Tobacco Documents Online

Acta Oncalogica 78 r1989 Fasn 5 FROM THE DEPARTMENT OF CAtiCER' EPIDEMIOLOGY. RADIUMHEMMET. THE DEPA:RTMEIrT OF EPIDEA<fIbLO- GY, INSTITUTE OF ENVIRONMENTAL MEDICINE. KAROLINSKA IhST1TUTE., STOCKHOLM. AND THE DEPART- MENT OF LUNG MEDICINE. HL'DDItiGE HOSPITAL. HUDDINGE, SWEDET:. SMOKING AND PASSIVE SMOKING IN RELATION TO LUNG CANCER IN WOMEN C. S%FASsOK, G. PERSHAGEN and I.KLOMINEK Abstract In a population based case-control study the association be- twcen female liing cancer, and'some possible etiological agents was investigated: 210 incident cases in Stockholm county. Swe- den. and 209'age-matched population controls were intervie..ed~ about their exposure expenences according to a structured ques- uonnaire A strong association between smoking habits and lung cancer risk was found for all histological subgroups. Relative risks for those who had smoked daily dunng at least one year ranged between 3.1 for adenocarcinoma to 33.7 for small cell carcinoma inia comparison with never-smokers All histological types showed strong dose-response relationships for average dailyy cigarette consumption, duration of smoking. and cumula= tive smoking There was no consistent effect of'parental smoking on the lung cancer nsk in smokers. Only 38 cases had never been regular smokers and the risk estimates for exposure to environ- mental tobacco smoke were inconclusive. The high relative risks of'smalllcell'and squamous cell carcinoma associated with smoking may have implications for risk assessments regarding passive smoking. Key uardsr Lung carcinoma, smoking. environmental tobacco smoke,, case-control- Carcinoma of the bronchi and lungs (lung cancer) is a common and highly lethal malignant disease. The domi- nating role of smoking as causative factor is established through numerous studies. The incidence is generally much higher among men, but in the USA lung cancer is now replacing breast cancer as the leading cause oficancer mortality among womem(l). Among Svvedish women the trend for annual increase of lung cancer is second only, to malignant melanoma of the skin (2). Many studies have shown that adenocarcinoma consti- tutes a greater proportion of the lung cancer incidence in females than in males (3). The difference can partly be explained by differences in smoking habits between the genders, but there are some indications that a similarr pattern can be seen, ininon-smokers (4~ 5). During the last few years several studies have indicated that 'passive smoking' or exposure to environmental to- bacco smoke (ETS) may be of etiologicaliimponance. The findings have recently been evaluated (6)h Most of the studies have focussed on the effects of ETS exposure during adulthood, but some data suggest an effect of childhood exposure both in smokers and non.smokers (7, 8)., To further investigate the effects on women of~smoking and other possible etiological factors for lung cancer, such as ETS and radon exposure in the home, and: possible protective effects of some dietary, components, we per- formed a population based case-control study. The first part of the study,, addressing the risks associated with smoking and ETS,,is presented in this paper. Material and Methods The study included Swedish-speaking women living in Stockholm county between 1983 and 1986. Persons in the county with suspected or newly detected'lung cancer are as a rule referred to one of three clinical departments of lung medicine (Karolinska, Huddinge, and Sodersjukhu- set), or to the Department of Thoracic Surgery (Karo- AcceptedSor publication 6 September 1988'. 623

624: ~ ~ C. SS'E!+SSON~, G. PERS~HAGE2: AND 1i. KLOMlNEK linskaY for further investigation and/or, treatment. To be included in the present study, the subjects should be in a physical and mental condition that allowed an intervie.+ lasting between one half~to one hour. Suspected and confirmed cases were interviewed'in the hospital wards. For inclusion in the study the diagnosisshould be confirmed microscopically or by unambiguous chest radiograms, in conjunction with a typical clinicali course. The majority of interviews were made before the diagnosis was confirmed. When a case was confirmed and included in the study, a population control, born on the same day was chosen at random from the population register in Stockholm county. If~she could not be traced or, refused to panicipate, she was replaced by another wom- an„who was selected and contacted in the same manner. The controls were interviewed by the same persons that interviewed the cases. The control!interviews