to have smoked upon reinterview (8). Although a detailed personal interview yields more accurate smoking,histories than reliance on hospital charts, if is still likely tJ;at, even when subjects are directly'interviewed and moree so when various proxies are used, some misel ' ucation of smokers as nonsmokers occurs. Lee has argued that random misclassification of smokers as non-smokers coupled' with a tendency of smokers to marry smokers could account for the observed association, of a spouse's smoking and! increased'.lung,cancer risk in non-smoking spouses (9). Assuming a 5% misclassification of'smokutg subjects, a RR' of 20 for active smoking, no true elI'ect' of passi've.e smoking, and a between-spouse smoking concordance of 3.45, Lee demonstrates the effects of such a bias. These include an apparent effeci of passive smoking (RR = 1.75) and the creation of a large proportion of'true smokers among the self-reported non-smokers with lung cancer. 188 I 11 I

Misclussif~cation of self-reported ETS erposure. A study by Pron et al. (12) suggcsts that mi' ~~~i!ication of self-reported: ET'S exposure may be euensive. They examined the reliability of responses in 1 117 control subjects who 1 had participated in, a study of passive smoking and who were reinterviewed on average six months later. Responses to an initnal question about exposure to ETS (yes/no) were more reliablt for exposure at home than at work (Tablo 3). Reproduabdity ofquestions concerning exposure to a spouse's smoke (yes/no) was high for both sexes, with the reliability being generally lower for other family members. Quantitative measures ofi E'TS exposure, ie, number and duration of exposures, were generally less reliable than qualitative (or dichotomous) , meastu es. In general, non-smokers gave more reliable information on all parameters of E'TS exposure than smokers. Unfortunately the study by Pron et al, did not examine the reliability of responses among cases as well as among controls. In case-control!studios particularly one must be concerned that the case's reporting of exposure maybe influenced by his diagnosis. In a study of lung cancer occurring in non-smokers, this could take the form of cases probing past exposures more intensively than controls and over-reporting exposures to ~ ETS, since some cases may feel compelled to find an explanation for their disease. On the other hand, it is also possible that cases might mini*ni~r their exposures out of an unwillingness to. blame a spouse. MJsclQSsification due to use the spouse's srnokingltabits. Using the presence of a smoking spouse as an indicator of ETS exposure can lead to serious misclassification of exposure. Based I on a survey of nearly 38,000 never- and ex-smokers, Friedman et aL (13) reported that the sensitivity and speci6dty of using the presence of a smoking spouse as a~ predictor of actual ETS exposure were quire poor. Thirty-nine percent of men and, 47% of women married'to smokers reported zero hours of exposure at home. Conversely, 49% of men and! 41% of women marrned to non+smokers reported some ETS exposure. CONFOUNDING Confounding is another major problhm ~ area for the evaluation of epidemiologic studies of ETS and lung cancer and one that, has received relatively little attentionl Several studies suggest that a variety of factors could' act as confounders of an ETS -limg cancer association. Friedman (13)' found that age bore a strong negative relationship to reported ETS exposure. Hours per week of ETS exposure were associated with alcohol consumption, marijuana use, being,currently unmarried, and, in a U-shaped, fashion„with 'no college education." Koo, Ho; and Rylander (14) examined a wide variety of behaviors of the non-smoking wives of smoking and non-smoking husbands in Hong Kong. They concluded that in general wives with husbands who had never smoked had I healthier lifestyles than wives with smoking husbands. Specifically, the former were of higher socioeconomic status, were more conscientious housewives, ate better diets, and had higher indices of family cohesiveness as well as better health status. A third study, by Sidney et al. (15) reported that dietary B-carotene intake was significantly lower in non-smokers exposed I to passive smoke at home than ! in non-smokers who were not exposed, after adjustment for ag e, sex, race, education, status, body weight, and alcohol intake. 189

They'concluded that dietary B-carotene intake was a potentiall confounder of the relationship between, E"'S and lung cancer. Other potential confounders included: occupation, domestic radon exposure, a, history'of exposure to therapeutic x-rays, and keeping pet bird5 in the home. This last is raised by a recent study from the hfe'therlands which found t}iat the odds ratio for lung, cancer among people who kept pet birds in their home was 6.7 (95% confidence interva12:2-20.0) after adjustment, for active smoking and vitamin C intake :(16). This'siudy did not assess ETS exposure among the subjects. THE AN1M1CAN HEALTH' FOUNDATION STUM7(' Since 1983, a study of ETS' and lung cancer in never smokers has been in progress at the American Healt;h Foundation. Alllung cancer cases interviewed in the context of'a large, multi- center study of tobacco-related diseases who report never having smoked~ more than one cigarette per day for a year are given a dtstailed i ETS questionnaire. For each case, 2-3 hospitalized controls who have diagnoses not known to be associated with tobacco use and who are also lifetime non-smokers are interviewed. Controls are matched to cases on age (+/- 5 years)y sex, race, hospital, and date of interview (wiuhin 3 months). The items in the questionnaire inditde exposure :m utero; in childhood (specific family members who smoked, years of exposure and average number of hours of'exposure per day, and a subjective rating of the intensity of exposure), in adulthoodt aC home (specifac family membersrwho smoke(d), number of cpd smoked by eaeh, years of exposure, number of hours per day, subjective rating of exposure, and where a spouse smoked;,whetber he or she smoked in the bednoom); in the workplace (number of' hours per week, years of exposure, number of smokers within ten, feet of subject, rating of exposure) ' for up to four different' jobs; and in various forms of'transportation andi in social situations. Ln, addition to ETS questions, information is obtained on demographic factors, occupation,, alcohol consumption, medical'. history, diet, and other factors. To date, this study has accrued a total of 90 lung cancer cases and 247 matched controls. We plan to continue recruiting subjects for the study in order to reach a~ sample size of 150 cases. Table 4gives a breakdown of the histology of lung,cancer by sex: Preliffiinary analyses of the data do not' indieate any striking,ETS exposure differences between cases and controls. Tables' 5 and 6 give crude odds ratios and confidence intervals for overall' exposure in childhood, adulthood at home, and in1 the workplace, in males and females, respectively. With the possible exception of exposure in childhood and among women, there is little suggestion of excess risk duetio1ETS. A fuller analysis of'tihese data, including adjustment for covariates, is in progress. CoYVeLUSrOM Epidemiologic studies of ETS and1ing cancer generally suffer from smalllsample size. Given the small magnitude of the observed RR associated with passive smoking,and!the probltzms associated'with multiple histologic types bias, miscla.ssification, and confoundIng, increasing thee sample size is one way to attempt to answer the ETS-lung cancer question with greater certainty. A case-control study of 10,000 lung cancer cases (7,500males andl2,5(90~females) could'be expected to I I ci 0 F I 190,

