Tobacco Documents Online

;~';~.;....~...~ ...... Ca.r., C+rrn Jud Co...;hr. 1. 15 - 21 Passive srno '' g and diet in the etiology of lung cancer among non-smokers Annz Kal2nddi, Klea Kzcsouyanni, Nelty Voropoulou, Gcorge Bastas, Rodolfo Saracci, and' I3irnitrios T'richopoulos (Received 4 April 1990; received nr rcvised fornr 1 May 1990; .uccpted 3 tti1ay 1990) A taae~mauoi study was undertahcn in Athear to esplorc rhe tok of pauive smoitinj and dier in the auasrion of lung onca, by histologic type. in non•smohiag womea. Among 160 women with iua; cinoet admirtrd to one o.i se.ea major hm*.1 in Gnater Ameas bn.esn 1927 'and 1989, 154 were iaterriewed in person: of tbose inceme+rcd, 91 were (ife-bng oon-aaolters. :Staong 160 idearifud coaauls with ur.vtutes orother oxAoDedic eondiuoas. 145 wcze iauetvirned in peraon; of'thoae inceMew•ed. 120 were life-lon uan smokers. allwa7 rto"o-smoleiri; „, womsn to. s~ker was asoaitnd with a:dati.e rislt fur luag taacer of 2.1 t4S X.°ooqfdeat~ I 1"~.I j;"attmbsr of opreitei smt~ed" _ daiIy by the Dttabsnd md yests o~ei:po.wts to hnsband"t .mokin= we:e positi`dY. buc noe atai~dy. ii~rrd to limg taoaer 13ae ds~ naW no eriden¢ of any ascociatiun with exposute to smolcint of other houtehold rtumben. aad the association wirb espoturo to pasai.e smohirt= at woric wa" amall and norstarisatally sifnitusat. Dicury dsu collocted throust s scmi-qvsnsinmc food-frcyueacy qaeaciuansire iadicued thar higa tCawmptioa offtuits srsr invetsniy selared m the rick of luo= tsaoer (tbe re6uve adt betwera auetae qusrtila wu 0?7 (Q u.10 -0.74)). Ndther regcnbka nor :nr orhae food group Ead an additional peocectis ef-tecs: Furthenaote, the appaxnr protective eFtat of veYtubbM wu nor due to euomaoid rinmip A coarcnt 2nd wss onir parily explained in tsrrna of rinnvn G. 7bc asttoaatioas oC lung oneer ssk wirh pasnve smoking and tacbuced Etuit innke .Ma:e indepradcnr and did aor w,nfouad eacA other. Passive smokin= was uwcnced wich an iqenaae of tDe tit'"oE'aII~hitxotag'e iTiX"s' of csncer, althongh the dsruioo wxi roorr jttodtsr'foi"~deaooovnoma: Key words: Lung cance:, passive smoking, diez, air poilurim viiamin~ A, vitamin C, csroccne. Yntsoduction The association berwcca involuntary exposure to tobacco ex-smo7ccrs among non-smokers (rx-smokers arc ac smoke and lung cancer was first reported in 1981;i" by incressed risk of lung cancer and are more likely to be the end of 1989'more than 20 epidemiologic studies had married to smokers rrlarive to life-long non-smokers), or ed this issur.;-10 In about onc-chitd of thtse to confounding t$cct(s) of unspecified facrot{s). One sntdirs there were stacisucally sign.ificant positive such faccor could be nutrition nutrition poor in sssociaaoru bessveen passive smoking and lung cancer arounoid sources of vitamin A) since passive smokca risk, .vllereas in another third there were positive but may be less health-conscious than non-exposed pcaons. scatistically non-sigaificanc relarionships, and in the The present study was undertaken to examinc the role remaiaing third die associuiona were minisnal or non- of diet and passive smoking in the causzrion of lung ocisteru. Overall the association berween passive smoking cancer in non-smokers, by histologic type. Special and lung =nccr is highly significant and, for pruucal emphasis wss given to the csclu.sion of ex-smokers from purposes, ehznce can be exdudcd as a possible explanz- any analysis concerning llfelong non-smokers. don. On the basis of biologic plaus7biliry and tpidami- ologic evidcnce, causality apprsn the most likely Matezi~ and mCthods cxpEnacion of the empirical association, but this view has not been uniformly acceprcd': It has been suggcszed All' women hospitalizcd during an i8-monch period E thac the usocixcion may reflect misclassif~auon of (1987 - 89) in scvcn hospirzls of the Grcarer Athcns ucs D1l /y7wCt4 JK(i11oUyQWer., WorGfloYJfJY. BCd BGtldl 7Ia in ve DtepJflfAew o/H,ygiene osd Epidinsiolojy. Urirrrtiuy oj'AtE'enf Mrafca/ Sc,b'ool, A'drnt, Greru. DA Surtrca ir o: tdr Lecnusao,ull{grscy for Rasemrn on Ci..ev. LpoA, franoe. Dr Tnc+iopauliu i, is tbe Depsursewr of Epidrmiofeg7HnmJ Scoool of Prb/ic Hral:b. Rrpriur rrqyralr ldould bw rsvt to Dr Tric.oapordoJ a: 63) H1nr~iayror ArcwYr. Biurow, :1Li 02113, USrt.. Tor ru..orb nut ryppo.ud lry tbr Iwtrnt,urua~! .dgrncy for Ruso/Gb ox Crarr =d :cr Commutios of ti.r Flrropeuu CumAwsiurJ. !t u pat of ,r6c FZlROPASS ca!l, borarr:. .fEC p.ojrcu_ 15 0 1710 Rap,d Cc+!am".aswni o( O.fard Lcd