were made during a personal visit (58 17e) or by telephone. A hospital control group was also included in some ofl the analyses pertaining to ETS-exposure. This group was selected among those patients with suspected lung cancer, who were interviewedl but for whom the subsequent, in- vestigation ruled out this diagnosis.. The cases and! hospital controls were interviewed dur- ing September 1983-December 1985. The last population control was interviewed one year later. The time lagg between inter, ievcs of casts and~ population controls was mainly caused by the interval beiween admittance of a patienrto a clinic and definite confirmation or rejection of the preliminan diagnosis. Eighty-six percent of the study subjects were interviewed by two,physicians (CS and'JKii The remaining subjects were intervievxd by two other physicians. A structured questionnaire was used for the interviews. Incontained questions about frequency of consumption of food-stuffs rich in vitamin A, carotenoid6, and vitamin C. exposure to ETS, smoking. andi data on all dwellings in which a subject, had lived for more than two years con- tinuously. Exposure to ETS was assessed through ques- tions about domestic exposure during childhood as well as domestic and work environmenti exposure during adult life. The criterion for being classified as a smoker was that the subject should have smoked daily for at least one year. Statistical evaluation was made with the computer pro- gram EPiLOG (9). Relative risks (rate ratios) were mainlyy estimated by stratified analyses with the extension for trend of the Mantel-Haenszel procedure (10; 1',1). In the trend analysis the exposures were scored 1„2, 3, etc. In the matched analyses the exact method for computing confidence intervals (CI) described by Miettinen was used (12). In the unmatched analyses Cornfield's method was used (13). For some of the analyses multiple logistic re- gression models were used'~as well (14). Significance inter- vals presented imthe article are two-sided and'95%r Cl are used throughout. Results. The stLdy finally included 2I0 cases and 209 population controlk, In addition, 191 interviewed patients were shown not to have lung cancer. For 9 patients primary lung cancer could neither be confirmed nor excluded. Seven subjects refused interview and 5 could not be inter- viewed~ because of their, medical condition, One hundred and seventy-five (84"lc) ~of the population controls were first hand choices. One control did not have a corresponding case, since the case had to be excluded during the analysis, when an autopsy revealed pnmary carcinoma of the colon with pulmonary metastases and not primary lung cancer. For two cases no controls willing to be interviewed were found.. Table I shows the microscopical classification of, the cases. All but two were histologically or cytologicall} confirmed. In one of these an autopsy showed character- istic macroscopical changes of malignant nature, but un- fortunately a microscopical investigation was not made. Adenocarcinoma was most common and constituted ap- proximately one-third of the cases. The age distribution was similar in the different histological groups. Table 2 displays the diagnoses for the non-lung cancer patients along with their smoking status. Malignant dis- ease other than lung cancer was the most common cause and constituted approximately one-fourth of this group of' patients. An additional 17pauencs in this group had a: malignant disease although not directly associated with their respiratory ailments and consequently nouthe reason for their hospitalization. Table 2 also: shous that the pro- portion, of smokers among the population controls was smaller than for the non-lung cancer patients., Smokrng. Alllanalyses pertaining to smoking were made using the population controls only. The relative risk for 1ung, cancer for those who hadi ever smoked; vs. never- smokers was calculated both in a matched analvsis and in an unmatched analysis adjusted for age. The two methods yielded similar results, e.g. the risk estimate for all lung cancers was 5.8 (CL• 3.4-10.3) in the matched analysis and e4 (CI: 4.0-10.5) in the unmatched. The highest risk was seen for small cell cancer (33.7, unmatched) and the loa- ;est for adenocarcinoma (3.1, unmatched). The mean age for the cases, who were never-smokers, was higher than for those who had ever smoked (66.3 vs:. 61.7, p=0.009l. Table 3 shows the dose-response relationships for dif- ferent types of lung cancer with average daily cigarette consumption as exposure variable. Subjects who had stopped' smoking more than two years prior to the inter- view (for population controls two yearsbefore the inter- view of the corresponding case): were classified as ex- smokers. Smalli cell an& squamous cell carcinomas showe&the strongestitrend of increasing relative risk with increasing smoking intensity. For some of the estimates the CI was very widt because of the small number of never-smokers among the cases, especially among those

