Tobacco Documents Online

Ihlernational Journal of Epidemiology C International Epidemiological Association 19911 Vol. 20. No. 1 i - Printed in Great Britaim Smolking and Other Risk Factors for Lung Cancer in Xuanweli, China ZHIYUAN LIU't,,XdNGZHOU HE"'AI1JD ROBERT S CHAPMAN'•' Liu Z'(Institute of'Environmental Health andlEngineering, Chinese Academy of PreventivelMedicine, 29 Nan WeiiRoad, Beijing 10050, China), He'X andiChapman R S. Smoking and other risk factors for lung cancerin Xuanwei, China. /nrer- national,Joumal of Epidemiology 19911 t 20: 26-31. In Xuanwei County; Yunnan Province, lung cancer mortality rates are among,the highesCin China in both males and' females. Previous'studies have shown a strong association of lung cancea,mortality with indoor air pollution from 'smoky' coal combustion. In the present case-controt study, 110 newly-diagnosed lung cancer patients and'426 con- trots were matched with respect to age, sex;pccupation (all subjects were farmers), and village of residence (which provided matching'with respect to fuel usel. This design allowed assessment of'known and suspected lung cancer risk factors otherthanithose mentioned above:Data fromimales and ifemales were analysed'by conditional logistic regres- sion. In females who do not'smoke; the'presence of lung cancer was statistically significantly associated with chronic bronchitis (odds ratio (ORI = 7.37, 95% eonfidence,interval ICI): 2:40-22.66) and familyhistory,of lung cancer (OR'4.1'8; 95%' CI: 1.61-10:851. Females' results' also'suggested an association of lung Icancer with duration of'cook'ing Ifood (IDR 1:00.9.1 B and b4.70), but norwith passive'smok'ing (OR 0.77; 95% Cl: 0.30-1.96).Jn mates, lung cancerwas significantly associated with ehronic:bronchitis (OR 7:32; 95% CIf 2:66'-20:18);,famity histony oflung cancer (OR 3.79, 95% Cl: 1L70- 8:42); and personal history of cooking food IOR 3136; 95% CI`. 1127-8:881:Jn males aidosc,res'ponse relationshipof'lung eancerwi'thsmok'ing index (years of smoking'amounPof smoking) was shown by risks of 1L00;,2'61, 2.17 and'A:70. (•:xanlinati(on (nl ('hinu\c natitnnwitlr iancvr tn(ml:lfilv sGati!stics rcvcal. that lung cancer nhntalihv rates inl Xuanwei C'Itunly^. Yunnanl I°novincc; :u'c':Inltnng tllc hiLhctit in Chin:t.' From 19731hruu)at IK)75-:mnu,llizcd' m:tlU lung 'c:utccr (Id•.:uh ratc.,,agc-adjitstcd tlo tlic Il)(t-l China pco{lulatiun- were 27.7 and 6:2;; per 11N)(NDU) in Xuanw'ci and China( rc%)xcDivcliv: Corresponding morltalities in females were 25.3 '(tnd 3.2 per I(Ifll(ANI, Marked geographical variation in,hmg,canccr mcrr- tctlitv cxists within Xaumnvei.'fhc ccxlntv'c:tn bc din•ided into,high-, mcdium- and low•murttditv arcas- in which age-adjustcu( Ilmg canccr m(trtalitiex in'hbtth sexes arC 126.11, 2(f.') und fi.(I Incr 1tN)Q1fN), rcv(tcctivclv: Xua'nwci,resiilcnts have tratditiromtllv burncdithrr.u tv(res (tf' fuel. 'snltokv' coal. 'smrokrlbsi coal. :mc(, woud, for re>i(Ii:ntiallhcating and ('uctkinL. l!ntil tlhc' IF)tilk fwcllw':I, nc:ulv_ .IfwtrVs hurncdiin;ln (r(tcnt (unvvn- tiliurdlhrc pit inrthc flinnrcnf thc d<vcllinc'.: nlnin r<Mtnl'; 'litr!nulc. nf'I~.ncirnrnnr.rnuJ IA1slt[t. :md'. I n¢nrccrfnC.. Chincx'Ac:nlcmvof 1'rrccntlcr M,.diuin,-. ?4.17:rnA\',-i It-ad11/MN7:o/, blunn .. t i};1.nviruttmtcnl:J I'nrl"cliun ihLi k..a..ui 6'Iirr:rngtr.P:r,r t,'.!:( ~ 27711 1 tl!SA. ~I'rc.cnl:rJJrc... Illnrl.mwlags(iwu!r, Iti,drn•n.a.l .ind'\11'nLr:rl ..ni.r. Nair„.i.rl I.rll„"n. ltl,lt•. 