Tobacco Documents Online

[C4NCER RESEARCN i6..4801--j8'.0-. September 19861 Thl~ f'c;~',ci `.c1bt? Passive Smoking and Lung Cancer among Japanese Women protectEd Dy, cocy•,5~',' kw (Tide i7'J.S. Code;. Suminori Akiba,' Hiroo T;ato; and William J. Blot' Radiation EJjrm RrxorcA FoanQation. HiruJAima, Japan IS. A., H. K./, and karionalCawter lnstinae. Berbesda:.NarydanG20E92(,14'. lD.J, .kBSTR'ACT A case-control studa cooducted in Hiroshima and tiitgasaki, Japan, revaltd a 50% increased'risk of IWtg cancer among tsoasmoking women whose husbands smoked. The risks tended to I.ncreue with amount smoked by thehusband, being highest among women who worked outside the home and whose husbands were beavy smokers, and to decrease with cessation of ezposure.,Tbe findings provide inceotive for further evalua- tioo of the relationship betweenpassire smoking and cancer among nonsmokers. [NTRODL.'CT1ON As part of a case-control investigation of lung cancer among atomic bomb survivors conducted primarily to evaluate the interactive roles of cigarette smoking and ionizing radiation (I)4 data were collected on the smoking habits of the subjecr's spouses and parents. Herein we report the effect of exposure to such passive smoking, focusing on married women who had never smoked themselves.. MATERIALS AND METHODS Since 1951 a cohort oG 110.000 Hiroshima and Nagasaki atomic bomb survivors has been followed by, the RERF,' formerly called'the Atomic BombCasualty Commission (2). During the period 1971 to 1980. 525 newly diagnosed cases of primary, lung cancer (Eighth Revision ICD 162.1) were identified among cohon members. The cases were ascertained from the Hiroshima and !ragasaki Tumor and Tissue Registries. the RERF mortality, surgical. and autopsy files, and Hiro- shima University medical records. The diagnosis was based on biopsy or surgical pathology findjngs for 25`b, on autopsy findings for 28%, on cytology for 4%, and on radiological/clinical findings for the re- maining 43%. Since the cohort represents a fixed population that is aging over time and is older tlian the general populition, the ages at diagnosis were higher than usual for lung cancer in Japan: the means were 72.1 for males and 70.2 for females: the ranges were 36 to 94 for males and 35 to 95 for females. Controls were selected from among cohort members without lung cancer, 2 for each casc in Hiroshima and 3 for each nse in Nagasakj. The controls were individually matched to the cases with respect to yr of birth (t 2 yr), city of residence (Hiroshima or Nsgasakj); sex, and whether or not they were among the 20`'ro of the cohort participating in the program of bienniali medical~ examinations giveo at RERF. In addition, controls were matched to cases on vital status. Since most.of the cases had died, most of the controls were also deeeased_ The deceased controls were chosen according to the above-mentioned matching criteria, plus year ofdeath (2 3 yr), and'tJtey were seliened from among all causes of dnth except cancer and,chronic respiratory disease.,The distribution of the controls series is as follows: ali.e, 13%; deceased from, cerebrovascular diaease, 2646; from ~ coronary heart dis- ease, 13%; from other circulatory d'nease, 12%;;from acute respiratory Received IoJI/a5:rcvised 4/24 ri6:aoeepted 5/19/a6. The.costs ofpubliotion of.this artick wtrt defrayed in part by the prymentt of paae charges. This artickmust therefore be hereby marked adsrtisen.rrrt in accordance .Yitn 18 U.S.C. Section. 1734 solely to indinte this f.ct. ' Towhom requests for reprints andcorrespondenae from outside the 1JNited Sutesvlwuld'be addrrssed, at Departmentof, Epidtmioloqy and SutiT Radiation Effects Research Fouodatioa.. 52. Hijiyama ParL Hirostiima 730. Japan (S. A.):.and from.the United Sutes, a.Epialemiolop'. aod Biostatistio Propam. Nitional Caocer Iasatute, tandow Buildlna. 3C16., BetEe.da M'D2U892 (w'. J. B.): r The abEreviations used are: RERF. Ridiarcion, EftavReaurrE Foandatioa: OR:.odds r*tiolsk Cl. confidence mterr.l(s). disease. 9 a: from.digestive disease.,Fic: from accidents. 