A. lGrlaaa'ua er al. with a dr.'inirc diagnosis of lung cancer formed the case seties, The hospitals included 111 three cancer hospitals in this area, the only hospital for chest disrases in Athens, andthc tiuee largest university geneml'hospitals. Women were included when there was a positive histologic or cqaologic aamination or when broachoscopy was con- sidered diagnostic of priruuy broacEiogenic earrinoma: A rotai of 160 czses were idestiried. ControNs were 160 women hospitalized in the orthopedic departments of the sarnc hospitals or the nearby hospital'for orthopedic diwrde:s, to which most accident cases from Greater rkchcas and the surrounding area are admitted. Controls wce rzndomlq seiected from those admitted within z week aftec the idenriucation of z corresponding cnse and had to be 35 ye3r5 of age or over. Among the control women, 102 had fiactures and the remaining 58 had other traunnacic or orthopedic conditions, All cases and controls were inrerviewed in pcnon in the hospiral wards, as soon as a denniie diagnosis was e:wblisiud; by one of five incerviewea who each intcr- viewed the sarLe proportion of cases and controls. There were no rc:~uals among cases but sis were too ill to be iatc-nieweci- Among coatrols. 12 were in a condition that did' nor permit incerview, and three re:used to paruci- pare. In the inte:viesvs, paticncs were asked to iadicacc in detail their lifelong smoking historics, their cxDosure to passive smoking-froaa tlieir husbands, from othe. household members and at work-as well as a aumber of other demographic. socioeconomic, and mcdical charaaerisda. Subjects were also asked to estimate the avcrue frequency of consumption (per month; per week, or per day), before the onset of the present diseasa, of 47 food' ircros or beverage categories. These items were selected from an extensive list of 120 items, using the criterion d= ttle seiectcd items shouldcover,collectively, more than 809b of the intake of each of the enugy.geacr- sting nucrients as well is of vitarnin A. This aiierion was established on information from control groups in a nurtbcr of case-conrrol studies undertaken in Athens to explore the cclc of diet in the csusacion of cancer ar vwarious sis•.cs." Lifetime exposure to air pollution was controlled in the analysis on the basis of inforrnarion about the lifelong residential and employment addresses of all subjects. The arms or residence and work were divided into five estcgones according to their esurazcrd ourdoor au-pollu tion levcs. For the Greater :4chens area. au-pollurion levels by borough were calculated on the basis of the mean yearly mcasurements (1983 - 85) of smoke and NO.. as recorded' in 14 monitoring stations dispcrscd throughout the area. A [inc for zero air poliution was drawn : lr the highest points of the surrounding moun- rsins: For each borough, the calcuLatcd air-pollution level was the average of the measiucmenc., of .he thrce nrarest stations, or the two nearest and the zero zir-pollution line, weighrcd' by the inverse of the distar.cc from the borough's center co the measurement points. Boroughs were then divided into four categories, category 4 being the most polluted with daily smoke values fscquently in excess of 400 µg/m3 and caregory I beir.g the less polluted with daily peak.smoke values rarely exceeding 100 µgl m3. Psst residences in rura!! or se-zii-urban areas (population less than 10:000) were considered as category 0, whereas past residences in other cities of Grcece were classified in categories 1 or 2 according to the recorded or presumed levels of air pollution. Finally, for every individual, a time-weighted sum was csleulated assuming 40 hours per week working time for individuals working outside the home. For housewives, their home residence forraedthc only basis for estimating their air- poIluuon exposure. When all subjecrs had their air- pollution cxoosure zssessed, they were distributed into four groups based on the marginsl auartiles of the air pollution index distribution. Since it is possible that the hospital czrchrnent arczs are larger for cance: patieZts than for patients with minor fractures and traumas, the possible air-polluuon associations wcre cakcn into account only in order to concroll for possible confounding (gencrated by either genuine causal c$rec:s or through selection forccs); and not for assessment of causality. Among the 154 cascs. 