Tobacco Documents Online

I ~ 'Involuntary Smoking and Lung Cancer: A Case-Control Study' Lawrence Gartinkel,2 Oscar AuerbacM,3 and Lou Joubert2' ABSTRACT-tn a case-control study in 4 hospitals from 1971 to 1981, 134 cases of lung cancer and ~ 402' cases of' colon-rectuM cancer (the controls) were identitied in nonsmoking women. All cases and controls were confirmed by histologic review of slides, and nonsmoking status and exposures were verified by interview. Odds ratios (ORyincreased with increasing number of cigarettes smoked by the husband. particularly for cigarettes amoked at home. The OR for women whose husbands smoked 20 or more cigarettes at home was 2.11 (95% confidence limits: 1.13, 3.Q5): A logistic regression analysis showed a significant positive trend of increasing risk with increased exposure to the husband's smoking at home, controlled for age, hospital. socioeconomic class, and year of diagnosis. Comparison of women classified by number of hours exposed a day to smoke in the last 5 years and in the last 25 years showed no increase in risk of lung cancer,-JNCI 1985; 75:463-469! Much interest has been expressed in the past several years in the reported relationship of involuntary or passive smoking and the development of lung cancer. Hiravama (1), in a prospective study in Japan, reported a 2:1 RR for nonsmoking women married to smokers as compared to the RR for nonsmokers married to n3n- smokers. Trichopoulous et al. (2), in a study in Greece, found that nonsmoking women with husbands who smoked ha& am OR about 2.5 times as high as that of women with husbands who never smoked, an& the OR rose to 3.4 in women whose husbands smoked more than one pack of'f cigarettes a day. In another case-control study, Correa et al. (3) found nonsmoking women married to smokers with a lifetime consumption of 41 or more pack years had an OR 3.5 times as high as the OR of women married to nonsmokers. In an analysis of data from the American Cancer Society's prospective study, Garfinkel (4) found little if any increase in RR of lung cancer for nonsmoking women married to smokers (4). The RR was 1.34 for light smokers and 1.10 for those whose husbands smoked 20 or more cigarettes a day. Kabat and Wynder (3) in a pilot study found no extra lung cancer risk in women and in nonsmoking men, exposed to smoke at home, but nonsmoking men exposed at work showed a slight increase in risk. Kabat and Wynder measured' exposure both by the spouse's smoking habit and the subject's report of direct exposure. An investigation by Sandler et al. (6) of relatively. young cancer cases in North Carolina found an overall 1.6 OR (smokers and nonsmokers) for exposed vs. nonexposed cancer cases. Exposed cases were those in w.hich the husband smoked. Exposed nonsmokers had a higher OR than that of exposed smokers. In nonsmokers the OR were elevated and statistically significant for cancers of the cervix, breast, and endocrine glands. NOttr:t 11A ntsWW " RN" ttpr ccpytl Me' (Tttw 11 w, Ct Occurrence of cancer in the latter two sites previously had not been associated with cigarette smoking. Lung cancer, also showed an elevated OR but was not statisticaIly significant; however, the number of cases was small. Two papers in Hong Kong by Chan and Fung (7) and by Koo et al. (8) show very little difference between cases and controls with respect to invoiuntary, smoking and lung cancer. Except for the two prospective studies, all of these studies were based on relatively few lung cancer cases in female nonsmokers; the number of cases ranged from 22 to 77 in various case-control studies. In a previous paper, we pointed out that in any study of involuntary smoking and' lung cancer, categorizing nonsmokers by the smoking habit of the spouse may lead to error in classification of'exposure. In the United States particularly, there may be many women, marriedto nonsmokers, who