Tobacco Documents Online

t.L..u.'.{-lL., ~.~JMJ~I.-/r.n~ I.,i . 2 1. Passive smoking and diet in the etiology of lung cancer among non-smokers Anna Katandidi. Klica Katsoupan,ni, Nells Voropoulou, George Bastas, Rodolfo Saracci, and ~ Diimirrios Trichopoulos fReceined SAprrl 1990: received 6t rtvirtdform 1 Mtry 1990. otcepttd s'~ Afcy 1990) A casa.tontaol study was undettaken in A'thens to tsplore'the role of'paxine smoitinlt andldiet in the causation oflung cancer, b)• histolo=it trpa in non.fmokinit women: Amon)t, Kto.+omett.rith Inn)t canaer admittcd to one ofse+ecn major htYtpitAls in Gnater Athens bktween 1987 and 1989. 254 werc inteeniewed in peeson: of those:interriewed. 91 were life-tonlt non-smokcrs. Among 160 identified oontrots with fraCtures or other onhopedic eonditions. 145 werc inter.ic*ed in person: of those interviewed. 120 wete lifr:-IonF, non.smokers. Mattialte of'a non-smokina .roman to a imokn was aeociated with a reiarire ri..k fbr lon)t cancer of 2.1 (9S'Y• tonfidente interval lC71'1,.1 -1.11. atunberof cixatr!ttt:s'uooked i dajtr bs the husband and years of exposure to husband's smoking were positi.ely;, but not significantly, related to lunE cancer rnstt. There wts no: evidence of anr asociation with exposure to srtwkinj,of other housoh'old'tnembers: and the assodation .vith,exprrstrre to passire smoking at work was small and not atatistinll,r significant. Dietan• data collected through a nemi.quantitative food-frcquenc7 qurctiprutaite indnrated that hjgh oonstunption of fttia wxs inrenely telated to the tisk of lung tarsocr (the rekarive risk betkren eacrense quutiks .ru 0.27 (Ci 0.lI0-4:7Y)). Neithtt .otttables nor any other food group had in additional'prorecti.-e effeot:,furncermore. the apparent protective effect of .elrCtabtbs warw not due to carotetwid ritxmin A coatent and wz• onis partll• explained in tetms of'.itarnin C. The atccociations of lhng cancer ridt +vith p2si.e s+noking and reduced ituit intake were independent .nd did nor confound each othcr: PauiTC smok'init,was icroc+ated with an incteue of the risk of' ali histologii types of nncor. although the ert+ation wqa more, modest for adenocnreinocaa. Key roordr.. Lung cancer. passive smoking. diet. air pollucion. vitamin A. vitamin C. carorcne. Introduction: The association between involuntary exposure oo tobacco smbke and lung cancer was first reportcdlin 1981:°`' by the end of 1989 tmore than 20 epidtmiologic stuciies had examined this tssue.'- i° in abour one•third, of these srudies there were statistically significant positive associations between passive smoking and lung, cancer, risk. whereas in another third there were positive but statistical y non-significant relationships, and in thee remaining third the associations were minimal or non- existent. Overall the association between passive smoking and lung,cancer'is highly signifocstnt and. for practical purpose's, chance can be cxclud'ed a.t a possiblc explana, tion. On the basis of biologic plausibility and epid'cmi- olbgic cvidence, causa'lit,• appears the most likeln- explanation of the empirical association, but this view has not been uniformly aeceptcd.,It has been suggcstcd'! that the association may reflect rnisciassification of ex-smokers among non-smokers (ex-smokers are art inerca';ced risk of lung cancer and are more l'ikelv ro br. married to smokers relatuvc to life-long no'n.smokers). or to confounding effect(s) of unspecified factpr(s)'. One such factorcould' be nutrition (e.g.. nutrition poor in carocenoid sources of vicamin A) sincc passive smokers may be lcss hearlth-conscious than non-exposed persbns. The prescnt studi• was undertaken to examine the role of diet and passive smoking in the causation of lung cancer, ini non•smokers, by histologic typc: Special emphasis