'rlc b7rrl ti~rv~lh (~.i.,,A inuu_ ,\rg„nnr. II r+l-0la I~1.\. .~ such lint Inl>':Iri .lill t~i l,lv u.cr(, tht,u h tVtr usc (tl ventihltcd stc>vc. i> incrcauin):, 1ti'itntrn h:t\c kcncrallk hccn res(rtmsihlc lA,rstirrlung:nul tcneling IFto d(nntcstir firc andlc(ot,kihc (hnd. though nlen ai these res. (wnsihilitirs in sttntr (nmilles: Tobacco smoking is,vc,rc rare in Xu:mwciis'(tmcn., vtic wclmc,n's' 1(tn,; cancer rmtes aru comparable to mcn's. Also. a surccv aflp:tst fuel usc shonccd that'in tile high-; mcdium- and I( u'-m(nrt.!Iitn' :Ircac of' Xunn- wci„ the pcrccnl:rtie.s of families u.ing' smc,ky coal before 1958 tt'orc `7.(t": , 6O.I /6 an(11(,:1"% re.Tccr« ivclv: C(trrc.pulnuling percentages ofl families using wto(td were I.-7!Yb. Ir).°l'%;i and 67:1'io. lhilurnr ro:i•:cn- trallinns of brnz(n(it)pt'rcnw: ((bAP} avcr:tge(I 627'utJ1(l0rnctresI (Int )iintlltelliCh-nlunt:!lit~ arta.and -76ut;111/0 m' in tlhc lo\e-m(nrtalihv arti;l. In addi6iun inul<tor pc+lliltic+n s,Imldrs trr,nrtha hiLli-nxortalitn':orca exhibited highi•r Anlrs-tc.t' mutn_rnicitv th:ln tli(tsa. frutn thc 11,~~-nutrt'.Ilitv arcal'' Adl' u~f tlhr.rc ohcur- V:ISiutltA'h':!6C.SCt't4'(I!tt,sl!_llcs'l :In a]St,t'iatlttll hBt'w'cClll indtnur snu,kc c(,ull Iiurnin_ and lunc canccr, in ~d11111\1'CJ.. , ` Thc c:!sc-contrin; .tu(tv rcl?cnrtcitlhrrc was dcsignir(d tu >ultlnlcntr•nt ciisti.nL inliorntnliutn I,n' tln'>r.>;ing t9tc inllurn.r 01 furt ,!. trlhur tltanl(ilul lVlnr un tllu utit'tlr- ri•niv (nf Iun,_ (:Inaroin Xu:lnhcii. Suih f;liao,re, incluil-

RICK FACTORS FOR LUNGCANCER IN CHINA ing, tobacep smoking, family, history, of' lung cancer„ history of'chronic bronchitis;,andicooking habits, have: been associated with lung cancer in areas other than, Xuanwei; but their importance as contriburtors to lung cancer in Xuanwei has not yet been systematically' determined. The present study also allowed compari- son of the relative impact of these factors in males and females. MATERIALS'AND METHODS In Xuanwci„93.4% of the total populatimn1werc fttrm- ers in11982. Because of this, and because lung cancer mortality in.Xuanwci farmers is high~j'this study was confined tmfttrmers. Conccntrating the target popula- tion tended to incrcasc the validity of the finctings:' Between November 1985 and December 1986, we identified 112 cases of newly-diagnosed lung cancer at Xuanwci hospitals and clinics, After exclusion of two paticnts with unknown addresses, 110 lung cancer patients (56' males and 54 femalcs) i wcrc included in data analysis. Of thcse, l'9 ('17°h) had bccnidiagnosed on the basis of cytological/pathological findings„ and the remaindhron the basis of clinical histories and X-rays. Controls were selected to match lung caaiccr pat icnts with respect to age (;+2,years), sex; and village of resi- dence: Because fuel use habits and dwelling,types are similar within individual Xuanweiivillagcs„this design was expected to provide effective matehing,with rest pect to indooriuel type and dwelling type. Such matching was,sought because it would increase the effectiveness with which facturs other than fucl'. type could be assessed. Cases and controls were matched on village, with as many eligible controls included as possible. Therefore, we seiccicd more con- trols for each case in a largc village than in a small vil- lage. The numbers of controls pcrcase varicd ifrom onee to five. Aftcr, exclusionof 26 eontrols hecausc of crro- ncous questionnaire responses, 426 contrrrls werr included in data analysis, an average of 3.K2 controlss per case: There were 9 cases with I control, 15'cascs5 with 2'controls:, 15'with 3 controls, 13 with 4 contrcols, and 58lwibh 5'control ., A'standi+rdiied questionnaire of'.the closccl~question type was developed. Study factors inelU.dcdl tobacco use histrnrv; fcimily, and personal medical hi.