Tobacco Documents Online

ENVfRONMENTAI. TOBACCO SMOKE AND LUNG CANCER 231 6Department of Health. ExecutiveYuan. Republic of China. Nealth Smaitrics. Vol. ff. Yh.f votLs1ics. 198t-199Z Taipei-. Deparcment of Ilealth, 1982-1997. 7 Ko YC, Lee CH. Chen MJ rr a1. Risk factors for primary lung cancer among non-smoking women in Taiwan. fnr J Hpidentinl 1997;26: 24-31. BAnnua/ Report of Tobacco. Alcohol Cansumptian lnvestigatinn in Taiwan Area. Bureau of Tobacco and Alcohol Monopoly. Taiwan Provincial Government. Taipei, Taiwan. 1974-1996. 9 Ko YC, Cheng-Li SC, Lee CH «a/. Chinese food cooking and lung cancer,in nonsmoking women. AmJEpidrmiu12000 (in press). sOSTATA. Statistics Sofrwarq Release 5.0. Santa Moniea. CA: STATA Corporation, 1998. f t Breslow NE, Day NE. Statistical Methods in Cancer Researrh. Vol. l. The Analysis of CaseControl Studies. Lyon: International Agency (or Research on Cancer. 1980. IZRothman 1U. The estimation of synergy or antagonism. Am J Epidemial 1976;103:506-11. 13 Brurzi P. Green SB, Byar DP, Brinton LA. Schairer C. Estimating the population attributable risk for multiple risk facton using case-control data. Am J Epidemiot 1985;122:904-14. 14Ger LE Liou SH, Shen CY Kao Sl, Chen KT. Risk factors of lung cancer. J Formosan Med Assoc 1992;91:5222-31. s5 Nyberg, F, Agudo A. Bof(etta P, Fones C. Gonzalez CA, Pershagen G. A European validation study of smoking and environmental tobacco smoke exposure in nonsmoking lung cancer cases and controls. Cancer Causes Control 1998;9:173-82. fb Brownson RC, Alavanja MC, Hock ET. Loy TS. Passive smoking and lung cancer in nonsmoking women. Am J Public Health 1992;82: 1525-30. 17Kabal GC Slellman SD. Wynder EL. Relation betsveen exposure to environmenlal tobacco smoke and lung cancer in lifetime nnnsmokers. Ani J Epidemiol 1995;142:141-48. tePeun GE. Burch JD. Hosve GR. Miller AB. The reliability of passive smoking histories reported in a case-control study of lung cancer. Am J Epidemiol f 988:127:267-73. f9Cummings K,bL Markello SJ, Mahoney MC. Marshall JR. Measure- ment of lifetime exposure to passive smoking. Arn J Epidemiol 1989; 130:122-32. 20Janerich DT. Thompson WD. Varela LR er a( Lung cancer and expos- ure to tobacco smoke in the household. NEngfJMed 1990;323:632-36. 21 Coultas DB. Peake GT. Samet JM. Questionnaire assessment of lifetime and recent exposure to environmental tobacco smoke. Am J Epidemiol 1989:130338-17. . z~ Brownson RC. Alavanja MC, Hock Ei Reliability of passive smoke exposure histories in a case-control study of lung cancer. 1nt J Epidemiof 1993;22:804-08. 21 Stockwell HG. Goldman AL, Lyman GH rt al. Environmental tobacco smoke and lung cancer risk in nonsmoking women. J Narl Cancer Inst 1992;84:1417-22. Z4Fontham ETH. Correa P. Ryenolds P et af. Environmental tobacco smoke and lung cancer in nonsmoking women-a multicenter study. JAMA 1994;271:1752-59. 2sZaridze D. Maximovitch D, Zemlyanaya G, Aitakov ZN, Boffetta P. Exposure to environmental tobacco smoke and risk of lung cancer in non-smoking women from Moscow. Russia. fnr J Cancer 1998;75: 335-38. 26 Boffeua P. Agudo A. Ahrens W et al. Multicenter case-control study of exposure to environmental tobacco smoke and lung cancer in Europe. J Nail Cancer (nrt 1998:90:1440-50.