are exposed to the smoke of others at work or in other areas. Conversely, some individuals married to smokers may, suffer acute effects from inhaling smoke and consciously avoid such exposure. A survey of 38,000 subjects by Friedman et al. (9) confirmed this hypothesis. About 40% oE women nonsmokers an&50°b of men nonsmokers who were married to nonsmokers were exposed to the smoke of others for some periods of time during a week, and 47% of nonsmoking women married to smokers reported that they were not expose&to tobacco smoke at home. In the study reported here, we record the smoking habit of husbands (total No. of cigarettes smoked and No. smoked at home); as welI as the number of hours a day the subjects were exposed to the smoke of others at home, at work, and in other areas. Other causes for concern are establishment of the microscopic diagnosi's of primary lung cancer and AaaREVIATIONS USED: CLaconfidFncc limits; OR=odds rauo(s);RR=relative risk{yl.~ . t Received April 22. 1985; accepted June 211; 1985 2Depanment of Epidemiology and Statistics„ Amerian, Cancer Society,, 4 West 35 SI.,,New York, NY t0001. sVeterans Adminissration Medical Centn. East, Orange. NJ.,and University of Medicine and Dentistry of New Jersey. Newark, NJ. 'We thank Dr. Robert V. P. Huner. St. Barnabas Hospital. Ltang- aon, NJ; Dr. Herbert Derman, Riverside Methodist Hospiul, Colum- bus. OH: Dr. Jerry, Rothenberg, Morristown Memorial Hosptul, Morristown, NJ; and Dr. Douglas Smith, Middlesex General Oniver- sity Hospital, New Brunswick. NJ, for gnnting us access to medical records and pathologic matenal. We also thank the following indtvid- wls for making valuable suggestions regarding the manuscnpt: Mr. William Haens:ell,Dr. E. Cuyler Hammond. Dr. Geotfrry Howe, Mr. Edward Cew, Mr. Hkrberr Seidman, and Dr. Steven Stellman. We alsoo thank Ms. Nancy La Vnda and Mr., Henry Vasquet for assistance in processing the data. 463 JNCI. VOL 75.NO S. SEPTEMBER 1985

464 Gatiink.l, Auerbach, and Joub.rt verification of the smoking history. In a study of the histologic type of lung cancer in relation to asbestos exposure, 49 of 774' men and women with a discharge diagnosis of microscopically proved lung cancer were recorded as nonsmokers in the hospital chart (10). After review of hospital records, histologic sections, and interviews, only 10 cases remained who had died of primary lung cancer and who had never smoked. One- half of the others had smoked at some time and, one-half the confirmed nonsmokers had a primary cancer other than that of~ the lung. It is apparent, therefore, that more studies on in- voluntary smoking are needed, with particular attentionn given to obtaining microscopic proof of primary lung cancer and more detailed information about exposures to cigarette smoke. METHODS To have available enough subjects for a case-control study of involuntary smoking, we obtained access to the records of 4 hospitals-3 in New Jersey an& I in Ohio. In each of these institutions, we identified all lung cancer cases in women recorded' during 1971-81. In 2 hospitals the cases were selected from the Tumor Registry; in 11 hospital, they were selected from the surgical index in the pathology department; and in the other hospitals, records from the pathology laboratory were checked against the medical records diagnostic discharge index. No case was selected that had been diagnosed prior to 1971. Cases with cancer of the colon-rectum served as controls. Colon- rectum cancers have been shown in epidemiologic studies not to be related to cigarette smoking. Charts then were located and reviewed, Cases that were diagnosed clinically only or by cytology, or as sarcoma or lymphoma of the lung, were excluded. Those that occurred in smokers (or ex-smokers), according to hospital records, also were set aside. Only those charts in which the patient was specified as a nonsmoker, or