91 had been life-long non- srnokcrs (1<Ss than 100 c1garCtes in their llreIInie); amonq them. 44 were dugssosed histoiogically (48%), 34 cytologically (38%),, and 13 (14%) through broncho- scopy. Among the 145 controls, 120 had been life-long non-smokers. The znalysu was confined to lifr-long non- stIIokem Three sources of passive smoking esa:mincd' in the presear study weze: husband'ssmokiag; smoking of othcs household members; and exposure to smoking at work. Exposure to husband's smoking was considercd to srart at the time of marriage or when the husband scirred smoking (whichever came second) and to end when the husband stopped smoking or died, or the couple separated (whichcvcr esmc firsz): Change of husband was considered equivalcnt to change in husband's smoking habics, whereas single women were considered is unccposed to husband's smoking. Yeaa of esposure to CJ husband's smoking and average numbcr of cigareszc ~ smoked daily by thc husband were separately e:camiaed ~ in the analysis. N Exposure to the smoking of household members othr. (A chan the husband was assessed by mulciplying the year< a woman livcd in each of her homes throughout her life. with the number ofsmokcrs in the corresponding homc (cscluding the husband);and by summing chcse prodlu terms. Subsequently, all womeZ were di5c:ibuted inec four groups; one containing those who had never bee: .G

ocpomed' to passive smoking from mcmbcrsof their houscflold; u1d three corrsponding to the certiles of inc.-essing household cxposurc. Finzlly, exposure to passive smoking in the workplscc was calculxt.ed is the time-wcighced'surn of exposure to smoking at worlt, the exposure being based on the number of smokers aalong people working in the same ciosed space. From the semi-quancitzuve food-frequency question- naire, energy intake and intakes of vitamin C and of vitamin A arld its constituclts (carotene andreanol) were estimated by multiplying nce nutrienc content of the selected portion size for each spcafied'food item. by the frequency thar the food'was consumad. and adding these csrimaces for 211 food iceaLS. Food-composition data were based primaaIy on values obc.lli.ncd-4om the Univeruty of Massachusetts Nutrient I3acz Base.' 1' Anal;vses were also undertaken in order to eslmine food consumption (catne: thin nutrient intakes). Subjects were distributed into marginal'quarsiles by cotal frequency'of consumption of food items belonging to spcafu food groups (e.g., meacs. fruits, and vcgetxbles). Multiple logisuc-regression~models were used for the sntistica examination and summarizyion of the dsta: In the arlalvbcs, a core modcli was used which included age (as a cac soricill variable in ten-ytar grouos), years of schooling ('quznatzcively). and intervicxe_ (four iadicator varisbles). All'I confidence intervals shown are 95 °.'e intervals. Analyses were done using the GLIM stacistic:al package (Numerical, Algorithms Groue Inc., Release s. 1978). Results Table I shows the distribuaon of cases and controls by seleeud'demographic chzrncccrosria. There are no signifuant differeaces with respect to age, years of schooling, current residence zod occupation, even though. etie.evariables were conuolled for in subaequent multi. vatiarc analyses. Table 2 shows the distribucion of cases and'conrrols by selected pzrameters of espasure to pasuve smoking. There is cvidcncc chac ezposure to pas3ive tmoking Is associated with intseased ii.sk, but the differences are not large enough oo be interpretable_ w i r h out concrolling for confounding effecu Table 3 ~ .tompares the d''utribuQon of cases and controls by lifdong exposure to rPutdoor air pollution. The two distributions are ilmost iilcnticaE Fcnally, in Table 4 the distribusion of cases and