%knryi rdrrnr- estic:fucl use history, indcxv fuel use history, personal history of cooking fiorrd! dwelling typc, cthnic group (nationality), and socioeconomic and educational levels. Af'ter strict interviewer training and field tcst~ ing, this questionnaire was administered'directly to all lunb,canccrpaticnts and controls. No interviewer or study suhjpct knew the purpose of the stucdy, and hypnthcses. 27' A summary index (if tohacco smoking was davcl-- oped for each subject. The smoking index,was calcu- lated by'multiplying',the duratitnnof'smoking'(in years) by the amount' of tobacco smokedi (in kilograms per month). A subject was considered to have a positive family'history oflung canccrif at least one relativrwas reported to have had the disease. The relatives included subjects' parents, siblings andlchildren, and parents' siblings. A subject was considered to have a positive history of chronic bronchitis if, he or she had been diagnosed by aidoctor to have this condition, or reported cough forat least three months per yearfoeat least two vears before the vear of interview. A female subject was considered to have:been exposed topassive smtoking if there was at leasuinne smoker (mninlv hus: band) who lived in the same household. To assess thc,effcets of individuallindtpendencvari- ablcs, unmatched, unadjustedludds ratios (iORs) were ealculated.`Crnnfidence intervals were caiculated'using Miettinen's menhod." Dose-response relationships wure examined for variables related to smoking and cooking. Trends within these relationships were tested by extcnsionof, the Mantcl-liacnszcl procedure.' To develop adjusted estimates of ORs associated with selected factors and interactirtns, conditlonall logistic regression models were also constructedi fbr ma9esand'.fcmales:" In these models„all variables werce dichotrnmtous„assumingvalues of 0'or 11. The selectedl risk ftrctors and interactions were trcated as indcpen- dent variables, andi the presence or absence of Iung, cancer was trcated as the dependent va,riahlc. These analyses were perforrned using the PECAN program.° "" RESULTS Distributions of characteristics in cases and controls are presented'by cencter, in Tahlc 1. Age,,family, size, cthnicgrowp, birthplace, cducationalllevcl, and dwell- ing type were comparahl'e in cases and controls, so these factors were'not considered further in dataianaly- sis: The effcct'of active tobacco smoking was not eval- uaterJ in females, since onlv one female (a control su0ject) ,hadlevcr smokcd tobacco. The village match- ing proviiledleffeccive matching on fuel tti•pe,because fuul-use hntnits (typa and avcragc mmount)'were similar in cases and controls. Crudd: and adjusted ORs forsmoking andlcooking habits are presented with;95'So confidencc intervals for malrs in Table 2. No relatimnship between lungsancer, and' ever haviiig s'mcnkcd was ohservcd.. T.herc was a suggestion of nurnottmic di>se-resptonse relationships of lung canccr'vith the age at which.smoking heg;tn, duratiinn of'smoking and amount smoked'hy month.

