Tobacco Documents Online

1t:1. 1 (',:nc'rr: 40. 0E>•-t-0(1y t!`)<7{ ,=' Ia~i .a1an ft l.iss, Inc. . ...., (~yy Pub1,Ga7Wn o/ the tntelnnUonlt Umon AQ.,n{t Cance[ t Publ,cat.on de t'Umon Imematwnale Conue h Cancel LUNG CANCER AMONG CHINESE WOMEN 1'u-Ta_nc GAnt. William J. aLOT'. !L'ci ZHENGr. Abby G. EKStioW`. Cheng Wcn Hsu`, Lynn 1. Lt:vln2, Rong ZI+nNGr and Joseph F. Fx.al1%1F%'t. 1tt.- 1SJNl,r4hai Cancer Insrirute. Jh:utglwu. /'rl>izle's Republic „f'Cu:ui: 'A'cuionuf Cancer hrsriturc, l3ethrsda. MD 20892, USA; u:ul 'Sfktnghui Chtst Hus,piwl. Slurn,4hat. f i•vp(<•'s Republic cJ* G:inu. A case-control study involving interviews with 672 female lung cancer patients and 735 population-based controls was conducted to investigate the high rates of lung cancer, notably adenocarcinoma, among women in Shanghai. Cigarette smok- _ ing was a strong risk factor, but accounted for only about one- fourth of all newly diagnosed cases of lung cancer. Most pa- tients. particularly with adenoEarcinoma, were life-long non- smokers. The risks of lung cancer were higher among women reporting tuberculosis and other pre-existing lung diseases. Hormonal factors were suggested by an increased risk associ- ated with late menopause and by a gradient in the risk of adenocarcinoma with decreasing menstrual cycle length, with a 3-fold excess among women who had shorter cycles. Perhaps most intriguing were associations found between lung cancer and measures of exposure to cooking oil vapors. Risks in- creased with the numbers of meals cooked -by either stir frying, deep frying or boiling; with the frequency of smokiness during sooking; and with the frequency of eye irritation dur- ing cooking. Use of rapeseed oil, whose volati_les following - high-temperature cooking may be mutagenic, was also re- ported more often by the cancer patients. The findings thus confirm that factors other than smoking are responsible for the high risk of lung cancer among Chinese women and pro- vide clues for further research, including the assessment of cooking practices. t'hc- a~.e-a~ijuatcd annual lung cancrr incidence rate amc~ng 1'cmalcs -ln Shanghai aurinl; rhe 1970's and 1980's has been atxwt 20 per 100,1XX) potpulation, one of the highest rates in China and in the world (National Cancer Control Gffice, 1980; Gao, 1982: Waterhouse et al., 1982; Shanghai Cancer Regis- trv. 1983).. Elevated lung cancer death rates have also been - observed among Chinese women in other parts of the world, ln.luding Hong Kong (}:un4 /v /t/., 1984: Koo et ul., 1985), Singaport (A4acLennan er ut.. 1977) and the United States tFraumeni and tilason. 1974: Hinds et al., 1981; Green and - Brophy, 1982). The high rates of this cancer are surar-isin¢ t since few Chine.r u•nm- on the basis ut' diagnusis, histologic type, and the site of the tumor within the lung. Two seni©r pathologists and 4 scnior clinicians were appointed to review the diagnostic information from all cases collectcd in the study, including X-ray films, cytologic and histologic slides. Female controls were randomly selected within S-year agc strata frona the general population of the Shanghai urban area. The number and age distribution of the controls were deter- mined in advance from the number and age distribution of lung cancer cases reported to the Shanghai Cancer Registry during the period 1980-81. The selection procedure involved randomly choosing a neighborhood committee from among the approximately 1,300 committees in urban Shanghai, then randomly choosing a household group within the committee and ascertaining from existing rosters the names of all females in the appropriate age range. Among these persons, 2 were randomly selected. If the first was absent during the period of study or could not be interviewed, the second was accepted as a control. Tables of random numbers were used in the random sampling. The cases and controls were interviewed by trained inter- viewers. A structured questionnaire was used to obtain infor- mation on demographic characteristics, exposure to tobacco, dietary and cooking practices, medical conditions, family his- tory of lung cancer, tnenstrual and reproductive factors, job history anJ othcr variables. All completed questionnaires and metlicaal abstracts wcre checkecd by a field supervisor, and the information was then abstracted on coding sheets for key-g punching and computerization in the United States. Statistical analyses of the collected data were based on mul- _ _ tivariatc techniques for case-control data (Breslow and Day, 1980). Logistic regression analyses were used to estimate summary relative risks (RR) of lung cancer associated with various faCmrc -E+,.- a:.-r_ _ I' - Css£Zzszoz