controls by frequency of consumpcion of specified k'cs.,d zroups and nutaicnrs is presented. There is no cicar or suggestive difference between cases and eontrols with respect to any of the indicated nuairional vtriables,,cxccpc for ccreals (P - 0.04) and fruits (P - 0.1:I ). The 2uociation with cereals is positive buc is not biologiczlly , cediblc. is not supported in chc literature. LYng concsr ~ng uon-rnso.44m Tnb(e 1. Dix~on of 91 aon-smaking wornea with lung nncrr and: 1?0 aoo-smolang cnrapaaswn womcn by uiaccd' dcmographic chuaccerisuet (peresnsages in patrnthesea) CAuaexmdc Cues Concals P Age <50 yaa 1'5 (16!5) 17(14.2) 30 - 59 yna 18 (19.8) 22 (18.3) 0.36 60-69 yeaa 27 (29.7) 31(25.3) ,0• Yeau 31 (34.1) 50 (A1.7) 5eisoo(ing <1 year 181.19.8) 27 (22.5) 1-6 yrsu 53 (58.2), 72 (60.0) 0.42 7 + reaa 20 21 (17.3) Currrac nesdmrr Greua Acuas 48 (52.7), 67 135.8) Odur vrbzn 11 (12.1) 9 (7,.5) xmi-urbnn 7 (7.7) 9 (7.5) 0.99 Rural 25 (27.5) 35 (29.2). Oocupacon E.er employed 67 (73.6) 80 (66.7) H&jwwiie 24 (26.4) 40 (33.3) 0.88: Stirinl sracus 'c•xr marricd' 83 (91.:) 109'(90.8) Singlt 8 (8.8) 11 (9.2) 0.99 'P rnhu for lincsr cead. Tab1e 2. Dism'burion ad'91 non-uanlaing aromen with (ung anca and 120 non-smoking eomparison wromen by seiertcd parsmerea of csPowte ro pau;vc smokiag (pezscncsga in pucnthacs) r'ti+s^•^=IIc. Casa Coaunls P Husband's jmoking e;,rueasiday W'Per smokd 26 (28.9) 46 (39.7) 1-:A 34 (37:8) 39 (33.6), 0.16 21-40 22 (24.4) 2= (19.0) 41. 8 (2.9) 9 (7.8) Husband's rmoting DiIItJO° of °Paw`r never >rooked 26 (28.9) 46 (39.0) <20 ycats 15 (16.7) 21 (17:8) :0-29 rtars 15 (16.)) 20 (16:9) 0.07 So-39 resa 17 15 (12:7) 40 •V 'O 17 (18.9) 16 (13,6) _ ~--- odtir ho-hotd expeaue Nonc 15 (16.7) 26 (22:0) Low ( tsr aenm7e) 2902.2) 26 (22.a) 0.60 Medium 24 (26.7) 2?,(22.9) FiiCh (3td terrilc) 2s (24.4) 39 (33.0). &pmure u work Fiousa+rift 24 (27.0) 40 (33.9) a/ in irnsi 32 (58.4) 6b (57.6) 0.13 Some ti (1C6) 10 (8.5). 'P hluc for Gncu crcnd. 17

.i. K:4rudidl cr a1. TabIF 3. Disuibution of 91 aoo-srriokine worarn with 1un` aoccr and Tabk 4- continued 120 non-smukint comparison womcn by indeY oi liftlong atPo:urc zfl outdoor sir polluaon (paccnagc' in partnrlxses) or avrricnt 1 (LOw) 2 3 4 (High) linC111, Air polfutioa indeu Cx+a Controlt lu qusrtik: verT lo.r. 32 (35.2) 43 (35.0 "ad'.quarulc:low 20(22.0) 26 (21.7), 3rd qtssrtile: modetace iH (19.8) 22 (18.3)' 4th qttartilcr high 21 (23.1) 29 (24.2) P fw 4aesr aend- 0.99. Tablb 4. Distribenion of 91 non-smokirl8 women wirit lua; ancer sad 12o nonsmakin8' comparison wromcn by appn:ciasisc margiml qsattilrs or teaila of frequency of wnsumption of sQeais?ed fboc Ztoaas and nutricnts (pcrecnapu in pareacheses) Food Yroup Quart::es P for or Aucricnr 1 (Low). 2 1(r6gn) Lin- trend. Ceres(s csses 27 (.'9..') 16 (17.6): 34 (37.4) 14 (,19.4) comau)s bi (35:8) 34 (28.3): 33 (27.5) 10 (B.3) 0.04 Fontaes cs:rs , (7.?) 18 (:9.8)', 53 (58.:): 13 (1ti'.3) canctois 14 ('11.7) 27 (222.5); 57 (47.5) 22 (18.3) 0.57 sugzts csxs 28 (30.8) 24 (,'6..5), 26 (28.6) 13 (14:3) contro)a 44 (36.7) 30 (25.0)', 31 (25.8) 15 (12:5) 0.40 Pu11es CL"s 50 (54.9) 36 (39.6)', 5 (5.5) controls 81(67.5) 29 (24.2) 10 (8.3) 0.26 vqcnblrs uxr 27 (29.7) 22 (24_2) 18 (19.II) 24 (26.4) controls 34 (28:3) 36 (30.0), 29 (]4.2) 21 (17:5) 0.44 Fntin car+ 35 (38.5) 19 (20.9) 15 (16.5) 22 (24.2) controis 22 (18:3) 44 (36_7) 24 (20.0): 30 (25.0) 0.11 Meats. fish. .yBs osn 26(23.6) 23 (25.3) 21 (23.1) 21 (23.1) conuols 39 (32.5) 27 (22.5), 31 (25.8) 23 ('19.2) 0.57 blilk and milk products cascs 21 (23:1) 29 (31.9) 16 (17.6) 25 (27.5) controlS 27 (22.5) 32 (26-7) 30 (25.0) 31 (25.8) 0.76 Faa and oils csscs 21 (27.5) 24 (26.4) 33 (36:3) 9' (9.9) controFa 35 (29.2) 30 (25.0) 33 (27.5') 22 (18.3) 0.66 cah ux: 70 (i6.9) 10 (11.0) 14 (12.1) canrrn& 4'. (79.2) 16 (13.3) 9 (i.5) 0_{ 4 Food Qoup Qusitilcs P for pti= non.ylppholic bcseraga aend tsxs 44(48.4) 30 (33.0), 17 (,18:7) caaaois 69 (57.5) 26 (21.7): 25 (20.3) 0.52 Psnmin A mea 23 (25.3) 21 (23.1)' 20 (22:0) 27 (29.7) controls 30 (25.0) 32 (26.7) 32 (26,7) 26 (21.7) 0.4s wool (preiormrd)) cam 22 (24._) 23 (25.3)', 1.1 (.3.1) 25 (2,^.5) controls 30 (x5.o) 30 (27.0); 32 (26.7) 211 (23.3) 0.72 ($.arotene cases 25 (27.5) 19 (20.9)', 20 (2=:0) 27 (29.7) controls 2S (23.3) 33 (I7.5)', 33 127.5) 26 i21.7) 0.63 v'nataia C cssea 30 (33.0) 16 (1'.6): 23 (25.3) 22 (24:.) ;pnuois 22 (12.3) iK (31,7) 29 (24.2) 31 (2),8) 0.30 Total lerurti^y, ca.