28 TAtuc I Factor INTERNATIONALJOURNAL OF EPIDEMIOLOGY Corrtpari.ton of luna cancer cans and'controls. Xuaewri: China, 1985-1986 Males Females Cases Controls' C1ses Controls A.varage'age (years) . 52 50 52 52 No. of people in',fatttity now 5;6 5.4 5.6' 5:4 No. of people in fatm'ly 2U'yean ago 5.8 5.5 5.9 5:5 Han inationahty (X.) 94.6 96.9 98.2' 97A 8otn im Xoan+rei (%) 100, too 100 98.0 TYet,storey, dwd)ing (%)'. 98.2 99.1 100. 100, Atttount of; tunoky' coat btvnr(tottslyear), 4.2 4.2 4.0 4.1 Amountof'.rood burnt (tonsfy'cu) I 0:8' 0:9 0.8 1L0 Taatt2 Oddsrarivt(OR)iand9.5%eonfidencrinrervalr(C!)forlrug cancer in malts aeaording;to tamoking and cooking. Xwnwri, Chieia,, 1985-1986 Factor Cases Controls ORe" ORa'• 959D CI Ever-smoker No 4' 19 1A0' 1100 Yes 52' 205' 1120 1.26 0,30- 5.26 Age of starting to smoke (years). Never 4 19 1.00 1.00 >20 20 80 1.19 1.10 0:25- 4:93 Q0 32 125 I 1.22 1.39 0.32- 6.06' Trettd'(p value) (p>0.05), Years of smoking 0' 4 19' 1.00 1.00 <35' 30 146 0:98' 1.07' 0:25- 4.59' a35' 22 59 1.77 1.71 0.36- 8.12 Trend (p value): (p>0:05) Amount of smoking (kg/permonth) Never 4'. 19: 1.00 1.00 - G0.5'. 25 93 1.28 1.41 033- 6! 09 0.6-1.0 20 93 1.02 1.09 0:211 4;82 >110 7' 19 1.75' 1.91: 0:32'-1t.40 Trend i(p value) (p>0.05), Smoking indext <2 4 30: 1.00 1.00 2- 24 99 1.82 2.61 0.69- 9.82' 20- 16 74 1.62 2.17 0.55- 8.6i'1 35- 12 21 4.28' 4.70. 1.03-21..40: Trend (p.value). (p<0:05)1 Often cooks4ood No 44 200 1..00. 1.00~ Yes 12 24 2.27' 3.36 1.2?- 8.88 However, none of these relatio(lships was statistically significant. In contrast, a Statistical'ly significant dose- response relationship of lung cancer with smoking index was observed. The adjusted OR in men who often cooked food (at least once a day) was 3.36 (95% CI: 1.27-8.88). The adjusted ORs were slightly, larger, than the crude ORs. Crude and adjusted ORs for cooking and passive smoking are presented for females in Table 3. No dose- response relationship of lung cancer with age at'whichi the womanlbegan+to'ooo'k food was'observed, but,the. OR'associatediwith the age at which the woman began to cook food (11-15 years old) was significant. Adjusted ORs associated with the duration of cooking weremuchilargcr than crude ORs. There was a sugges- tion of dose-response relationship of lung cancer with the'duration ofcooking food for the adjusted iORs. No relationship of'lung',cancer with passive smoking',was observed. Odds ratios for family'his2ory'of lung cancer and per- sonal history of chronic bronchitis were signiftcantdy associated with lung cancer in both~sexes but'duration of using an unventilated fire pit was not (Table 4). All conditional'; logistic regression ORs were larger than crude ORs in Table 4. DISCUSSION This study was intended'to supplement previous's'tud; ies which had shown a strong association of indoor smoky coal combustion with lung,cancer in Xuanwci County." Full understanding of llung,cancer, aetiology in Xuanwei, and comprehensive risk assessment of the effect of smoky'coal ittse„require'systematic'assessment T^aua 3 Oddh Ra6or (OR) land 95% confidence inee.rals (CI) for /weg cancer iri females according to cooking and passive smoking, Xuanwei, China, 1985-1986 Factor Cases Controls ORc•' ORa"' 95% CI I. Age oPstarting to cook > I5 13 73 1:00 1.00 I1-15' 30 69 2.44 2:37 1.09- 5.15 c10 11 60 1.03 1.25 0.45- 3.49 Trend (p value ) (p>0.05) Years af cool-ing .. 630 7 53'1.00 1.00 3.1r44'. 28 85, 2:49: !',.IBI 1.76- 47.49 ;D45' 19 64 2.25 14.70 t.61~-13X:03 Trend (p value): (p>0.05): , Passive smoking , No 9 26 1.00 1.00' Yes 45' 176 0.74, 0.77 030•- 1.96' 'ORc = Crude odds ratio. "ORa = Odds ratio after adjustment by, conditional logiitic'ORc=C>•udeoddsratio:, regression,fwothenriskfactors. "' ORa = Odds ratio after adjustment by conditional logistic Smoking index = Years of Smoking'Amount of smok',ing:., regression for othenrisk (uctors. I,!