and measures of exposure to cooking oil vapo-s. Risks in• creased with the numbers of meals cooked tL either stir frying, deep frying or boiling; with the frequency o( smokiness during cooking; an4l with the frequency of eye irritation dur- ing cooking. Use of rapeseed oi/, whose volatiles following hi&h•temperature cooking may be mutagenic, was also re- ported more often by the cancer patients. The findings thus confirm that (actors other than smoking are responsible (or the high risk of lung cancer among Chinese women and pro- vide clues (or further research, including the assessment of cooking practices. t•hc a:c a.ljuac.f annual lung ng :.tn«r utcidencc rate amOnt; (cmalr> in Shangh.u .iurin}; the 1970's and 1980's has lvcn atx)ut 20 per ilX),lX0 population. one of the highest rates in China and in the world (National Cancer Control Gffice, 1980; Gao. 1982: Waterhouse er al., 1982; Shanghai Cancer Regis- tn-. 1983).• Elevated lung cancer death rates have also been obsen-cd among Chinese women in other parts of the world. in.ludini Ho~ng Kon4 (Kung er ul.. 1984: Koo er (il., 1955), Singapore (htacLcnnan er rtl., 1977) and the United Statcs tPraumFni and Mason. 1974: Hinds er al., 1981; Green and Brophy. 1982). The high rates of this cancer are surprising since few Chinesc women smoke cigarettes (Deng and Gao-, 1985). punhermorc, in Shanghai (Zheng and Gao, 1986) and elscv.•here, hospital records have indicated that most of the lung tumors arc adsntkarcinomas. a type of lung canccr less >tr0ct:hy related to smoking (Luhin and 9lot, 1993). To explore rcasoru for the high rates of lunL caro.-er among women in Shanchai, the Shanghai Cancer Institute, in collaboration with the US National Cancer Institute, carried out a population- - hased case-control im•estigation. Herein we report the results of this study, quantifying the role of smoking and evaluating a ~ variety ~~f su,pcctc~l risk factors. M.aTE1:1.aL .9N1) METHQDS All nc-wIv diagnosed cases of primary lung canccr (9th f:evision ICD 162) during the 2-year period February 198-3 f=cbruary 1986 aamong female residents of urban Shanghai aged 35-64 years were identiticd by a rapid reporting systetn for lung cancer established for this study. The system was built upon the existing Shanghai Cancer Registry, the oldest in China (Gao. 1982). Trained staff contacted medical facilities in Shanghai to ascertain new cases, so that interviews could the approximately 1,300 committe_s in urban Shanghai, then randomly choosing a household oup within the committee and ascertaining from existing rosters the names of all females in the appropriate age range. Among these persons, 2 were randomly selccted. If the first was absent during the period of study or could not hc intc-rviewed, the second was accepted as a control. Tables of random numbers were used in the random sampling. The cases ses and controls were interviewed by trained inter- Viewcrs. A structured questionnaire was used to obtain infor- mation on demographic characteristics, exposure to tobacco, dietary and cookinb practices, medical conditions, family his- tory of Iung cancc;r, menstrual and reproductive factors, job hi.tur> and other variables. All completed questionnaires and- - medical abstracts were checked by a field supervisor, and the information was then abstracted on coding sheets for key- punching and computerization in the United States. Statistical analyses of the collected data were based on mul- tivariatc techniques for case-control data (Breslow and Day, 1980). Logistic regression analyses were used to estimate summary relative risks (RR) of lung cancer associated with various factors, after adjusting for age ( E 55, 55-59, 60-64, 65-69), smoking (non-smoker; smoked less than 20 years or less than 10 cigare(tcs/day; smoked 20 or more years and 10- 19 cigaretteslday; smoked 20 or more years and 20 or more cigarcttcs/day), education (no formal education, primary school, secondary school and higher) and other variables, and to evaluate statistical significance. Population attributable risk (PAR) estimates for smoking, adjusted for age, were also derived (Whittemore, 1983). RESULTS A total of 765 lung cancer patients were identified during the 2-year period and interviews conducted with 672 (88%). \Ye excluded the 93 patients who died, including 38 ascer- tained by death certificate only. There were no patients who refused interview. Forty-thrce percent of the cases were diag- nosed noscd by tissue biopsy. 38% by cytology, and 19% by repeated Address reprint requests from China to Dr. Gao and from other coun- tries to Dr. Blot. I be rapidly scheduled (typically within 2 weeks of diagnosis). The :t;tff revieuxd relevant medical records, abstracting data Reccived: March 30. 1987 and in revised form May 29. 1987. ts9EZz9zoz