in which the smoking habit was not recorded, were further investigated. All the slides for these cases and controls were pulled from the files (an average of -15 slides/case) and were reviewed blind (by 0. A.). In a small sample, slides for cases and controls were reviewed a second time to check consistency of the findings. Another sample of slides for smokers with lung cancer, and for subjects with diagnoses of sites other than lung or colon-rectum, also were selected for histologic review and were mixed in with, the slides of nonsmokers. If slides were missing or not available, or of too poor quality for accurate diagnosis, the blocks for the case were located and new slides were prepared. An interview based on a standardd questionnaire was obtained for all cases and controls, along with micro- scopic proof. The interview was with the woman if she were still alive or with next of kin if she had died. Seven interviewers did all the questioning: three did interviewing in all 4 hospitals. About three-quarters of the interviews were with the patient or with spouse or children. All other, informants had known the sub- ject for at least 25 years and were able to suppl'v the necessary informationi All interviews were reviewed by the supervisor. W,hen the information was incomplete, another respondent was contacted. A second interview was obtained in about 10% of the cases and controls. Women who had never married and who lived with another member of the family were classilie& according to their relative's smoking habits. Therefore, the word "husband" as used in this• paper means husband or, ecohabitant living in the same household. Of the cases, 57% were married and living with their husbands at the time of the cancer diagnosis. The interview included questions on current smoking habits of the husbands of the cases and controls up to the present time or to the time of death;,on the number of cigarettes smoked per day at home, and the number of years they had smoked. The interviewer also asked about the average number of hours a day the woman had been exposed to the smoke of others at any time during the past 5 years, during the past 25 years at home, while at work and in other areas, and during her childhood. Women whose husbands smoked cigarettes only occasionally were counted as not exposed; occasional exposure at home, work, or in other areas also was counted as not exposedi We matched b lung cancer case to 3 colon-rectum cancer cases. Controls were matched to withim5 years of age and were from the same hospital. In most age groups there were many colon-rectum cancers in women of the same ages for matching purposes. The colon-rectum cases were checked for histologic proof in the same way as were the lung cancer cases, and the smoking interviews were obtained by the same interviewers who obtained the lung cancer interviews. The interviewers were not told the diagnoses, nor did they know the hypothesis of the study: Several different analytic procedures were used: The Mantel-Haenszel procedure for obtaining a point estimate of the OR with a 11:3 match was employed„as adapted by Pike and, Morrow (11), with CL as shown by Miettinen (12). To compare subgroups of exposures, the matching was broken, and OR and CL were computed by the Mantel-Haenszell method. In addition, a logistic regres- sion model was used, with estimation and testing procedures as given by Breslow and Day (13): To permit comparison with previous studies, the subjects' exposures to cigarette smoke were classified in several different ways: 1) exposure over the last 5 years, 2) exposure over the last 25 years, 3)iexposure to cigarettes smoked by husband, and 4) exposure to cigarettes smoked by husband at home. RESULTS Table 1 shows the process through which data for 134 cases of lung cancer in nonsmoking women were obtained from the four hospitals. Of 1,175 women Pisted as having lung cancer, 892 (76%) were smokers or had smoked in the past„according to hospital records. Of the 283 remaining women, 36 (12.7%) were proved histologically to have other than liung cancer upon JNCI. VOL 75. NO. 3. SEPTEMI+ER 1965