6<: and from other causes. I4 e. Interviews were sough't during 1982 with all cases and: controls: or their next of kin- who lived'in Hiroshima and tiagssalu, The interv)e„" ers were aware that the study concerned lung caneer, but they were not told of' the case-control, status of the stud} subjects. A structured questionnaire was tued to obtain, histories of cigarette smoking and demographic. medicali oocupational, and other fbctors. If the indnrdual was married, inquiry was made about the smoking status of the spouse., including the average numbtr, of cigarettes smoked per day, age started smoking, and, for those who stopped; the age oficessation of smoking.. Using this information, togetherwith the numbers of yr the husband and wife lived together, an index of e uposure to the spouse's smoking was calculated. In addition, a single question was asked regarding whether the subject's mother and/or fatherr smoked when the subject was living at home as a child- OR were calculated as measures of the association between lung cancer and passive smoking and other factors (3), Estimates of the OR, and corresponding significance tesu. were obtained by a conditional logistic regression analysis for matched'data (4). Tests for trend use& consecutive integers for levels of the ordered'eategories. Because there were a priori hypotheses that passive smoking might increase lung cancer risk. alllsignificance tests for passive smoking effects were one- sided with 90% CI used for interval i estimates of the OR. Because interest focused on spouse smoking patterns, eliminated from the analyses were the one case and'6 controls among males and the 4 cases and 7 controls among females who were never married. Among the married individuals, almost all l had been married to only one spouse. Among those with more than one spouse, information was availablF only for the most recent. Also excluded from each,tablt were individuals with missing data for the variable being studied. RESULTS Interviews were obtained for 428 cases and 957 eontrols, respectively, 81 io~ and 82% of the eligible cases and controls. The two primary reasons for nonresponse were the refusali of next of kin to answer questions about, their deceased relatives and the decision not to attempt to locate next of kin for subjects who had moved outof Hiroshima or Nagasaki. The distribution of informants is given in Table 1, indicating that the informa- tion for most of the subjects was provided by next of kin. The type of trspondent„however, was similar for cases and controls. Table 2 shows the lung cancer OR according to the smoking status (smoker versus never smoked)' of the subjects and their spouses. In both sexes there was an increased lung cancerr risk associated with direct smoking. As indicated, almost all (93 %)~ of the male lung cancer cases were smokers, but only a minority (38%) of the women with lung cancer in this population were reporte& to have ever smoked. Although not shown, the OR increased with the numbers of cigarettes usually smoked per day during adulthood'for both men and womena Among males who smoked I to 9, Wto 19,,20 to 29, and 30+ cigarettesper day, the OR were 1.7, 1.8, 3.4, and 9.7; respectively (P for trend < 0.01); Among females who smoked I to 9, 10 to 19, and 20+ cigarettes per day, the OR were 1.9, 2.0, and 4.9 (P for tren&< 0.01). Table 2 shows that among female nonsmokers married to smokers, there was an elevated risk for lung cancer (OR - 1.5; 90% CI - 1.0 to 2.5; P- 0.07). Although similar increases associated with smoking habits of spouses were observed for female smokers and for male nonsmokers and smokers, sufTi- 4804 i