.-?.Nl t K 1\ CHINIiSF. "'O\tlil 605 1\NI.F I- KI : tl l'~ t K~~\\ O1 f l~~• t~ \\tl.K a»(A'IMFa) t11TH \l MISI H 01 r'll.,\KI r 1*1 N lMUtl;I 1) LI(K 11.AY ANf) 17t'Ij.\11U\ UF l\1()F>I\(i ._.__..'~.'.--..--. _--' -----'-------'-~--.--_~ - - - - ------- 1)ui.aip•n .1 .nu,lurl; ?r.•n~cr YCJr. > 1t .. ) liara r<(•u, ._ . . ... _.-.-~~---- ~----'--'----•- -'--------'-- -'- -- ti l. KK: u5 " CI l r Cmnd, RR' Vt : CI . tu ;t, .t5 tc! la ry I I :i)- - -- 1~ I.s t).9-2 .2 34 29 2.4 t..t-s.t St, 33 3.2 2.0-5.1 8.9 2.0-40.2 %.1' 10 14.1 7.1-28.0 .\.au•toJ :or .r;: \,m~•m,•:Cr, jrc u•Cd i> thC rctcrCna gnwp. TAHt.F: 11 - KFt. ~I lt t K1JkS tN t-I-'a. CAa'rK Attt).E; N(ri-St,tuKlNt: No%1t.S kSStX IkTI P N Il H 1 FAK1 ll' t-11 tt I rH ASMOK1 VG HUSBAND lcu. l,.N . ?U 57 qq 1.0 20 'y bl 93 1.1 0.7-1.8 30-39 78 107 1.3 0.8-2.1 ~ 40 45 : 6 1.7 1.0-2.9 3.Jjua1N 1'.•t aEc snJ c>Ou:auun \-ra\ tilcn;. Alrnmt: the 542 interviewed cases pathulttlgically or :\tt)lt>>•;icallv diaLnosed. atlentxarcinuma was the prcdomt- nant crll~typc :kct~unting for 61 5c ttf all rases. 2?% wcrc aquanl.)u. ~:arcinttnlaJ. b Y tt'cre i)at-Cell (or small-cell undif- I.rrntisat-a) .an.cr:, and 11 % were mn.rtures and other cell l~. t\ A total tt( 71S controls were intcrviewctl. Among these 71 ty.7''H were "srcond" controls, chosen mainly because the tir.t selected control had moved from the Shanchai urban area ~ _ - - or \%a; f,.aunti to Ik outside the eligible age range. The distri- huttons by agt• education and marital status were generally .intllar htt\t'cen cases and controls. More controls than cases 02% +'s. _'0 F). htmcvcr, %%-ere in the oldest age group of 65- h9 \tars. hut all suhaauent analyses were age-adjusted. Ci,%:ttr<*trr .crrnoki,lp C'1Lart:ttc anntl,in,: \%;IZ associated with a:i>'niticantlv in- .- . -- .. - . -rea.cti risk tit lung canccr, even though onl\ 35 % of the lung cancer patients (compared to 18% of the controls) had ever .m0ked. There \taa a 3.3-folti excess risk (95% CI = 2.5- s'-) of lung cancer amonL smokers. but risks were higher for squamous-cell cari inttma_ (RR = 7.2. 95 `7o CI = 4.6-11.1) and oat-cell cancer (RR = 7.2. 95% CI = 3.6-17.0) than for .t.icnt•.•art=inont.+ (RR = 1.5, 95% Cl =. 1.0-2.1). The RR for ever, lung cancer risks among non-smoking women tended to increase with the number of years a woman lived with a husband who smoked, the RR reaching 1.7 among those with 40 or more compared to less than 20 years' exposure (Table II). The risk in this heavily exposed group was even higher (RR = 2.9. 95 % CI = 1.0-8.9) for squamous- and oat-ecll t:arcinuma. Previous lung diseases Since lung cancer in its early stages may be confused with other lung diseases, we excluded non-malignant lung diseases occurring within the 3 years preceding interview in evaluating the el'fcet ot' prior lung disease upon lung cancer risk. Table lll shows that previous tuberculosis, pneumonia and emphy- sema were significantly associated with lung cancer risk'even after adjusting for smoking. Although some individuals rc- ported having 2 or more of these diseases, the excesses for cach persisted when those with multiple conditions were ex- t:ludcd. Further analysis (not shown) indicated that the effect ot' tuberculosis was not related to the use of isoniazid or streptomycin. While tuberculosis and pneumonia were related to both squamous/oat-cell carcinoma and adenocarcinoma of the lunL, emphysema and chronic bronchitis were associated - only %vith the squamous- and oat-ccll types. Cooking pruc7i<'cs Surhean and rapesced oils were the oils used most often for rcxiking in Shanghai, with over 95% of women reporting the use ttf huth products. Rapesc-ed oil, howcvcr, was reported as the most ot'ten used cooking oil by 52% of the cases compared tt) 45 % Of the controls. The overall increase in risk associated with rapeseed compared to soybean as the most often used oil was 1.•i (y5% C( = 1.1-1.8). Table IV shows that the excess lung cancer risk associated with use of rapeseed oil existed at each level of reported frequency of eye irritation when cook- in):, a subjcctive variable representing severity of exposurc to Ss9C rz. z9zUz

