Tobacco Documents Online

Lung Cancer in Nonsmokers NO'TICE This rnater;ad~ may be prote::ted~ by r,;; ight bw (Title 17 t;.S, Code). GEOFFREY C. KABAT, PnDd AND ERNST L WYNDER, MD Among 2668 patients with newly diagnosed lung cancer interviewed between 1971 and 1980, 134 cases occurred in "validated^ nonsmokers. The proportion of nonsmokers among all nses was 1.9% (37 of 1919) for men and 13.0% (97 of 749) for women, giving a sex ratio of 1:2.6. KreyberQ Type ll (mainlyy adenocarcinoma) was sssore common among nonsmoking eases, especially women, than among all lung cancer cases. Comparison of cases with equal numbers of age-, sex-, tace-, and hospitsl-matched nonsmoking controls showed no differences by religion, proportion of forei4n-born, marital'ststtts, residence (urban/ rural), akohol consumption or Quetelet's index. Male cases tettded to have higher proportions of profes- sionals and to be more educated than controls. No differences in occupation or occupational exposure were seen in men. Among women, cases were more likely than controls to have worked in a textile- related job (relative risk - 3.10„9596 confidence interval 1.11-8.64); but the significance of this finding is not clear. Preliminary data on exposure to passive inhalation of tobacco smoke, available for a subset of cases and controls, showed no differences except for more frequent exposure among male cases than controls to sidestream tobacco smoke at work. The need for more complete information on exposure to secondhand tobacco smoke is discussed. Cancer 53:1214-1221, 1984. (S~ MAA4~ / ALTHOUGH LUNG CANCER risk is strongly associated with cigarette smoking, lung cancer does infre- quently occur in nonsmoken.'s Several features distin- guish lung cancer in nonsmokers from that occurring in smokers. First, most cases of lung cancer in nonsmokers are found in women.2-3 Second, the distribution of his- tologic types of lung cancer differs between smokers and nonsmokers. In smokers the epidermoid type predomi- From the Dirision of Epidemiolo8y; Mahoney Institute for Health Maintenance, Amencan Health Foundation, 320'East 43rd Stttet. New York, New York. Supported by National Cancer Institute contract N0){P-0S684 and grant CA-3261 i7. Address for reprintx Geoffrey C. Kabat. PhD. D+vision of Epide- miology. Mahoney, Institute for Health Maintenance. Amenan Health Foundation. 320 East 43rd Streeti New York. NY 10017:, The authors thank the following cooperating institutions and indi- viduals for their valuable eontributions Memorul Hospital. Dr. David Schottenfeld: Manhattan Veteran's Hospital: Dr. Norton Spritr. Long lsland-Jewish Hillside Medical Center. Dr. Arthur Sawitaky:Uhiversity ofAlabama Hospiaal. Dr. William Bridgers; Birmingham Veteran's Hos- pital, Dr. Herman F. Lehman; Loyola University Hospital (Chicago). Dr. Walter S Wood: Hines Veterut's Hospital (ChieaBo), Dr. John Shary: Hospital of tha Unirenity of Pennsylvania, Dr. Robert M. Levin; Jefferson Medical Colkse and Thomas Jefferson University Hospital, Dr. J. E. Colbert; Allegheny General Hospital (Pittsburgh). Dr. Stanley A. Briller. lJniversity, of Pittsburgh Eye and Ear Hosptal, Dr. Lewis H. Kuller, Pittsburgh Veteran's Hospital. Dr. Eugene N. Myerr Molfitt Hospital (San Francisco); University of Californu atSan Francisco and County 'Hospital (San Franciseo). Dr. Nicholas Petrakis: and St. Luke's Hospital (San Francisco). Dr. Richard A. Bohannan. The authors also thank Ms. Margaret Mushinski for her collaboration in the early sta8es of this study. Ms. Nancy Vtotsos for prolp amming assistance. and Mr. Monte Hewson and Ms. Maria Nanfaro for manuscript preparation. Accepted for publication August 31. 1983. nates, whereas in nonsmokers adenocarcinoma is more common, especially in women.2-S This article presents data from a case-control study of nonsmoking patients with histologically confirmed di- agnoses of primary lung cancer with respect to histology, demographic factors, residence, Queteltst's index, alcohol consumption, previous diseases, occupation and occu- pational exposures, and, to a limited extent, exposure to the tobacco smoke of others. Due to the small number of cases and controls on whom we have information on passive inhalation, the data presented here on thatques- tion are in the nature of preliminary results. A discussion of previous studies concerning this issue emphasizes the need for obtaining more detailed information on side- stream smoke exposure an& related variables. Methods All cases of primary cancer of the lung occurring in cases who reported never having smoked on a regular basis' were extracted from an ongoing case-control!stud,v. oftobacco-related cancers conducted in a number of cities between 1971 and 1980t and described prcviously.° For each case, the hospital chari was re-examined in order to confirm the diagnosis and the absence of smoking ~ ra • Ourdefinition ofa nonsmoker was someone who had neversmoked W as much as one citarette.,pipe.,or apr, per day for a year. CJ t The majority of the ases (and matched controls) were interviewed ~n at Memorial Hospital in New York City,. 30 of the 37 male cases and ~; 70 of the 97 female ases. a~+. 1214

