Tobacco Documents Online

Em•ironnrental tohnrro amoke and kmg eanrrr average number of hours per day working with smoking co-workers. A variety of measures reflecting different aspects of exposure to ETS were developed. The simplest index was the classification or exposure to ETS as a dichot- onious variable. As information was obtained both in childhood and adulthood (?23 years old) and at home and at work, separate indicator variables were created to represent each of these situations. Duration of domestic and occupational exposures to ETS was also assessed. Intensity of exposure at home during adulthood was measured by the average number of cigarettes smoked daily by subjects' husbands (99.6% of women were married), and intensity at work was measured as the average number of hours working with co-workers who smoked. Potenriol confounding variuhle.c We elicited information on level of education, incomc, first-degree relatives with lung cancer (parents, brothers, sisters), dietary habits, variables measuring indoor air pollution from Chinese-style cooking, and occupation. Most cooking in Shanghai is Chinese-style stir-frying. This represents a potential health hazard as it has been found that fumes will be produced when cooking oil is heated to about 280 °C [16], and that exposure to the oil fumes may be associated with the risk of lung cancer [10, I I]. We thus created an indicator variable for subjects who reported that fumes were visible during cooking at least three times per week. The extent to which the kitchen filled with smoke was defined as the following: "none" meant that no smoke (from heated cooking oil) was ever visible during cooking and "considerable" meant that the kitchen was on average filled with smoke such that vision was affected during cooking. All other conditions were defined as "somewhat". Diet was assessed by inquiring about the usual frequency and amount of consumption of 91 food items representing the niajor foods consumed by residents of Shanghai. The questions were used to assess the average consumption 5 years before the interview, ignoring recent changes. The monthly amount (grams) and monthly frequency of consumption were obtained for each food item. Daily intake of individual nutrients including vitamin C, beta-carotene, protein, and fat, was calculated by multiplying the amount of each food item consumed by the per-gram values in the National Chinese Food Composition Tables [17j, and then summing across food items. For each job held for at least 2 years, subjects were asked to describe their occupation. industry, duties carried out. and to provide the number of years of 609 employment. Job and industry titles were coded ac- cording to the system used in the 1982 population census of the People's Republic of China (Ig). The following occupations were considered as being at potentially high risk [19]: hat and shoe makers; timber processing; wood, bamboo, hemp, rattan, pahn and straw products makers; paper manufacture and paper products makers; stone sharpeners: welding and flame cutters; and painters. An indicator variable was created for subjects whose occupation of longest duration was in any of these high-risk occupations. We also devel- oped variables representing cumulative duration of employment in each of these occupations. It has been found that employment in the textile industry may be protective for lung cancer [20], but there were too few subjects to assess this. Statixtietrl analrsi.c We included in the unconditional logistic regression [21/ models all variables that were accepted or suspected risk factors for lung cancer, and the same set of variables was used in the analysis of each histological type. To gain insight into patterns ofexpostu'e-response and to provide a finer level of control or confounding, we made use of Generalized Additive Models 1221. Briefly, these models allow the nonparametric fit of continuous independent variables using nonparametric smoothing techniques. We used locally weighted regression smoothers (LOESS) because of their flexibility in specifying the amount of smoothing and in portraying interactions. For variables modeled parametrically we also performed two-tailed tests for linear trend by comparing the ratio of the estimated regression coeflicient and its standard error to a standard normal distribution [21]. Results Of the 706 eligible cases (including smoking and nonsmoking women). 57 had died before they could be asked to participate in the study. In-person interviews were completed for 649 (91.9%) cases. One hundred and eight (16.0%) of 675 control subjects were alternate control subjects. enrolled mainly because the primary control subjects had moved out of Shanghai and could not be traced. We excluded the 145 case and 74 control subjects who reported that they had smoked at least one cigarette per day for at least 6 months, thereby leaving 504 lifetime nonsmoking lung cancer patients and 601 nonsmoking control subjects. `Interviews were conduct- ed with surrogate respondents for 14 (2.3%) and 102 (20.2%) of control and case subjects•, respectively. Fifty-