SMOKING AND LUNG CANCER.IN wOMEN' Table I Lung cancer cases among women in Stockholm counrv according to histological rcpe of7umor and'diqenostii 1Krlficatlon' Histol: id Cytol. id No microsc. id Total Mean a e ev ence ev ence ev ence n' c7c g ( ears) C/IC n °h n 9n y All cases 148 70j5 60 28.6 2 1.0 210' 100.0 62.5 Squamous cell 41 77.4 12 22.6 0 0.0 53 25.2 63.7 Small cell 30 66.7 15 33.3 0 0.0 45 21.4 62.6 Adeno 55 76.4' 17 23.6 0 0.0 72 34.3 ' 61.6 Other 22 55.0 16 40:0 2 5.0 40 19.0 62.3 t. Table 2 Diagnoses and smoking among female non-lung cancer parients intert:iru.rd at departments for pulmonan diseases in Stockholm counn as toefl as proportion of smokers among population con- rrols' Diagnosis n % Smokers c7c Malignant tumor 47 24.6 59.6 Breast 13 6,8 46.2 Gynecological 11 5:8' 54.5 Pneumonia or other respirator> infection 35 18.3' 54.3 Unspecified' pulmonary infiltration 23 12.0 65':2 Benign tumor or cyst 22 111.5 40:9 Pleuntis B 4.2 50.0 Tuberculosis 5 2.6 60.0 Sarcoidosis 5 2.6 20.0 Haemoptysis 5 2.6 100.0 Chronic bronchitis 5 2.6 100.0 Bronchiectasis 4 2.1 75.0 Atelectasis 4: 2.1 25.0 Other specified diagnosis 24 12.6 67.0 Unspecified diagnosis 4 2.11 50:0 Total 191 58.1 Population controls 209 42.6 with squamous or small cell cancer. Nevertheless, the lower limit of the CI was 2.9 and 6.9 respectively for the lowest, smoking category within these groups of cancer. Only one control belonged, to the highest exposure cate- gory, why the risk estimates for this category became very imprecise. Average daily consumption of cigarettes was highly correlated to cumulated smoking (r=0.90, Cl: 0.88--0:92)' and to duration of the smoking habit (r=0:73, Cl: 0.68-0.77). As a consequence of the high correlations the dose-response relationships were similar for these expo• sure measures. The inf]uence of the age at debut of daily, smoking on relative risk is shown in Table 4. The risks are adjusted for 625 duration of smoking. No statistically significant associ, ation could' be found between smoking debut and risk although almost all point, estimates of relative risk were higher for those starting before 25 years of age than after. Analyses were also made with simultaneous adjustments fonage and intensity of smoking.,The results were similar in both~types of analysis, although the risk estimates were somewhat higher when the latter type of standardization, was used. This could be expeeted as persons with an early debut also had a higher cumulated exposure within eacli age stratum. Other age stratifications were analysed. bur' the results,were similar to those presented. The effect of smoking cessation onirelative risk of lung cancer is shown in Table 5. Few subjects. espectall~ among the cases, with an average daily consumption of more than 10 cigarettes had ceased to smoke. The data indicated a considerable decrease in risk already within 10, years of smoking cessation compared to continued smok- ing: There seemed to be a stronger effect of' smoking cessation for squamous and small'cell carcinomas than for the other histological types. Enuironmenral tobacco smoke. Risk estimates of lung cancer associated: with ETS were mainly calculared' for easesand controls who had never been daily smokers, but for exposure to ETS during childhood calculations were also made for smokers. To increase power, the risk esti- mates presented for never-smokers were calculated withan expanded control group consisting of' population con- trols and those non-lung cancer patients. who did not have any malignancy. The estimates arrived at when using onl., the population controls were quite similar. The carcinoids and the microscopically unconfirmed cases were excluded from the risk calculations pertaining to ETS. Table 6 shows estimates of relative risk for smokers associate&with exposure to ETS from the parents. Sub- jects with a smoking father only, were classified as ex- posed toJow levels while subjects with a smoking mother were classified as exposed to high levels regarrdless of the smoking status of the father. No significantly increase& relative risk was seen in~ any, of the exposure groups.