ENVIRONMENTAL TOBACCO SMOKE AND LUNG CANCER 225 1974 1916 197a 19a0 1982 19" 1996 19" 19901992 1994 1996 ~ Years Figure 1 Smoking rate (%) and age-adjusted mona9ry rate (per 100 000 women) of lung cancer for women in Taiwan. 1974-t996 the National Department of Heahh,6 there have been about 1350 female deaths from lung cancer per year in Taiwan in recent years (1991-1997). However, only 9-10% of women suffering from lung cancer had a history of cigarette smoking.7 The prevalence of smoking in females aged >16 years has remained at the low level of 2-5% for a long time (1974-1996), but the mortality trend over this period, in contrast, has increased steadily (Figure 1) 8 Although cigarette smoking is considered to be the most important cause of lung cancer, it cannot fully explain the epidemiological characteristics of lung cancer in Taiwanese women, who smoke rarely but contract lung cancer relatively often. In Taiwan, annual nationwide surveys reported a high smoking prevalence (55-62%) among mene and, as expected, this has lead to a large amount of involuntary exposure of children and non-smoking women to ETS. Despite the fact that the authors in the previous study found that exposure to fumes emitted from cooking oils, when not reduced by an extractor such as a fume hood, appeared to be an important risk factor for lung cancer in non-smoking Taiwanese women,7•9 the role of ETS among female lung cancer patients should not be dismissed. The present case-control study was undertaken to investigate the relationship of passive smoking and lung cancer, especially for non-smoking Taiwanese women. Risk assessment for ETS was focused particularly on the effects of cumulative exposure in childhood and.adult life, and their interaction. Materials and Methods Study population This study used a case-control design with one case matched to one or two controls. The source population was geographically defined as all women resident in the greater Kaohsiung area, which comprised 1 I urban, 9 suburban and 26 rural communities, and has a total population of about 2.1 million inhabitants. The cases were drawn from Kaohsiung Medical University Hospital, which is a highly regarded teaching hospital in South- ern Taiwan, and is accessible to patients from all socioeconomic groups. A system of rapid case recognition was introduced for the determination of lung cancer cases, so those patients could be recognized and selected into our study as soon after diagnosis as possible. In this system, all newly hospitalized women sus- pected of suffering from primary carcinoma of the lung (lCD-9, code 162) were traced from hospital medical records and quickly verified histologically by a pathologist. A total of 295 eligible cases of lifetime non-smoking women suffering from lung cancer were identified between January 1992 and June 1998. Of these, 14 patients had died or been discharged by the tinre tite interviewcrs visitrd the wards, 7 were too ill to participate, and 6 refused to be interviewed. In all 268 (91 0) non-smoking female lung cancer patients were interviewed. The controls were derived from the same geographical areas as cases. They were also lifetime non-smoking women seleaed from hospitalized patients at the same hospital with conditions unrelated to tobacco use. The controls included patients with eye problems (cataract and glaucoma), bone fractures, and women undergoing physical check-ups. Identified by the same system of rapid disease recognition as the cases, the first andlor second eligible controls were selected within 3 weeks of the case being identified and matched to case on age (m 2 years). Only one control patient was matched to each of the first 78 cases during 1992 and 1993, due to administrative and budgetary limitations. All remaining cases were matched to two controls, except for 13 cases for each of whom, because of old age, only a single match could be found. Of the 492 non-smoking age- matched women identified as suitable controls, 445 (90%) agreed to be interviewed for the control group (eye problems: 52%; bone fracture: 27%: physical check-up:2l%). Interviews A structured questionnaire was completed for each case and control patient in a face-to-face interview by two trained inter- viewers. The questionnaire was designed to collect information on demographic characteristics. smoking history- lifetime occupations, dietary factors, history of lung diseases. cooking practices, cooking conditions. air pollution in general inside the home and ETS. A field supervisor checked all completed questionnaires and relevant medical records, which were then transferred to coding sheets for computer analysis. Data specification Lifetime smoking history of study subjects was collected through personal interview by the trained interviewers. Subjects' spouse or other next-of-skin were asked through a short interview for verification of their smoking status. People who did not smoke as much as one cigarette per day for one year, or > 365 cigarettes over their lifetime were considered lifetime non-smokers. Information about cases and controls exposed to ETS was collected for three categories. These included childhood (-18 years) exposure at home, adult life (319 years) exposure at home and workplace exposure. In each time period at home, any patient living with a regular family member who was a smoker for at least one year was considered to be a 'potential' passive smoker. Among these, a passive smoker was identified as . a patient whose family members had smoked in her'presence', as some Chinese smokers do not smoke at home in the presence of their family. After establishing passive smoker status, a series of detailed questions was asked for each active smoker identi- fied about such circumstances as: duration, starting year, stop- ping year and the number of years living with the smoker. For patients who had a regular job. the exposure was measured by asking patients whether they were exposed to tobacco smoke generated by co-workers in the workplace. Each job held for >_5 years was assessed separately. Patients who reported being passive smokers were also asked to indicate the number of years the co-worker smoked in their 'presence' and the number of active smokers in the workplace. Three indicators of tobacco smoke exposure were constructed for these three categories of