.imilar betu•ccn cases and controls. More .ontrylIs than ca:es t i'_ 'r c's _'-0`.f ). hcmever, were in the oldest ag. roup of 65- a9 Nears, but all subsequent analyscs were age-adjusted. C-:_:trcttc .nn,l.m,: %%a: a_..ck:iatc.l \~ ith a si8niticantly in- .'rcaNcd rislk of lung c•tnccr, even though onl\ ' ?.S 4of thc lum_ cancer patients (ccmil+arcJ tc, Iti c cit the controls) had ever ant0kcd. There %%a: a ',.i-(0l.1 excess risk (94 h' C1 =2.5- 3._) of Iuni cancer anionk smokers. but risks were highcr for syuamous-«Il carcinoma (RR = 7.2. 95% Cl = 3.6-I1.1) an.f oat-cell cancer (RR = 7.', S15 Y Cl =1.6-.17.0) than for a.icnocar:inoma (RR = 1. 5. 95 5-c Cl = 1.0-2.1). The RR for all lung cancers combined tended to rise with increasing num- tkrs of ctgarettts smoked per day and with increasing duration ot smoking ITahlt 11. The excess reached 14-fold for females y ho smoked 20 or more i Ig:IrtttcJ pcr day for more than 30 \cars. Similar trends ca:ted for aclrncxarcinoma and for syua- ~ - - - -- - mou..oat-ccll cancers, but the magnitude of the increase was considerably greater for the latter (not shown). \t'c calculated population-attributable risk (PAR) estimates for :m.ikins: in each age group. The PAR rose with age, from g% tc, _7 % to '8 k to 40 n at ages < 55 55-59. 60-63, and t,5-09. reslkctively, primarily because the prevalence of smoking rose with age. In total, we estimated that 24~'ro of all fcniale lung cancers in Shanghai were due to smoking. Rr.srve• smokiirg No significant increase in risk was observed for overall eah,isurr to environmental tobacco smoke during childhood (RR = 1. 1, 95 % C1 = 0.7-1.7) or adult life (RR = 0.9, 95% C 1 = U.b-.l.J). For these calculations, exposure was said to ki::ur if the :uhjeit had ever lived with a smoker. When c.rx)>ure \.a. defined in terms of husband's smoking, how- 969Czz9ZOz only with the squarttuus- and oat-cell types. C-cwkrng prctc'lic'c+s So~ bean and rapeseed oils were the oils used most often for ccxokim.: in Shanohai. with over 959'c of women reporting the both use crf ~th rrcklu.t:. Rapeseed uil however. was reported as the most uttcn used ccxtking oil by 5_ h of the cases compared ttt o(' th< controls. The overall increase in risk associated with rapesrcJ compared to soybean as the most often used oil - - was 1.4 (9.5 % CI = 1.1-1.8). Table IV shows that the cxscss IunL cancer risk associated with use of rapesced oil existed at each level of reported frequency of eye irritation when cook- ing, a subjcrtive variable representing sevcrity of exposure to cooking vapors. The calculations for this Table excluded the few women who never cookc-d, and employed as the reference group women who most often used soybean oil but never or rarcly reported eye irritation. Table IV also shows that risks of lung cancer were independently related to eye irritation, with the highest risks (RR = 2.8, 95% CI = 1.8-4.3) among those using rapeseed oil and frequently reporting irritation. The patterns \%•ere similar for squamous/oat-cell cancer and adenocarcinoma. >Ve also observed, after adjusting [or eye irritation, a 60~'r higher risk for lung cancer among women who reported cunsid-erablc- or somewhat smoky conditions in their homes when cooking, another rough measure of exposure to cooking vapors and to house vcntilation (Table V). In addition, the risk ratios increased with the number of different dishes per week prepared by stir f-ryiFlg, deep frying, or boiling (Table VI). In contrast, no significant case/control differences were associated with the type of fuel used for cooking. The RR and 95% CI associated with coal, gas and wood as the usual fuels were 0.9 (0.7-1.3), 1.1 (0.7-1.5). and 1.0 (0.6- 1.8) respectively. There was no trend in risk with increasing Ycars of use of coal. the most common cooking fuel in Shang-

t-06 GAO tST nL. T.\IILF: I1I - I:LLAI1\'li RI\KN UI- l.tiN(S ('AN('IiR ASS(X'IArI•1) 1\'IfII l'RI:VIIIIIS LUN(J IH\I:AJf.C ly,A IunF .f..ea..es C.wr.Js All .a.ca RR' y~q CI -1'.luanuwv •utr•ell RR' 93111 CI Auknu• carcuwvna RR' 93% talnc 554 318 I.(1 - 80 1.0 - 229 1.0 - 1'uN-rrult+.ia 61 SO 1.7 1.1-2.4 20 2.0 1.1-3.7 42 1.6 1.0-2.5 Pncunxrnw 35 0 1.9 I.'---3.O Ih 1.8 t).9-3,Ft 26 l.5 0.9-2.7 l-n(I•h% >clnu I\ 17 2.U 1.11-3.7 I') 4.5 2.(1-.1O.3 6 0.7 f).3-2.0 Chronic hr.Tnchttt. 86 1 12 1.2 11.8- 1.7 35 1.4 0.8-2.5 33 0.8 0.5-1.3 without cmpMscnia Others 30 30 1.3 0.7-2.2 8 1.7 0.7-4.2 13 1.0 0.5-2.0 '.ill ruks adlu.t<~) t.K acc. .du; atlon a:r! .m..king and rilatis c tu r.mkn r Ith Mf rC(wN1(d tHIPr lung Jiaa.c. Pcr.oms with nxxe thin one past lung di.ca.c arc included In ca.