610 three percent of these interviews were conducted with the husbands of subjects. Two hundred and ninety-one (57.7%) lung cancer cases were diagnosed by tissue biopsy. 96 (19.0%) by cytology, and 117 (23.2%) by repeated X-ray films combined with other clinical evidence. Among the 387 pathologically or cytologically confirmed lung cancer cases, adenocarcinoma was the predominant histologic type (76.5%). The remaining cell types were distributed as follows: 48 squamous cell carcinomas (12.4%), seven small cell carcinomas (I.8%), one large cell carcinoma (0.3%), and 35 carcinomas of mixed cell types (9.0%). The distribution of selected variables in case and control subjects is shown in Table I. The age distribu- tions of case and control subjects differed because a higher proportion of cases who smoked were excluded, L. ZlrnqC et al. thus altering the original age-freqitency matching. Hence, the higher odds ratios for ages 35-54 years and 60-64 years in Table I do not reflect the underlying age-- incidence curve for lung cancer in Shanghai. Table 2 shows odds ratios (OR) and 95% confidence intervals (Cl) for lung cancer by different indices of exposure to ETS, adjusted for the variables listed in Table I and for respondent status. Because of the high correlation between vitamin C and beta-carotene (Pear- son correlation coefficient = 0.90) we found it su8icient to include only vitamin C in the regression models. Age, income, and daily intake of vitamin C were modeled nonparametrically by LOESS using two-thirds of the data around each data point to carry out the smoothing ("span"). For childhood exposure there was a sugges- tion of increased risk only among cases diagnosed with a nonadenocarcinoma (OR = 2.4; 95% Cl: 0.9-6,4). Tnhle !. Distribution of selectcd variables for noncmoking lemalc lung cancer caw• subjects and control subjects, Shanghai. China. 1992-94 Characteristic Controls (n ° 601) Cases p7 = 504) OR' 95 %('1" No. Percentage No. Percentage Age (years) 35-54 12x 21.3 149 ]9.i 1.7 I.'---'.4 55-59 128 21.3 88 17.5 1.0 (/.7-I A 60-64 159 26.5 141 28.0 1.3 11.91.8 65-69 186 31.0 126 25.0 1 Monthly household income (yuan per person) 560 146 24.3 . 89 17.7 1 60.1-83.3 155 25.8 133 '6.4 1.4 1.0-2.1 83.4-115.0 145 24.1 132 26.2 1.5 1.6 2.1 > 115.0 155 25.8 150 29.7 1.6 1.1-2.3 Family history of lung cancer No 582 96.8 463 91.9 1 Yes 19 3.2 41 8.1 2.7 1.5- 4.7 Employment in potentially high risk occup,rtions` No 518 86.2 389 77.2 1 Yes 83 13.8 115 2?.8 1.5 1.I-2.1 Average dail) inlake or ritamin C(mg) 548.8 151 25.0 167 33.2 1 48.9-66.0 I 5n 25.0 124 24.6 0.7 0.511.9 66.1-83b 150 25.0 10] :0A 0.6 n.4-n.8 > 83.6 150 2511 112 "_' 0.6 0.4 0.9 Average daily intake or bew-carotcnc (pg) <-1514.5 151 25.0 154 30.6 1 1514.6-2237.7 I 50 25.0 121 '-4.0 0.7 11.5-IA 2237.8-2849.3 1511 25.f/ I I Il 21.8 0.7 (1.5.-1.1/ > 2849.3 150 2511 119 J16 /l.7 (1.5-I .11 Extent to u'hieh kitchen Illled uith smoke None 280 46.6 177 35.1 1 Slight 260 43.3 241 47.8 1.5 I.I-1.9 Considerable 61 10.1 86 17.1 2.3 1.6 3.3 " Odds ratios (OR) adjusted for agc. " Cl:confidenccinlervul. ` This cnriuble representk employmenl in the longest job that subjeds hxd in nny or Ihe f011ou'ing n-prinri "highcisk" ocaupatiom: ha( and shocmnkcrs: timbcr proccssine and wood. bamboo. hcmp, rnttun. palni and straw ProducU makcrc: papcr mxnufncture and paper product makers: stone sharpeners: welder and Ounu euucrs: and painters. t

616 30. Coultns DB. Samct JM. McCarthy. 1F- er nL I19901 A Ixrsonal monitoring study tn ns.sess workplace expr»urt lu environmansd tobacco smoke. Arn .1 Pn61ir Fleadtlr 80. 9Sti 9911. il. Muram:nsumm M. Umcmura S.Ok, dn T. rv ul. (1984) Cqilmning uf Personal ecposruc to tobacco smoke willl a nndy deccloP.d nicolinc pcrsnnal nronitor. F•rnrirnn Rr.c l5: 2 I%??7. 72. Tomxlis L. ed. (199(1) C'mnrr: Cruc.r.v, Urrurrrnrr miJ Cururvl. Vol. I(10. Lyon' Inlernutinnal Ageucy I'or Restnch on Cmtcer. 31 SiemiaEycki J (1991) Ri.rk Purtnn Inr ('nurr.r in r4r II'rrrAplrrrr. Boca Raton. FL: CRC Press. 34, Brackbill R, Fratier T. Shilling S(1953) Smnk ing char:mleristia ol' US workers. 1978-1980. Aru J!nJ dfrA 13: i41. 35. Stcllman SD. BolTctta P, Barhnkel L(1958) Smoking habils of 800,000 American men and women in rclnlion tn their rkcupn- tions. .4rn J bid Ffed 13: 43-58. ?6. C+rtdcna. WM. Thun MJ. Auslin H. rr o/. (1997) Enoironmenral tobacco smoke and lung nmc¢r murtality in the .4merican Cancer Societys cancer prcvention study II. C'~rmrr ('mr+r.t Cauru18: 57-64. L. Zholeg et al. 37. Ilrownttm RC. Alav:mj;l MC. Hock ET. Loy TS (1992) Pnssire smoking and Itmg cancer in nnnsmokine. rmnen. Am J PnMir Ilr•rrlr/r 82: 15'-5-15J0. 73. Co)kkmd KT, (7nxkmcar H. McMichacl AT. Ilolhrook RH (1977) Bias duc to miula>silienlion in the estimnlion of rclalive risk. dur J F.piArmiul I(15: 488-495. 39. Brownson RC. Alavunja MCR. Hock El' (1993) Reliahility of Passive smoking cxlxxure Itislorics in a ensr <ontrol study of lung cancer. Int J F.'pirA•urinl22: 504--8118. asse.- 41/. Coullss DB. Pcake GT. Sanxl JM (1989) Questionnaire manl nr lil'etintc and mcenl csposurc lu environmenud loharcn smoke. Am J EliiJerrriul 130: 33g-347, 41. I4on G E, Burch UJ. H uwr G R, M iller A B(1988) The reliability ol' paviv¢ mtokine histories reporltti in a cnsccontrol slntly nf lung cancer. Am J F.pidruriid 127: ?67-?71. 4'-. US N:ainn;il Inslitutc I'or Oc<upalinnal Sal'ety and Health (1991) L7nrirnnruemnl Tahrrrrn ,S'rnnkt in the IRuAp/urr: Lung ('mvr•r uuJ Ullrrr llerddi Ejlrrtc Cin:innnli. OH: NIOSH Publicztion nn. DHHS (NIOSH)91J(44.