+rs '_3 (Z5:}) 25 (2.'.5)~ 18 (19.8) 25 (27:5) controls 30 (25.0) 27 (22,5)i 3'5 (29.2) 28 (23.3) 0.93 i:s noo particularly marked, and may well be cxplained by ncc multiplicity of comparisons made; it was nor further explored'. By conrrur, the negzrive sssociztion with consnmpuon of fzuin is biologically csedfble given their high content of vic2min C and some caroteaoids, and has been found in many otlier srudics. Among the non-associacions, cocal' energy intake deserves puticuiar art•encion beC13IICit indicates that, in quancitatave cerau, chcrt is no diffcrcnri3ll diersry reporting berween cases snd controls. The associarinn of lung cancer with exposure to passive smoking rhrough marrisge to smoking husbands was further ccsmined by multiple logistic regression;, conuolling for age, years of schooling, and interviewer. 3?le rdarive sisk (RR) coarrsscing womcn married ro C ~ smobcss with ctiose m2uied w non•smokrrs was 1.92 with C! 1.02-3.39. The edfect of the average number of 1:4.1 cigarettes smoked daily by the husband and'the duration (n of marrisge to smokers were avaluaced in two different ~.l mod'els, controlling for rhe same core vxliables as sbovc N and introducing, altemativel'y, the daily number of (~ cigarettes and slsc duraaon of maariage as quanurnuve rerau interacting with the smoking muus of the husband ~ (the 'gau' device).12'"'lhe lung cancer risk among rloa- smokers increated by 16 % for every 10: ycus of eicpostlre to husband's smoking arld' by 6% for every addiriotial psck of cigarettes smoked daily. These estimates arc low and statistically non-signifianr- probably bccause, while 19~

12 -i@ U .r, •:< ehe smoking starus of the husband is w.tlidly2seertained. the qttannrat.'ve aspects Of the eX=7osuTe Zre difflNlr cD yssess accurately. Similar mode?s were used to uuss the egccts of exposure rc tobacco smoking by othcr household'mem- bers or at the workplace. The results were qualirnavclyy sunil2r to those presented in Table 2. Tberr was no evidence ofany egcct from exposure to smoking of other household masabers. where.s the e5f'ecc of exposure to passive smoking at work was very small and not scuist- ically significzat (the RR between c*=c quartiIcs was 1.08 (0-24-4.87))l Controlling for air pollutionhad no effect on any o£ the multiv,triate analyses. Tablc > shows multiple logistic rrgrenion-derived RR for lung cancer among non-imokers berween e=e=e quartilts of selected food groups and nutrienu. The RR estirnates are adjusced' for age, years of schooling, interviewer, and total energy intake. There is liule, if any. conioundinq 5ctwcrn the two indicated food groups (ve,etabics and'fruits), or smong the three intiicnrcd nutrients. Tncrcore, the data suggest that fruits, but noc vegetables, protea against lung cancer, and cnat vitalnin C alone cannot emlftin all or most of the protective cffect a.ssec7xted with fruic consumption_ There is also eridence that retinol (preformed vitamin ?i): i•sr from beir-g protective, may actually be associated'with an inccased risk for lung czncer in this' series. Finally: exposure to husband's tobacco smoking, and fruit consumption, were siinultaneously introduccd in aaother model (together with age. yesrs of se.yooling, inrerviewe:, and'tocal energy intake) to cxplorc whecher the passive smoking erfrct is confounded by inadequate intake of fruits, and vice versa. Tnere is no such evidence. In fact. the RR associated with exposures to husband's tobacco smoking increased from 1. 92 co 2: ? I and the relative risk issociaced'with hign vr. low eonsumprion of fruits dfcse.cscd' from o. S3 to 0.2', .lnrroduction of cereals to the last modc!' had no effcct aviiereas the study wzs Tabk S. Multiple logiuic rngrasir,n.dcrived re!2rive riskior iuag on¢cr uqong eon•smoken between esaatae quu:iles of srlectec :ood ¢oupr ar nutticnrs Food jroup or nuaienr $clui.e risk becween ccreme quarmica', Coatidrncc incerral" P value lT-Carocene 1.01 0.G-0-1.s) 0.96 Rcrina( (prrturmrd) 1.31 0.18'- 1.