626 C. 5VENSSON;G. PERSHAGEN AND. ].. KLOMINEK Table 3. Relatiue risk' (RRl jor lung cancer among tuomenin Stoekh'olin counn. in relation to aerragr dail) cigarene consumption' Never- k Ex-smokers Current smokers p for trendi ers smo n n RR >0-10'cig/day, >10-20 cig/day >20 cig/day CI) (9517 c n RR (95 C/r CI ) n RR' (959 CI) n RR (95c7c CI)' All cases 38 30: 2.6 42 4.6 81 12:6 19 59.0 <1 4 x 10-` (1.4-5~1) (2.5-93) (6:5-25.2), (7.6-) Squamous cell 5 6 4.0 10 9.7 28' 36.2 4 96.0 <3:8x 10-1` (ILO-16.9) (2.9-45.9) 010-168.9Y (6.9-1 Small cell 2 5 9.1 13 33.7 20 72.11 5 215.8 <1.8x10''" (1.4~9.7) (6.9-265:3) 0 1.9-452.6) (18.3-) Adeno 22 12 1.8 12 2.2 22 5.4 4' 19:7 <1.7x10- (0.8-4.3) (1.0-5.8) (2.4-13:2) (1.7-) Other 9 7 2.5 7 3,6 11' 7.5 6 82:5 <8.O x 10-' Controls 120 36 (0:".1) 30 (1.1-13.4) 22 (2.2-243) 1' ' The estimates are adjusted for age. Subjects who had stopped smoking more than 2 years before the interview (for controls 2 ytars before the interview ofithe matched casc) are classified as ex-smokers. 7 Ex-smokers not included~in calculations of linear trend The exposures were seored 1, 2. 3 and 4. ~' Upper confidence intervals not given because of imprecision of estimates due to the small number of individuals in the high exposure stratum. Table 4 Relatiix risk' (RR) for lung cancer among women in Stockholm counn associared'witk age at debute of dailY smoking" >25 years 19-25 years -18 years p for trend' n RR (95 9r Cl) n RR' (959c CI). All cases 32 58 2.0 52 1.2 0.9 (0.8-5.3) (0.5-2.8) Sqpamous cell 9 18 2.0 15 1.1i 0.9 (0 6-7.3) (0.3-3:8) Small cell 7 18 2.2 13' 1.3 0:9' (R:6-8.4)i (0.4-4.9) Adeno 10 I I 1.6 17 1.3 0.6 (0.4,.6:0)'. (0.4-4.4) Other, 6 11 2.2 7 1.0 0.7 (0.5-9:9)'. (0:2-4.2) 1 Controls _ 18 14 21 ' Stratified analysis adjusted for duration of smoking. Subjects who had stopped smoking more than 2 years before the interview (for controls 2 years before the interview of the matched'case) are excluded. = The exposures were scored 1, 2 and 3. although all ri'sk estimates exceeded 1.0 in women with smoking mothers. In never-smokers, adenocarcinoma constituted the dominating histological group with 22 (57.9 %) of the total of 38 carcinomas. There were only 5 squamous cell and 2 small cell carcinomas, making specific analyses of these histological groups unfeasible. Table 7, shows risk estimates for different, ETS exposure variables among never-smokers. Most of the point esti- mates of the relative risk were greater than unity but the CI were wide due to the small number of cases., There were no significanc wends. Multiple regression analysiss yielded risk estimates very similar to those presented in the table. There was a significantly increased 'risk' of being ex- posed to ETS in the home, if the subject herself was a

0 0 SAtOKING.AND,LL'NG CANCER IN wORiEN Table 5 Relariue risk(RRi for lung cancer among uiamen in Stockholm counr, associated with smoking cessation fsc) compared to current'smoking (04 vears afier cessarion) ,' Current smoking 3-10 years since cessation >10 years since cessation p for trend n n RR' n RR (959~ Cl) (959 C1,) All cases 142 16 0.6 14 0:3, 0.0004 (0.2-1.4) (0.1-0:6), Squamous cell 42 5 0.5 1 0.0 0.0006 (0.1-1.6) (0.0-0:4) Small cell 38 2 0.3 3 0.2 0j001 (0.0-1.3) (0.0-0.7) Adeno 38 5 0:5 7 0.5 0.06. (0.2-1'.5) Other, 24 4 0.7 3 0.4 0.08' (0.2-3,2) (0:1-1.6)i Controls 52 13 24 Stratified analysis adjusted for age and average daily cigarette eonsumption. ' The exposures were scored l. 2 and'3'. Table 6 Relative risk fRRi for lung cancer among euer smoking uromen in Stockholm counh in relation ro parental'smoking during the first, decade of li(e' Uaex- d Father smoker Mother smoker p for trend' pose n n RR n RR (95% C1); (95% C1) All cases 94 57 0!8 19 1.8 0.9 (0;3-1.4), (0_5-7.0) Squamous cell 27 17 0.7 4 1.3 0.6 (0.3-1.7) (0,2-8.8I Small cell' 25 13 0.7 5 2.1 1.0 (0.3-1.8) (0:3r14.0). Adeno 23' 19 1.1 8' 3'.0 0.3 (0.5 2.5) (0.6-21.6) Other 19 8 0.4 2 1.1' 0i08 (0.1-1.3) (0.0-20.0) Controls 45 39 5 Stratified analysis adjusted for age and average daily cigarette consumption.. 