No: 3 LUNG CANCER 1N N0NSMOKERS - Kabat and Wynder 1215 . throughout the patient s lifetime. The histologic type of lung cancer was obtained from the pathology reportor the discharge summary foreach.case. Those cases in whom the diagnosis was not: primary lung cancer or in whom there was an indicatiom of smoking, even in the remote past; were excluded from the study. Those remaining in the study are referred to as "validated" nonsmokers. A control, was matched to each case on the basis of age (t5 years), sex, race (with 5 exceptions$), hospital, date of interview (±2 yean); and nonsmoking status. Controls were selected from a large pool of hospitalite6 patients who were interviewed over the same period as the cases and who had diseases which were not tobacco- related. The distribution of diagnoses among the controls was as follows: men, 62:1 % other cancers, 24.3% benign neoplastic disease„ 13:5% non-neoplastic disease: women,. 59.9% other cancers, 14.4% benign neoplastic disease, 25.8% non-neoplastic disease. All subjects were interviewed in the hospital with a standardized questionnaire including questions on de- mographic factors, occupation, occupational exposures,, tobacco smoking, alcohol use, Quetelet's index (kg/crn? x 10,000)a and history of tobacco-related diseases. Two different versions of the questionnaire were used over the 10-year period, the first from 1971 to 1976: and the second from 1976 to 1980: Differences between the two ques- tionnaires included a longer list of occupational exposures in the later version, and a longer list of previous diseases in the earGer questionnaire (diabetes, gout, bronchitis, emphysema. hypenension, asthma, pleurisy, pneumonia, bronchiectasis, and tuberculosis) than in the later vetsion,, which included' only four questions on previous diseases (chronic bronchitis or emphysema, asthma, diabetes, and elevated blood pressure): Alcohol consumption was assessed in current drinkers and exdrinkers (combined) relative to never-drinkers and occasional drinkers (combined). Occasional drinkers were those who consumed less than I ounce of whiskey equiv- alents of alcohol per day of beer, wine, and hard liquor combined. AlcohoC intake was categorized into three lev- els: (1) never/occasional drinking. (2)' 1 to 3.9 oz/day, and (3) 4+ oz/day. In ad'dition, a number of questions on exposure to passive smoking were introduced in an addendum to the main questionnaire in 1978, and the addendum was re- vised in 1979. Thus, information on passive smoking was obtained on only a subset of the subjects, for men; 25 of 37 cases and their matched controls; for women, 53 of 97 cases and their matched controls. This number of responses was obtained for those questions included in both, versions of the addendum, whereas the number of TAeLE t. Histologic Type o( LunS Cancen in Never Smokers and Smokers Men Women (Nb.) (%) (No+ (%)'. Never smoken Kntyber8 type 1 13 (35.1)i 20 (20.6) Epidermoid/squamous 13 (33.1): 16 (16.5) LarSe ceilPiianrcell 0 4 (4.l) Kreyberg type ll 20 (34.1) 72 (74.2) Adenocarcinoma 16 (43.2) 60 (61.9) Alveolar 4 (10:5) 12 (12.4) Mixed (Kreybers I & 11) and undi(fcrentiated/ anapUutic 4 (10:8) 5 (5.2) Total 37 97 Smokers' Kreyberg type 1 1187 (63.1i) 341 (52.3) . Kieytierf type ll' 600 (31.9) 279 (42:8) Mixed (Krevbers 1! da 11) and undifRetentiated/' anaplastic 95 (5:0) 32 (4.9) Total 1882 652 ' A more detailed breakdown by histolo`ic type is not presented for smokers because this information.wu not coded: For the nonsmokers this information was retrieved manually: responses was smaller for the question "Does your spouse smokeT', since this question appearedlin only one version and since it was not answered by those subjects who were not married, widowed, separated, or divorced (see Table 3). Differences between cases and controls were assessed by the chi'-square test for independence.''and by the Man- tel-Haenszel extension test for linear trend:' Point esti- mates of the relative risk with test-based 95% confidence intervals were calculated following Miettinen's method.° Results For the 10-year period, 1971 to 1980, among 1919 cases of primary lung cancer in men. 37 (1.9%) occurred in validated nonsmokers, Among 749 lung cancer cases in women, 97 (13.0%) were validated nonsmokers. This diRerence in the proportion of nonsmokers in men and women is highly statistically significant. X2(1) - 137.21„ P < 0.001. Histol'ogic Type Table 1 shows the histologic type of lung cancer for nonsmokers and smokers by sex. Among male smokers with lung cancer there were nearly twice as many Kreyberg type 1§ cases as Kreyberg type 11(1187 versus 600), while $ One oriental malt case was matched to a..white eontrol: two hispanic and two onental female cases were matched to white controls. ; Kreyberg type I includes squamous eell. oat,ttlC small cell and large cell arLinomu: Kteyber8 type 11 includes sdenocarnnoma. bronchtolxr, and alveokar carcinoma.