Involuntary Smoking and Lung Cancer 465 TABLE 1.-Lung cancer in women who neurr nnoked Recorda oJ 4 hospitals. 19?1-81 No. of women examined Status At hospitals: l T % , ota A B C D Microscopic proof of~lung 243 93 276 663 1.175 cancer on hospitaU record Smoker 200 70 182 440 892 Nonsmoker or smoking 43 23 94 123 283 100:0 habits not stat'ed° Reinterview revealed; 15 3 41 54 113 39:9 smoker Reinterview revealed 18' 14 45 57 134 47.3 nonsmoker No microscopic proof of 10 6 8 12 36 12.7 lung cancer ' 68%of the hospital records listed patient as nonsmoker: in 32%of the records, smoking habits were not stated. review of slides by one of us (0. A.), and 1i13'(39.9 q) were found to be smokers upon reinten•iew. Only 134 (47.3%) were lifetime nonsmokers with histologically proved primary lung cancer. They Were the only cases therefore suitable for this study. Among the colon-rectum cases. there were many fewer that were misdiagnosed-onl) 1.4%. The age distribution of thetasesand controls is shown in table 2. More than half were 70 vears of age or older, and 22%, were 80 years of age or older at the time of diagnosis. The histologic diagnosis of lung cancer cases was as follows: 65% adenocarcinoma, 16% large cell, 8% squamous cell, 4% oat cell. 3% alveolar cell, 3% mixed, and 1% too undifferentiated for classification by cell type. Table 3 shows the OR and~CL for risk of lung cancer, according to the 4 methods of classifying smoke exposure. The OR ranged from 1.13 to 1.31. All 4 methods resulted in lower 95% CL of less than I and were not statistically significant. Table 4 shows the average number of hours per day that cases and controls were exposed uo other people's TABLE 2'-Ape diatribution:oJlung aanaer caeea and controts A Cases Controls ge. yr No. % No. % 40-49 5 3.7 17 4.2 50-59 28 20:9 86 21.4 60-69 28 20;9 88 21.9 70-79 44 32.9 121, 30:1 80-89 24 17.9 82 20.4 ?90 5 3.7 8 2.0 Total 134 100.0 402 100.0 TABLE 3:-OR/or matched grovpa of wio+nen for riek of lunp cancer from erposure to amoke, as rllsaaijied in 4 ca.tegoriea Classification Risk of lung cancer for women OR CL ExposedAo smoke over last 5 yr 1.28 0.96-1.70 Exposed to smoke over last 25 yr 1.13 0.60-2.14 Husband4moked 1.22 0.97-1.71 Husband smoked at home 1.31 0.94-1,83 smoke for the lasr 5 years an& for the last 25 years. The women exposed during the last 5 years had an OR (adjusted for hr exposed per day) of 1.28 (95% CL: 0:98,. 1.66) and those exposed for the last 25 years had an OR of 1.12 (CL: 0.81, 1.42). No increasing trend with increasing exposure was. apparent in either group. In the 5-year exposure group, the OR went down with increased exposure, but the OR in each ofthe exposure groups was not statistically significant. Table 5 and text-figures 1 and 2 show exposure classified by the husband's smoking habits. The OR for women married to smokers was 1.23 (CL: 0.94, 1.60); for those whose husband smoked at home it was 1.31 (CL:. 0.99. 1'.73 )Wbrne?14;W7tlisC'ltusbands smoked 40 or mor F_ cigarettes a=day fiad an OR of -1.99 (Ch::--lall,-3. ). ~ ~,,,,NgmeA„WbqK,Kusban4unoked 2f1>oLqwre cigarettes at . ~ home had an OR of 2.11 fC>t.: 1.13. 3.95). These were the onfy specific smokiing groups in which the OR were statisticallw significant. Jfhe Mantel extension test fo7r TABLE 4.-Ntnnder oJcasea and eowtrola e:poeed to s+noke of others during 5 and 25 yr 6efore diagnoria Exposure. No: of hr/day Total Variable No: of None 1-2 3-6 27 Total women Last 5 yr No. of cases 80 15 25 14 54 134 i No.,of controls 263 31 59 49 139 402 OR 1.00 1.59 1.39 0.94 1.28 w 95% CL 0.90-2.72 0.96-2.03 0.69-1.28: 0.98-1.66 W Last 25 yr W No. of cs,ses 42 17 45 30 92 134 ~ No: of controls 136 72 109 85 266 402 ~ OR 00 1 0 77 1 34 14 1 1.12 ~ 95% CL . . 0.60-0.99 . 0.96-1.87 . 0.83-1.57~ 0.81-1.42 f"A S J,-,Ct. VOL 75. NO S. SEFTEMBER 1985