612 Trrh(rJ. Associations between expo.cure to environntented tobacco smoke at home in childhood and Ihe risk of lung cancer among nonsmoking wommn in Shanchai. China, 1992-94, cxcludinc subjects with exposures in adulthood Number of years of cxpn>urc Controts (n=178) Cases OR'' 95%CI" AII lung cancers (n = 129) None 114 85 1 1-19 33 ?O 0.9 0.5-1.8 2(l23 31 24 0.9 0.5-1.9 p-V;due for linear trend 0.84 Adenocarcinoma (n = 76) None 114 51 I 1-19 ii 12 1,0 0.4-2.3 20-23 31 13 0.7 0.3-1 .7 p-Vnlue for linear trend 0.80 Nonudenocarcinoma (n = 2(1) None 114 9 I 1-19 33 5 24 U.6-9.o 20-23 31 6 3.3 U 9-12.0 p-Value fnr linear trend 0.06 Unknown cell type (n = 33) None 114 25 1 1-19 33 3 11.4 0 1 - - 1.7 20-23 31 5 1.2 n.3 -0.0 p-Value for linear trend (1.80 " Odds ratios (OR) ucre aQiusted for age. intake of vitamin C. respondent status, smokhuss of the kitchen during cooking. family history or lung canoer, and potentially high-risk occupations. " Cl:conhdenceintervat. L. Zhnng et al. were carried out for 20.2% of case subjects and only 2.3"/0 of control subjects, we excluded all proxy respon- dents and found that the results were essentially identical to those presented above (data not shown). Discussion Our findings of an association between the risk of lung cancer and occupational exposure to ETS are consistent with several other investigations [I5, 23-28]. A number of case-control studies among nonsmoking women have reported positive associations: two American studies found for ever exposed to ETS odds ratios of 1.4 (95"/u Cl: 1.1-1.7) [23] and 1.3 (95"/o CI: 0.7-7.9) [24]: a study in Japan found a 20% excess risk (OR = L2; 95% Cl: 0.7-2.0) for working with at least one co-worker who smoked [25]; an investigation in Greece found an OR of 1.4 (95% Cl: 0.8-2.5) [15]; and two recent European studies reported OR of 1.2 (95"/a Cl: 0.9-1.5) (26] and 1.6 (95% Cl: 0.9-2.9) [27]. A summary estimate of the OR for ever exposed to ETS in the workplace is 1.4 (95% Cl: 1.2-1.7) [28J. We found a strongcr association of lung cancer with occupational exposure to ETS than with domestic exposure. One possible explanation is that women were exposed more intensely at work than at home. In our study 69% of subjects exposed to ETS at work stated that they were exposed at least 3 hours per work-shift (typically 8 hours) and 67% of these subjects worked TnMr 4. Associations between expn<urc to environmental tobacco smoke at home in adulthood and the risk ol' lung cancer among nonsmoking women in Sh:mghai. China. 1992-94. excluding subjects who only had exposures at work or in childhood Exposure index Controls (n=410) Histological type of lung cancer All iypes (n - 3241 Adenocarcinoma (n n 184) Nonadenocarcinonta (n = 52) Unknown cell type (n = 88) No. OR" . 95% CI" No. OR' 95% Cl" No. OR" 95% CI" No. OR" 95% CI" Number of years hving wit None h a husband 114 who s 85 moked I 51 I 9 I 25 1 1-20 86 81 1.1 0.7-1.8 47 0.9 0.5~-1.6 17 '-.5 1.0-6.8 I8 LI 0.5-2.3 21-35 102 74 1.0 0.6-1.6 45 LI U.6-1.8 1'- 1.5 0.5-3.9 17 0.9 0.4~-1.8 > 35 IOS 83 1.1 0.7-1.8 41 0.9 (1.5-1.5 14 I!! U.7-5.1 28 1.2 0b-2.5 p-Value for linear trend 0.69 0.78 0.30 0.66 Number orcigarettes per d None ay smokrd b 114 y subje 85 cts' hus I bands 51 1 9 I 25 I I-10 90 89 1.4 0.9-7.2 50 1.2 0.7-2.0 I S 2.3 0.9-6.1 23 1.3 0.7- 2 .7 11-20 174 123 0.9 0b1.4 64 1.4 n.8-2.2 24 1.7 0.7-4.1 35 IA 9 0.5 1.9 > 20 32 28 1.4 0.7-2.6 19 1.7 0.8-3.5 4 2.4 0.6r9.3 5 Q7 01-2.5 r-Vnlue for linear trend (1.89 0.88 0 24 0.66 " Odds ratios (OR) were adjusted for age. income, intxke or vitamin C. respondent staus, smokincs or the kitchen during eonking. family history or lung cancer. and potentially high-risk occupations. " Cl:confidenee interval.