1216 CANCER Marcii 1 1984 VW. 33 TAd1.E 2. Disuibution of Background Variables in Caes and Controls Men ca.e Women ConuoH Cae conuoFs (No.) (%) (rw.) (%r (no.) (x) (No.) (%)I AV sr9 13 (33) 12 (32) 12 (12) 13 (ls) 30-39 II (30) 12 (32) 26 (27) 24 (23) 60-69 7 (22) 10 (27) 29 (30) 34 (33) 70+ 6 (14) 3 (i) 30 (31) 24 (23) Tonl' 37 37 97 97 Rtbpon Prouxanr 2 (6) S(14) 27 (21) 34 (36) Catholic 16 (46): 14 (40) 31 (32) 36 (3a) Jie.ish 17 (43) 13 (37) 3E (40)' 24 (25) OUa 2 (6) _ 3 (9) 0 (0) 1 (I) Toul 33 35 96 96 Yr of eduaticn 1-11 3 (S.4) 6 (16.2) 311 (39.2) 29 (29.9) 12 7(16:2) lI (29.7) 25 (27:e) 37 (3i.l) 13-I3 6 (21.6) { (21.6) 14 (13.3) 15 (13.5) 16+ 20 (56.1)12 (32.4) 16 ()73) 13 ((3.3). Taal i 37 37 97 97 Occupaionaa stasus Ihofmona( 22 (39:3) 14 (37.a) t (t.2) Ii( (11:3): SkSlled 6 (16.2) 7 (1i.9) 26 (26.i) 35 (36:1) Snniski)kd 2 (5.4) 9(24.3) 6 (6:2) 6 (6.2) Unskilled 3 (i.l) 2 (3.4) ! ().3) 3 (32) Houar.irc 0 - 0 - 38 (39.2) 21 (2l.9) Aetieed/uaemp(oyad 4 (3.3), 3 (13.3) 11 (11.3) 12 (12.4) Toul 37 37 97 97 among female smokers the numbers were more similar (341 versus 279). This difference is statistically significant, z2( I) - 25.91, P < 0.001. Among male never-smokers, there were 13 Kreyberg type I versus 20 ICreyberg type 11 cases, while among females, there were 20 Kreyberg type I versus 72 Kreyberg type II cases. Although the number of male nonsmoking cases is small, the difference between men and women is statistically significant, X=(1) - 3.90, P < 0:05.. Furthermore, the difference between the proportions of Kreyberg I and Kreyberg U in never- smokers compared with smokers is statistically significant in both sexes (for men, x2(l) - 10.54, P < 0'.005; for women, X=( l)- 35.46, P< 0.001): Age Table 2 gives the age distribution of cases. Male cases are significantly younger than female cases (X2(3) = 11'.30, P < 0.025). The mean age for men was 53.9 years (SD [standard deviation] 14.3) compared with 61.6 (SD 11.3) for women. This younger age of male cases appears to hold for both Kreyberg I and Kreyberg 11 types: the mean age for Kreyberg I and Kreyberg 11 lung cancer in men was 52.8 and 53.6 years, respectively, while in women Kreyberg I had a mean age of 63.7, and' Kreyberg II had a mean of 61.0 years. Education Kreyberg II cases appeared to be more educated than Kreyberg I c,ases in both.sexes (data not presented). Case-Control Comparisons There were no differences in male cases and controls by religion, proportion of foreign born, marital status, and residence in childhood, adolescence, and adulthood. Male cases were better educated (57% of cases had gone beyond college compared to 32% of controls), and a higher proportion were professionals (60% of cases compared to 38% of controls) (Table 2). Thesc differences did not reach statistical significance. Female cases and controls did not differ significantly on proportion of foreign born, marital status, education, occupational status, or residence in childhood, adoles- cence, or adulthood. There was a nonsignificantly higher proportion of Jewish women among cases compared to their controls (40% versus 25%) (Table 2). In both cases and controls, the proportion of urban dwellers incrsased' from 70% in childhood to 80% in adulthoodL History of previous diseases: No case-control' differ- ences were found for history of chronic bronchitis, em- physema, diabetes, asthma, pneumonia, or hypertension in males. In females, there were similar findings, except more female cases had a previous history of pneumonia than controls: 16/40 cases versus 3/38 controls (X2(1) -10.9,P- 0.001). Quetelet's i>'rdex• Quetelet's index was calculated using the subject's weight 5 years prior to diagnosis for 22 male cases and their matched controls and for 50 female cases and contrcas on whom this information was available.. No difference was seen between cases and'' controls of either sex. Al'cohol.• No significant differences in alcohol intake were found between cases and controls of either sex. Occupational exposure: No differences in occupational exposures were observed between male cases and' controls. In females, the only significant difference was that 14 eases reported working in a textile-related job compared to S controls (relative risk, 3.10: 95% confidence interval 1.1 1-8.64). Of the 14 female cases„2 were diagnosed with Kreyberg I, 11 with Kreyberg Il and I had miAed-type lung cancer. For those cases and controls interviewed between 1976 and 1980, information on the duration of exposure to occupational and environmental substances was available. There was no difference in the mean num-