was gia•en to the exciLrsion of ex-smokers from anv analvsis concerning lifelong non-smokers. Ma terials antd' methods All women hospitalized' during an 18-month period (t987 - 89)in seven hospirals of the Greater Athens area Dru KaLandeiri; Karror,rovrrrr: Yyro'porrlotr, m.diSrttat an inrhe fltprrrmrnr o,rlly,t•rrnr sftd F,nrdnniologr. L'mrrrtrryo/Arlienr A9e.lrcel Sctioo+: rhuSr>ts: Gnece. Dr Saraca tr ar rhrJrrret+iaurnv.tl iSgrncl Fi•RruenJ~ oe Crneor. Lyon: rnm;c. Dr rrich%.oiv/or rr rh tJc Utpurrratar,o; Eprdrro,orogr. hirrvrad Sehool'of Pa64e Ff,raltb: Rrplrn/ ha1ML!!J sdor/d br rrutt'to Dr Trtiliopcrlos ar ( '1f3 H><aungron.bvrnvc Boston. MA 0?113'. USA, Thr.eserrch war srPporrr/h'~ r3r ln:rrna+iwralAgrar) for Relraroh oo Canctr and tlie Commtrrron o/tkr Etrroliean Conrmunrt;[s. 11 it Aurr of the FUROPASS collab'orarira ECC prvucr 15 ' t'r!r0 ksyid C:-munitauons of (r.tcrd Lid

A. KaGrndedr et al., with a definite diagnosis of'lung,cancer formed the case series. The hospitals irncluded all thrce cancer hospitals in this arca, the onlp hospital for ches diseases in Athens, and the three largest university general hospitals. Womem were included when there, was a positive histologic or cytologic examination or when bronchoscopy was con- sidcred diagnostic of primarv bronchogcnic carcinoma. A iot,a~l of 160 cases were identufoed. Controls were 160 women hospitalited in the orthopedic dtpartments of the sarne hospitals or the nearby hospital for orthopedic disflrdrrsm to which most accident cases from Greater Athens and the surroundit2g,arta are ad'rnitted. Controls were randomly selected from those admitted within al week aficr the identification~of a corresponding case and had' to be 35 years of age: oi over. Among the control women. 102 had fractures and the remaifling; 58 had other traumatuc or orthopedic conditions. All I cases and controls were interviewed in person in the hospital wards, as soon as a definite diagnosis was established, by one of five interviewers who each inter- vicwcd the same proportion of cases and controls. There were no refusals among cases but~ six wece too ill to be interviewed. Arnong controls. 12 were in a condition thac. did not permit imterview, and three refiue&to parvici•pate. !n the intervikws: patients were asked to indicate in, detail their lifelong smoking histories. theirc exposure to passive smoking.-from; their husbands, from i other household members and at work-as we11'.as a number of' other demographic, soc'roeconomic. and medicall characteriscicsi Subjects were also asked to estimate thee average frequency of'consumption (per month. per a~eck.. or per day), before the onset of the priesent discase. of 47 foodiitems or beveragrcategories: These,items weree seiccxed from an extensive list of 120 items, using the criterion t,har the selected items should cover, collectively. more than 80% of the intake of each of the energ.,-gcncr- acing nutrients as well as of vitarnin A. This criterion was established on information from controll groups: in a number of casc-controi studies undertaken in Athens to explbre the role off diet in the causation of cancer at variou s : sites. ri Lifetime exposure to air pollutionwas controlled in the artalysis on the basis of infbrmation about the lifelong residential and employment addresses of all subjects: The areas of residence and work were divided into five categories according to their estimated outdoor air-pollu- tion levela. For the Grcater Athens arca, ait-pollution ltveis by borough were calculated on the basis of the mean yeardy measurements (19'83 - 85) of smoke and :r02- as recorded in 14 monitoring, stations dispersed throughour the area. A line for zero air pollution was drawn'at the highest points of the surrounding moun• tains. For each