-' 0.06 Vir3mio C 0:67 O:d2 - 1.05' 0.08 Vr=ctablh 1.09 0.44-_'_SS 0.86 Fruitx 0.31 0. 13 - 0:36' 0.02 •Cantroti ng inr igc. veara rdscltiaoline„inrervrco.cr. ar•.d eooi cnergy, inrslc'c. e,)spa., l.ursg csnccr srnong Aon-tmo.ten .uU atuall tu itlluw tilr.ttiitigfui Jsy-+Uttcuc uf intcirLurc effects, if any. The last mode! wae also applied separately for adeno- nranomas, on one hand, and for squamous, small- and large-cell cartanomas„ on the other, using in both in- stances the total'set of controls. The results nre shown in Table 6. It appears that the cEccu of passive smoking are more evident for squamous small, and large-cell car- cinomzs ra!<cn togetaer. than for adCnocarrinomas, although the difference is not statistictill~ signiucant. On the rnntrsry, the nutritional factor(s) associated with fruia consumption appear to be equally strong in~ both groups. Discussion Three major repora have conduded diar rile ezisting dac strongly support a causal relation between passive smoking and lung canccr.'e-ta Tacre have been also more than 10 epidemiologic studies assessing the role of nutrition in the etiology, of lung csncer. In a cricical revie.v; WiIlctr=7 sutTlrnarized the evidence y being re•rtarisably consistent in suggesting an inverse association bc:wcen cuotenoid'~ources of virnmin A and the risk of the discase. Since both exposure to passive smoking, and a dicr poor in fnur and vegetabks„ may rezlecr inadequarc healcil cducarion, it is conccivable. that each of the two ficrors could confound the relztion of the other to the risk of lung cance Tne present study suggests that this ~wz~chtc cue;,the effet:is_of p2Mire smoking and dur appear,to.be. indepcadcnt. Residual confounding on che b•asis of a conceivable association betwccn husband's smoking of high rar (rather than low tar) cigareaes and' inadequate fruit intake by his wife is unlikely, because high- sn& low-nr ciEarc:tcs confc- similar e-mosures in the context of passive smoking (tar iaraiie depends primarily on the filter used): Furthermore. the special e$ort to ccclude oc-smokea from the study of lung canca among non-smokers provides 2ssunncc thst the resultz Toble 6. Mclciple lo;iscic tegressien-cerived tslarive riaic (934L ranridence inre<.als) for lung nneer by hisrnlOSie type arttong non• smuiea acrordiut to bmbind's robxco smokios snnu. sad'to high ss. low quartilt oi uuic connunption' Hiswiogid t}peS Hiuband' uaal;er rt: non-tmoker Fruits consumpuon high iJ. low quutilc .ill lung anccr ?:11 (a,09-f..ltil 0.17 (0:10-0.74) Adreonranomas Squmowq stm.l1• i ::Di (0.9tl-S.24Y 0?2 (0i07-0.73) lsrgc-ttll! 2.18 (D.88 -'. 3: ); 0. 24 (0, 04 - t. 96) -CancrollinF for ige. ynn of'scharoling. irttcrvie*'er. and ronl cncr67r innkc. "For e i o>; rhe 41 csses hi>roloKic ~vpr -an nor' a+ausalt. 19

A. Kr~laaaiil:' et a1'1 of passive smoking do not rtsBcct bias gene.'ated from misclzssif caaon of ex-srnokcs among the non-smokers, The present study has advantages and disadvsacagcs. All women were interviewed in person by medically qualified interviewers in the hospital wardy; there were very few refusals, and most of the lung cancer cases wcre histologically or cytologically confirmed. However, the study is a hospital-bzsed clse-conuol investigation, and the possibility of selection bias cannot be escluded. Ncverthe9ess, the pxrt:apating hospitals admit more rhan anc-half of the lung cancer cases and accident paliean hospitali2ed in the Grratcr Athens area; csses and controls were very similar with respccr to the demographic and socio-cconornic characteristics e:amincd; the RR linkiag, active smoking to lung cancer (3.3) in women is very similar to the couesDondusg estimates dcrivcd from other larger studiese" and the air-pollutsion Endings in the present study, suggesting thar there is no associauon~betweea air polllstion.srld'lung cancer risk, are in Gile with the collcctivc evidence in the iaternationai literaturt. `"0' The appropriateness of using patients with fr-.cttues and other traumatic conditions as controls may be ouescioned on the basis of a postulated associ- ation br.ween diet and osteoporosis and the well-known link becwmz alcohol intake and risk of accident. The link bcrncea diet and osteoporosss