•b lung dlxa.c :atcgon TABLE IV - RELATIVE RISAS OF LUNG CANCER ASSOCIATED K'ITH COOKING OIL USED MOST OFTEN AND FREQUENCY OF EYE IRRITATION WHEN COOKING E.e uriutltn -hen avMing OIt used rtxwt on- Ca.cs C.xnrola RR' 95% CI Ner er or rarely Saytxan 140 '-1a 1.0 - Rapesced 145 193 1.2 0.9-1.7 Sometimes Suybean 70 72 1.5 1.0-2.3 RaRcsee-d 87 63 2.0 1.3-3.0 Frequently Soybean 59 56 1.4 0.9-2.3 ,Rapcsecd 90 50 2.8 1.8-4.3 Ttxal Soybean 269 342 1.0 - Rap(;seed 322 3()6 1.4 1.1-1.8 --- 'Adjuate.lloc age. education and sntoking. TANLF: % 1- RFL:\TIVF. RISKS UF LI:NG CA\CER ASSOCIATED N\'ITH NU\1BFR OF hIfFtRE\T ()ISHFS PER a FEK PREPARED B1 DIFFF.R1:\T METHt7DS l)F (`(k1K1\(: Nurr.t.rr u( dl.hr> (t c r~•t`.\ Ca.t. C.wlv.>I> RR' Yt% CI Stir jnin,q <-0 336 408 1.0 - 20-24 198 211 1.2 0.9-1.5 :5-_4 4A 47 1.2 0.8-1.9 > 30 34 15 2.6 1.3-5.0 cker,•n tn.%: 0 502 594 1.0 - 1 85 68 1.5 1.0-2.1 2 21 - 15 1.6 0.8-3.2 a? ~ 4 1.9 0.5-6.8 , i 499 TABLE )' - RELATIVE RISKS OF LUNG CANCER ASSOCIATED WITH FREQUENCY OF EYE IRRITATION AND HOUSE SMOKINESS WHEN COOKING E)c Irrlt'JINIn Hcw.c tR10kIneX% Caaes Crxurds RR' 95% CI Never/rarely None/slight 244 380 1.0 - Sumcwhat/ con.iderablc 55 55 1.6 1.0-2.5 8(tmetimes/ frequently None/slight 212 200 1.6 1.2-2.1 Somcwhat/ con.iderablc 109 60 " 2.6 1.8-3.7 'Adjuslcd for age. education and smoking. TABLE VII - RELATIVE RISKS OF LUNG CANCER ASSOCIATED WITH DIETARY INTAKE OF VITAMIN A Vaumin ()uartile kvcl of conwmption' c~/x~wre .anaAlc I (Hlgh) 11 11f IV Vitamin A 1.0 0.6 0.8 . 0.5 index (0.5-0.9) (0.6-1.1) - (0.4-0.7) Retinol-rich 1.0 0.9 1.0 0.9 lixxis (0.7-1.3) (0.7-1.3) (0.7-1.2) Car(r'tcne-.rich 1.0 0.6 0.5 • 0.5 foods (0.5-0.8) (0.4-0.7) (0.3-0.6) ' Risk relative to highe>a quartile or cra(sumption an(d adjusted foe age. education and smokina. 95% CI in aarentheses. .sA 1~4 1 n \/tI) Tt.;~ n.•.~..~~n•.~c... .I.oc pnPOn~lr.r ~ nr moln v v e zz9z0z

TANt.t: \ 1- RELATIVE RISKS E)h LUNG C.9.NC't:R .aSSEX'IAT1:D WITF/ NUMBER 017 DIFFFRE\T DISHES PER 1cF.eK RRF.PAREI) Bl' DIFFERENT SIFTIInDS [)F E'tktklM; \umNCr .1: JiJK.. Nr .ax\ Fw. <'.,nln.l. RR' 95 14 l'I $:ir fnirrjc <_ .(1 ??6 408 1.0 - 20-24 lyS 211 I._'- 0.9-1.5 4S 47 1 2 0 8-1 9 1 . . . ~311 .1J 15 2.6 I.3-5.0 l)eeri rn irtg 0 t()? 594 1.0 - I 85 68 1 5 1 0-2 1 , :1 IS . 1.6 . . 0.8-3.2 ?3, 4 4 1.9 0.5-6.8 84 ~ilra,k 43 96 124 1.0 - s-7 :%1 483 1.0 0.7-1.3 ti-I1 63 40 1.8 1.1-3.0 ;~ 1= 67 33 2.2 1.3-3.7 'Aa/1uuyJ f r al;c. cJucau.m and .nr.lmF, hai. frtp<asures to coal or othcr fuel fumes wcrc generally associated only with cuoking, since nearly all homes in Shang- hai were unheated. Dicr The women «,cre asked about their usual frequency of con- sumption during adulthood of 32 commonly~ eaten foods, in- clu~iinc the major contributors of vitamin A. Using Chinese food composition tables to estimate the rctinol and carotene content of each food and applying these estimates to its frc- .)ucncv ~ of intake, an index of vitamin A consumption in ret- inol-tquivalent units was constructed. The risks for lung cancer tended to be lower among those with low values of this index &s9VzZg~o Z TABLE %'ll - RELATIVE RISKS OF LUNG CANCER ASSOC~IATED WITH DIETARY IVTAKE OF VITAAifN A Vuit»in x r Ouartile kvcl tif conwmplinn' .W . %I c s'ariahic IIHiEhI 11 tll IV Vitamin A 1.0 0.6 0.8 0.5 index (0.5-0.9) (0.6-1.1) (0.4-0.7) Rctinul-rich 1.0 0.9 1.0 0.9 tixxls (0.7-1.3) (0.7-1.3) (0.7-1.2) Carotene-rich 1.0_ 0.6 0.5 0.5 fsX-.1.'. (0.5-0.8) (0.4-0.7) (0.3-0.6) 'Risk rclali.e to highcst quanite of conwmption and adjusted for age, education and snuoking. 931* CI in Mreniheses. (Table VU). This association was accounted for mainly by a lower risk among those with a reduced consumption of c-aFo- tenc-rich foods (the dominant source being dark green vegcta-. bles). No effect on risk was found for consumption of the rctinol-rich ti)ods (mainly fish, eggs and liver). The patterns were generally similar for squamous/oat-cell cancer and ade- nocarcinoma, and for smokers and non-smokcrs. AfE•,rsrnull rrrul rc•/woxlurtil•c fircturx The risks of lung cancer were higher.among .women "with shorter menstrual cycle lengths (Table VIII). -The association was primarily seen for adenocarcinoma, which showed a strong . dose-response relationship. Among women aged 55 years and over with a natural menopause. thc risk_of adenocarc_inotna'- tcnElcd to increase with the total number of menstrual cycles over their lifetime. Some increased risk of adcnocarcinoma was seen when natural menopause occurred at age 50 or later (RR =. 1.3. 95% CI = 0.9-1.7, after adjusting for menstrual . cycle length). No associations were seen with age at menarche, age at first pregnancy, or parity.