' RICKFACrORS.FOR LUNG CANCER,IN C Ft1NA 29 Tuan.E4 Adjustcdtod&ralio.rlin'R/and195%conjdcnrriiunnalc(iCl) forlrrngranccrirtnralntand/rnialrs:,acrordingrofpmiliaAhistorvoflung. cancer, historyafthkonir hronahitis and years taingumrnfitated firr pir, ,Xuann•ci,.China; /NS5-19Wb i Males Case Control ORa1 (95% CI)' Case Control Fcmatcs Tota1 Otta (95% CI) 1 Case Control ORa (95% Ct) Familial historyy of lung ancer, No 41 200 45 192 86 ?92 YRs 15 24 3:791(1.7n- 8.42) 9 10 4.18 (1.61-1n:85) 24 34 3:75 (2.05- 6.83) History otchronio bronchitis No 39 209 38 184: 77 393 Yes 17 15 7,.32 (2.66-2tL1B) 16 18' 7.37 (2.40+22.66): 33 33 7'61'(3.62-16.00) Years using', unventilated fire.pit'<45 22 107 ' 21 84 43 19,1 a45 34 1'17 1.78(0.46- 6.93) 33 118' 0.73 (0.20- 2:6n), 67 235' 1.12(0:46- 2.74) 1 'ORa . Oddts',ratio after adjustmentby,eonditional tirgistic'sogression dor other risk factors. not only of fuel,us'e, but'of other known ~and suspected risk factors as well. To achieve such an assessment, we chose a study design which provided' effectivc case- control matching,with respect to.fuel type and average amount used as well as to age and gender. Xuanwci is a very rural area. lnithe present study. almostall lung',eancer patients were'diagnosed by the county, hospitals': Only 17% of the lung,cancer cases were'based on cytolbgical/pathological findings,,t33% were bas'edi on clinical historics and X-ray findings. Ttiereftore, misclassification of the cases may exist in the study. However, other reasons may improve valid~ ity of lung cancer diagnosis in the study. Because of poor mt:dical' care in Xuanwei, most lung cancer patientshadlreached an advanced st'agcofthe disease when diagnosed, and loca'l doctors hadlwidL caperi- ence of this diagnosis because of the high lung cancer morbidity in Xuanweii In order to assess validity of the diagnosis, we also followed up a numbcrof'litngcancer paticntsfrom the study;,altnostall'of whom dicdwi'thin' six months of diagposis: In both malcs and females; the'current studi7 dis- closed consistent' and statistically significant associ- ations of lung', cancer with chronic bronchitis and positive family hi'storyof lung cancec Lung cancenwas also associated with the frequency of cotoking food (in mall:s) and the durationlof cooking food (in fcmalcs)„ even after the matching,on,fuci type inherent in the study'design. Not surprisingly, lung cancer was associ- ated with active smoking in males. No associationwith passive smoking,was observed in females. In Xuanwei, women arc generally responsible for cooking food~ so the variable 'cooks or does not cook food' could not'be'asscssed in females. However, the O',Rs associated with the variable 'ycarb of coa>king' suggested that lung cancer risk increased with increase in duration of cooking,(Table3). There was' no dose- response relationship between lung cancer, and age at which a woman began to'cook,food. This observation may be due to ttie'poss'ibility that'women whoreported cooking food at less than ten:y,ears old did not really cook food at that age. The OR' in males who often cooked'food'wasoventhree timesgreaterfihan in those whodid'not. It is likely that those'who cook food inhale more coal-smoke pollution than those who