PASSIVE SMO1atiG AND LG'tiG CANCER AMOtiG JAPANESE WOMEN Table I PerrtnsadedirrriDrrion oJ rrspondenu TabFr 5. Oddr ratios Jor bnt mncnansana.r.otv/no/uind -o~n ocrordiit8.to sub)act (`•t ) Ser of stud necenc7 oJarposr.r to ArsAwd!' nwluna y Tii f Mak Female mc o Case Control OR' 90% C1' exposun Respondent. C.ae Control Cue Control None 21 82 1.0 ' Self' Spouse 6 51 7 48 16 12 19 11 Not exposed within last 10 yd 31 87 13 (0.9- 2:4) Esposed.nthimlast 10 yr 40 85 1.8' (1.0. 3:2): Child hter-in la» D.u 23 11 22 12 33 18 33 17. (P for trend - 0.05) . a Others 10 ~ 10 21 19 ' Odds ritio .nd'90'S CI : from matched analysis. "Tbese 'es-passire amokers' nr• tbose -bose husbands quit smoking 10 or Totat I r I 00 I00 100 100 more yr prior to the diapwsis of lung ancer lor 10 or: more yr prior to the date of ickcaion for controls):or thorr -ho wen not livioy with their husbands becsuse n 264 595 164 362 ofsepsrstions, divorce. or his deatlt 10 or more yr pTior to the dia8nosis. Table 2 Oddr ratios Joe l itwj mwce ' r accog dinr to sao kiag natas oJiAe srAjen Table 6 Oddt ratios of litna mtacer .imont r•oRSnwklrat aornen acnvrdins to their oterpetion aed their Wu6ond.r' s+wokinj aarttt A a+td is/6rr tpou se H usbrnd's Sex of Subject Spouse Occupation smoking subieet smoker smoker Case Control OR' 90% Cf' of subieet sntti:' Case Control OR' 90% CI' Mak Nb, No" 16 101 1_0` Housewife' Never 6 20 1'.0` Yes 3 9 1.8 (0:5.5.6), Light I I 34 0:9 (0.4. 2.1) Yes No 190 388 3.4 (2.1.5.5)'. Heavy I5 35 1.5 (0.7; 3.3) )'es 51 86 4.2 (2.4.7:3)'. Whiiecotlu' Never, 7 23 1.0 (0.4,2.4) Female Nb, Na' 21 82 LO` Light 9 20 1:.71 (0.7:.4.5) Yes 73 188 1.5 (1.0. 2:5), Heavy 8 16 1.6 (0.6. 4.1) Yes No 8 14: 2.2 (0:9. 5.1). Yes 50 56 3.6 (2.1.6.1) , Blue collarr Nevee 6 21 1.1 ~ (0.4, 2.9) ' Odds ratio and 90% C l from ma tched anal ysis:. Light 5 22 0.5 (0.2:1.5) " Individual reponed ne ver to hav e smoked c igarette s. Heavy 7 6 10.4 (1.6, 66.7) Referent category. Table 3 Oddi rarioj for ArsAund'1 rnr cawc tura/ dai amorta n ly roRSrmp nsntok tion of rr +~o+nen cilarener ccording to ' Light. husband smoked less than 20 cigarenes/dal: tscavy, husb.nd smoked 20'or more ci8arenes/dq. "Odds nuo and 90% Cliftom matched analysis. `House.rife defined as woman who was employed outside the Some for so more than l0'yr. No. of e7pRttes husband usually smoked/dAy Case Control OR' 90% Cl' 4 Reference category. ` Ofl-icc and s.ks workers. 2Ezeludes 5 ates and 34 controls who were fivmcn. k" 0 21' 82 1.0 1-19 29 90 1.3 (0.7. 2131 20-29 22 54 1.5 (0.8. 2.8l 30+: - 12 23' 2.1 (0.7, 23). ' Odds ratio and 90% C7 from ttatabed analysis. (P for trend - 0.06) Tabk 4' Odds rotios for Gur oaurr awr,ont noRSnwtfit8 wonnn .ctordits8 to wsbatd's divacion of snsolung cqarmes wAik rwamrd Yrhusb.nd smoked ei8uettes Case Control OR' 90% CI' 0 21 82 1.0 1-19 20 30 2.1 (1.0.4.3) 20-39 29 a1 1:5 (0:8.2:7) 40+ 22 59 1.3 (0.7. 2.5) ' Odds ratb and 90% CI from matched analysis. cient data for detailed analyses of passive smoking patterns were available only for female nonsmokers. The data for nonsmoking women are categorized in Table 3 according to the number of cigarettes the husband usually smoked per day during adulthood. There was an increasing lWng cancer risk wirth increasing amount smoked per day by the husband, with the OR slightly exceeding 2-fold for women whose husbands were heavy'smokers. No monotone trend of increasing risk associated~with increasing duration of exposure to husband's smoking was found (Table 4). Risks according to time of exposure are examined in Table 5. The odds ratios were lower among "ex-passive smokers" than among women who had been exposed to their husbands' smoking within the past 10'yr. The reduction in risk with cessation of exposure remained after adjusting for the amount oficigarettes smoked per day by the spouse. As shown in Table 6, the risk of lung cancer tended to increase in relation to exposure to the husband's tobacco smoke for each of housewives, white collar, and blue collar workers. The highest odds ratio occurred for women who had blue collar jobs and were married to men who smoked one or more packs of ciga- rettes per day, but the numbers involved were small. The odds ratios from the matched logistic regression analyses presented in Tables 2 to 6 are generally similar to unadjusted odds ratios that can be calculated from the cross-products of the numbers of exposed and unexposed cases and controls, indicating tharconfounding in unadjusted analyses by age, city, vital status, utd yr of death (the matching factors) is not substantial. We also assessed whether the associations with passive