1*o... S LUNG CANCER IN NONSMOKERS • Kabat and Wynder 1217 ber of years of exposure in textile-related jobs (16 years) T.u+t.r 3. Exposure to Passive Inhalation Among a Suttset of cases and controls. Among the cases, the specific oc- cupations were the following; one seamstress, two dress- makers, one sewing-machine operator, one assembler and yarnwinder. one dress-shop worker, two salesladies who had done factory work, one apparel manufacturer, one clothing paeker, one typist, one washerene/housekeeper, one bookkeeper, and one housewife. Among the 37 male cases only a few (5) neported' ex- posures to substances of potentially etiologic intenest. An electronics engineer had~ 35 years of exposure to cleaning chemicals; a designer had 25 years of exposure to chem- icals and acids and 15 years of exposure to plastics and glues: a director of sales for a chemical corporatiow (a chemist) had 12 years of exposure to chemicals and acids; an upholsterer had 30 years of exposure to asbestos, rub- ber, and solvenu; and a machine shop anendant had 37 years of exposure to metals. grease, "and oil: Among the 97 female cases, in addition to exposure to textile work reported by 14, few reported'; other ex- posures. The assembler/yarnwinder who reported~ expo- sure to textiles also reported exposure to metals for 28 years; a machine operator had 10 ~ years of exposure to metals: an assistant medical techni ,ian had 10 years of exposure to chemicals and acids; a social worker had' 5 years of exposure to metals and welding: am electronic prototype technician had 14 years of exposure to chem- icals and acids, metals and solvenu; and a chambermaid had 23 years of exposure to ammonia. We looked separately at the small~number of cases who develope& lung cancer younger than age 40, eight men and six women. The occupations of the men~ included an accounting professor, an accounting clerk (who had been a teacher for 11 years), a neurosurgeon, a stock trader, a postal service clerk, a law student, a salesman, anda self-employed president of a supply company. None of the men reported any exposures. The female cases included two housevrSves, an assistant manager for the American Automobile Association, an electronic pro- totype engineer (mentioned above), a telephone operator, and a high school teacher. Only the electronic prototype engineer reported any exposures. The distribution of his- tologic types among these younger cases did not appear to differ from that of all nonsmoking cases. Passive inhalation: Of the 25 male cases and controls who were asked about exposure to other peopie's cigarette smoke at home, six male cases reported having been ex- posed compared~ to 5 controls (Table 3). Eighteen of 25 cases reported having been exposed to cigarette smoke at work compared to 11 of 25 controls. T'he ditference is just statistically significanU (P = 0.05). Mantel extension test for linear trend in the fnequency, of exposure (four levels) in cases and controls gives a chi-square of 2.88, P < 0.005. The number of male cases and controls who of Cases and Controls Men Women Cases Controls Cases Controls (Nb.) (%) (No.) (4c)~ (No.) (%) (No.) (%) At home' Yes 6 5 16 17 No 19 20, 37 36 Total 25 25 53 53 At workt Yes LB 11 26 31 No 7 14 27 22 Total 25 25 53 53 (P < 0:045) Spouse smoke; Ever 5 5 13 15 Never 7 7 17 10 Total 12 12 24 25 ' Current exposure on a regular basis to family members who smoke., t Current exposure on a regulan basis to tobacco smoke at work. #, Spouse's current or past smoking habits reported that their wives smoked was identical, 5 of 12 in both groups. In both groups the wives had smoked for comparable periods of time. No differences on exposure to passive smoking at home or at work were found in women, 16: of 53 cases were exposed'at, home compared to 17 of 53 controls„and 26 of 53 cases were exposed at work compared to 31 of 53 controls. Of the women, who were asked, about their spouses' smoking habits, no differences between cases and controls were found in the proportion who~smoked, 13/24 for cases versus 15/25 for controls. Again, years of smoking in the cases' husbands did not differ from years of smoking in the controls' husbands. Discussion Due to the powerful role of smoking in the etiologyy of lung cancer, other risk factors can best be studied in nonsmokers with confirmed nonsmoking histories. Thus, a key feature of this investigation is that in~order to "val- idate" the diagnosis of primary lung cancer (obtaine& from~the discharge summary or the pathology repon) and; the nonsmoking status of all study subjects (obtained in the original interview), we went back to the hospital rec- ords and abstracted information on diagnosis andsmoking history. If the chart indicated that the patient had smoked tobacco at any period of his or her life, the person was excluded from the study. In the rare instance that no mention: of smoking history was found in the chart~, the patient was included. Of the 156 cases of lung cancer in