226 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY exposure: (1) a dichotomous variable which classified patients as those who reported any exposure versus none: (2) total number of smokers; (3) total number of 'smoker-years', calculated by summing the number of years the passive smokers reported exposure to each active smoker. Among these indicators, exposure sources in childhood at home included father, mother and other family member, and in adult life at home- included husband, husband's father and other family members. Otherwise, 'pack-years' of exposure to husband's smoking were estimated by multiplying the number of packs of cigarettes smoked per day in the subject's 'presence' by the number of years he smoked while living with her. Information on other air pollution factors inside the house, such as Chinese incense, mosquito coiis and cooking fuels were also collected. Questions were directed at the type of material used, duration and frequency of burning and the age of patient at the time of exposure. Each housewife who had to cook for her family was asked about cooking habits frequently used and the type of exhaust ventilation systems employed in kitchens such as chimneys or fume extractors. Other questions were directed at factors such as lifetime occupation, daily diet, and history of lung disease. Occupations were classified into one of five categories: administration, hazardous industry, farmec house- wife and others, and a person's lifetime occupation was taken to be the job they held the longest. Daily dietary habits were assessed by measuring the frequency and quality of consump- tion of 12 food items, including meat, vegetables, fruits. pickles, smoked food, etc. Information on lung diseases including tuberculosis, emphysema, chronic bronchitis and asthma were separately collected. Statistical analysis We analysed the collected data using the statistical software package STATA.10 Analyses included the x2 test for comparing demographic factors and some multivariate techniques specially designed for matched case-control studies. To explore the distribution of risk factors, tables of case and control numbers were studied rather than matched pairs. Odds ratios (OR) and 95% CI for various associations were determined from matched pairs using conditional logistic regression models.ll Statistical significance of trends for matched pairs were calculated by categorizing exposure variables and treating scored variables as continuous. All OR estimates were adjusted for residential area, educational level and occupation, in order to reduce residual confounding, as well as for tuberculosis, cooking fuels and fume extractor, which were found to confound the association between passive smoking and lung cancer in this study. Inter- active effects of exposure to ETS during childhood and adult life (including exposure at home and workplace) were evaluated by assuming an additive or a multiplicative interaction relation- ship. Main effect variables for the two exposure periods and their cross-product terms were incorporated into the logistic regression models for the testing of interaction based on a multiplicative model. The synergism 'S' index proposed by Rothman and its 95% CI were computed to evaluate the empirical deviation from the additive interaction relationship.l2 Otherwise, the proportion of lung cancer cases attributable to tobacco smoke exposure generated in different life stages (population attributable risks proportion) was calculated according to Buzzi eral.'s method.13 Results Demographic characteristics, confounding factors and cancer types Cases and controls were generally comparable with regard to background demographic factors and occupations (Table 1). Table I Distributions of demographic and confounding factors among non-smoking women. Taiwan, 1992-1998 Cases n=268 Factor/Category % Controls n=445 % Age (years) __.... <41 5.6 4.9 41-50 146 . 15.3 51-60 24.6 24.9 61-70 30.2 31.5 >70 25.0 23.4 Residential area Urban 32.5 30.8 Suburban 44.0 48.5 Rural 23.5 20.7 Ethnfcity Fukienese 89.9 87.9 Mainlander 4.9 ....... . . . 4.9 Others i2 7.2 Religion ......... Buddhism 33.2 39.1 Taoism 8.6 7.4 Folk religion 53 0 47 2 . . ..._._.. Others 5.2 6.3 Marital status Married 76.5 79.6 Single 1.9 1.6 Widowed 20.2 17.8 Divorced . 1.5 ~ 1.1 Years of education No 57 1~~~~ 52 4 ...._.. _......,...._... ..._._ . . _..,...__._.. .... . . 1-9 33.2 35.7 >9 9.7 11.9 Occupation Administration 28 0 24 5 ......,.. ._.__._. ._ _._._ . _ _.__ . .. __ ._.. ...._ . , __._ ... Hazardous industry 4.5 4.0 Farmer 26.1 28.5 Housewife 26.5 29.2 Others 14.9 13.7 Iuberculosisa No 2So 93.3 _ .. . ._.. _ _._ 443 . ~ 99.6 Yes t Y 6.7 1 0.5 Cooking fuelsa No cooking or gas 44.7 61 1 Wood or charcoal 46.6 33.6 Coal or anthracite 8.7 5.3 Kitchen with fume extractora Yes 56.4 3-77 84.8 No . 7 43.6 6 P 15.2 ° P < 0.05 for companson of cases and controls.