466 Garfink.l, Auerbach, and Joubtrt TABLE 5.-Smoke ezpoaure before lunq cancer diaqnoais, as ctasaified by husband'e smokinq habita Husband's total smoking habita Variable ° Ciecrettesrday . Cigar and/ None <20 20-39 >40. or pipe Na ot caaes 43 11 32 30 18 No. of controls 148 45 102 52 55 OR" 1.00 0.84 1.08 1.99 1.13 95% CL 0.61-1.16 0.81-1.44 1.13-3.501 0.78-1.62 None° Husband't anakinQ habit. at home Cigarettes/day <10 10-19 >20; All'types of smoking 91 254 1.23 0:94-1.60 Cigar and/ All types or pipe of smoking Total No, of women 134 402 Total Nb. of women, No. of caaes 44 29 17 26 18 90 134 No. of controls 157 90 56 44 65 245 402 UR° 1.00 1.15 1.08 2.11 1.17 1.31 95% CL 0.84-1.58 0.76-1.54 1.13-3.95 0.80-1.70 0.99-1.73 °FiQures include single women living alone. Cohabitants living with single women were classified as "husbands." ~Mantel extension t,est for trend (one tailed): x= 2.31. P<.025. ° Mantel extension test for trend (one tailed): z= 2.35, P<.025. trend in both. groups was statistically sigrtifit:a~ (P.<.025, one-tailed test):_ . z Analysis also was done, for years of smoking. That' were sta unll W~QR.,tor„fht1se smoking ~cor ti~~iit~au f e5~smokzftaztd: 2J7ol iiriTiking ir T"iome , ut noiienT was ippaFeiiiffhose who reported smoking for 30-39 years and 40 years or more had much lower OR that were not statistically significant. Table 6 shows OR for exposure categories by age gToup; histologic type of lung cancer, identity of the respondent who was interviewed; and socioeconomic status. Data are for average exposure for the last 5 years„ for the last 25 years, by husband's total smoking habit, and by his smoking habits at home. OR generally were higher for those 70 years of age or over, for those with adenocarcinoma, when someone not in the immediate family, was the respondent, and for those in the lower or lower middle class. There does not appear to be a pattern of high OR for any of these subgroups in all 4 exposure categories. Some of the OR art statistically significant, but they usually, carry very wide CIL with them. Table 7 shows the OR for classification: of~ exposure of women to smoke at home„at work, and in other areas, as compared with those women not exposed' at all.. OR for exposure at work during the last 5 years was 0.88, for the !'ast 25 years, it was 0.93. The highest OR observed was 1.77 for exposure during the last 5 years in "other areas." None of the RR shown in this table are statistically significant. One of the questions in the interview was with regard to exposure to smoke in childhood. Those women who replied that they had been exposed in childhood had an RR of 0.91 (CL: 0:74, 1.12), LOGISTIC REGRESSION ANALYSIS An unconditional logistic regression model'was used- which included terms for age, hospital, socioeconomic status, and year of diagnosis-to account for possible confounding factors. Testing was done on each of the four exposure variables, three of which were used in the Mantel-Haenszel analysis: 1) exposure during the last 5 en tJ11E! 43 CafTnnL1 1" 044 c to _I-__I C1o.ar o 0" tt 32 sU 46 IOt b! I.is 0 I.n pYll/ 4L ML OIOR[a 1 t6 t64 T[xT-ncuRC l,-OR (or exposure to husband's total smoking habits. vo• sp0[ tM IO.M fo• t*s/MfM a a" CAaEl. 44 Lf 17 !M COKnq" I37 f0 5a 41 T[xT-F10URr 2-OR forexposure to husband's smoking habits at ~ home. JNC1. VOL 75. NO.S. SEr'TEEM6ER 1965 lY