1218 CANCER March 1' 1984 va. st our computer file of self-reported never-smokers, review of the hospital chart revealed that 13 were actually smokers or had smoked at some time, and 9 were not primary lung cancers. These 22 cases were excluded from the anal- ysis. Confirmation of the diagnosis and' nonsmoker status of the controls was carried out in the same way as for the cases. For none of the controls was the self-reported nonsmoking status contradicted by information in the eltart. The finding that more cases gave a conflicting response on whetherr or not they had ever smoked than controls (13 of 147 primary lung cancer cases compared to none of 134 controls) is of significance. This suggests that some lung cancer cases tend to deny a smoking history more than controls with non-tobacco-related diseases. In a study of the role of cigarette smoking in lung cancer, such denial of cigarette consumption or under-reporting, which may also take place, would tend to reduce the estimate of the relative risk. In a study of lung cancer in nonsmokers, the inclusion of cases with a smoking history (misclas- sification) would also reduce associations of the disease with other risk factors. Although we attempted to eliminate all smokers from among the cases and controls by using a conservative definition of nonsmoker and by excluding any subject with a history of smoking either in the questionnaire or in the hospital chart, it is possible that some subjects who reported never having smoked actually did smoke at some time. The current study confirms earlier findings that among lifelong nonsmokers lung cancer is exceedingly rare, and that the more conservative the definition of nonsmoker and the more detailed the smoking history; the lower is the proportion of nonsmokers found among lung cancer cases.3 Histologic Type As found in earlier studies, Kreyberg type II (primarily adenocarcinoma) is more common in nonsmokers with lung cancer than in smokers and, in both groups, Kreyberg type 11 is more common in women. The percentages of nonsmoking cases with adenocarcinoma in our study (43% of males, 62% of females) are in close agreementt with those from the American Cancer Society's prospec- tive study (46% of males, 59% of females, L. Garfinkel, personal communication, 1982). In view of the differences in design and method of selection of subjecu, this agree- ment suggests that these percentages may be representative of nonsmoking lung cancer cases generally. Sex Ratio In our nonsmoking cases there are 2.6 times as many females as males, even though the male-female incidence ratio for lung cancer is 2.4,10 and the male-female ratio among all lung cancer cases in our file is 2.6 (1919/749); The larger number of nonsmoking women with lung can, eer compared with nonsmoking men is presumabl,, due to the historically higher proportion of nonsmokers among women compared to men. Doll found no difference in the age-specific death rate from lung cancer among non- smoking males and females.' Similarly, Garfinkel" found no difference in the age-adjusted lung cancer mortality rate for nonsmoking men and women. Case-Control Comparisons Previous diseases: Our finding that female cases had a higher frequency of previous history of pneumonia compared to controls is difficult to interpret since we do not have information on the age at diagnosis or on the duration of pneumonia. Occupation: Earlier case studies of lung cancer in non- smokers have included occupations in males with -ex- posure to dust and/or fumes, i.e.. a carpenter, a joiner, a fitter, and a Oour miller among the 7 male cases in Doll's study;' two painters, a smelter, a blacksmith, a gasoline truck driver, a gasoline and oil delivery man and gas station attendant, a cabinet maker, a sawmill worker. and an engineer among 20 male cases in Wynder's study; a plumber/steamfitter and' an auto body and fender re- pairman among 8 male cases in the study by Wynder and Berg.' Among female cases, the occupations were less suggestive of exposure to inhaled substances. These studies interviewed small numbers of nonsmoking cases, and did not make use of a comparison group, Our findings of a statistically significant threefold excess risk of lung cancer among women who reported having worked in the textile industry is of interest. Doll, in his study of lung cancer among nonsmokers. lists occupations of more than 3 years duration in 7 male and 40 female lung cancer cases. Out of 31 women who had been em- ployed outside the home, 5 had worked as seamstresses or di!essmakers.' However, there is no clear relationship in our data between duration of exposure and risk of disease. The mean number of years of exposure was the same for cases and controls. Most importantly, it is not clear, that there is a single exposure or group of exposures that all, of the workers in textile-related jobs have in common. Furthermore, it should be emphasized that our oc- cupational data are limited since there was room only to code one occupation-that of longest duration-and two exposures. Occupational and environmental!exposures to specific substances were obtained by asking the subjects whether they had' ever been exposed for more than a year to any of a list of substances. Selfreported exposures of this kind are subject to information bias since awareness of such exposure could be expected' to vary with the in- 20ti3382243