ENVIRONMENTAL TOBACCO SMOKE AND LUNG CANCER 227 There were no apparent differences in demographic characteo- istics between the three sources of control patients in this study. The mean age of the cancer patients at diagnosis (61.5 ± 12.2 years) was a little greater than that of the controls (61.2 t 11 .5 years), but the age pattern of case-control pairs was, of c'ourse, closely matched. Only three variables (tuberculosis- cooking fuels and fume extractor) in our data set substantially con- founded the relationship between passive smoking and lung cancer. Non-smoking cancer patients suffered frequently front adenocarcinoma (67.9 % ), squamous cell carcinoma (17.9%), and small cell carcinoma (1 L.2 i), and rarely from other cell type carcinomas (3.0%). Environmental tobacco smoke Distributions of childhood and adult life exposure to ETS at home or in the workplace are shown in Table 2. A total of 43.7%. 62.7% and 10.2% of lung cancer patients reported at least one year of near-distant exposure to any source of passive smoke at home or in the workplace during childhood and adult life. The risks of contracting lung cancer for ever-exposed women were 2.1-fold (95% CI: 1.4-3.1), 2.0-fold (95% Cl: 1.4-2.8) and 1.5-fold (95% C1:0.7-3.5) higher than that of never- exposed women in these three categories of exposure (data not shown). Women passively exposed to tobacco smoke as adults had the largest average number of active smokers in the workplace, however, the largest average cumulative exposure in terms of smoker-years were observed in the household. In addition, comparing exposure sources over the course of a life- «me, the highest proportion of exposure occurred in the house- hold during adult life (62.5-62.7%) and a smoking husband caused the largest share of household adult exposure (72.9- 76.5%) for both cases and controls. Taking all exposure parameters into account, a non-smoking woman whose father smoked in her presence during her child- hood had a 70% higher risk of lung cancer than a non-smoking woman whose father never smoked (P < 0.05; Table 3). A similarly significant elevation of lung cancer risk (2.2-fold) was observed for non-smoking women exposed to the passive smoke produced by her husband in her adult years. There were no significant excess risks of lung cancer observed if a smoking father or husband did not smoke in the participant's presence. At an exposure level of >20 pack-years from a smoking hus- band, lifetime non-smokers displayed a --I.S-fold increase in lung cancer risks. There was no evidence of significantly elevated risks of lung cancer due to exposure to passive smoke produced by other family members or from ETS in the workplace. Lifelong cumulative exposure to household or workplace tobacco smoke in different life stages was compared. Patients exposed to passive smoke at the highest level in childhood were 2.2-fold (95 % CI : 1.4-3.4) and in adult life 2.6-fold (95% CI: 1.6-4.2) more likely to develop lung cancer than non- exposed women. Childhood and adult life exposure were both found to independently contribute to the risks of contracting lung cancer, even when exposure in the other life stage has been taken into account. A clear dose-response relationship was evident (Table 4). Interaction effects The synergistic effects of household or workplace exposure to ETS during childhood and adult life (measured in smoker-years) Table 2 Distributions of childhood and adult life exposure to envirmtmenlal tobacco smoke at home or workplace among nun-smoking women, Taiwan. 1992-1998 Exposure category Cases Controls Childhood exposure at homea . . . . .__ ....,....__.. ...._ . ... No. of ever/never exposure ~ 1171151 127/318 Per cent of any exposure ( 43.7 28.5 Exposed subjects: No. of smokers (mean ± sd) 1.9: 1.0 2.0: 1.0 Total smoker-years (mean z sd) 29.7 x 12.3 ~~ 30.8 x 12 7 7 Per cent of lifetime exposure 31.5 327 Smoker-years from father (mean x sd) 17.9 z 8.6 18.1 t 9.2 Per cent of childhood exposure (%I 60.3 58.8 Adult life exposure At homeb No. of ever/never exposure 1681100 207I238 Per cent of any exposure (%) 62.7 46.5 Exposed subjects: No, of smokers (mean m sd) L8 x 1.1 1.7 z 1.0 Total smoker y<ars (mean x sd) 41.0 z 24.9 36.3 m 18.8 Per cent of lifetime ex osure I%) 62 5 62 7 p . _..__. ...._. ._.. . ...._..._ . _..,.._... Smokcr-years from husband (mean x sd) 3 L.4 3 16.9 26.5 x 16.6 Per cent of home exposure (9e) 76.6 73.0 At workplace° - No. of ever/never exposure 201117 21/294 Per cent of any exposure (%) 10.2 6.7 Exposed subjects: No. of smokers (mean ± sd) 2.7 ± I.6 2.5 m 1.9 Total smoker-years (mean ± sd) 32.8 z 22.7 26.3: 19.3 Per cent of lifetime exposure (%) 6.0 4.6 Lifetime exposure No. ofever/neverexposure 189/791 249/196 Per cent of any exposure (%) 70.5 56.0 Exposed subjects: No. of smokers (mean x sd) 3.1 x 2.0 2.6 x 1.8 Total smoker-years (mean a sd) 58.3 m 34.6 48.2 x 28.0 Per cent of lifetime exposure (%) 100.0 100.0 a Exposure sources included father, mother and other family member. b Exposure sources included husbands faiher. husband, and other family members. r Women who did not have regular job have been omitted in this calculation. were evaluated by including interaction terms in the logistic regression equations. The OR for non-smoking women who reported childhood passive exposure to tobacco smoke were higher than for those reporting no such exposure for each level of exposure in adult life. A positive does-response association between adult life exposure and the risk of lung cancer were found both in women who had and those who did not have childhood exposure. The OR observed for the combined effects of childhood exposure with various levels of exposure in adult life were also compared with those expected on the basis of the additive or multiplicative interaction relationship. The risks of contracting lung cancer among non-smoking women exposed to ETS at the level of <41 smoker-years in adult life was found to be a little greater than risks expected by the additive or muhi- plicative no-interaction models. At expnsure levels of >40