T,Me 5. Asvtxintions between the risk of lunc cancer and different indices of exposure to cnviromnental tobacco smoke at work mnong noasmnking women in Sh:mehni. China. 1992-94' htdices Controls (n = ( 1) All types (n = 504) Adenocarcinomu (n = 296) Nonadenocarcinoma (n - 91) Unknown cell type (n = 117) No. OR" 95% CI` No. OR" 95 ; CI` No, OR" 95% Cr No. OR" 95 % CI` Exposed ut work No 474 368 1 209 1 63 1 96 1 Yc, 127 136 1.7 13 2.3 J7 1.8 1.3-2.6 28 1.7 I.0-?.9 21 1.3 0.7-2.3 Numbcr of hours per da,v exposed 0 474 363 1 209 1 63 1 96 1 12 4$ 30 I.(/ 0.(r1.7 21 1._' 0.7-1.' 7 l 2 l 3-4 49 43 1.6 IL.O-2.5 31) 1.7 1.0-2.9 S 1 1.0 O.i.?0 IU 1 1.1 0.5-2.1 >4 70 61 2.9 1.8--4.7 36 2.8 Ib-7.8 16 3,4 L6-7.1 9 2 1 9-5 0 0 p-Valucforliixartrcnd <0.(F)I ~(L(RII ().01 . 0.09 . . Numlxr of yenrs exposed U 474 i6g 1 20) 1 63 1 96 1 1-1'_ 35 43 ' 2.0 1.'-3.3 26 t.g 111...3.' 12 2.9 1.3.-6.7 5 1.1 0.4-3.1 13-'-4 49 48 14 0.9-2.3 28 1.4 0.5'_.4 10 1.3 0,6-3A IU 1.7 0.8-3,7 > 24 4i 45 1.8 1.1-2.8 t) 2.3 I 1-39 6 1.1 0.4-2.9 6 1 1 4-2 7 0 n-Valuc for lincar trend 0.50 U.17 0.25 . 0.34 . . Number of co-workcrs who smokcd 0 474 163 1 209 1 63 1 96 1 1 2 56 37 1.0 0.6 1.6 ?6 1.2 0.7 ? 0 5 0.6 0? 1.7 6 I.(1 0.4-2A }-4 41 42 1.7 I.I-'.8 29 '.2 1,3-};1 8 1.9 US-4.4 5 0.7 0.3-1.9 >4 30 57 3.0 I.S-4.9 32 2.5 14-0.i 15 3.7 1.7-M.0 10 3.7 1 6-R 7 p-Value 1'or linc;tr trend <0.001 <O.W I .<0.OUt 0.06 . . ' Includes all subjects. regardless of other souRes of cxpo.urc to cnvironmcnt;:l tobacco smoke. ~' Odds r,rtios (OR I w'en udiusted for age. income. intake ol'viCUnin C. res{xxalent st:nus,.ntukin<s<of thc kitchcn during nxrkine. I:rmily hictory uf lung c:mrtr-Mrtenti:Jly high-risk rxculcuions, and donxsGC ealxxure to cnvironment:d tnbaecu smok<. ` Cl: conlidence intcrvnl. .1 L5E6 50ZZS

614 with more than two smokers. In contrast, only 5°L. of subjects who were exposed to ETS at home had lived with more than two smoking household membcrs. In an American study [29]. it was found that 47% of non- smoking women who were married to smokers reported no exposure to ETS at home, suggesting that light smokers may not smoke in the home at all. It is unclear, however, whether these results arc applicable to our study population. In addition. exposure to ETS in workplaces without effective smoking policies ntight he quile high [30, 31). In an American survey [30], exposure to tobacco smoke at work was reported by 13 of 15 nonsmoking volunteers. Eleven of the 13 subjecls encountered an average of nine smokers per day, and the average number of hours of exposure was 3.4. Using personal air monitoring, Japanese investigators [31] detected a higher concentration of airborne nicotine in the workplace than in the home. As there are numerous accepted carcinogens in the occupational environment [32, 33) and smoking behav- ior may vary by occupation [34. 35), it is possible that certain occupational exposures may confound the asso- ciation between exposure to ETS and the risk of lung cancer, or they may alter the risk of lung cancer by level of exposure to ETS. We found that the variables we used to represent occupation had little impact on our estimates of risk. This is consistent with the results Of two other studies in which confounding by occupalion was evaluated, and minimal efiects were found [36, 37). However, as we did not evaluate exposure to specific occupational agents, there is still the possibility that these estimates may be biased because of residual confou nd i ng. We carried out the analyses by histological type because there is abundant evidence that active smoking is much more strongly associated with the risk of squamous cell and small cell cancer than with the risk of adenocarcinoma and large cell carcinomas [3]. The analyses by histological type largely corroborated this expectation as the OR for nonadenocarcinomas were generally higher than for adenocarcinomas, although these estimates were based on small numbers. In interpreting the histological-specific results Of this study, it must be recalled that a large proportion (23"/,) of lung cancer cases were diagnosed solely on clinical mid!or radiological evidence and that we were unable to carry out an independent review of histological diagnoses. It is thus likely that sonie secondary tumors were included and that histological types may have been misclassificd. We have no data to indicate the extent of misclassifi- cation, but this should be independent of exposure to ETS because, unlike lung cancer in active smokers, exposure to ETS is rarely documented at diagnosis. L. Zlrnng ct al. Including secondary lung cuncers in 'the case series would attenuate the association between lung cancer anti exposure to ETS [3g]. The miscla sslllcatlon of nonadenoctrcinomas as adenocarcinomas would result in an overestimate of the association between adeno- carcinomas and exposure to ETS, while the misclassi- fication of adenocarcinoma as nonadenocarcinona would resttlt in an underestimated association between nonadcnocarcinoma and exposure to ETS. if the true association between exposure to ETS and nonadeno- carcinomars was slronger than that between ETS and adenocarcinoma. In the present study. exposure to ETS was assessed with a structurcd questionnaire, which