No. 5 LUTaG CANCER IN NONSMOKERS - Kabat and Gl!Ynder 121:9 dividual, with educational ~ level, with different jobs, and between cases and controls. In onlv 7 of the 14 cases did the coded occupation mention textile work. The re- maining seven cases reported occupations not specifically associated with textiles, such as "typist." but reported exposure to textiles. Evidence from existing occupa- tional studies of lung cancer risk in textile workers is scant.'=-" No cohort study of textile workers appears to have been carried out: The apparently minor role of occupational exposures in our male cases is consistent with the high percentage of professionals (60%) among them. Although our data do not suggest an important role of occupation or ex- posure to specific substances, it would be desirable in the future to obtain more detailed and objective occupational histories on cases of lung cancer occurring in nonsmokers. Passive inhalation; The plausibility of a role of passive inhalation in lung cancer can be questioned on several grounds. Although sidestream cigarette smoke contains higher concentrations of toxic components than main- stream smoke,'s it is diluted in the ambient air to varying degrees (depending on the size and shape of the room, proximity to the smoker, and ventilation) by the time it reaches the passively exposed person. As shown by Auer- bach and coworkets,16 the changes in the bronchial ep- ithelium characteristic of smokers are rarely observed in lifetime nonsmokers. Nevertheless, the possibility that heavy exposure to secondhand smoke over a long period of time could lead to increased' cancer risk cannot be ruled out at present. Secause questions on passive inhalation were introduced in our questionnaire in 1978, we only have information on this factor for between 28% and 68% of our subjects depending on the specific question. We present the dis- tributions of responses to these questions as preliminary data since the numbers are small. Cases do not differ from controls except for the greater exposure to cigarette smoke at work reported by male cases compared to male controls. Those cases who reported passive inhalation exposure did not differ in their distribution of histologic types from unexposed cases. The difference between ex- posure to cigarette smoke at work between male cases and' controls could be due to information bias, although there is no indication of such bias in the responses to the other questions on passive inhalation. The studies which, to date, have addressed the issue of passive inhalation and lung cancer have differed in methodology, the population studied, the type of lung cancer studied, the degree of histologic confirmation, and in results. These studies are summarized in Table 4. They have been commented' on by a number of investiga- tors."-"'14 We wish to draw attention here to several points which are crucial in assessing a contribution of passive smoking to lung cancer and which need to be considered in future studies. First, the proportion of his- tologically confirmed diagnoses in the studies listed in Table 4 ranged from 35% (Trichopoulos et aL [20J) to 82% (Chan and Fung 1211). Given the difficulty of di- agnosing lung cancer, histologic confirmation is essential. Second, Trichopoulos e1 a1:10 excluded' adenocarcinoma and'terminal bronchiolar cases, whereas adenocarcinoma predominated in Hirayama's cases=T (personal commu- nication, 1981), in those of Chan and Fung,21 and in our cases. In the American Cancer Society study reported by Garfinkel," histologic type was obtained for lung cancer cases during the first 6 of 12 years of the study. Seventy percent of these cases had histologic confirmation but some of these were only identified as "carcinoma." Among the cases with confirmed histology and information on specific cell type, 46% of the male and 59% of female nonsmokers had adenocarcinoma compared to 23% among male and 46% among female smokers (personal communication). Since little is known about the etiologic significance of different histologic types and since the distribution of types differs in different populations, it is premature to restrict studies of passive inhalation to par- ticular types. Third, although histologic classification of lung cancer is imperfect, it is desirable to stratify by the major his- tologic types in the analysis if the number of cases permits since different histologic types may have different etiol- ogies. Finally, all of the previous studies used the amount and duration of spouse's smoking as the measure of ex- posure to passive inhalation: Focus on the spouse's smok- ing may fail' to provide an adequate measure of the sub- ject's exposure for a number of reasons: (1) a subject's actual exposure depends on how