borough, the calculated air-pollution level wzs the avcragc of the mcasurcmesots of the three nearest 116 stations„ or the tn•o nearest and the zero air-pollution linc: weighted by che inverse of the distance from the borough's center to the measurement points. Boroughs were then divided into fourcategoricst category i being the most polluted with dailirsmokevalues ftequently in! excess of' 400 tegJrn` and category 1' being the less polluted with daily peak-srnoke values rarely exceeding 100 µg/m'. Past residenccs in rura'l or semi-urban areas (population less than 10.000) were considered' as category 0, whereas: past residenccs in other cities of Greece were classified incategories 1' or 21 according to therecordcde or presumed levels of air pollution. Finally. £orev,ery indiwidlutal; acnnse-weighted sumwascakulatcdl assuming 40 hours per week working time for individuals working,outside the home. For housewives, their home residence formed the only basis for estirrtatiiog; their air- pollution expostire: When a11 subjects had' their air- pollistion exposure assessed. they were diszributed~into four groups basedi on the margiz9al' quartiles of the air polluuon.index d!istribution. Since iris possible that the hospital catchmenc arcas are larger for cancer patients than for patients with minocfractures and traumas: the possible air-pollution assodations were taken,into account only in order to controN for possible confounding (generated bv either genuine causal effcczs or through selection forces)L and not for assessment of causality. Among the 154 cases„ 91 had been life-long non~ smokers (less than 1100 cigarettes in theirlifetame); arnong, them, 44 were diagposcd histolpg#cxlly (4S'0k ). 54, cytologically (38'96), and 13' (14'%) through broncho- scopy. Among the 145 conuols: 120 had becn4ife-long non-smokers. The analysis wasconBned to lifez long non- smokets. Three sources of passive smoking exarnined in, the present study were: husband's smoking: smoking of other household'members: and exposurc to smoking at work. Exposure to husband's smoking was considered to start at the tirne of marriage or when the husband started' smoking (whichever came second) andito end when the husband stopped smoking or d'ied, or the couple separated (o+hichever carne first). Change of husband was considcred equivalent to change in husband's, smoking habits, whereas single women were considered as unexposed toi husband's smoking. Ycars of exposure to husband's smoking and& average number of cigucttes smokedidaiJy by the husband were sK,paratelay examined' in the analvsis. Exposure to the smoking of household members other than the husband'wa''< assessed by multiplying the vears a woman lived in each of her homes throughourher Iife, with the.numbcr of smokers in the correspbnding home (excluding the husbandj and bx• summing these productt terms. Subscquently, all women were distributcd intoo four groups: one containing, thosc who had never bcrn, i6l

Lumq oaxctr among non-tmoktra xposcd' to passive smoking from members of their tousehold: andl three corrsponding to the rcrtiles of ncreasing houschold' exposure. FinalH; exposurc' to )assi;+e smoking, in the workplace was caltulated' as thc ime-weighrcd Isum of exposure to smoking at work. the -xposure being based'on the number of smokers among xoplc working in the same closed!space.. From the scmi-quantitatave food-ftequettcy question- zaire. enetgy ' intake and I intakes of vitamin' C and of -icamin A and its constituents (carotene and rttinol)',werc !stiroaced by mulciplxing the nutrient contcnt of the. .clected portion size for each specified food item, by the 'requcncy tlut the food was'consumeds and addibg,thesr rstimates for all food items. Food+cvrnpositson dnta were »aed primarilv on values