is the only feasible method at present for assessing lifetime expo- sure. We were unable to test the criterion validity and reliability of the questionnaire used in the present study. However, sonie other studies have indicated that ques tionnaires are reliable instruments in evaluating long- term exposure to ETS (39-411. We carried out intervicws with surrogate respondents for about one-tenth Of the subjects. Proxy responses might result in misclassification of exposures because Of lhe surrogate respondent's lack of detailed knowledge of the subject's exposure. As the proportion of surrogate respondents was greater amongcasesubjects than among control subjects, the overall misclassification probabili- ties may differ between cases and controls. In logistic regression models an indicator variable to represent respondent status was included in order to control for differences between case and control groups of the distributions of respondent status. In addition, we carried out analyses that were restricted to self-respondents. and the results were similar to those obtained when all subjects were included and when respondent status was controlled for, indicating that using data from surrogate respondents did not bias the results of the present study. The present investigation has several major strengths. These include the population-based design and very high response rates. Few deceased cases were lost and the low refusal rates minimized possible selection biases. In addition to measuring exposure to ETS from different sources through life, we had detailed information on most key potential confounding factors. including in- door air pollution from Chinese-style cooking [10. 111, although this last factor did not affect the results. In conclusion, the findings of this study support the hypothesis that long-term occupational exposure to ETS increases the risk of lung cancer in women who have never smoked. Our results ttre consistent with the conclusions Of lhe US Environmental Protection Agen- cy [5) stating that ETS is a lung carcinogen in humans, and with the US National Institute for Occupational

603 The incidence of lung cancer in women living in China is among the highest in the world (6, 7]. It does not appear, however, Ihat tobacco smoking is a major risk factor for lung cancer nmong Chinese women, as the population-atvibutable risk proportion in Chinese wonien is only about 25% (S). The main reason fur this relatively low popuhition-attribulable risk propor- tion is that few Chinese women smoke [9] and those who do smoke do not consunte ns much as others [9. 10). As tobacco smoking is highly prevalent in Chinese men [9), exposure to ETS may play an important role in the development of lung cancer in Chinese wonun who never smoked. Although there have been a number of studies in which tite association between lung cancer and exposure to ETS in Chinese women has been investigated [10-12J. we felt it necessary to conduct the present case-control study as the data from these studies have been difficult to interpret because of several methodological limitations. First. dietary habits were not accounted for, although several studies conducted in othcr settings have indicated that certain aspects or diet may confound the association between lung cancer and exposure to envirrnmientnl tobacco snioke (13-I5]. Second, exposure to indoor air polltnion from Chinese-style cooking was not controlled for. while :ome investiga- tions have shown it to be associated with an increased risk of lung cancer [10, II]. Third. the effect ol' occupational exposure to ETS was not assessed in any of the Chinese studies. Subjects and methods Selertion of rn.ce suhtects Eligible cases were women who were permanenl residents of Shanghai and who wcre dia,=nosed between I February 1992 and 31 January 1994 with an incident, primary lung carcinoma (ICD-9 162). To be eligihle, subjects had to he between 35 and 69 year> old. Case subjects were identified I7om the popuhrtion-bascd Shanghai Cancer Reeisuy. This registry hus essentially complete coverage of all eases of cancer oceurring in the urban Shanghai area [6). For each eligible case the hospital chart w'as examined in order to confirm the diagnosis. and the histological type of lung cancer wus obtained from p,ithnlogy reports or hospital discharge summaries. For cases whose pathological or cytological reports were not available, clinical evidence and radiological findings were used ;ts the basis of the diagnosis. L. Zhnqe et ;tl. ScleNion nJ ronlrt+l .rtrhjvt'I.c Control subjects were selected r,mdomly from the general population by frcquency-matching to the age distribution of incident femede lung cancer cases in 1987-59. The Shanghai Residential Registry, which includes all permanent adult residents of urban Shang- hui, was used as the sampling I-rante. We randomly selected subjects from the personal identification cards retained by the Registry. Each card contained the name of a resident and address- date of birth, gender, and occupation. Using the expected age distribution of the rascs, we randontly selected a set of starting positions in the card tile and then selected the first two cards of wonten whose ages fell within predetermined 