• Involuntary Srnoklng and Lung Canc.r 467 TABLE 6.-OR jor smoke exposure cateyories. by age group. Aistoloyic type of lung cancer. idenriay of respondent, and .ocioeconomir atatv.e Smoke exposure Specification No. of cases Last 5 yr Last 25 yr Husband's smoking habits Total At home OR 95% CL OR 95% CL OR 95% CL OR 95% CL Age. yr <60 60-69 70-79 ?80 33 28 44 29 0.96 0.82 1.82 2.00 0.65-1.42 1.00 0:57-1.19 0.55 0:93-3.53 1s22 0.76-5.25 1,75 0.63-1.51 0:41-0.73 0.78-1.90 0.81-3.78 1.19 120 1.26 1.28 0.72 -1.98 0.66-2.19 0.79-1.99 0.72-2.27 1.30 1.42 1.43 1.10 0.75-2.26 0~70-2.88 0:85-2.39 0:68-1.79 Histologic type Adenocarcinoma 87 1.43 0.99-2.06 1.15 0.85- T. 56 1.33 0.94 -1.87 1.48 1. 01-2.,17 Squamous cell carcinoma 11 1.28 0.52-3.19 0.85 0.43-1.69 5.00 1.28-19:33 5.00 1.43-20.18 Large cell carcinoma 21 0.55 0.41-0.74 0.67 0.47-0.94 0.76 0.51-1.13 0.62 0:45-0:86. Mixed and other 15 2.29 0.57-9.10 2.67 0.41-17.35 0.81 0.48-1.37 1.00 0.53-1.77 Respondent Self 16 1.96 0.62-6.17 0.91 0.51-1.60 0.83 0.50-1.38 1.00 0.55-1.74 Husband 34 1.00 0.67-1.52 0.46 0.38-0.55 0.77 0.56-1.06 0~92 0.63-1.34 Daughter or som 48 0.92 0.67-1.26 1.41 0.85-2.36 3.57 0.84-15.28 3:19 0.91-11.19 Other 36 2.23 0.90-5.54 2.23 0.83-5.96 1.58 1.11-2.67 0.77 0.57-1.03 Socioeconomic status Upper and upper middle class 6 1.60 0.31-8.19 1.50 0.34-6.59 1.23 0:36-4.18 1.50 0.34-6.59 Middle class 75 0.78 0.63-0.97 0:92 0.71-1.19 1.15 0:84-1.59 121 0.87-1.69 Lower, and lower middle cl`ss 53 2.58 1.10-6.01 1.45 0.86-2.44 1.23 0:83-1.84 1.45 0.88-2.38 years, 2) exposure during the l'ast25 years, 3) husband's smoking at home, and+ husband's smoking outside the home. The latter variable was used rather than the husband's total smoking as a check of the validity of exposure to husband's smoke and was derived by sub- tracting the number of cigarettes smoked at home from the totaf number of cigarettes the husband smoked per day. Each of these factors was tested as a continuous exposure variable-the most powerful technique for detecting any true underlying risk. Table 8'shows the results of: this analysis. Exposure for 5 years and 25 years had negative coefficients. The test for TABLE 7.-Nrmber of ca.e+ and eontroL exposed to smoke of otAers at Aorne, at work, and in ottier areas Variable Last 5 yr No. of cases 80 37 14 13 No.,of controls 262 99 52 24 OR 1.00 1.22 0.88 1.77 96% CL 0.92-1.62 0.66-1.18 0.93-3.38 Last 25 yr No. of cases 42 73 34 19 No. of controls 135 204 118 43 OR 1.00 1.15 0.93 1.42 95% CL 0.89-1.49 0.73-1.18 0.89-2.26 Smoke exposure None At home At work In other areas cigarettes smoked by husband at home showed a positive trend of increasing risk with increasing exposure and was statistically significant, with a P-value (one tailed) of .032. The test for cigarettes smoked outside the home was not statistically significant. The table also shows esti- mates of RR at the 10 hours per day exposure leveliand at 20 cigarettes per day smoked by the husband. The RR from exposure to 20 cigarettes/day smoked at home was 1.70; outside the home, it was 1.26. RR from exposure during the last 5 years and during the last 25 years were less than l. A separate analysis that included respondent identity did not change the results materially. TA9LE B.-Gopistie regression model' for inuotuntary smoke exposure varia6les, on coatinuourdose-re+ponae 6aiu Variable Coefficient P-TILlueb (SE) Smoke exposure level RR` 5-yr exposure to -0.0069 0.422 10 hr 0.93 smoke (0.0035) ~ 25-yr exposure to -0:016 0 303 10 hr 0.85 ~ . smoke Ciprettes (0A31) 0.026 0.032 20 cigarettes 1.70 ra smoked at: home (0.014) ~ Cigarettes 0.012 0.127 20 cigarettes 1.26 smoked outside (0.010) home ~ • Model'includes terms for aBe, haspital, socioeconomic status. and year of diagnosis. 6 One tailtd. lV ~ ' Relative to the nonexposed woman. W JNCt. VOL 75.,taO: 3. SEPTEMBER 1965