smoking were consistent across the various strata de- fined bythe matching factors. The numbers of subjects in several of the categories became quite small with this fine a cross-classification, but the trends with husbands' smoking tended to be seen throughout, with no strong differences by age group or by city of residence. The trends were also apparent for each type of informant (self, husband, child, and other); in particular the elevated risk for heavy relative to nonexposure to husbands' smoking was detected when data were reported by the husbands or subjects themselves. Radiation exposure was also examined as a potentialiconfounder and effect modifier. No significant influence of radiation dose on the passive smok- ing association was detected, although the trends with passive smoking seemed stronger among the unexposed. Information on the histological types of lung cancer was unavailable for 43% of the cases who were diagnosed only on radiological or clinical evidence. We conducted separate anal- yses among those.vith and without a pathological confirmation of lung cancer and found increased risks associated with passive smoking for both groups. The OR among nonsmoking women 4905

PASSIVE SMOkiNG AND~ LUNG CANCER AMONG JAPANESE WOMEN n.arried to smokers w°as 1.4 for the cases and their matched controls with a histologically confirmed diagnosis, and 1.6 for those with a clinical'yradiological diagnosis. Among women with a histological diagnosis, adenocarcinoma was the predom- inant cell type, but the distribution of histological types varied by smoking status (Table 7), The percentage of squamous and small cell carcinoma was much higher among smokers than nonsmokers. Although based on small numbers, there were also more squamous and smali i cell cancers among nonsmoking females whose husbands smoked. Responses to the question on parental smoking while the subject was a child were provided for only two-thirds of the subjects. Among these the mothers of the subjects were reported to be smokers for 13% of the cases and 17% of the controls,. and the fathers. for 67°'ro of the cases and 66% of the controls. Hence there was no overall increased risk associated with parental smoking, nor was there any significant increase after stratifying by smoking status of the subject. Among male smok- ers, the OR for lung cancer associated with maternal smoking was 1.1. DISCUSSION The results from this case-control study suggest that there may be a moderate excess in lung cancer risk associated with` passive smoking. The odds ratios for lung cancer among non-' smoki.~ women tended to increase with amount smoked by 1.,:. _ _.,.. . _ t~eeu hus rtds,' a trend~ seen among,,housewixesv,asytfg*llcas women who worked outside the home:-The highest~odd.r.at}~os among nonsmokers were for women who worked m blue coUar jobs whose husbands were heavy smokers, women presumably with the highest exposure to environmental tobacco smoke. There was little association with parental smoking or with ex- passive smoking, suggesting that cessation of' exposure may lower risk. The findings are generally consistent with results of a national cohort study of mortality, among Japanese women (5) and of several epidemiological investigations conducted elsewhere in the world (6-8): Updated follow-up for the period 1966 to 1981 of the study conducted among an adult population selected from multiple areas throughout Japan, excluding Hiroshima and Nagasaki. showed a gradient in mortality with amount smoked' by the husband (9): The increase in risk reached 90% among those whose husbands smoked 20 or more cigarettes per day; a figure in line with~ the 2-fold excess for 30 or more cigarettes per day of smokers in our study. The similarity in results, despite different methodological approaches, suggests that the association between lung cancer and passive smoking is not an artifact of recall bias which can affect retaospective studies. Furthermore, we were unable to identify any strong confounding factors, including radiation exposure, that may have accounted for the passive smoking association. ft is noteworthy that a recent survey in Kyoto, Japan, found significantly elevated levels of cotinine, the major metabolite of nicotine, in the early morning urine of nonsmokers who lived in households with smokers or worked in offices/factories with Tabit 7 PemwRr hi.aolodicd,QiuuiMttiow ojlwq n.wnrs swaow: jrwrla accad,ne wo rAei.wd'Mnv t._s.