do not. %Vu et a!'repmrted that subjects exposed to burning coal' uscd'l for heating or cotoking, in, a, stove or fireplace during themajority of childhood and;the teenage years had a lung cancer risk 2.3'times higher than subjects who were not so exposed." Wanger al have',reported' similar results from China.'' Gao et a! reportedl increased risk of lung cancer in Shanghai women who cook frequently with rapeseed oil." This observationn raises the' question of whether, pollutants associated not onlywith thc,cooking,fucl but also with the cooking method promote lung cancer:. The observed association of lung cancer with posi- tive family history of the illness may be attributable in part to recall bias. However, our results are consistent with other studies"'j° which tends to reinforce the validity of the association. Our relative risk estimates for, positi ve family history'of lung cancer, 3'.79'in males and 4.18'in females, were also similarto previousstud- ies: The association may be due partly to the fact that a subject's relatives livedlinthe same environment asdhe subject for some time. klulvihilll has postulated that some abnormal ty;pesofgcne might increasesensitivitye to environmental carcinogcns.'"'Further research will be necessary to clucidate andldistinguish the roles of genetic and environmentallfactors in carcinogenrsis.

INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Tobacco smoki ng is,generally aceepted'to be a major cause of'lung cancer.'¢u We observed an association of lung cancer with tobacco smoking in males, and this observation tended to enhance confidence in, the results: However, the association was wcaker, than has been reported in many previous studies:Oj19'When con- sidcred individually„duration of'smoking,,amount of' smoking, and age at which smoking,began were only weakly associated with the illness. Only the smoking, index derived by multiplying duration by amount of smoking,was significantly associated with lung cancer. These observationsmay'be due partly torthe fact that only 23 (8!2qa) of 280 malesin this stvdy were lifetime non-smokers. In such a small comparison group,,even fairly small changes in the distribution of non-smokers between cases and' controls could have produced marked differences in observed ORs associatedlwith smoking. More,importandy many farmers in Xuanwei' smoke tobacco through a long bamboo cylinder partly filled',with water and'the passage of smoke through thee water may filter,out carcinogenic substances. Studies, comparing the composition of water-filtered to unfil- tered tobacco smoke are currently in progress. It is also quite conceivable that the large amount'of air. pollutants inhaled during indoor smoky coal burn- ing in Xuanwei partly overwhelm the carcinogenic effect oGtobacco smoking; For example, as mentioned above, the average indoorconcentration ofBAP in the Xuanwei region of high lung cancer mortality, was 627 ug/l00 m' in a recent survey: An individual inhal- ing 12 m' ofai'r per day'mightAhcrcforc inhale 9154 ug of BAP'in a year, if'he or, she spent'eight hours per day indoors. In,contrast, an individlral smoking 20 cigaret- tes per day might be expected to~ inhale only about 700 ug o6Bt1P in one yean2' Thus, it is notespeciallyt surprising that the ORs associated' with smoking in Xuanwei males were smaller than reported in other studies. Because unusual environmental conditions prevail in Xuanwci, it would not be advisable to generalize these ORs to other areas. Smoking is very rare in Xuanhvei females. In addi: tion„we observed no