5-year intervals. When a primary control subject (the first selection) could not be interviewed- we then contacted the alternute control subject. /n« •rrirn:. Subjects were interviewed either in a hospital, at home or at work. The interviewers werc retired nurses, and they were trained to adniinister a structured queslion- nairc that included questions on level ol-education and income, all occupations held for at least 2 years- rc;idcntial history, family history.of lung cancer, cook- in_ activities, dietary habits, tobacco smoking, alcohol use. and a detailed historp of exposure to ETS. Surrogate subjects (mostly next-of-kin) were interviewed when the patients were too ill. We excluded deceased subjects. All subjects were interviewed regardless of their smoking status, and only at the analysis stage did we exclude suhjects who were smokers. Although the study focused on nonsmoking women. it was decided to interview all subjects whether they smoked or not. We excluded from the analysis subjects who had smoked at least one cigarette every day for at least 6 months. This led to a slight alleration in the age distribtnion of case and control subjects. A.c.cr.ccnten( of expnsure in IiT.S' Expoaure to ETS in the home was assessed in childhood and adtdthood by asking the subject to enumerate all household members who smoked and, for each, die type ol tobacco product used, the average number of cigarettes smoked per day, and the number of years of smoking when the subject lived in the household. Exposure to ETS at work was assessed in jobs that lasted at least 2 years. For cach of thesejobs we inquired about the lotal number of years working with smokinc co-workers. the ntnnberof smokingco-workers, and the

Enrirnnmenrrrl mhnrca .anake rrrrrl hrng rancer 611 7irh7e 2. Associations between exposure Iu environnuntal tobacco smoke and the risk of lung eancer among lifetime nnnsmoking women in Shanghai. China. 1992-94 ETS exposurc circumstance Controls (n = 601) Histological type or lung eancer All Iypcs Adenocarcinoma Nonadcnocarcinoma Unknown cell type (n = 504) (n = 296) (n = 91) (n = 117) No. OR" 95 % Clb No. OR" 951% CI" No. OR" 95% Cl" No. OR" 95%a CI" None` 114 85 I - 5I 1 9 I 25 I In childhood" only 64 44 0.9 0.5-1.6 25 0.9 0.4-1.5 I I 2.4 0.9-6.4 8 0.8 0.11.9 In adulthood4 only 217 192 1.4 0.9-2.0 109 1.2 0.8-1.9 34 2.2 I.0-5.0 49 1.3 0.7-2.4 Athomeonly 162 132 1.2 0.8--1.7 70 0.9 U7-L5 24 2.0 0.9-4.8 38 1.3 0.7-24 At uork only 22 24 1.9 0.9-3.7 13 1.5 117-3,5 5 1.7 I.I-d3.1 6 24 0.8-7.0 At work and at hnme 33 36 1.9 I.I-3.5 26 2.2 I.1d.i 5 2.2 0.6-7.5 5 1.0 t1.3-3.1 At both ages 206 183 1.2 0.8 1.8 111 1.2 0.7-1.8 37 2.1 (1.9--t.8 35 1.0 0.5-1.8 At hnme only 134 107 1.0 0.7-L6 63 1.0 0.6-1.6 19 1.7 0.7-4.2 25 0.9 0.4-1.7 At work only 24 29 IJ 0.9-1,4 20 1.7 0.8-3.5 7 3.1 1.0-10.6 2 0.8 0.2-3.6 At work and at home 48 47 1.6 0.9-2.7 28 1.5 0.8 -2.7 I I 2;t L0-7.9 8 1.4 0.5-3.6 Husband smoked All subjects` 377 322 1.1 0.8-1.5 187 LI 0.8-1.5 59 1.1 0.7-1.8 76 1.2 0.7-1.8 Selected subjects' 116 89 1.1 0.7-1.7 48 0.9 0.5-1.6 17 „ 0.8-5.9 24 1.1 0.6-2.4 ' Odds ratios (OR) were adjusted for age• income, intake of vitamin C. respondent stenus, smokineas of the kitchen during cooking, family history of lung cancer. and potentially high-risk occupations. " Cl: confidence interval. `- The reference category. " The cuto0 between childhood and adulthood was 23 years of age. ` Compared with subjects whosc husbands had not smoked. r Including muried women only: subjects with exposure to ETS in childhood or at work were excluded. When subjects were exposed to ETS as adults, risks were elevated for all types of lung cancer combined (OR = 1.4; 95% Cl: 0.9-2.0) and for cases diagnosed with a nonadenocarcinoma (OR = 2.2; 95% Cl: 1.0- 5.0). The main contribution to these excess risks was fron7 occupational exposure to ETS: the odds ratio for all types of tumours was 1.9 (95% Cl: 0.9-3.7) and OR greater than two were found for nonadenocarcinomas and lung cancer with unknown cell types. The patterns of risk among subjects exposed both in childhood and adulthood were generally similar to those who were exposed only as adults. Exposure to ETS from spouses has been the most commonly reported measure of exposure in the literature [3-5J. We found slight increas- es in the risk of lung cancers for exposure to husbands' smoking (Table 2). Table 3 shows the analyses by duration ol' exposure to ETS during childhood. No associations were found except for the nonadenocarcinomas, in which the OR increased with increasing numbers of years of exposure. We also found no evidence of associations by number of years living with a husband who smoked or by the average number of cigarettes smoked per day (Table 4). Table 5 shows adjusted ORs for different indices of occupational exposure to ETS. As there were few subjects whose sole source of exposure was in the workplace, exposure-response relationships were assessed for all subjects. Across all subtypes of lung cancer, risks increased by increasing average numbers of hours per day of exposure to ETS and by average numbers of co-workers who smoked. On the other hand, increasing trends were not found for number of years of exposure. We carried out a number of sensitivity analyses to determine the robustness of our results. We evaluated the potential confounding effect of occupation by using duration of employment in each of the high-risk occupations and found similar results to those analyses that included simple binary variables representing ever employed in these high-risk occupations. We also evaluated statistical interactions between exposure to ETS and employment in high-risk occupations. cooking- related indoor air pollution, and consumption of vita- min C. After inspection of risk coefficients, likelihood ratio statislics. and standard errors, we concluded that there were no meaningful statistical interactions in the data. Lastly, as interviews with surrogate respondents