t JNG C.4NC E?R IN CNINGIE3 wOMFN T.\81_E \'lll - KEL:)TI%'1: RISk:S nF LU\(7 CAtiC•NK ASS(X-IATF:U )S'ITH LENGTH OF MENSTRUAL Cl'CLE 607 Ltnilh,d n.rn.iruil e.:tr 1,Lt.1 (II .a,e. ..._ _ C.vun.l. RR' V5$ CI S4wnxw.; ~ua.Al -. ... KR' Vt% Cl AJentr carcialmo RR' 95% Cl > 33 43 60 1.0 - 12 1.0 - 18 1.0 - 30--i3 ?-,'- 327 1.6 1.0-2.6 62 0.9 0.4-2.E) 124 1.9 1.0-3.5 26-29 241 268 1.6 1.0-2.7 51 0.8 0.4-1.9 127 2.1 1.1-.3.9 <26 98 78 2.2 1.3-3.7 23 1.6 0.7-3.9 54 2.9 1.5-5.7 'AJIV>ICJ itlr Jgt. hfU:JIHon. .111NE,Ing Jnd Q`luLIr11J H I11iMIRlatltM. TAKLF IX - K[I AT-I\ E RISKS UI Ll'`G CA\C•EK FUK «'t)Alli\ LVI:K EMI'l•OYEt) IN MAIUR OCCUPATIONAL GRUUE'S 4).:ulult,•n :J1cF.w.' (•a.a. CPtMnJ. RR' 95% CI 1-11 Professionals and technicians: leaders of 0 state tirganizatit)ns. Pany and mass tarea- niZationX and CntCrpri5e unitS Office and related peraunnel; sales \surker, Service workers - -- Agricultural, liorestry, anintal husbandry and tishcn• \curkcrs - - . - Pr(tiluctian, transpuaatiun and other rclated N'urkcrx Never N•Orhed 75 96 159 160 24 21 436 471 1.0 0.7-1.4 0.7 0.5-1.0 1.0 0.8-1.4 1.1 0.6-2.1 61 75 1.1 0.7-1.G- . --- - ,,. N',wncn empltaN in n>,trc tEun tWW tw.-utulituul Cal.gan arc inclu.krJ in cu:h iatrgtlry in which thcy worked. The cttxlinE xhcme iras based on the nystcm usejd in'(he 1992 Pt"lation Cen.u..+i thc Parplr.'s Retx,blic of China IP.ywlatiym Census UI'lice, 148t1.-'Adjuateal li)r age, efuc•rtion anJ >nxtkrnj. = Occ•uFarinn Most \.•untcn rcpt)rtcd working outside the home, but ca.c:/ control differences according to major occupational categories were small (Table IX). No major occupations were associated v.-ith increased risk of lung cancer. A decreased risk, however, \, as observed for women ever employed in the cotton textile intiustry, the largest cmployc-r of women in Shanghai. There N'as a slight increase in the relative risk of lung cancer amung women ever employed as cooks (RR = 1.2. 95% Cl = 0.6- '_. 1), hut few worked longer than ?0 years. Fau)il\' 6isrr)n• The cancer patients reported about the same frequency of lung cancer in their mothers (1.0%) and fathers (1.7%) as the cc.ntr-ol: ( 1.0% and 1.5%, respectively). The RR, adjusted for agc, c-ducati(an and sn)(\king, associated with having a parent with lung cancer was 1.1 (95% Cl = 0.6-2.3). More sibs • . r . . . .. . - . :,.. bined_ (Blot and Fraumeni, 1986), an overall 30% exce3s of .. lung cancer (RR=1.3; 95% Cl e 1.1-1.5) was found among non-smoking women married to smokers, with the RR reach :-. :. . ing 1.7 among those most heavily exposed. Although the causal significance of the relation ofprtor:lung . disease to lung cancer remains to be'clarified, the high ~rEVa ...: Isnce of previous pulmonary infections may havo contrtbuted ,: in part to the high incidence of lung:eancer' among_8hanghat women. Earlicr in, this century, non•mali gnant lung disease was one of the leading causes of death in China {Kan,rY981) With the advent of antibiotics and improved living conditions, the incidence and mortality of-chron(c, lung diseases; ~~icu > larly tuberculosis, declined. ` N(:vertheless, a,substanttaI~ppr tion (38%) of the women with lun cancer'in this study repOrted prior lung disease, including 12% who were long-fertnaut•vt vors uf tuberculosis, whereas significantly lower.percentages ' of controls reported thvse disestses. To some extent'it ts;posst sGsE%z9zoz

VlI-lX t'r.tluctiun. tranaxm:1ti()n ar related wU(Ia'r\ i(hcr 436 471 1.1 0.9-1.4 t1 Never worked 61 75 1.1 0.7-1.6 \~,wtkn ctoplo.W tn mon thun .Kk v:ututtonil iatrgan urc tnauJ.d in .jh<ut.pn tn w•htch th.}• w.~rl.~d. Thc crxling achemc wah Nast4 sm the syaom used in tht N~; R~,Lttnvt Ccn.u..H thr Pcoplc'. R:puMh: . i Chnu (Pupulatton C.n,.u.. Uli~... Iva,s).-'AJju>tN lior agc. cducation and snxtking. Ot•c'ttncuirrn \fo:t \%urnen relx,ned working outside the home. but casC/ control differences according to major occupational categories were small (Table IX). No major occupations were associated with increased risk of lung cancer. A decreased risk. however, was observed for women ever employed in the cotton textile industry, the lartest employer of \.•omen in Shanghai. There was a.light increase in the relative risk of lung cancer among women ever employed as cooks (RR = 1.2. 95% Cl =. 0.6- _ ?. I), but few worked longer than 20 years. Faurtl\• histt>'n• Thc cancer patients reported about the same frequency of lung cancer in their mothers (1.0%) and fathers (1.7%) as the controls (1.0% and 1.5%. respectiv~cly). The RR, adjusted for agc, education and smoking, associated with having a parent with lung cancer was 1.1 (95% Cl = 0.6-2.3). More sibs were reported to have lung cancer, but the numbers affected were small (6 cases, 3 controls; RR = 3.0, 95% CI = 0.7- 12.5). Only one child, of a control, had lung cancer. bined (Blot and Fraumeni, 198f), an overall 30% excess of lung cancer (RR=1.3, 95% Cl = 1.1-1.5) was found nd among non-smoking women married to smokers, with the RR reach- _ - ing 1.7 among those most heavily exposed. Although the causal significance of the relation of prior lung disease to lung cancer remains to be clarified, the high preva= lenct: of previous pulmonary infections may have contributed in part to the high incidence of lung cancer among Shanghai women. Earlier in this century, non-mali nant lung disease was one of the leading causes of death in,~hina (Kan,- 1981). With the advent of antibiotics and im roved living conditiotu, the incidence and mortality of chronic lung diseases, particu- larly tuberculosis. declined. Nevertheless,' a substantial. por= tion (38 %) o f the women with lung cancer in this study reported prior lung disease, including 12%`:wh©.wcre long-term survi- vors of tuberculosis, whereas significantly; lower percentages of