7 Regardless of smoking habits of the father. ° The exposures were scored 1, 2 and 3. t / Discussion ; AU subgroups of lung cancer were strongly associated with smoking. Due to the small number of never-smokers. 627 among the cases, especially among squamous and small cell! caneers, and of heavy smokers among the controls, the confidence intervals were wide. The magnitudes of the risk estimates were greater than, bun not, incompatible with, results from previously published studies on,female lung cancer (15-IB)i Contrary to previous studies (15: 17, 19) no clear associ- ation between early smoking debut and! risk was seen. although most of the point estimates were greater than unity when smoking debut after age 25 was used as refer- ence category.. In light of the clear dose-response rela- smoker. The point estimate for cases was 4.0 (CI: 1.7-9.3), and for controls 3.0 (CI: 1.5-6.2): For controls there was also a significantly increased 'risk' of being exposed to ETS on the job if the subject had ever smoked (RR 1.9, CI: 1.0-3.7). For cases the corresponding point, estimate was 1.1 (Cl: 0.5-2.6). 42-599105

628 C. 5VENSSON, G. PERSHAGEN AND J. K),.OMINEK Table 7 Rrltaritx risk'(RR) for lung cancer among never smoking women in Stockholm rountY in relation to different measures of exposure to ETS' Cases Controls. RR 951, Cli Exposure from the parents' Unexposed 19 98 1.0. Father smoker 12, 71 0;9 0.4-2.3 Mother smoker 3' 5 33 0:5-18.8 d (pSor trend':, 0.6) Exposure as a ult. Unexposed 10 60 1.0 At home or at work 17 90 1.2 0.4-2.9 At home and at work 7, 24 2.1 r 0.6-8:1 f i (p for trend': 0.4) et me exposure Li Unexposed 7 35 1.0 As child, or adult 13 88 1.4 0.2-2.5 As child and adult 12 51 1.9 0.2-3.7, (p for trend'.' 0,5) ' Stratified analysis adjusted for age. " Age 0-9 years. 3 The exposures were scored 1. 2'and 3. tionships in the other studies as well as in studies on men, the presennfindings were unexpected. A possible explana- tion could be that those who started to smoke at a younger age inhaled less deeply or that they to a greater extent smoked cigarettes with filter tips. The observed decrease in the relative risk of lung cancer after smoking cessationi is in agreement with previous observations (16. 18, 20). Approximately one-third of all cases were classified as adenocarcinoma.. Among the never-smokers adenocarci+ noma constituted almost 60%r of the cases. Among the currenn smokers the corresponding figure was 27 1Zc. The proportion of adenocarcinoma among the never-smokers is in good agreement with several previous studies on female lung cancer (5, 16, 21-23). The results pertaining to ETS in the present study were not conclusive. The small number of never-smokers among the cases could be one important reasonfit should be noted, however, that most of the point eslimates of . .: .- relatiW.nslR%er!,t,greatet` than tonity whicK agree with trestilt3i"!`roatprevious'studies on ETS exposure and with tislr`estitrlates'eonctxning'active smoking (6, 24). To reliably estimate the risk associated with ETS, it is essential to identify a sufficienr number of never-smok- ers. In the present study; only 38 of the 210 cases had never been daily smokers.. Four of these were excluded from the calculations of risks associated with ETS, since they had carcinoids or tumors which were not confirmed microscopically. A post hoc calculation of power for de- tecting a 50% excess risk associated with exposure to ETS in the home, showed that it was in fact only about 10%. For, detecting small risks, it is essential to minimize misclassification of exposure. The variablescharactenz- ing exposure to ETS used in this study may noobe optimal in this respect. Both intensity and temporallaspectsof the exposure are probably of importance for the outcome. It is very difficult, however„to retrospectively quantify ETS exposure. The tolerance for tobacco smoke differs be- tween individuals, and it is not improbable that this can influence their exposure estimates. If such individual in- formation bias exists, it is uncertain whether h leads to non-differential or systematic misclassification~ There are also difficulties involved in assessing the relative