ob'tained'frorn the L'tliversit)r. )f Massachusetts Nutrient Data Basea' Analyses wcre ~ undcrtaken in'order to examine food consumption rather,tti'an nutrient intakes). Subjects were distributed mairginal quartilcs by'total ftequcncv of consumption -)i food items :belongin'g to specific faod groups (e.g.. rneats. fbuits: and vegetables). Multiple log{stic•regression models were used for the statistical cx3m'ination andlsummarization of't}ic data. In the analrsrs; a core modei'was used which irlclild'ed age ('as a ca'tegorical variable in tentyear gr~oups), years of schooling (quanritativelv)„ and interviewer ('fri•ur indicator variablcs)„ All' confidence intervals shown are 9'5'% intervals. Analyses were done using the Gll3i statistical ps.cicagc (Numerical Algorithms Group Inc.. Relcase jl. 1d?8). Etesulics rzble 11 shoas the. distribution of cases and controls by ,elected demographic characteristics. There ue no 1'ifocanr differences with respect to age,, years: off scnooling. currenr residrncK and occupa'tion, even though t. c variables were controlled for in subsequent multi- vardate analvses. Tablti, 2 shows the distribu.ciorllof cases and controls by selected parameters of exposure to passive smoking. There is evidence that exposure to passive smoking is associatedi with increased risk, bun thee difkncnccs are not large enough to be interpretable a•ithout, controlling for, confounding effects. Table 33 compares the disuibution of cases and controls bv lifelong exlnosure to outdoor air pol0uriom The two distributions: are almost identical'L Finaldv, in Table 4' the distribution of cases and controls by frequency of consumption of specified food groups and nutrients is presented. There is no clear or suggcstive diffe'rence between cases and controls with respect to any, of the indicated nutritional variables. except for cereals (P' - 0,.04). and fruits (P' _ 0.11): The association with cereals is pt!sitia•c but is not biologically crediible. is notsupporoed in the literature. Tablc 1, Distributton of Oll nomsmokinR .romen.rith lung:caneer and 14201 non.smoking compsrispn women bv selected demographic chamccensnes (petcentages in parcntheses) Characteristie Coses Controls Ps • r,'2e < 50 years : 1s (4G.5)' 1'7(14.2) 50 - 59 gean )8 22 (18:)) 0.36' 60-69 yeon 2 % (29.7J: 311(25.8) 70'• vears. 31 (,'34.1'): 50 (M1.7) Schooling <1 year 18 (I19.8) ' 27(22.5) 1-6+xars 53 (158.2), 72 (60.0) 0.42' s• vears 2o11R2.0), 21 (1;:5) CurteM'residesxe Greater Athen! 48 (ISZ.7)' 6'7, (55:8) Other urban I1 (112.11), 9 (7.5) $emi-urban 7 r7.79 9 (7.5) 0.99 RuraJ I 25 (27:5) 3S'(;0:2). Occupation Ehcs emplo,ved' 6;'('3'.671 80(66.7) FBousewife 24 (26.4): 40 (33:3) 0.88' Marit.l status. Ever mamed 8b(91',2)i lo9 (90.8) $ingle 8 (a.8'), rn (9.2) 0.99 'P value for linear trend. Table 2. Distribution of 92 non-smoking, women with lung cxncer and 1:0 non+smoking compuison «omen by,seiecmd pusmezersofexposun to ps.eiive,smoking lpercerstzges in panenthrse!) Charsccetutie Gues Conttols P lAusbaAd's smoking Cigarettesridlg never smoked 26 (2819) 46 (59.7). 1-20 34 ('37.8) 39 (33.6)', 0:t6 211 - 40 22 ( 24:4) 22 411- . 8 (8.9) 9 ('.8), WOsbandis smking, Duration,of exposure never smoked, 26 (28.9) 46 (39:0)', <20 vears ('16:7) 21!(17,.8), 20'-29 yett! 1'5 (16.7) 20 (16.9)', 0:07 30'- 3"l years 1", 718M 16 (12..')', 40 - vtatS 17 (18.9) 116 (13.6)', Other, household exposure None !5 C)6'.7) 26 (22.0), 1:oR+ ( Ibt tertilc) 29 (32.2i 26 (22.0), 0.60 Atedium 24 (26:,') .7 (22.9)'. High (ird tertile) 22 (24.6) 39 0,,01 ' Exposure u work Filouse~ife 24 (27.0) 40 (33.9) Minimal 52 (!58.4) 68157-6) 0.13 Some 13 (I14L6) 10 (8.5)' 'P'.•elue for Enear trend. 17