controls reported these diseases :To.some,extent it ts:possi-- ble that recall or ascenainment-bias may- eontribute ~to ahe : associations observed with prior lung diseases:'The elevated '. risk df lung cancer following tuberculosis.. however; ts consts ; tcnt with recent studies in other coutttries; and is not explained _` DISCUSSION 1'hc high incidence uf lung cancer among women in Shang- hai. together with the low prevalence of smoking in the general popul;uion. led us to consider a numbcr of possible etiologic factorc. While cigarette smoking was an important cause of lume cancer, showing a clear dose-response trend, the majority of lung tumos, particularly adenocarcinomas, occurred among non-smokers. Environmental tobacco smoke may account for some. but probably few, of the cancers among non-smokers, since there wa. little or no association with ever having lived with a .maker. Amone non-smoking women married to smokers. ht\\~e\~er, there~was an upward trend in risk associated with - - increasing years of exposure. This latter finding is consistent µith rcpuns in other-parts of the world. When data from ncartv a dozen studies evaluating passive smoking were com- by cigarette smoktng.or treatment with isoniazid, a pulmonary, . carcinogen in laboratory animals (Howe et a1.;:1979; Hinds et (il.. 1982: Bakris c•t Eil., 1983). Emphysema was also significantly related to lung cancer, after adjustment for smoking habits, with the excess limited to squamous- and oat-cell carcinomas. This finding adds to the evidence that chronic obstructive pulmonary disease enhances the risk of lung cancer (Skillrud et af., 1986), even when controlling for smoking practices. Also noteworthy is the ele- vated risk associated ciated with prior.pneumonia, especially :since an association with lung adenocarcinoma has previously;been reported among women in Los Angeles (Wu et al., ,1985). While pneumonia typically occurred during adulthood in'.our study, the finding in Los Angeles primarily concerned child- hood infection. Little evidence was found to implicate the type of fuels used for cooking in lung cancer risk, consistent with findings from 0 0 0 : oo4.Czzszoz

608 GAO kT nL. (Koo cr al., 1983). The risks of lung cancer among Hong Kong Shanghai women increased. however, with various measures iaf exposure to cooking oil vapors. These included the number of different dishes prepared per week by either stir frying, deep frying, or boiling: (he frequency uf eye irritation when cuoking: and the smokiness of the house when cooking. In Chinese wok cooking. regardless of the method used, oil is usually pi)ured into a wok and heated to high temperatures before meat or vegetables are added. Even boiling may entail some exposure to cooking oil vapors, since oil is often added to the water before heating. Consequently, the living quarters may become smoky during cooking, with opportunity for exposure tt) inhalable cooking oil vapors. The plausibility of the hyfx)thesis that lung cancer may be related to c.x)king oil vapors, particularly from rapeseed oils used in Shanghai. is enhanced by recent experimental investi- gations. !n one study the mutagenicity of products from cook- ing oil was assayed by the Ames test (Qu et al., 1986). The extracts of condensed volatiles of rapeseed oil, refined rape- seed oil, and sovtkan oil heated at about 270°C were all pKi,itive in tr.ter'strain TA95 activated with S9. The mutage- nicity ntcity of the extract from rapeseed oil volstiles was stronger than that from sovbean oil volatiles. There was no evidence of mutagenicit.~.• in the oils thtmsehes, either heated or unheated. -- In another study, the extracts'of condensed volatiles of rape- seed oil enhanced the yield of micronuclei in polychromatic cn•thrtkvtcs of the bone marrow of mice, with a clear dose- reslxm.e relationship, retlecting damaLe of chromosomes and - cell geneotoxicity b~- r.tpeseed oil volatiles (Chen. 1987). Al- though these tests often correlate with carcinogenic potential, no bioassay studies have vet been carried out, to our knowl- edgc. If the effect of rapesced oil smoke on lung cancer incidence is real, the problem is of great importance to popu- lati.ons of eastern central China and-other areas of the world where the ~~il is often used for cooking. Chinese rapeseed oil, \h hich is pressed from seeds of Brassica carnptstris, contains ab.lut 50% erucic acid (Chinese Academy of Medical Sci- ences. 1981 ), in contrast to rapeseed oil with <29o erucic acid (Canbra oil), which was recently approved for sale in the United States (Federal Register. 1985). Several studies have shown that the risk of lung cancer is elevated by a 1ow intake of foods containing vitamin A, partic- ularlv as its precursor beta-carotene (Colditz et al., 1987). Although reported mainly in Western countries, this associa- - __ (ion has also been noted among Chinese women in Singapore tti1acLennan er al., 1977). However, we found no evidence of a protective effect among women in Shanghai, where intake this unusual finding in females, which was observed also in a case-control study of lung cancer in Hawaii (Hinds et al., 1994). However, the protective effect of carotene-rich foods was mainly confined to current smokers in one large-scale study (Ziegler et (qt., 19$6), and thus the effect may be less evident in Shanghai where few women smoke. A clue to hormonal factors was suggested by an association between menstrual cycle lengt1i and lung adenocarcinoma. A 3-fold difference in adenocarcinoma risk was found among Shanghai women reporting short (<26 days) compared to long (> 33 days) menstrual cycles, with only a weak trend for squamous- and oat.