0m.rr, (.uev:,anJtSpruu/ II/.UU/bI4.1YY4. '<.~' 1999 A7ia~rr AruJrvrrir PuAti.d.rrs. PrinrrAin rhr .\'mhrr/rrnd.. A case-control study of lung cancer and environmental tobacco smoke among nonsmoking women living in Shanghai, China 607 Lijie Zhongl ", Mark S. Goldberg" ', Yu-Tang Gao2 & Fan Jin2 rJoint Department.r qf Epidemiology. Bio.stari•stics ruul Occupational Herrlth, McGill Unirersiqr, Montreal. Canada: '-Depurtnrent qJ EPirleminlagr. Shanghai Carc•er lnstiune, Shanghai. People's Republic njChinrr: lEpidemiology anrl Bin.craristic2• Uuir• Research Ct~nue on Humau Herdrh, /NRS-brs7itru Arvnernrl-Frappier. Unirer.sitr ojQuebec, 53/. Roulevard c%s Prairies, Laral, QC, Canada H7V 187, Tel: (4-i0) 687-5010: e.ri 4338: Fax: (450) 686-5599; Email: nrark('ri•pnluir.epi.ruc'gilLc'a I"Author (or cnrre.s7)oudenrel Receieed 26 O<lobcr 1998: nettpltd in reviud form 21 June 1994 Kc.r Irords: case-control study, environmental tobacco smoke, epidemiology, lung neoplasms, passive smoking. Abstract . Gurorlucrion: The incidence of lung cancer in women living in China is:unong the highest in the world but it does not appear that tobacco smoking is a major risk factor for lung cancer. As tobacco smoking is highly prevalent in Chinese men, exposure to environmental tobacco smoke (ETS) may play an important role in the development of lung cancer in Chinese women who never smoked. We conducted the present investigation because previous studies did not account for dietary habits or indoor air pollution from Chinese-stylc cooking and they did not assess the ellect of occupational exposure to ETS. afc•rhods: A population-based, case-control study was conducted to evaluate the relationship between lung cancer and exposure to ETS among nonsmoking women living in Shanghai. China. Five-hundred and four women diagnosed with incident, primarylung cancer between February 1992 and January 1994 were identified through the population-based Shanghai Cancer Registry. A control group or 601 nonsmoking women was selected randomly from the Shanghai Residential Registry, and was approximately frequency-matched to the age distribution of the lung cancer cases. Information on lifetime domestic and occupational exposure to ETS was obtained through face- to-face interviews. Adjusted odds ratios (OR) and 95% confidence intervals (CI) were estimated by unconditional logistic regression. Results: The OR for ever exposed to ETS from spouses was 1.1 (95% Cl: and the OR for ever exposed to ETS at work was 1.7 (95"/o CI: 1.3-2.3). Furthermore, the OR increased with increasing number of hours of daily exposure to ETS in the workplace and with increasing number of smoking co-workers. No associations were found for exposure to ETS during childhood. Cnnclusinns: The main findings of the present study are that long-lerm occupational exposure to ETS, both alone or in combination with exposures at home, conferred an increased risk of lung cancer among women who never smoked. The inconsistency of the results regarding exposure to ETS at home and at work may have been due to lower exposures at home, Introduction It is widely accepted Uiat active smoking is one of the principal causes of lung cancer [1-3]. Over the past two decades there has been considerable intcrest in thc potential health eFl'ects of environmcntal tobacco smoke (ETS), and over 37 epidemiological studies have been conducted to investigate the relationship with lung cancer in lifetime nonsmokers (3-5J. It is plausible that ETS causes lung cancer, as the same carcinogens found in mainstream smoke are also present in ETS. While there is a wide range of variability in the estimates of risk from the epidemi- ological studles, collectively there appears to be an excess risk of about 30% for nonsmokers exposed to ETS (4. 5J.