is tcnuous, however, and there is no cvidence diat increased intake of fruil or vegetables can lead to osreoporosis or inereasc the probability of a tr3umatic lesion through anorher mechar.ism: Thete is also no reason to believe thu alcohoi intake deserves important consideration in studies of mostly elderl,v Greek women, and there is no obvious link bcrwr-n high alcohol intakc on the one hand, and low exposure to passive smoking or high intake of fnurs, on the other. The association between diet and lung cancer has beea exhaustivcly and' critically reviewed by Willclt.14 Ir appesrs that physiologic considerntions and some results from 2aimal studies have poiatcd' to vintmin A zs a possible protcctive factor against severnl' nsmors, in particular lung canccr. Most of the epidemiologic studies, however, provide no evidence thzt retinol plays an important role in the etiology of lung cancer, whercas the majoricy of these studies indi.catcd thu a high intake of fruiu and vegctablcs is inversely related to the oecurrcnce of the disezse.1D It has bccn assumed initially ehat the e{.t-'cct of tnurs and vegetables could be esplzuncd through their high content of carotenoidk. some of which are eventually converted into vitamin A. Eiowevcr, cnrotcnoids may have ocher actions that 2re not shared bv vitamin A - notably qucnchinQ singler oxygen and free radicals char could uthcrwise initiate harmful biochcmical reactions like lipid pcroxidation.'' Gradu- ally, 1LonSensus has begun to develop ttut carotenoids (atad in particulzr, beta-carotcne) are itaportanr - themselves racher than as precursors of virunin~ .f Although this view is certainly compatible with cY empirical evidence which strongly suggests that fttlia sr. vegetabies protecc agairul lung cwrer, it is not the on credible hypothesis. As Wiilett" and other authors ha• warncd. it is possible char othcr components of certai fruits arid vegetables, such as rsroteno;ds unrckted t vitamin A, or indole compounds,'a are the protecti% fiaot(s): Thc findings of the present studV point towztt this alre:native intezprecztion of the overall cvidrsc: They indicate ciaat fzuirs, rathe• chan vegctables, are tt impol protective facrors and'Ihat their effect, if rez is mediated ncitiir_ through bcM-cuotene. nor cxclu: ively through vitamin C. It is possible, of course, that the results of this stud~ notwithstanding thr'r scatisticai signi.ticance, are duc t chance. It is also true chat :he utilized sern.i•quanticaciv food-frequc.zcy queuunnairc was limiced'to only 47 foc Itelns and that the avallable nutncnIt dYtaDase 7V3S nc specirically developed to represent the Grzcs diecu ~ intakes- Nevcrrindcss. both issues cannot readily csplai the pattern of assccistions secn in Table 5. Furtaermom these findings are compatiblt with the relatively to, incidence of lung cancer in the Greek population - population with the hi;hcst per caaica cobacco coc sumprion in the world, but with a vc:v, high fru: consumntion as -Mc11'.'-3 The t"rndings of this study with respect to pasuvv smoking and histologic type ane compatible xith finding previously reported. It is now accepted that active smok ing causes all histologic forms of lung cancer but tisa the association is weaker for ad'cnocucinorna." Passiv scnoking has been linked to adenocarcinoma°•''s'as we: as to other forms of lung clnccr," but the associarioi has been demonstrated less consistentlv for th forma.=s=' On the contnrvl the dictary wociaao: noted in the present study does noc appear to dcpcn, on histologic type, in line wim the evidence emergia. rJ from ot6cr.'s though not all,prrvious studies. O N References ~ t. Trichopoulos D, CCahndidi A, Sparos L. a4acManon B',A Lung cancer and passivc smoking. Gtrr f Canccr 1981: 2- 1 ~ --i. r^ ~ W;re,y+-, T fVnn.cmnlrin¢ wiven of heavy smnkcnt ha Wv a higner risk of lung cnncer: a study, from Japan. Br .4le. ~ f i)81':28.2: 1°3-S. 3. Trichopoulos D. P-43sive smoking and lung cancer. Th lpscn Lecnue 1987. Sund f'Suc .11iJ 1988; 16: 7S - 4 4. 5arscci1 R. Riboll E. Pissive smoking and lung cancec :urrcntcvuicncc ind ongoing nudiascchc InccrnarioM Agency for Research on Cincer. M:iution ARe.remrb 198, '°: 1 li? - ?7.. 5. Lun 'Cl-?. fiung ITm. Wmg CM. et J. Smoking. passiti