ENVIRONi511;NTAL TOBACCO SMOKE AND LUNG CANCER 229 Table 5 The interanion effecu fur lung cancer associmed with chddhood and adult life cxpusurc iu cnviruumenlal tubaccn .moke iETS) among nomsnwking women, Taiwan, 1992-1998 Adult life exposure' No 1-20 21-40 >40 Factor/ smoker-years Cases/ Controls ORb 95% Cf Cases! Controls ORf' 95 % Cl Cases/ Controls ORI' 95 % Cf Cases! Cnntrol, OR' 95% Cl Childhood exposure ~No~ ~79/196 1.0 13/26 1.2 10.5-2.6) ~ ~ 34/59 1.4 I0.R-2.51 25/37. i.7 10.9-3.3) Yes 18131 1.4 (0.7-2.9) 9116 1.8 (0.6-5.31 30/41 2.1 I1.1-4.0) 60/39 4.7 (2.4-9.4) Exp. By Addi.e 1.6 1 8 2.1 Exp. By Multl 4 ..._ 1.7 2.0 .. 2.4~ ~ ~ S(95%CI)° 1.3(0.1-248) 1.4 (0.2-6.7) 34 (1.0-10.8) ' Adult life exposures included home and workplace exposure. t' Odds ranos were adjusted (or rnsidential area. education. occupatinn. mber.ulusis. cooking fuals and fume ec[racaor. ° Expected odds ratio under adduive no-inturacrinn model, d Expected odds ratio under muhiplicaiive no-interaaiun model. ` Synergism index based on an additive modei. In the exploration of the cause-effect relationship and risk assessments between environmental exposure to tobacco smoke and lung cancer, the most commonly reported index of passive smoking in earlier epidemiological studies has been the presence or absence of a smoking spouse.16.17 This type of exposure index has been frequently queried and found to be inaccurate,°,Ig because of the lack of consideration of passive smoke exposure which may occur outside the household, as well as that which may have occurred during childhood. Cummings et af.19 have tried to measure the association between ETS exposure occurring during childhood, adult home exposure and workplace exposure. Unfortunately, little association was found, and reliance on measurement during a single life stage (e.g. adult life) or from a single source (e.g. spouse) to define lifetime exposure, probably results in the misclassification of a substantial number of subjects. On the other hand, the adverse health effects associated with passive smoking have been found to be different between exposures in different life stages. Janerich « aL20 and Wang et aL4 in their case-control studies indicated that exposure tn passive smoking during childhood may predispose a person to develop lung cancer later in life. regardless of adult exposure. Evidence from these studies sug- gests that the effect of passive smoking on lung cancer should be assessed on the basis of exposure from different life stages. No appropriate gold standard, not even urine cotinine level, is available to verify the validity of questionnaire measures of lifetime involuntary smoking. However, questionnaire responses with good reliability are believed more likely to have good validity.Coultaserat.21 andBrownsonetal.22comparedresponses on smoking behaviours of each household member between original and second interviews conducted within several months of follow-up. Concordance was high (>84%) both [or parental smoking status during childhood and for spouses smoking status during adult life, although the agreement for ETS exposure measured by quantitative methods were not as high. Methodological study concerned with the reliability of various exposure sources between subjects' passive smoking reports and exposure reports by those of surrogates also showed that agreement of exposure generated by parents or spouse was relatively high compared to that generated by other household members or co-workers.19 For studies of the health effects of lifetime passive smoking. for which parents and spouse were considered separately as the major exposure sources during childhood and adult life- high reliability on the relevant responses could reduce the probability of exposure misclassification in different life stages. . In this study, all possible snurces of tobacco smoke exposure were assessed from childhood and throughout adult life. The authors employed an index of passive exposure, which entbodies both the numbers of active smokers as well as the years smoked by each smoker to assess the effect of exposure accumulated in various life stages. The exposure index used concurrently verified that the subjects with a smoking relative or co-worker were in fact exposed to passive smoke (e.g. that the active smoker smoked in their presence), as not all subjects are substantially exposed to passive smoke when living with an active smoker. This is an important valid qualification of environmental exposure to tobacco smoke. In fact, if we carried out our data analysis on the basis of binary yes or no categories for the smok- ing family member or co-workers, most of our positive findings concerning the relationship between passive smoking and lung cancer weakened or vanished. This type of misclassification of ETS exposure may pardy account for the lack of significant positive findings in previous studies. The relationship between involuntary exposure to tobacco smoke in different life stages and lung cancer has been increas- ingly noticed in recent smdies.4•2o.22-26 Risk assessment studies concerned with exposure during adult life have consistently found that a smoking husband significandv elevated the risk of lung cancer at the highest level of expostve.''2-'-' a finding con- sistent with this study. In cuntrast. the risk panern of contract- ing lung cancer is nut so compatible for exposure occurring during childhood. Brownson rt 1122 2 and Fontham er a1..2't in studies with large sample sizes. fuund no excess risk of lung cancer from parents' snwking even thuugh cumulative house- hold exposure up to the level of >18 smnker-%ears was exam- ined. However, Janerich .