associationioClung cancer with passive smoking in females. Such an association has been!reported inlseveral previous jmvestigations:'"" In non-smoking women in Shanghai, Gao er al observed a limitedlassociation of 7Ung,eancer witlh passive smok- ing; in that study the,relative risk ranged from 1.0 in women living,less than 20 years with,a smoking hus- bandlto 1.7 in those living with a srnoking husband for at least 40 years." Hbwever Koo et a1 have not' observed a consistent association of7ung,caneerwith, passive smoking in Chinesewomen.'"'Thcse authors also stated that correlates ofpassive smoking such as diet, and socioeconomic status can act as important confbunders when the health risks of passive smoking are evaVuated'I The heavy indoor air pollution in Xuanwei mayaiso overwhelm the carcinogenic effect of passive smoking. The effectofpassive smoking on, lung,cancer may depend on local environmental fac- tors and results obtained in a~given region may there- fore not, be applicable to other regions. In summary, this study was undertaken to sup- plement existing evidence showing a strong association of lung eancerwith indoor useofsmokyeoal imXuan• wei. Our results disclose important associations of lung cancer with factors other than fuel ty,prand therefore indicate that these factors must be considered in any comprehensive, quantitative risk assessmet:c of lung cancer in Xuanwei. Our results also confitm indirectly that'smoky coal pollution is an important determinant of lung caneor in Xuanwei'. A separate case-control study, which will allow simultaneous direct analysis,of the effects of indoor air pollution and other knownl.and suspected lung cancer risk factors in Xuanwei; is currently in progress. ACKI*1O W LEDGEMEh1TS The authors are grateful to Chaofu Huang andlDeyi Shan, who helped'with data collection. REFERENCES' ' Office.of:research~.ofcancerprevention.and treatment; Ministryof: Health: : Data.of caneerr mortaliry, in the Ptvpft's. Republic.ofChina. Beijing, 1980. = tsf umford U iL, ,He X Z. Chapman R,S; etal.Lung cancer and indoor air pollution inXvanweii China. Soitnce1987; 2-45t.217-20: 'ChapmaniR'S, Mumford 1 L, Harris DBIa.a1.'Theepidcmiologyof' lungcancerin Xuanwci, China: Current progress, issues, and research strategies. Arch E,rviron :Htalrh 1988; 3: 180-5. ' Schlessedman 17. Case-control'studitsr Dcsign, Conduct, Analysis. New. York: Oa[ord.. University, Press;,1982.. °Lilicnfold A' M;.Lilienfeld D. Foundarionr of tpidtmioloty.New York. Oxford IW niversitv ~Prcss; Second, 1980. •'Miettinan O.S: Estimahility and estimation in ease-referent studics: Am 1 Epidtmio!'1976; 103: 22G-35: ' Mantel N; Chi square tests with one degree of freedom: extension of the-.Mantel-Hbenszel procedure. 1Am Stat Asaoc196b; 5'g:696-700! ' E3reslow~N E, Day N E. Staniseital me,rhods in cancer rtsnacch.; Vol !. The analysis of care-controlttudits Lyon. IARC'.Sciendific, Puti4iationNo 32. 1980. ' Gail M H, Lubin 1. HI Rubinstein,L V.Likelihoodcalcvladons for' matehed case-controt Istudiesand!survival studies withried I deathitimes. Biometrika 198t;.Gb:70r7, 10 Storer.B E. Wachotder S', B'reslow NE. Maaimumlikelihood fitting of general irisk,modelsto stratifieddita. ApplSr.ari.rr 198?; 32: 172-81. " Wu A.H, Hcndencon(t E:.Pike M C„Yu.M!C. Smoking:andooher risk facrorsdorirrngcanecnin women. JNCI 1985; 74c :747-51.. " Wang X Z. Lungg cancer and I air, pollution indoor and outdoor. Environmennand Htalrh' 1983; 1:.7-12. "Gau.Y7,.Blot W.1, ZhengW;ar al..Lung,canceri among,Chinese women. !nt l ,Cancer 1987; .4f1d 604-9.