much time the smoking spouse smokes in his or her immediate presence; the spouse could' be a heavy smoker but spend very little time at home;; (2) in addition to the current spouse's smoking habits, those of former spouses may be equallyy important; (3)~ the subject may live with other relatives who smoke; (4) exposure to tobacco smoke at work can be a substantial proportion of a person's exposure; (5) exposure in cars, commuter tnins, buses, and in other situations, such as restaurants, movie theaters, ete., could be significant. It is for these reasons that we have recently revised our questionnaire to include detailed questions which will give a more complete picture of the subject's exposure, both in respect to different environmental set- tings and to duration of exposure for each specific com- ponent. If passive inhalatiom in nonsmokers is associated with increased' lung cancer risk, by what mechanism does it exert its effect? Since adenocarcinoma is the most com- mon histologic type of lung cancer in nonsmokers, one could hypothesize that inhaled sidestream smoke increases

1220 CANCER March 1 1984 Vol' S3 TABLE 4. Summary of Studies of the Role of Passive Inhalation in LunB Cancer in NonSmokers Author/, type of uudy/ population No. of ases Histology Hirayama (198 1)22 174 deaths in muried Out of a sample of 23 axs. Proepective/ eonsmokin8 women 17 were adenoarcinoma Japanese w/lun8 ca among nonsmoking 91.340' nonsmokin8 wira aged 40+ married women YeaR Garftnkel (1981)" 195 deaths from lung a Analysis of data amon8 male from two eonsmokers: 564 prospective deaths from lung a studies/ACS among female population and nonsmokers (ACS):' Dorn study of 168 litne a deaths .rcterans" among nonsmokers (porn) Histolopic confirmation ofda in 69% of ases in 6rst 6 years of ACS study. Among lung cancer ases with confirmed detailed histology. 46% of male and 59% of female nonsmokers had adenocarcinoma compared with 23% of male and 46% of female smokers (personal communiation) Trichopoulos rr al. 40 female nonsmokers 14 cases were histologically (1981)2' Case- w/lung a other than confirmed: 19 were eontrol/white adenoca or term,nal! cytoloqicaily confirmed: 18 kmale raidents bronehiolar were clinialiy confirmed: of Athens. eaduded adenoamnoma Greece and terminal bronchiolar Chan and Funj' Only two nonsmokers IS of the 84 femole cases were Case-control/ out of 208 male lung squamous or epidermoid' Hong Kong a cases: 84 a: 38 were Chinese nonsmokers out of adenoarrinoma: 15 had no 189 female lung ca hiuol)Vc verihation patients Findings A dose-response relationship was men between the nonsmoking wives' risk and the htstiands' smoking habit: wives off easmoken or of t-19 ciss/day-smokers had RR - 1.61: wives of smokers of x20 cip/day had RR - 2.08 No sifnificant increase in lung ct risk seen in nonsmoking wives of smoking husbands compared with nonsmoking wives of nonsmoking husbands RR of lunf a associated w/ having a husband who smokes <I pock/day was 2.4: RR associated w/ having a husband who smokes > I pack/day was 3.4. (z' for linar, trend' - 6.45: P < 0.02) Among nonsmoking women the proportion of cases whose spouse smoked was sli8htly lower than that of controls (34 of 84 or 40 .~5% rs 66 of 139 or 47.5%): Among nonsmoking women. there was no significant difference in the proponion of cases who used kerosene fuel in cooking compared with controls. Ca: ancer: dit: diag<toaic ci8s: ci8arettes: RR: relative risk: rrr rerstu. • Chan WC. Colbourne MJ. FunB SC. Ho HC. Eronchial ancer in the risk for this type. Volatile components of cigarette smoke, including volatile nitrosamines, are more likely than respirable particulate matter to reach the periphery of the lung. Current findings suggesr most lesions in non- smokers are located in the deeper portions of the lung. Nonsmokers exposed to ciQarette smoke in enclosed spaces are reported to have increased levels of carbon monoxide in their blood,21-2D which suggests that other Comments Exposure index was based'on smoking habits of husbands Exposure index was based on smoking habits of husbands Exposure index was based on,smokin8 habits of husbands and former husbands It is unclear what' question was used regarding inhalation since in an earlier paper'. the question ~ is given as -An: you exposed to the tobacco smoke of others at home or at work?": whereas here reference is made only to "smokin8 habits of spouses." No information is given on how many subjects were marned Hong KortB 1976-1977. Br J Cance 1979: 39:182-192. ~ volatile components could reach the lung. It would be W. important to know in this regard whether the location of ~ lesions in the lungs of nonsmoking lungrancer cases with ~ exposure to passive inhalation differs from that among ~ smokers. N In addition to the etiologic factors discussed in this 4~, artide, other possible explanations of the occurrence of U1 lung cancer in nonsmokers should also be considered.