I I 0 I J I I N r yield approximately 150 male and 250 female never smokers, based on estimates of the frequency of lung, cancer among never smokers (2% ~ for males and 10°.'b for females [7[). Table 7 shows the sample sizcs ncceasary in,eaeh group (assuming,eqlLal numbers of cases and controls) to deteu'RRs between 1?5 and 200,,with a one-tailed alpha of 5%' and 80% power, given varimus proportions of exposed controls. While it is highly unlikeIy that such a study would i be funded solely to assess the effects of LTS exposure, the study could' be designed ~ to make an important contribution to the radon-lung cancer issue as well. Spe 'c~f'icaitj!, studies of domestic radon exposure have also suffered from small sample sizes and have produced variable and unstable estimates of the ris t of radon exposure in never smokers. In addition, there is a need to better assess the interactive effects of active smoking and radon exposure. Since ETS and radon exposure are both risk factors for lung cancer, and since one may confound, or interact with, the other, a large study designed to measure both factors ass reliably as possible would have considerable scientific merit. 191 I

Prospective Studies Hirayama (1981) Carfinkei (1981) Case-Control Studies Trichopoulos, et al. (1981) Chan & Fung (1982) Correa, et al. (1983) Koo, et al. (1983) Kabat & Wynder (1984) Wu, et al. (1985) Garfinkel, et al. (1985) Lee, et al. (1985) Akiba, et at. (1986) Dalager, et ai. (1986) Pershagen, et at. (1987) Lam, et at. (1987) Koo, et at. (1987) Table I Epid emiologic Studies Re/a ti ve Risk 1.63 1.18 2.1 0.75 2.03 1.54 0.79 1.2 1.12 1.03 1.48 1.5 1.28 1.65 1.55 95% C.L 1.25 - 2.11 0.90 - 1.54 1.18 - 3.78 0.44 - 1.30 0.83 - 5.03 0.90 2.64 0.26 - 2.43 0.6 - 2.5 0.74 - 1.69 0.41 - 2.47 0.88 - 2.50 0.8 - 2.8 0.75 - 2.16 1.16-2.35 0.94 - 3.08 ,, TIBEM2O%

~ _ _._.. ~... ... a.A W i W Table 2 Cell Type Related to Spouse's Smoking w Study Histo%gic rype N 4dds Ratfo 95% C. l_ * Adenocarcinoma 16 1.02 0.33 - 3.16 Dalager et al. Squamous & Small * (1986) Ce{I Ca . 14 2.88 0.91 - 9.10 Other 18 1.31 'k 0.48 - 3.57 Squamous or 20 3 3 1 -1 1 4 1 Pershagen et al. Small Cell Ca. . . . (1987) Other 47 0.8 0.4 - 1.5 Adenocarcinoma 131 2.12 1.32 - 3.39 Lam et al. (1987) Squamous Cell Ca. 27 0.85 0.35 - 2.06 Small Cell Ca. 8 3.00 0.53 -16.9 Adjusted for gender, age, and study area. zZgE2:V,J2o2:

Table 3: Reproducibility of ETS Exposure Data C?u+estiorz Kappa value Ever lived with regular smoker? 0.66 Ever exposed j to smoke at work? 0.46 Nlo. of resident smokers? 0.515 No. of jiob sites re poirtsd? 0.37 i Duration, of residential exposure? I 0.45 Source: Pron et al., 1988 194 I i r I I I I I

Table 4 Histology of Lung Cancer Among Never-Smokers Ma/es N ( % ) Females N ( % ) Squamous & 5 (13.5) 10 (18.9) Small Cell Ca. Adenoca. 25 (67.6) 26 (49.1) Large Cell Ca. 5 (13.5) 6 (11.3) BAC 1 ( 2.7) 7 (13.2) Other 1 ( 2.7) , 4 (7.5) 37 53 tZSEzzJzOz

Table 5 American Health Foundation Study Males. Cases Controls OR 95% C.1. €xposed in Childhood: No 15 36 1.00 Yes 21 69 0.73 0.34 - 1.59 ~ Exposed in Adulthood- at home- .41 No 23 68 1.00 ----- Ye s 13 32 1.20 0.54 - 2.68 Exposed at Work (ever): No 16 4 5 1.00 ----- Ye s 21 80 0.98 0.46 - 2.10 siSuzzqz0z