RICK FAC110RSIFOR LUNG CANCER'JN CHINA ~. 31).I J, Jensen i Epidemiological review of lung eanccr in man. In: Mo1v E. Schwaehl D', Tomatis L(eds). Arr pollunon and cancer in mon„Lyon, IARC, 1977; .'kuhata 0 K, Lilienfeld A M. Familial aggregationof'lung cancer in ih'uman. JNCI 1963; 30: 289=312. VY L„EMon iR'C, Chen V' W„Bailey-Wilson J'E: Rothschi W'H. lncreasatl familial irisk for lun JNCI 19t16: 76d 2l7 22 ca r g .. nce ~g9oCman T E H l I t i D D M46 ihiU7 I Fa ili i v ra asa oger m a resp ory ,,,.. ~" tract eancer. JAM 198P: 21'l: 102f~3 A . ~ , ~lwYvihillJ J. Hosi fadon iinAuman lung tumourr an exampk of eco- ~e. - penetia id onaobgy, JNCI 1976; S3: 3-7:. ;lineb L A, Etnster V L, Warner K'E, A'bbotts Jl Laaszlo J: Smok'ing ' ' ' a and lung anoer: an ovorview: CancerRe 1984; ( ,12:,p( 1)i . 5940--58. '4/Yeder E L„Graham E A. Tobacco smoking as,a possible eliolog, , ial factor in bronahogenic carcinoma; , JAMA 1950: 143: : 329-36. M IL, Goldstein H. Gerhardt P R. Caneer and tobacco smok- ing: Preliminaryneport:JAMrA 1950; 1N3: 336-g. R. Hill A B. Smoking and carcinoma ofrtAe lung: Preliminary report. Br Med J 1950; 2: 739-4g. C): Heahh Juuandr of,f6e human environRnenr: Geneva. 1972' os D;, Kalandidi A, Sparros L, MacMahon B. Lung anaer and passive smoking. Inr J Cancer 1981; 27: 1-4: n Correa P, Pickk G W; Fontham E. Lin Y„Haenszel M1V.' Passive smok ing and lung cancer. The Lancet 19g3clid S95'-7, "Garfinkel L Passive smoking andleancer-Americamexperience:. Prev Med '19&t; 13: 6914. ="Blot W 7, Fraumeni J'F Jr. Passive smoking and lung cancer. JNC! 1986; 77: 993-1000. ' Koo L C. Ho l H-C. Saw D. Active and passive smoking among , female Irhng anar Patienu and knntrols in Hong Kong. J Fsp C4n Cancer Res 1983; 1; 367-75'. "Koo L C, Ho J H+C, Saw D! Is passive smoking an added risk factor for lung cancer m Chinese women? J ErP CJin Cancer Rts 1964:3:277-93. ' Koo L,C, Ho J H-C, Saw D, Ho C-Y. Measurement of'passive smoking and estimates of risk for lung aneer, among non- smoking Chinese females. Inr,l Cancer 19g7; 39: 162-7- " Koo il: C, Ho J IH-C, Rylander R. Life-history correlates of envinon- meoul tobacco smoke: Astudy on nonsmoking Hong Kong Chinese wives with smoking,venus nonsmoking husbands. Soe,Sa Med 19gg; 2b: 759-60: (Revised'version received September 1990) '