•r al.2i) and Srockwell rr a1..23 as in our study, identified an about nvufnld risk uf luug cancer when the subject had ?20 shwker-years of household exposure during childhood. Wang er r.l.a found a 3.1-fuld increased risk of lung

228 INTERNATION.4L JOURNAL OF EPIDEMIOLOGY Table 3 Odds ratios (OR) and 95% Cl for lung cancer associated with sources of envirommcntal tobacco smoke in childhood and adult life among non-smoking women. Taiwan. 1992-1998 No. Exposure categorya Cases Childhood exposure at home Father Non-smoker 136 Absence 36 Presence 96 Mother ........,__.,..__ .._ . Non-smoker 260 Absence Presence 6 Other family member .._. _.___...,. . Non-smoker 167 Absence ~ 47 Presence 54 Adult life exposure at home Husband Non-smoker 82 Absence 40 Presence ~ 146 ~Pack-yearsc 1-20 55 21-40 53 >40 38 Husband's father Non-smoker 217 Absence 17 Presence 34 Other family member Non-smoker 136 Absence 72 Presence 60 Workplace exposure Co-workers Controls ORb 95% CI smoker-years 436 1.0 2 7 0.9 (0.3-3.1) 288 l.0 94 0.8 (0.5-1.3) 63 L4 (0.8-2.2) 192 1.0 89 1.2 (0.7-2 0) 164 2.2 (1.5-3.3) 89 1.5 (0.9-2.4) 51 2.5 (1.5-4.2) 25 3.3 (1.7-6.2) 352 1.0 43 0.5 (0.3-L1) 50 1.2 (0.7-2.2) 252 1.0 ~ 128 1.0 (0.7-1.6) 65 I 5 (0.9-2.3) 400 1 0 _ . . __. ...,... ._. 24 0.7 (0.3-L5) 2t 1.2 (0.5-2.4) Non-smoker 236 ....._.,.._ . .. ... . . .. .. .. .. ... _ . . _ Absence 12 Presence 20 a Smokers who smoked cigarettes in the presence of passive smokers were classified as 'presenc<'. otherwise 'absence'. 6 Odds ratios were adjusred for residential area, education, occupation. tuberculosis, cooking fuels and fume extractor. c The measure was (or women whuse husbands smokrd cigarettes in their present. Women whose husbands were non-smokers were used as reference group. smoker-years- in contrast, the increased risk apparently departs from the risks expected by the two interaction models. In the additive model, the significant synergism'S' index for exposure >40 smoker-years indicated that childhood exposure has modified the effect of exposure generated in adult life when women were exposed to the highest level of tobacco smoke. However, no statistical evidence suggests that the data significantly deviated from the conditions of multiplicative interaction relationship (Table 5). Table 4 Odds ratios (OR) and 95% Cl for lung cancer associated svvh cumulmi, cxposure ta emironnuntal tobacco smoke in different life stages among non-smoking wonsen. Taiwan. 1992-1998 Life stages/ No. Caxes Controls ORta 95% CI OR26 95% Cf Childhood exposure at home No 318 1.0 1.0 1-20 27 33 1.8 (0.9-3.6) 1.5 (0.8-3.0) . >20 90 94 2.2 (1.4-3.41 1.8 (1.2-2.9) P-trcnd 0.001 0.011 Adult life exposure° No 97 227 1.0 1.0 . .1-20. 22 42 1.3 (0.7-2.5) 1.2 (0.6-2.4) 21-40 64 ]00 1.5 (0.9-2.4) 1.4 (0.9-2.2) >40 85 76 2.6 (L6-4.2) 22 (1.4-3.7) P-trend 0.001 0.002 Lifetime exposure ~ ~ No 79 l96 1.0 ...1-Z0 16 _ _ 33 1.3 (0.6-2.6) 21-40 54 90 1.6 (0.9-2.6) 41-60 ~~ 43 59 2.0 (1.2-3.5) >60 76 67 2.8 (1.6-4.8) P-trend 0.001 ' a Odds ratios were adjusted for residential area, education, occupation, tuber- culosis, cooking fuels and fume extractor. 6 Odds ratios were adjusted for residential area, education, occupation, tuberculosis, cooking fuels. fume extractor and exposure in the other life stage. e Adult life exposures included home and workplace exposure. Population attributable risk proportion Passive smoke exposure generated in different life stages was assessed simultaneously to determine relative attributable frac- tion. The risks due to childhood and adult life exposure to ETS separately accounted for 18,4% and 25.5% of total attributable risks, even after confounding factors were adjusted for. When exposures in these two life states were combined, the popu- lation attributable risk proportion of lung cancer was elevated to 37.2% for non-smoking participants. Discussion In the present study, the authors identified a woman as a non- smoker by personal interview, and confirmed this from her family members. Although the subject's urine specimen was not available for the validation of smoking status. the proportions of non-smoking women among case and control subjects were very similar to those estimated by previous studies conducted in Taiwan 9.t4 Otherwise. Nyberg er a(. suggested that bias from smoker misclassification is likely to be insignificant, if smoking status of study subjects were validated by their next of kin.ls Women undergoing physical check-ups were selected as a part of the hospital controls in our study. Because the demographic distributions of such controls were comparable to that of the other two groups of controls, the possible selection bias result- ing from these controls should be limited and not influence our conclusions.