-c;.Il cancers. A relation of short menstrual cycles to breast cancer risk has been suggested in data from Sweden (Olsson rt til., 1983) and, to a Icsscrextcnt, the United States (Sherman et a1., 19fi2). Some increases in lung adeno- carcinoma risk were also associated with late menopause and with a high estimated total number of menstrual cycles among, women aged 55 and over having a natural menopause. A role of hormonal factors is also suggested by the observation that among non-smokcrs adenocarcinoma affects proportionately more females than males (Lubin and Blot";19g4); and by'the.. findings of estrogen and progcstcronereccptoFS in some lung` adenocarcinomas of women (Chaudhuri- e.t al.; 1982): We discovered no relation to oral' corttraceptives•. or .replacetiient estrogen therapy, but use of these compounds am_ ong .women in the study group was rare in Shanghai. It'seertu unlikely that thc menstrual pattcrns of Chinese womencontribute.greatly; to their high lung cancer risk, but the internal consistency of the trends suggests that future studies of lung'caticsr in China and elsewhere should examine endocrine hypotheses in more detail. This large population-based case-control gtudy of lung can-- cer in urban Shanghai has confrmed that cigarette smoking is a strong risk factor among Chinese women, but only accounts for about one-fourth of all newly, diagnosed cases.. Causes of the remainder are unelear, but occupational factors did itot appear to be important, nor did familial. tendency to lung cancer. Our data suggest, however,' that prior lung diseases, °. hormonal factors, and cooking~ practices may be ittvolved:, Most provocative are the associations with` cooking oil :vola-~ tiles, and further investigations are needed,to evaluate'.-thetr' contribution to the high lung -cancer rates among'Chinese women in various parts of the world. ; AcKKOwtsEDcF..MEnrrs This work was supported in part by the NCI. We thank Dr. To4Czzszoz

inougn tnrsc tests ottcn corrctate witn carctnogcnte potenttat, no bioassay studies have yet been carried out, to r knowl- edge. If the effect of rapeseed oil smoke on lung cancer incidcncc is rcal. the problem is uf great importance to popu-s lations of eastern central China and-other areas of the world Whcre thce oil is often used for cxiking. Chinese rapesecd oil, which is presscd from seeds of Brassictr cattrpesrcis, contains atkiut 50% crucic acid (Chinese Aeadcmv of Medical Sci- ences. tyS 1). in contrast to rapeseed oil with <29o erucic acid (Canbra oil). which was recently approved for sale in the United States (Fcderai Register. 1985). Several studies havc shown that the risk of lung cancer is elevated by a low intake of foods containing vitamin A, partic- ularly as its precursor bcta-carotene (Colditz cr al., 1957). Although reported mainly in Western countries, this associa- tion has also been noted among Chinese women in Singapore tMacLennan rt al.. 1977). Howcver. we found no evidence of - - a protective effect among women in Shanghai, where intake -~ of fresh, carotcne.rtch, dark grccn vcgetablcs is high by world standards. In fact. a positive association was observed between carotene intake and lung cancer risk in females (in contrast to no association in males). We have no ready explanation for . . elsewhere should examine endocrine hy ~ potheses in more dctail. This large population-based cas ontrol study of lung can-, ccr in urban Shanghai has confirmcd that cigarette smoking is . a strong risk factor among Chinese women, but only accounts for about one-fourth of all newly diagnosed cases. Causes of the remainder arc unclear, but occupational factors did not appe:ar to be important, nor did familial tendency to lung cancer. Our data suggest, however, that prior lung diseasu, hormonal factors, and cooking practices may be involved. Most provocative are the associations with cooking oil vola- tilcs, and further invcstigation,r arc needed to evaluate their contribution to the high lung cancer rates among Chinese - - women in various parts of the world. ACKNOWLEDGEMENTS This work was supported in part by the NCI. We thank Dr. . 13.J. Stone, Dr. Shu Xiao-ou and Ms. R. Parsons for comput- ing support. Ms. C. Chen for computational and translation . assistancc, and Dr. B. Henderson for advice and helpful suggestions. REFERENCES B.KRts. G.. MrLOrULOS, G.P.. t:oRCHIk. R., EzDtvu, E., Ro, 1., and 1'owx. B.. Putm.mary scar ia_ rcinunu. Cunc•er, 52, dy3-J97 (1983). BLOT. \i'.J.. and FRAt aEm. J.F.. JR.. Passive smoking and lung cancer. J. nat. Cancer Inst., 77, 99z-1000 (1986). BRESLoN'. N.E.. and DAY. N.E.. Statistical methods in cancer research. The unuih:sis of casa•control studies. L1RC Scientific Publication. 32, pp. 19<-246. IARC. Lyon t 1980). CHAUDWRI, P.K., THOMAS. P.A.. WALRER, M.J., BRIELE. H.A., GUrTA, T.D.. and BFATTIE. C.\V., Steroid receptors in human lung cancer cyto- sols. Can<'er lett., 16. 3?7-33:. (19R2). CHE., T.D_. Mitronucteus test of condensed vulatiles of rapeseed oil. Tumor. 1987 (in press). CHI.-qESE f1C AUl:MYUF MEDICAL SCIENCES . facxW coinpoSttion tables. Peo- I+le's Health Publishing Co.. Beijing (t981). CoLrnTi. G.A., STAMPFF.R, M.J., and Wtu.ET, \V.C., Diet and lung cancera a review elf the epidemiologic evidence in humans. Arch, intern. Mcd., 147, I57-160 (1987). DENG. J., and GAO. Y.T., Prevalence of stnoking among 110,000 adult '_ residents in Shanghai urban area. Chinese J. : prev. <Med, 19,' 271-274 (1985). FEDERAL REGISTER. Vol. 50, No. 18. US Government Printing Office, - -- Washington. DC (1985). FRAUMENI, J.F., and MASON. T.J., Cancer mortality among Chinese Americans. 1950-1969. J. nat. Cancer lnst.; 52, 639.-665 (1974). ,' GAO. Y.T.. Cancer incidence in Shanghai during 1973-77. Nat. Cancer lnst. Monngr,, 62. 43-46 (1982). GREEN, J.P., and BROPHY. P., Carcinoma of the lung in non-smoking Chinese women. West. J. Med.. 136, 291-294 (1982). - zo4,EZZ9zoz