=-aY' rnwkiwr aauu Ceu type t%) Squamoum or Adeaaaatinomr 5obien HusE.ad small tt.tl or tnrVr.ceu rrmokn ®oker. oaac.er caaaer. No No 0 100 Yes 16 r 94 Yes 58 42 smokers (10). The cotinine concentrations among nonsmokers living with 2-pack-a-day smokers were roughly equivaltnt to the cotinine levels of smokers of less than 3 cigarettes per day. Precise estimates of the lung cancer risk associated with, this level of smoking are not available, since not many smoke so few cigarettes per day. However. 3 well-known prospective studies of mortality among smokers [the American Cancer Society study involving nearly I million volunteers (11), the 16- yr follow-up of 250,000 United States veterans (1'2): and the 20-yr follow-up of 34,000 British doctors (13)] foun& relative risks of lung cancer of 4.6, 4:8, and: 7.8 among 1 to 9, 1 to 9: and I to 14 cigarette-per-day smokers, respectively. Linear interpolation between these valLes and the base-line level of 1.0 for nonsmokers would yield estimated relative risks for 11 to 2 cigarette-per-day smokers of nearly 2-fold, about the same order of increase observed for "heavy" passive smokers in this study. Hence, if the Kyoto results (10) are applicable elsewhere,' and if urinary cotinine levels reflect levels of exposure to the carcin- ogenic substances in tobacco smoke, then the observed magni- tude of the increased lung cancer risk among passive smokers in Japan seems not greatly out of line with what might be expected based on their exposure to environmental tobacco smoke. It should be noted' that the risk ratios for lung cancer asso- ciated with direct smoking (as shown in Table 2) were lower in this case-control study than typically found in cau-control and cohort investigations in other countries (14). The lower OR among smokers in part arises from our selection, in order to minimize respondent bias, of controls matched to cases on vital status, which led to the inclusion of some controls who died of smoking-related diseases. However, lung cancer risk ratios gen- er'ally similar to those in this study were also reported in the prospective study of Japanese adults (9). Because of the lower relative risks of lung cancer among smokers in Japaa, differ- ences in the OR betweendirect and passive smokers are not as high as in western countries. lindeed, we found OR for 'heavy' passive smokers to be nearly, equal those for women who were reported to be light smokers themselves. While such similarity was unexpected, characteristics such as the size and style of midentiall units might result in a higher env'tronmentaJ-to- direct tobacco smoke exposure ratio in Japan (and thus less of a difference in OR for lung cancer between passive and dii'ect smokers). This in fact is suggested by the comparison of the cotinine analyses between Japan and Great Britain (i0; 15); where the ratio of cotinine levels in passive compared to direct smokers was considerably higher in Japan. Our finding that lung cancer risk among nonsmokers may be less closely related to duration of exposure to tobacco smoke, the major determi- nant of lung cancer risk among smokers (13), than to intensity and recency of exposure also may be noteworthy. Such a differ- ence might contribute to a higher ratio in Japan of lung cancer risks in passive compared to direct smokers, since the current prevalence of smoking is higher in Japan than in either Grean Britain or the United' States, but the marked temporal increase in smoking began later (9, 16). The present study did not rrplicate the finding of a case- control study in Louisiana which showed a higher risk among male smokers whose mothers had smoked(R): Although we did find higher percentages of smokers among both cases and controls and among both men and women whose parents had been smokers, thers was no elevation in the OR among smoking ' Thett s some Qoestion about ttx'v aeeerittiraNlity. siac: oounme . kxb r,mon8 beavy pasnt &mokms in Kyoen-ert .bour ooe-motE tDee k•eb m axrur smokm, ie wevtwt.ton6outooe-fifonb in a mcrot Bnoa6 sruQy. (15). tnl+oth snbdri bo,.e•v:,tite rnnary k~b Lavxs.edLc propornooton esam.urd D- smotint ei0oatrs. #{3{?(~~xiri4A,<