Linrg cancer s»rong xon•rmo4srr lv, to ivc ud tot zN Lin -tc, ow -a ~n- Zilt tive ngs .ok- hac sive sdl ion the cion cad ~ing .G B. : 27: have 31ed The •-9. nccr: ional :t 89: smoking and' histoiogical rypes in lung cancer in Hong Kong Chin= women. Br f Gmur 1987: 56: 673 - 8. 6. Koo LC. Ho JH-c. Saw D. Ho C-y. Mezsuremcats oi paaivc smoking and cstitnxcs of lung cancer risk among non- smokin; Cnincsc fcmala. Grr f Cjxcer 1987: 39: 162 - 9. 7. Gao Yt. Bloc WJ. Zheng W: er al: Lung canca among Chinesc women. lAt f Gncrr 1987; 40: 604 - 9= 8. Inoue R. Hinyama T. Passive smoking and' lung canca in women. In: Aoki M. Hisamicai S. Tominaga S, eds. SnsoExg mrd Hecltb 1987, Arasterdam: Escerpn >ytedlca,. 1988 ~ 283 - 5. 9. Gcn GY. Liang ZH. Zhang AY', Wu GL On the relar:on- ship bcrwcen smoking and female lung onccr. In: Aoki' b(. Hisamicni S, Tominags S, eds. Ss+okiag asd Htultb l987, Amsterdzrn: Excerpn Medicz. 1988: 483-6. 10: Shimizu H. Sforishiu'd,Mizssno K, et a!. A ase-centrol ssudy of lung cancer in non-smoking women. Toho,iu J Ezp bfad 198l3; 154: 389 - 97. 14. I:aevuyanni K. W'rllezt W„Trichopoulos D. er a~ Risk of bresst cancer among Greck women in rolation to nucricnc intake. Gracrr 1'983: 61: 131 - 5. 12. Micrdncn O.. Totorrtird L'pudrmcclogJ. New York: Wilry. ?rf cdicsl. 1985: 216- v4'. 13. Trichopoulos D:,Hsich,C..'sfacylahon B, et al. rlgc x any birth and bnsas cancer risk. 1ar f Caccsr 1983; 31: '01- 4. 14. Adnzionxl Researrh Council. Commictec on Passive Smoking. inviro,rmenca.r Toeacco, Smo.4.e: MeancriAg Fxposurer arrdAuruing F1.n1rb Efficrr. Washiag:on DC: National Acadcrnv Press. 1986. 15. Surgeon Gcncral': The Heslth Conreqrnneii of lua.tzacrp SmoRiny: . Rtoon. Rockville. Matyland: US Deputmcnr of Health and Human Scrviccs, Public Hoalth Sc:vicc, Cearea forDiseasc Contro6; Center for Health Promoeion and Educuion. Office on Smoking and Health. 1985; DHHS (CDC) 87.8398: 332. 16. World Health Organizacion. Tobacco Smo.Eeicp, LARC Monographs on the Evxluacion of the Cattinogenic Risk of Chemicalt to Humansc Vol. 38. Lyon. France: World Health Orgsnuaaon, LkRC, 1986. 17. Willcct W. Nluiriona! Epidrirr.ology. New York: Oxfocd L`nivcrsiry Pms, 1990: 292 - 310. 18: 5ucgeon General. The H'talrb Corrseqsrewcer oJSmo"g for WorAen: A Reporr. Rockvilld, Maryland: C.'S Depart- mern of Hcalch and Human Services„Public Health Scrvice. Office of the Assisnnt Seurnry for Halrn. Cfficc on Smoking and Health, U5 Gvvcmmcnc Ptincing Office, 1980:0-326-003. 19. Venm JE. Air pollution as a risk factor in lung cancer. Ars f F,pid.mio! 1982; 116: 42 - 56. 20. Hacas:ei W. Loveland DB, Suken MG. Lung cancer mortalicy u rcl2ted to,rtsidcncc and smoking nisaoars. L White malcs. fNCI 1962; 28: 947 - 1Q01. 21. Pem ~R, 'Ihe marJeed'differences between carocenoids and n:dnoids: methodological implicirions for biochcraid epidemidogy. Csncer Sirn 1983; 2: 237-40. 22. Waccenbcrg LW,, Loob WD. Inhibition of polycyciic aromacic hydroestbon-induced neopl:uia by natunlly occurring indblcs. C:rrcer Ror 1978: 38: 1410 - 3. 23. S'iaYo A. Dalla-Vo ,ia P. Skalkid'u Y, Kauouyanni K, Tric3sopoulos D. aa Epalriarfor, of <6e Effusrsaresu of Toboeco-co:uro!lsgizlasiae Policiu in EEC Alember Ststes. 194 -1987. a report to the Commission of the European Communities. Athens: Europe Against Cancer, 1989; 1:,° U - 6+. 24: Surgeon Genrzal. e?educing rnt Heslrb Coruequcnces of SmoRr»g: 2) Yearr of Progrerr: a Rrporr: Rockvillc, A13n-iand: US Depurmenr of Htalrh and Human 5ervices. Public Hr.arh Scrviccs, Ccnrasfor Disc.x Conuol, Cmrcr br Chronic Disasc Prevention and Health Promocion,. Office ior Smoking and Hcalcb, 1989: DHH5 Publkuion (CDC) 89 - 84 11. 25. Brownson AC, ReiiJS. Kccfc 1J. Ferguson SW, PiirilJA. Risk factors for ad'enoarcinoma of' ehe lung. Am f Hoiderr.io! 1987~: 125: 25 - 34. 26. Dalagvr NA. Pickir LW, Mason IJ er ei. The rrlacion of passive smoking to lung cancer. Gnc.r eZes 1986: 46: 4808 - 11. 27. Patshagen G. Hruber Z. Svensson C. Passive smoking and luag mncer in Swcdish.womai. Anr f fPJJimiol 1987; 125: 17-24. 28. Wu AG. Hcndcrson~BE, Pike MC, rsat. Smoking and oshcr risk raccora for lung cancer in womicm fNCt 1985; 74: 747 - 51. I. L N. sssivc 71