A: Kslcndidi et ali ibution ot91 non-smokinR.rornewaith lung cancer and T.bie 4. ctsntrnued. s Di t s bk . 3 e 120 non-smakin((,cornpari»n,women by indez of lifelong eaposure to outdts'or ait pollunon (percentapes in plrentheses) Air pollution rndex Gses Controb lut qtnotilt: vet)• low $2'(3'S'.2) 43 (!;6;8D 2ndIquartilc: low 20 (.22.0) 26dR'1.') 3rd'guartik: moderate 18(19.8) 22 (18(3) 4th quartile: high 21 (23.1) 29'(21(:3) P for linear trend  0.99. Table 4. Distribution of 91 non-stnokinR .romen with ilirnF cancersnd 120 npt u:wkin8 compuisots Wonton.by, appn°mnMte raArvml quartslls' or tettiles af,frequt;nov of consumption of specified food ,Srotsps utd nutrients ppereenta(rrs in parentheses) Food group Quarciles P'fot or nutrient 1' (LOrc) 2 3 s,(Hi`h) hneat rrend Cereal! eases 2" (29.")16 (17.6) ' 341(;7,4D 14 (15.4)) controls 0 (3154) 34 (28.3) 35 (2M)1ID (6.3) 0.04' Potatoess cases 1R (1918) 33(58.221, 13 (14.3) control.t 14,(11.7) 2', (22.5) 57'(47M) , 22 (M3) 0.57, Sugars cases 28:(30.8)' 24 (26.4) 26'(28.6'1 13 (1a;3) controls 44 (36.'), 30 (25.0J 3'1 (25.8). 15 (12:5) 0.40- Pubxs 1 cascs 50 (3e.9) 36 (39:6) 5 (5.5) controls 81 ($7.5) 29(2412) 10 (8,31' 0.26 Vegetablcs vascs 2,? (29.') 2: (24') 18(191) 24(264) controls 3a'(28.31 36 (30.0) 29 (24.21 21 p1"'.5i) 0.44 Ftuiis cues 315 (3'8.5) 19 (20.9) 17 O~i~S) 22I(2412) controls 22 (15.3) 44 r36.%i 24 (20:0), 30 (2SL0) 0:11 N1eAts, fisli; eggs cuts 26 (28.6)' 23 (2513') 21 (23.1) 2( (23'.1) , contro:s 39 (32.5), 2: (22.51, 31 (23.6) 23'(19.2) 0.57 Milk .nd milk prodisMS' ctses 21 (23.1) 29 (31.9); 16 (1".6) 25:(2?.5) controls :' (22.5) 32 (26'.') 30 (25.0! ;I ,(2'5'.8110 7.6 Fau and oils caxa :S (22:51 23'(26Ai 33 ('36:3) 9, (q.9) controls ;5 (29.21 30 (25.0) 33 (2%:5) 2S (1$.31 0a66 C;ola eases '0(G6:9) )K'+(11.0) 11 (1Y:11 controls 93 (77.2) 1G'(13.3) . 9: ('.5D 0.44 18 Food!xnoup Quart@les. P for or nutrient 1 (Lowt 3 4 ( HW linear trend Other non-alttsholic bereW. cases 44 (48.+t) 30 (33,01 17, (,1'8.!) control} 69 (J_51 26 (21.7) 2'S (20.8) 0:12 Vitattlitl . A oasss 2!I (25.3) 21123A) 20 (:2.0) 2' (29.'), controls 30(25.0) 32 (26.: ) 26 t21.'q a;+ts. Rerinol t,prtformedl cases 22(24.21 23 (25;31 21i(23,1) 25 (:'.S), controls 30(25.0) 3b (23:01 32 (26.'1 2F (:3'.3I11 0:72. qT.cerocene rarcs 25 (2'.'=5) 1'9 (20:9) 20 (:2:0) :: (29.TJ controls 28 (23.3) :3 (27:1'n 33;(27:5) 26 Q1.7)i 0;68' Vitt:min C caasctt 30 (33.0) 16 (I)t.6) 23 tRS:3D 22 (:.1.21') concrols 22 (i18'.3D 38 ('31117) 29 (24.2) 3'3 125.81' 0-30. Total energy cases 25 (25.3)' 25 (.7iJ) l8 (119.8) 2l'(:".5Y controls 30(2S'.0')' 2" (-7.5'J 33 (29.2) 26 (23.3)', 0.9'3'3 is not particularly marked. andimay well be explained by the multiplicitv of comparisons made:, it was not further explored. By contrast. the negative association w'rth consumption of fruits is biolbgically credible given their high contenr of vitamin C and some carocenoids. and has been found in many other studies. Among,the npn•ascsociations'. total energy intake deserves particular attention becatLxe it indicates t'}lati, in quantitative tCrms. there is no diffcrentin!1 dietary reporting betweenir.ses and controls. The association of lung,cancer with exposure to passive smoking through marriage to smokireg husbands axs flurhrr exarnint:d by multiple logistic regressien, controll'in'g,for age„years of'schooling; and interviewer,. The relative risk (IRR) contrasting women married to smokers with those married i to non-smokers was l'.92 with CI 1.p2'- 3'.59. Thc cfftict of the average number of cigarettes smokedidailM1 by the husband and thediaration, of marriage to smokers were evaluated in two different modcls. controlling for the same core variabl'es as above and' introducing. albcrnatiicely, the dnilr number of cigarettes and the duration of marriage as qtaantitacivee terms interacting with chesmoking,status of the- husbutd' (the 'gacc device).1" s The lung cancer risk arnong non- smokers increased by 16% for cverv 10 3•ears of'exposure to husband''s smoking and by 6e16 for everv Additional pack of cigarettes smoked dxily: These estitnates arc low and statistically non-significant - probably bcrawse, while 1