468 GartinkN, Auerbach, and Joub.rt DISCUSSION In a previous paper (4) the problem of classifying involuntary smoking on the basis of the husband's smoking habit was discussed. It was pointed out that questions directed at ascertaining a quantitative estimate of the number of hours a day that subjects were exposed might be a better measure than the total number of cigarettes that the husband smoked, inasmuch as not all of the husband's smoking was done at home. In the present study we classified the exposure both ways: by the number of hours per day the subjects were exposed to smoke of others and by the husband's smoking habits. We also recorde& the respondent's estimate of how many cigarettes a day the husband smoked at home. In this group of women, husbands who smoked cigarettes smoked an average of 27 cigarettes a day: of which 11.5 cigarettes on average (43%) were smoked at home. O{ course, all cigarettes smoked at home were not necessarily smoked in a room where the subject could have been exposed. In this study, the husband's smoking at home was related to the women's lung cancer, whereas number of hours of exposure a day to all sources of tobacco smoke was not related. A potential source of error was the hospital's report of whether the subject smoked or not. In this study, 40% of the women with lung cancer, classified as nonsmokers (or smoking not stated),on the hospital record, were smokerss at some time (itable 1); Another 13% did not have primary lung cancer. It is apparent, therefore, that in any study of Variablp None <20 No. of cases Nonsmokers (in study) 43 11 Smokers (originally called nonsmokers) 21, 9 Total(unscreened) 64 20 No: of controls Nonsmokers (in study) 148 45 Additional.controls° 119 38 Totsl(urtacreened) 267 83 OR 1.00 1.01 None <10 No. of cases Nonsmokers (in study) 44 29 Smokers (originally called nonsmokers) 23 22 Total (unscreened) 67 51 No. of controls Nonsmokers (in study) 157 90 Additional controls° 126 75 Total(unscreened) 283 165 OR 1.00 1.31 TABLE 9.-Hypothrtieal OR reaultinq from comDi>,ing aromrn in study u•ith umm.en oripinaUy classiJied as nonsmokers but who attuatly amoked of cases who were smokers. The balance were distributed according to the smoking habits of controls in the study. involuntary inhalation and lung cancer, the smoking histories of the subjects have to be confirmed as well' as the extent of their involuntary, exposures. Smoking histories of husbands were obtained for the 113 women who were smokers. The distribution by smoking habit is shown in table 9. As we might have expected„smokers are more likely to be marru6 to smokers than are non- smokers. The table shows that 43 of 134 women, or32.1%, of the cases included as never smoked in this study had husbands who never smoked; but only 21 of 173, or 18:6%, of women who smoked and were mistakenly classified as nonsmokers in the hospital record had husbands who did not smoke. Among the controls only 8:5%of women who were called nonsmokers (or smoking was not stated) were smokers. The table shows the effect on the OR, when one assumes that 8.5% of the additional controls needed for the 1:3 match had husbands with the same smoking distribution as the husbands of cases who were smokers, and that the balance had the same distribution as that of' the 402 controls included in the study. The OR for the husband's smoking increase to 1.61 overall and are as high as 1.63 for the 20-39 cigarette a day smokers and 2.32 for the women whose husbands smoked 40 or more cigarettes a day. For exposure to the husband's smoke at home, the OR are 1.66 overall„ 1.53 for women whose husbands smoke L0-19 cigarettesa day, and 2:85 for those whose husbands smoke 20 or more a day at home. Thus the inclusion of women whose smoking habits have nott been reviewed greatly increases the OR. Husband's total!smoking habits Cigarettes'day Cigar All Totals andror types of 20-39 >40 pipe smoking 32 30 18 91 134 43 24 16 92 113 75 54 34 183: 247 102 52 55 254 402 90 45 47 220 339 192 97 102 474 741 1.63 2.32 1.39 1.61 Husband's smoking habits at home Cigarettes/day Cigar All Totals andLor types of 10-19 z20 pipe smoking 17 26 18 90 134 22 30 16 90 113 39 56 34 180 247 56 44 55 245 402 52 39 47 213 339 108 83 102 458 741 1.53 2.85 1.41 166 • Upon reinterview. 8.5%of the controls were found to be smokers. They were distributed according to the smoking distribution of husbands N JriCl. VOL 75. !1O: !. SEPTEMBER 1985