230 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY cancer associated with tobacco smoke exposure before the age of 23 years. Because an exposure ot 20 smoker-years is equivalent to living with one smoker during the first two decades of life, ahigh but not uncommon level of exposure.Z0 even if the differences in findings among the studies can not he reconciled, the impact of childhood exposure in enhancing tlte risk of lung cancer in adult life should not be dismissed. Our non-smoking women exposed to various levels of ETS in their childhood were found to contract lung cancer in proportion to their exposure, with significantly elevated risks (2.2-fold) at the highest level (>20 smoker-years). In contrast. significantly elevated risks (2.6-fold) in adult life were reg- istered at exposure levels >40 smoker-years, but not between 21-40 smoker-years, although as exposure increased so did the risks. Similar relationships between passive smoking in child- hood and incidence of lung cancer were observed by studies conducted in New York20 and Florida.23 and the existence of a threshold for childhood exposure to tobacco smoke was sug- gested. Otherwise, the exposure level which results in significant elevated risk may be easier to reach in childhood than in adult life. The increased risk associated with a one-unit increase in smoker-years was somewhat larger for childhood exposure than for exposure during adult life (OR = 1.35 for childhood; OR = 1.27 for adult life), when smoker-years was treated as a continuous variable and risk estimates were adjusted for ex- posure in the other life stages. Because of imperfect metabolism, detoxification and immunity in children,4 susceptibility to ETS in this non-smoking female population may be greater in child- hood than in adulthood. It is interesting that children exposed to household tobacco smoke are not only more susceptible to various respiratory syndromes,l but also to lung cancer later in adult life. fn addition, it has been proposed that women with a history of non-malignant lung disease are exposed to an increased risk of lung cancer.l7• 18 According to Janerich et al.,Zo childhood respiratory diseases may initiate certain changes, which may lead to lung cancer later in adult life. Our findings indicated that there was insufficient statistical evidence to explain the difference between observed and expected OR for the joint effects of passive smoke in childhood and adult life based on the multiplicative interaction relation- ship. However, the synergism 'S' index indicated that the interaction structure between exposures in these two life stages started to deviate significantly from the additive model at ex- posure levels >40 smoker-years. Although only about 32-33% of lifetime exposure to ETS occurred in childhood, this, based on an additive interaction model, appeared to modify the risk effect of subsequent tobacco smoke exposure during adult life. Otherwise, the appearance of a 2.6-fold of relative excess risk in the combined exposure of two life periods at the highest ex- posure level (>40 smoker-years), as well as 3.3-fold increase in risk observed from a population-based case-control study for non-smoking women who were exposed to both childhood household smoke and had the highest level of adult exposure (>48 smoker-years),24 may suggest that ETS exposure occur- ring early in life potentiates the effect of high doses of exposure in adult life in determining the development of lung cancer. There was a lower proportion of workplace exposure to tobacco smoke in this study (6.7-10.2%) than that reported by Kabat er a1.17 and Fontham et a1.24 (57.0-63.2%). This is not surprising since, for women, a large proportion of their occupation categories did not give them any chance of exposure to ETS (housewife: 27-29 %), or made it easy for them to escape the exposure (farmers: 26-29%). Exposure in the household has been incorporated in household exposure in adult life (others category; 14-15%, mostly women who work in their home). In fact, the number of active smokers and years exposed were examined for women who reported workplace exposure; the mean value of 2.5 smokers and 11.3 exposed years respectively was found in our control group, which is not far from such exposure observed by other studies.t9 In a female population, such as Taiwanese women, who smoke relatively rarely, but live in an environment with high male smoking prevalence, the population burden of lung cancer due to ETS is relatively important. In ourpresent study, about 18.4% and 25.5% of the aetiologic fraction of lung cancer were separately attributed to environmental exposure to tobacco smoke during childhood and adult life. Tobacco use by the children's father (58.5%) and wife's husband (72.9%) were found to account for the majority of exposure in a respondent's childhood and adult life, respectively. Otherwise, a 37.2% population attributable risk proportion due to lifetime exposure to passive smoke was detected, and therefore it is reasonable to expect that prohibition of smoking at home or in public places could yield considerable health benefits, for both children or tton-smoking women in Taiwan. fn summary. environmental exposure to tobacco smoke in childhood and adult life both contributed independently to the risk of contracting lung cancer. Children were found to be more susceptible to ETS than adults and such early exposure appeared to modify the effect of subsequent tobacco smoke exposure in adult life. Smoking prohibition would be expected to protect about 37% of non-smoking Taiwanese women against lung cancer. Acknowledgement This study was supported in part by a grant from National Department of Health, the Executive Yuan, Republic of China (Grant No. DOH-82-DC-001) and National Science Council (Grant No. NSC 84-0412-B-037-049, NSC 85-2331-B- 037-065). References t Surgeon General, Department of Health, Education, and Welfare. The Health Consequences of Involuntary Smoking: A Report of the Surgeon General. Pub. No. (CDC) 87-8398. Washington, DC: US Government Printing Office, 1986. ZBoard on Environmental Studies and Toxicology, Committee on Passive Smoking, National Research Council, National Academy of Sciences. Environmental Tobacco Smoke: Measuring Exposure and Assessing Health Effects. Washington, DC: National Academic Press, 1986. ;Environmental Protection Agency. Respiratory Health Effects of Passive Snoking: Lung Cancer and Other Disorders. Washington, DC: EPA Office of Research and Development, December 1992. 4 Wang FL, Love EJ, Liu N, Dai XD. Childhood and adolescent passive smoking and the risk of female lung cancer. tnt I Eridemial 1994; 23:223-30. S Chen CJ, Wu HY. Chuang YC «alL Epidemiology charaaeristics and multiple risk faaors of lung cancer in Taiwan. Anticancer Res 1990: 10:971-76.