PASSIVE SMOKI1iG AND LL'tiG CANCER AMONG JAPANESE WOf.1EN 1 Japanese men or womem associated with maternal or paternall smoking. However, it was often difficult for the respondents too provide information on parental smoking, and data on this exposure were missing for about one-third of the subjects. One of, the eoncerns in this study was the adequacy of data provided by surrogate respondents. Only a minonty of the patients could be interviewed directly because ofithe often fatal outcome of lung cancer and the need to include cases diagnosed as early as 1971 in order to assemble sufficient numbers of subjects for analysis. The distribution of respondent types was comparable between cases and controls so that response bias is unlikely, but the possibility of poor quality information for both cases and controls existed. We could evaluate this possibility, however, since many of the cases and controls had provided information on their smoking habits in routine RERF surveys conducted in the 1960s when all study subjects were alive (1, 2). The data in Table 8 indicate very high concordance in the identification of a female as a nonsmoker or smoker by a next of kin in 1982 and by the woman herself in the 1960s. In addition to providing some confidence that the data provided by surrogates are adequate, the confirmation of nonsmoking status by a next of kin argues against the possibility that Japanese womemtend4o report themselves as nonsmokers when they actuall} smoke. The 1982 survey revealed a higher per- centage of male smokers than reported earlier, but the increase was both for self as well as next-of-kin interviews and may reflect an aetual increase in smoking prevalence over time. Questions about the smoking habits of spouses were not asked in the surveys in the 1960s. so that self versus surrogate report- ing on this variable cannot be assessed directly. In our study, however, there were no significant differences in the passive smoking trends according to respondent type. in partieular, an increased OR was seen for nonsmoking women whose husbands were heavy smokers when the data were reported bythe hus- bands themselves. Another concern in this case-control study was the reliability of the diagnoses of lung cancer. Forty-three % of the cases were diagnosed solely on clinical and/or radiological evidence. The percentage was high in large part because the cohort being followed was elderly, and surgical or biopsy, procedures were less likely to be performed on older patients. The OR associated with passive smoking, bowevet, were similar when calculationss were restricted to histologically eonfirmed cases. We also cal- culated OR after deleting 23 cases and their matched controls for whom a diagnosis of possible or probable lung cancer was made only on radiological grounds and who had stuvived3 or more yr (all were in fact Gving as of January 1984), since the diagnoses for at least some appear to be questionable. Little change was noted. Smoking has been shown to induce a1t types Table 8 Carqoa-iwn oJnwatiV aarr from the 1992 nx-oownd mdy ad JtERF ar..eyi i. 1964 w 1968 The numbers of p.ired respomd for the 4 aes-mformam wmaoriea below are 58. 679. 