Em•irruuorr/lul Irrhurc'n suroke and Gmg cwlcc'r Safely and Health study indicating that ETS is a potcntial occupational carcinogen [42). Although the individual excess risk of Iung cancer from exposure to spousal ETS is small, this exposure may have u significant health impact on Chinese women because of the gre:u number of male smokers and the ubiquity of ETS in China. Acknowledgements This research was supported by the National Natural Science Foundation of China. The authors thank Dr James Hanley, Dr Maric-Elise ParenL and Dr Jean- FranFois Boivin for their thoughtful comments and suggestions. Dr Goldberg gratefully acknowledges re- ceipt of a National Health Scholar award front the National Health and Research Development Program of Health Canada and support from the Fonds de Ia recherche en sante du Quebec. The work was conducted in the Department of Epidemiology, Shanghai Cancer Institute. Shetnghai, People's Republic ol' China. References I. US Depurnncnt of Health. Educntiml, and Wclfare (1964) StuuA- nrg and Br•u6hr Rrpnrr nf rLr Arlri.cnnC'anrnilrrr rn rhr .Stnyrnn Gt•nrrul n,f rlrr PuhlirHetdrh Sr'rcifr. Rock,'illc. MD: US Dcparl- ment of Health, Education, and Wcll:rrc. Public Health Scrvicc. Center for Disease Control, PHS Publication noa 1101 ?. US Dcpartmenl of Health and fiuman Services (1982) Thc Hrrrlrlr C'rntcrquouct'r of SrurrJ'iqqr Crnrnv. A Repnr'r uf dk• .S'urRran 6rrural. Rtxkville. MD' US Del*urtntcnl of Health and Humtm Serviccs. Public Hc:Jlh Service, Office on Smoking and Health. DHHS PuMicnlion no. (PHS) 82-50179. 3. Internntion;d Agency for Research on Cancer (1986) l.4RC- Munqervph> un Ihc 6cnluurinn uf Cnrciungrrrir Riak nf C'Irrrnicul.r rullrnnurrr: Tuharrru$utnkirr,e IVo7unrr 1Kt Lyon: World Health Organiz;uinn. hHCrnational Agency for Rev;nch oil Cancer. 4. National Re.e:uch Council (1986) F.nriroHUrrarrr! Trrhrn rn .Sinukrc .Hrn:raing E'rpruores and .as.cnr.ring Jiraldr F/kxv. Wcrshinplon. DC: Ntuional Acadcnly Press. 5. US F.ncirrmntennd Proteetion Agency (199]) Recpirrrnnr Hoalrh EIJr'rts u! Pnccirr .Smnkiu-,, l.Gnrg ('rnrrer wu! OrhrvDicurdc'rr, Ollicc nl' Rcuxrih and Developmenl. Ollicc ol' tlcalth and F.n.ironnrcnlal Assessnrcm. (EPA fi0(1 fi`HI IIOBF) 1Vashingtnn. DC': US EPA. Ollice nf Research und Decelopment RD-689. Deccmhe.r. 6. International Agency for Research on Cancer (1986) In: Muir C. W;nerhouse 1, Muck T, er ul.- cds. ('nnerr !nrlrlrnre in F'ire ('rurrinrwrr. IARC Scicnlilic Publication no. 38. Vol. V. Lyon: Internntiun;d Agency for Research on C'nncer. 7. Jin F. Dorsa SS. Zhcn_ W. Blol WJ. Fraumcni 1F Jr. (iao 1'T (1993) CCancer incidence Irends in urhun Shanghai. 1972 19ti9. 1N 1 Cnrra•r' 51: 764 77U. 615 8. Gao 1"G Blot WJ. Zhcng \V, Fraum¢ni 1F Jr. H:u CW (19SF) Lung cancer and smoking in Shnnghai. hrr J lipirlrnrin! 17: 277 2xn. 9. Dcng J. Gan YT (195i1 Prevalcnce of smoking nmong I It1.IXNt atlull residents in Sh;rnghni urhm :rrc2 ('Liursr .l I'rrr •IhJ t9:2n 274. In. Wu-Williams AH. Dai aU, IBot N'J. rraf. (199n) Lung canca ammng womtn in north-cast China. Rr J Cimrrr 62: 952-9\7. 1 I. Ko YC. Lee CH. Chen MJ. er n! (1997) Risk faclors for primar, lung rmccr:unong nnn>mnking teonxn in Taitcan. /rrr J Epidrurinl 26: 24 .31. 12, Knn LC. Ho 1H. Snc D, Ho CY (1987) Measuremmrts of passicc smoking and cstim:ues ol' lung cnncer risk ammog non-smoking C'hine.v fcmales. !u!.! Cmrrrr 39: 162-169. 13. L.c Marchand L. Wikens LI2. Hankin JHL er rA. (1991) Diaarc pallern< of 1'cnralc nonimokcri reilh and nilhmn erposure to environnuntal tobacco smnke. Cnmrr ('nme.< ('unrra! 2: I I-16. 14. Sidncy S. Qaln BJ. Fricdman GD (1989) Dietary intake nl" carotcne in non.mnkers with ;md nithom pazsive sntokin_ al hontc- :Inr J Epidrwinl 129: 1,103 13119. li. Knlandidi A. Katsouynoni K. Vorolioulou N•rr ul. (1991)) Pussite smnking and dict in thc clinlo_y nf lung canc•er ;m)ong nonsnrok- cr<. C'uutpr C'nurrt Cunlrul I: 15 -'I. 16. Qu YH. Xu GH. liuang F(19X6) Ames test in Ihe emissions of hc:ncd edible oil. ('nmrr (In Chincsc) 6: >g-6(I. 17. Chinese Academy of :bkdiad Sciences (19N1) Fr.nd ('umpatlrian 7,06 tln Chinese). Beijing: Pcoplc's Health Publishing Co. IX China Report (1985) 10 perotinl zmipling tabulation on 19s' population cen.w of the Peopl:s Republic of China. Foreign Bru;nlctsl Infonnation Serrices. 19, Levin LI. Zheng W. Blot Wl, rr nL (1938) Oeeupalion and hrng cansrr in Shanghai: a cux-comrol slud)•. Br J!nd 11rd 45: 451r 45x. 20. Levin LI. Gao Y.T. Blot WJ. Zhcrm W. Fr.unncni 1F Jr (19871 Decreased risk of lung cancer in Lh< cnltnn IcVilc industry of Shanghai. C'unrer Rec 47: 5777-57$I. ?I. Breslae NE. Day NE (19C0) Sxnlslicn! blrrhmLc in (.'nrrmr Rrcrnrr4: Thr AurJrsic u/ ('u.ce-Cunaul .Audirs. IARC Scicntilic Publications nn. 32. Lcon: International Agency for Rcscarch on Cancer. pp. 19??46. ". Hastic TJ. Trhshirnni Rl (1993) fieu~ ~ub:ed :Iddirirr• S1w1cLr. Nen York: Chapman & H:rll. 21. Fonthanr ETH. Corrca P, Reynolds P, er rr/.I I994) Environmental tobacco smoke and lung cancer in nonsnwking nomen: x nwllicentcr studv. J.1ll.{ 271: 171?-I719. ]4. Wu AH. Henderson BE. Pike MC. Yu MC 119M5) Smoking nnd othcr risk factors ror lung cancer in ceomen. .1 Srrrl Cwrrr•r hnr 74. 747751. ?5. Shinritu H. Mori>hila M, Mizuno K. rv nl. (199R) A cax-control study' of lung cancer in nonsmoking wontcn. TuL(#Au J E'rp rllyd IS-1: 3R9-197. _'6, lioll'cna P, Agudo A, .Ahrcn. 1F- rr ol. (1998) Mulliccntrr c:rsc contr.rl slud) of caposure to cmirnnmmntal tohaccv smokc anJ lung cmccr in Eurolu. l A'w! C'runrv Inrr 9(I: 144/1-14N). ]7. N;'bcrg F. Agrcnius \'. Svartengren K. Srcnsson C. Pershcrgen G (1998) Emironmcnl:d tobacco smoke and lung carocer in non- smokers: Doci lime since ceptvure play a rnle' F.PiJrvnioloAr 9: Jnl?03. 28. Wells AJ (1999) Lung canccr from srnuking al .rnrk. dm.l l'rrhllt !lrahlr 88: IIpS IIQ9. 29. Frictlman GD. Pclilli DB. Batcol ItD (198i) Prerdcncc and correltllts of pn: ive smokingi. :lor J Pu61kHer.Nh 77: 401 4115