The classification used4n this study might be aiticized because some women married to ex-smokers could be counted in the same exposure category as a woman exposed to smoke up to the time of he~final illness. However, alllpatients who have gone through diagnosiss and treatment for lung cancer had some period of time when they were not exposed to others' smoke„ either before or after treatment. We believe that the classifica- tion we used was indicative of the "usual amount of smoke to which the person was exposed."'To determine the experience of a"Ipure" nonexposed group, 17 cases and 56 controls in this study were identified who were not exposed to the smoke of others during the last 5 years, during the last 25 years, whose husbands never smoked at home or elsewhere, and who never were exposed'to smoke in their childhood. These cases and controls were compared with all other subjects. The OR was 1.14 (CL: 0.81, 1.59). In conclusion, we found an elevated risk of lung cancer, ranging from 13 to 3 1%, in women exposed to the smoke of others, although the increase was not statis- ticallv significant. The women who were marrie& to smokers of 40 or more cigarettes a day or who were exposed to the smoke of at least 20 cigarettes a day at home showe& a risk twice as high as that of women not exposed at all. This result is consistent with the dose- response risk of exposure to the husband's smoke shown in some case-control studies (2, 3). A dose-response relationship was confirmed in a logistic regression analysis. The lack of a relationship when exposure was classified by hours exposed to smoke oGothers may have occurred because this variable does not accurately mea- sure intensity of exposure. There is no consistently InYoluntary Smoking and Lung Canc.r 469 higher, risk for certain age groups or by histologic types, or by exposure at home or at work. Exposure in other areas carried a higher OR, but this finding is difficult to interpret. REFERENCES (1) HIRAYAMA T. Non-smoking wives of heavv smokers have a high nsk of lung cancer:, A study from Japan. Br Med J 1981; , 282:1'83-1 BS. (2) TRteHOrorLos D. KALAr:DtDi A. SPARROS L, et al. Lung cancer and passive smoking. Int J Cancer 1981t 27:1-4. ()1 CoRREA P. Fon-rHAM E. PieKt-t CW. et al. Passi4e smoking and lung cancer. hncet 1983: 2:595-591i (4) GARrFlNKEL L. Time trends in lung nncer monality among nonsmokers and a note on passive smoking. JNCI 1981; 66: 1061 ~ 1066. (S) KASAT GC. wvwDtR EL. Lung cancer in nonsmokers. Cancer 1984; 53:121 4-1221. 16) SAnnLtR DP„E%'tRsot: RB, M'it.cox A). Pauive smoking in adulo- hood and cancer, risk. Am J. Epidemiol ~ 1985. 121:37-48: (7) CHAt. WC, Ft,r:6 SC. Lung cancer in nonsmokers in Hong Kong. lni Grundmann E, edl Cancer campaign. Vol: 6. Cancer epidemiology: Stuttgart and NtW York: Fischer Verlag, 1982: 199-202. (&)' Koo LC. HojH=C. SAW D. Active and passive smoking among (emale lung cancer patients and controll in.Hong Kong J,Exp Clin Cancer Res 1983: 4`.367-375.. (9) FRIEDMA] GD. Prrrrn, DB. Bswou RD. Prevalencr and correlates of passive smoking. Am J Public Health 1983: 73:401-405. 110) At'ER6AGH 0..GARFINKEL L. PARKSVR. et aIL.Histologic type of lung cancn and asbestos exposure. Cancer 1984; 54;3017-3021. (1)), PtKE MC. MoRROw RH Statistical analvsis of patient~control studies in epidemiolog.: Factor under investigation on all-or- none variable. Br J Pre. Soc Med 197q;,24:42-44. 112r MiETra.tn OS EstimabilitN and estimation in ase-referent studies. Am] Eptdemtol 1976, 103.226-235. (13) BREStow NE. D+i NE- Statistical methods in cancer research. \'ol. I. Analysis oficase-comrol studies Lyon: IARC. 1980 JNCt. VOL 75. I.O 3. SEPTEMBER 1965