Nb.5 LuNG CANCER IN NONSMOKERS • Kabat and Wynder Exposure to ionizing radiation in the course of radiation treatment could be responsible for some cases. Also, Auerbach and coworkers26' have suggested that lung cancer could arise in nonsmokers secondary, to healed' tuber- culosis scars, although this is unlikely to account for many cases.2' Another possibility is that lung cancer in non- smokers, especially adenocarcinoma, is estrogen-related since it is more common in women than in men. It has been shown that adenocarcinoma of the lung frequently contains estrogen receptors.2' Still another possibility is that carcinogens of nutritional origin could be carried to the lung by the blood. These possibilities deserve epi- derniologic exploration. REFERENCES 1. Doll R. Mortality from lung cancer among non-smokers. Br I Cancer 1953; 7:303-12: 2. Wynder EL Tobacco ua cause of lung cancer, with,special ref- erence to the infrequency of lung cancer among non-smoken. Penn- aytfvcnia Mtd'J 1954; 57:1073-1083. 3. Wynder EL Bery JW: Cancer of the lung among ttonsmokers: Special reference to histologic patterns. Cancer 1%7; 20:1761-72: 4. Vincent RG. Pickren 1W, Lane WW et a1. The changing histo- pathology of1un8 oncer. A review of 1682 axs. Cancer 1977; 39:1647- 1655. 5. Ruffx P. Hirscti A, Marteau D. Bi`non J, Chretian J. Etude etiol- ogique et histologique de 448 ancrrs du poumon:,Ann Med lntrrn 1981;,132:12-15. 6. Wynder EL Stellman SD. Comparative epidemiology oftoba¢co- related cancers. Cancn Ra 1977; 37:4608-4622: 7: FleissJL Statistical methods for rates and proportions. New 1!ork: John Wiley and Son, 1981. a. Mantel N. Chi -square tests with one degree of fieedom: Extension of the Mantel Haenszel procedure. J Am .Ster Aksoc 1%3; 59:690-700. 9. Miettinen OS. Estimability and estimation in ase-referent studies. Am J Epidtmial 1976: 103:226-235., 10: American Cancer Society. Facts and Figures. Chicago: ACS, 1981. 11. Garfinkel L Time trends in lung cancer mortality among non. smokers and a note on passive smokinj. J Ntrtf Cancer Inst 1981; 66:1061-1066. 1221 12. Williams RR. Stegens NL Goldsmith JR. Associations of cancer site and type with occupation and industrq from the Third National Cancer Survey interview. J h'atl Ccncn /nst 1977; 59:1147-1150. 13. Heyden S, Patt P. Exposure to cotton dust and nespiratory disease: Textile workers. "brown lung,- and lung cancer. JA'MA 1980; 241: I797- 1798: 14. Heyden S, Fodor JG, Industrial cancer education and srnenine for 19,000 Canon Mills empioyees. J Chron Dts 1981; 34:225-23I i. 13. Brunnemann KD, Adams ID, Ho DPS, Hoffmann D. The in- Buence oftotircco smoke on indoor atmospheres:ll. Volatile aed tobacco. apecific nitrosamittes in main and sidestream smoke and their contri- bution to indoor ai'r pollution. ProceedinPs ofthe Fourth Joint Conference on Sensing of Environmental Pollutants. New OAnns. Louisiana. 1978; t76-880. 16. Auerb.eh Q„Garfinkel L Hammond EC. Chantes in.bronchialn epithelium in ntlation to cigarette smoking. 1955-1969 rrrsys 1970- 1977.,N'Eng! J Med 1979; 300:381-386. 17. Correspondence Br Med J 1981; 262: 28 February: 733. 21 i March; 985. 4April; 1156. 25 Apnl; 1393,283: 3 October, 914. 18. Hammond EC. SelikofLlJ. Passive smoking and lung cancer with comments on two new papers. Environ Res 1981; 24:444-452. 19. Lee PN. Passive smokinR Fd Chtm To.cicol 1982; 20:223-229: 20. Trichopoulos D, Kalandidi' A. Spuros L MacMahon B. Lung cancer and passive smoking. !nt! J Cancer 1981;,27:1-40. 21. Chan WC. Fung SC. Lung cancer in non-smokers in Hong Kon`,. In: Grundmann E. ed. Cancer Campaign; vol 16. Cancer Epidemiology. Stutteart. New York: Gustav Fischer Vertag. 1982; 199-202. 22. Hiiayama T. Non-smoking wives of heavy smokers have a higher risk of lung cancer. A study from Japan. Br Aftd'J 1981; 282:183-185. 23. Harke HP. The problem of "pastive smokine". Muenchtner Medtanischc Wochnuch'nJt, 1970; 112:2328-2334., 24_ Russell MAH. Cole PV. Brown E. Absorption by non,smokers of arbon monoxide from room air polluted by tobacco smoke. Lancer 1973; 1(7803);576-579. 25: Aronow WS. Effecu of passive smoking on anpna pectons. N Engf'J Med 1978; 229:2i-24.. 26. Auerbach O.,Garfinkel L Parks VR. Scar cancer of the lung Increase over a 21-year-penod. Cancer 1979, 43:636-642: 27: Hinds MW, Cohen HI. Kolonel LN: Tuberculosis and lung ancer risk in nonsmoking women. Am Rev Respir Du 1982;,125:776-778. 28. Chaudhun PK, Thomas PA, Walker MJ, Briele HA, Du Gupta TK. Beattie CW. Steroid receptors in human lung cancer cytosoks: Cancer l.euns 1982; 16s327-332. Vilter Symposium: Lysapbomas April 12,1984 - This symposium will be held at the Westin Hotel; Cincinnati„ Ohio. Direct inquiries to: Orlando J. Martelo, MD, FACP, Director, Hematology-Oncology Division. 6367 University of Cincinnati College of Medicine, 231 Bethesda Avenue, ML 0562, Cincinnati, OH 45267 (513) 872-4233.