O Inicrnatlonal Epidermobyual As•ooaiion 2000 Prinicd in G(ar linlam laerrner,.m.dloumal n/EpiR;nuNrgy 2010,21 .22-I-211 Lifetime environmental exposure to tobacco smoke and primary lung cancer of non-smoking Taiwanese women Chien-Hung Lee,'1b Ying-Chin Ko,ab William Goggins,c Jhi-Jhu Huang,d Ming-Shyan Huang,d Eing-Long Kaoe and Hwei-Zu Wangf . Background For a female population with a high lung cancer moaality rate, such as Taiwanese women, who smoke relatively rarely, but live in an environment with high male smoking prevalence, the risk and population burden of lung cancer due to environmental tobacco smoke (ETS) are relatively important. Methods An age-matched case-control study was designed to investigate the effects of cumulative environmental exposure to tobacco smoke during childhood and adult life on lung cancer risk among non-smoking women in Taiwan. InformatiotT`6-n passive smoking from all possible sources and life periods were obtained from interviews with 268 and 445 lifetime non-smoking cases and controls. Conditional logistic regression and synergism 'S' index were applied to the data to assess the independent and joint effects of passive smoking in different life stages while controlling for possible confounding variables. Results Risks of contracting lung cancer among women near-distantly exposed to the highest level of ETS in childhood (>20 smoker-years) and in adult. life (>40 smoker-years) were 1.8-fold (95% CI: 1.2-2.9) and 2.2-fold (95% Cl : 1.4-3.7) higher than that among women being never exposed to ETS, and the two variables accounted for about 37% of tumours in this non-smoking female population. Children were found to be more susceptible to ETS than adults and such early exposure was found to modify the effect of subsequent tobacco smoke exposure in adult life based on an additive interaction model. Conclusions Environmental tobacco smoke exposure occurring in childhood potentiates the effect of high doses of exposure in adult life in detetmining the development of lung cancer. Smoking prohibition would be expected to protect about 37% of non-smoking Taiwanese women against lung cancer. ....... ..._....._... _...,.... Keywords Lung cancer, environmental tobacco smoke, case-control studies, epidemiology, effect modification __._.. . ...,..._......_ ..._...__. _............_. ._.... _ ._..._.. _ . ..__..._._.._ ..__ .... Accepted 27 October 1999 ....__... .__..._......,..._........__. ....._...._ ....._ __. . __.... ....__....... _.._.... In 1986, two landmark repons by the US Surgeon Generall and the US National Academy of Science2 concluded that the involuntary inhalation of cigarette smoke by non-smokers can cause lung diseases, most notably lung cancer. More recently. ' Graduara Institute of Medicine. bschool of Public Hcalth, Kaohsiung Medical University. Kaohsiung. Taiwan. ROC 807. ` Massachusetls General Hospital Bioslatistics Centeo 50 Staniford Street, Baston MA 02115, USA, dDepartment of Inlernal Medical. °Departmenl of Surgery Medical. IDepanment of ophthalmology, Kaohsiung Medical Universily. No. 100 Shih-Chuan lss Road, Kaohsiung. Taiwan. ROC 807. Reprint requests to: Ying-Ch:n Ko. School of Public Heallh. Kaohsiung Medical UNversity, No. 100 Shih-Chuan Ist Road. Kaohsiung, Taiwan, ROC 801. Emaili ycko@mail.nsysu cdu uv the US Environmental Protection Agency has examined about 30 epidemiological studies from around the world and reached a similar conclusion.3 However, in these studies, the most common measure of exposure to environmental tobacco smoke (ETS) was the amount of tobacco smoked by the spouse of a non-smoking wife during their adult life together. Only a few studies assessed exposure by taking all possible sources and life periods into account. According to Wang er aL4 unless exposure to ETS is assessed throughout life, including childhood. important damage to children's health may remain hidden. Lung cancer has been the leading cause of cancer death for women in Taiwan since 1986. The cumulative mortality rate ranked as ninthhighest in an international comparison of 18 countries or areas.s According to annual cancer reports from 224