impor- tance of domestic exposure compared to exposure in the work environment. The high risks found for smokers withia lbw consump- tion in this study, and particularly for squamous and small cell carcinomas, have implications for the assessment of lung cancer risks associated with:ETS. On one hand, they suggest that relative risks of 3 or even higher for, squa• mous and small cell carcinomas in heavily exposed indi: viduals may not be unreasonable. On the other hand. they make control:of confounding by smoking a cntical issue. A poor control of confounding would be expected to primarily give rise to increased risks of these histological types. ACKNOWLEDGEtv1ENTS This study has been supported by the Jubilee Fund of the Swedish National Bank. We also want to thank Dr Margareta Lingner for va)uble help in the collection of data. Request for reptinrs: Dr Christer Svensson. Department of Epidemiology. Institute of Environmental Medicine. Karolinska . Institute. P.O. Box 60208, S-10401 Stockholm, Sweden, REFERENCES 1. US Department of Health and Human Services. The health consequences of smoking: A report ofi the surgeon general. Office on smoking and health. Rockville. Md 1982. 2:National Board of Health and Welfare. Cancer incidence in, Sweden 1983. Stockholm: Allmanna Forlaget. 198ti: 17. 3. IARC. Tobacco smoking. IARC monographs on the evalua- tion of the carcinogenic risk of chemicals to humans. Lyon;, IARC, 1'986; 38: 203. 4. Lubin,JH, Blot WJ. Assessment of lung cancer risk factors by histologic category. J Natl Caneen Inst 1984: 73: 383. 5. WynderEL, Mabuchi K, Beattie EJ Jr. The epidemiology on lung cancer. Recent trends. J' Am Med Assoc 1970: 213:, 2221. 6. NRC. Environmental tobacco smoke. Measuring exposures and' assessing health effects. pp. 223-249. Washington. DC: National Audemy Press, 1986. 7. Correa P, Pickle LW. Fontham E„ Lin Y. Tockman, MS. Passive smoking and,lung cancer. Lancet 1983. 2: 677. 8:, Sandler DP; Everson RB,, Wilcox AS. Passivt smoking and adulthood and cancer risk. Am J Epidemiot 1985; 121: 37. 9. Epicenter Softwar!e. Epilog. Version 2.0. Epicenter software. Pasadena. Ca 1984. 10. Mantel N„Haenszel W. Statistical'aspects of the analysis of

data from retrospective studies of disease. J Natl Cancer Inst 1959: 22: 719. 11. Mantel N'. Chi'-square tests with one degree of freedom: Extensions of the Mantel-Haenszel! procedure. J Am Stat Assoc 1963; 59< 690. 12: Miettinen OS. Estimation of relative risk from individuallyy matched series. Biometrics 1970; 26: 75: 13: Cornfield JO: A statistical problem arising{rom retrospective studies. ln: J. Neyman. ed. Proceedings of the Third Berke- ley Symposium on Mathematical Statistics. Berkeley: Ca: University of California Press. 1956: 4: 135., 14. Breslbw NE, DayNE. Statistical methods in cancer re- search: The analysis of case-control studies. Lyon: IARC 1980; 1: 192. 15. Benhamou E, Benhamou S. Flamant R. Lung cancer and women: Results of~a French case-control study. Br 1 Cancer 1987; 55: 91. 16. Lubin JH. Blot WS_Berrino F, et a]. Patterns of lung cancer risk according to type of cigarette smoked. IntJ Caneer, 1984; 33: 569. 17. Wu1AH. Henderson BE. Pike MC, Yu MC. Smoking and other risk factors for lung cancer in women. J NatI I Cancer Inst 1985; 74: 747. 18. Wynder EL. Stellmani SD! Impact of long-term filter ciga- rette usage on lung and larynx cancer nsk: A case-control study. 1 Natl Cancer Inst 1979: 62: 471., 19. Joly OG. Lubin,JH. Carabolloso M. Dark tobacco and lung cancer in Cuba.,J Natl Cancer Inst, 1983; 70: 1033. 20. DoIIiR, Peto R. Mortality in relation to smoking: 20 years' observation on male British doctors. Br Med 1 11976: 2: 1525. 21. Beamis JF 1r, Stein A, Andrews JL Jr. Changing epidemiolo- gy of lung cancer. Increasing incidence in women. Med Clin North Am 1975; 59: 315. 22. Pershagen G. Hrubec Z. Svensson C. Passive smoking and lung cancer in Swedish women. Am J Epidemiol 1987, 1:5. 17. 23., VincentTNl Satterfield JV', Ackerman LV. Carcinoma of the lung in women. Cancen 1965; 18. 559. 24. Pershagen G, Review of epidemiology in relation to passive smoking.. Arch Toxicol 1986; (Suppl 9): 63.