45, and 92; re.pectirely. Sex of lnforrnant 1964-t9d8 eunent 1982 mrokine wrus. ( 4n f. sub)ect ia 1982 amok'er Never Smoker Male Selfi No. 18 14. ves 0 68 Surroyate No. 12 13 1'es 1 74 Femak Self. No 87 0 Yes 0 13 Surrvtate No 65 3 Yes 0 32 of lung cancer, but its effecti is greater for squamous and small cell~ carcinoma than adenocarcinoma (17). W hether passive smoking might have the same predilection for squamous can- cers is not clear, but our limited histological data (Table 7) are consistent with this notion. It is ofiinterest, that the highest OR for passive smoking has been reported from a case-control study in Greece (6, 18 ' 19) where the cases were limit'ed' to lung cancers other than adenocarcinoma. in summary, the results of this investigation suggest that exposure to envi>•onmental' tobacco smoke may increase the risk of lung cancer among nonsmokers. The findings, from one of the two areas of the world where the possibility of a passive smoking hazard was first postulated, add to an accumulating body of evidence on the issue. While the total evidence is not definitive and not all studies show significantly positive asso- ciations (20-22), the results are suggestive enough to warrant further evaluation in larger studies where passive smoking exposures can be more fully quantified. ACKNOW'LEDGMENTTS We thank Dr. Robert Hoover and Dr. Joseph Fraumeni. Jr., for 6elpfui sMestions, Dr. B. J. Stone and Dr. J'ay, Lubin for advice and computer assistance, and Ttieresa Pino and Michele Rasa for taanu- Ksipt preparation. REFERENCES 1., Blot. W. J., Akiba S.. and Kato. H. loninn8 radi.tion and lun8 aocer. a review inclYtdinB prcliminary rssults from a au,eone>'of stud) amon8 A- bomb survivors. 1n: R.,Prentioe aad D. Tbompaon (edf.). Atomic Bomb Suevivor,Dau, pp 235-248. Philadelpbia: SIAM. 1984. 2. , Bse(ie, G. W.. Kano, H., and Land. C. E. StudieY of tAe martallty of A-b'omb ~~. wrvivors. R.diac. Res.. I8: 813-649. 1971. 3. Breslow. N. E., and Day.,N. E.,The analysis of wse<ontrol studies LARC (lot. ASnr<y Res.,Cancer) Sci. Publ.. 32:, 1-281. 1990. 4. Lubin. J. A computer pro8ram for the analysis of matcbdd easeconaol snMies: Comput. Biomed. Res..,11: 138-143, 1981. 5. Hirayama. T. Nbn-smokin8 wives of Aeavy, amokers bave a dijber.nali of to8 onar. a zndl from Japan. Br. Med. J!„Id2i 183-185, 1981. 6. Triciopoubs. D., Kal.ndidi. A.. Spartoi L., and MacMabon. B. LunB cancer and passive smokin8. Ioa,J. Cancer. 27: 1-r. 1981. 7. Corms. P.. Pickle. L. W:, Fontaam. E., Lin. Y.. and Haenszel, W. Paasive amokine and'limf csncer. tM nat: 2: i95-597, 1983. 8. Gvfmkel, L. Arerbacb. 0.. and Joubert. L. tsrolnntary smoking and litn8 eancer. a oanecontrol study. J. Nad: Canar lnst. 7d: 463-469. 1985, 9. Hinyama, T. Passive smoking and lunTS cancsr. oonainency of anociation. Laecst. 2: 1425-1426, 1983. 10., M.tsukura, S., Taminato. T.. Kitano. N.. Seino. Y:, Hamadi H.. Ucbi6ahi, M., NilkyjimR H:,,aod H'u•tu. Y. FJfects of enviroomentat toti.cno smoke on urinary counine escrrtbn m aooasmokers: evidence for passive smoking. N. En81. J. Med.. 311: l28-832. 1984. 11: Flammoed, E. C. Smoking in relatxio to the death rates of onemillion men anQ wamra. Narl, Cancer Inst. fNoooQ:. 14: 127-204; 1966. 12: RoBot, E-, and Murr>+y.,J. L Smoking and causes of death among US eterans: 16 years of obae"we. Public Health Rep.,,95: 213-22P, 1980: 13. DolL R., and Pno. R. Mortaliry m relation to smokins: 20 yean obaetvation on mak Bririah doctors. Br. Med. J., 2: 1525-1536. 1976. 14. Surgeon General. The beahL consequeooes of smokin8: esocer. Wasbington. DC: Dep.rtment of Health aodHuman. Services. 1982. 15. Wald: N. J.. Boreham. J.. Bailey. A., RiteAie. C_ Haddow. J.. and Kaisht. G: Uriaatyy coeininr as marker of lsrsatEiinB otDer peopk's tob.cco xmoke. Laoc:r, 1: 230-239, 1984. 16. Don; R:, and Peto: R. The muaes of oncer. J. Nail. Canczr lnst-, 66: 1197- 1312, 1981. 17. Lubin. J.. H., asd Blot. W. J. As.raameni of lung cancer risk factorss by bistolopccateaory. J..Natl. Cancn Inst.,.7J: 383-389..1984:. 18. Triebopoulos, D.. K.laodiQi. A.., and~Sparros. L..LunB cancer and:p+nire smokin8:conclusion ofGrak srudj. Laoaet_ 2: 677.678, 1983. 19... Trichopoalos. D. Passive smoking and lung canccr..Laecct.. 1r.684. 1984. 20.. Garfinkel. L. Time rrends m.lun8 cancer mortaliry among ooo-emokers and a note oo psv.e tmokina.. J. NGtI. Caocer lnaL. 66: 1061-1D66. 1981. 21. Koo;,L.,C., Ho;,J'. H.,.and Ss.. D, ls psa'rw smokin8 an added risk factor for kn8 oetxr m Cbiaese wmen. J. Eap. Clm.. Cancer Res.. J: .277-283,. 1'984. 22, Kalia. G. C:..ed W yndkr. E. L. Lung catocv in nonsmoken. Casar. (Pbilat ). , 53: 1214-f`22I, 1984. 4807 2 0 2 351 04,10, 123s -