the smokii:g,status of thc,husband is validh• a.seertained,. the quantitative aspects of the exposure are difficult to a5sess acc)nrateln'. $imilar models were usedl to assess the cffects of exposure to tobacco smoking by other household mem- bers or at the workplace. The results werc qualitacivel}', similu to those presented in~ Table 2. There was' no evidence of any effet:t frorn exposure to smoking,of other houschold~ members. whereas the effect of exposure to : passive smoking, at work was vcry, small'and not statist- irs!!l}y significant (theRR between extreme quartiles was 1.08 (0.24 - 4.87)). Controlling forairpollution.had no effect on any of'the multi-mriate analyses. Table 5 shows multiplclogisuc regression-derived RR'for lung cancer among non-smokers between extrtrme, quanciles of selexted food groups and nutrients. ThaRR atimaoes are adjusted for age, years of schooling,. interviewer, and total energy intake; There is litde; if iy. confounding between the two indicated itz';odgroups (vegetables andl fruits) or' among the three indicaced nutrients. Therefore. the dttta suggest that fruits. burnot vegetables. proteer against 11ang caneer. and that vitamin C alone cannot explain ail ormost of the protective effect associated with fnrit consumption. There,is also evidence chat retinol, (preformed vitaminA). far from being, protective. may actually be associated with an increased risk for lung, cancer in this series. Finalh', exposure to tituband"s tobacco smoking, and fruit consumpcion, a•ere, simultancously, introduced' in another model (together with age. years of schooling, interviewer, and total energy intake) to explbre whether the passive smoking eff"cct is confAunded i by inadequate intake of ftuits, and vice versa. There is no such evidence. In fact. the RR associated with exposures to husbandrs tobacco smoking increased from~ 1.92 to 2:11 and the lative risk associsted with ~high.vs: low'consumptiwn,of, fruits decreasedifrom 0.3'3 to 0.27. Ir,troduction of'cereals the lisc model had' no effect whereas the study was Table S. Multiple )oRistie rcgression-deti.td relati.r risk forlitng cancrr amonE,nonsmokers bc:ween eztreme quutiks of iclctted food .;roups or nutrients Foudieroup or Relative ri4k Confidence P value nutrient berweon extreme qvarciics' inrervalt' 9!Carorene I'.0:' 0.(+4- ).59 096 . RCtinol (pre.`ormed), 1!.31 0:98-1.7' 0:0fi, Vitunin C: 0.6': 0.J2'- 1.0b 0:-0B veretabler Von o:va-r.68 o:80 Fyits 0.3!i' 0.131-0.86 0.02 'Controllrnt[ for aCe: t'ears of s<hoolinR. intorviercr. and toaal cnerFy inuke; "95'v6 Lyng, ca..cer• among ao>r-smoher•s too small to al'low meaningful assessment of interactive, effects: if anv. The lhtsrmodel was also applied'separately foradenor carcinbmas, on one hand, and forsquarnousr small• and largc•call carcinomas. on, the othcr, using , in both in• scances the total'set of controls. The'resu'lts are shown in'Table 6. It appears that thceffecrsof pa,aive'srnoking ue more evident for squamtsua small- and IargeKCil'car- cinomas taken' togpther: than for adenocardaomas.. although the differenee is noo statistdcally significant. On . the contrary. the nu trition al fnctor(s ) usociated'with fruit consumption appeu to be equally strong in both groups. Discussion Three rnajprrepoms have ct,nclwdied'that the existing data strongly support a causal relation between, passive smoking and lung cancer."-i4 There have beenalso more than 10 cpidiemiologic studies assessing thc tolc of nutrition in the etiology of lung, cancer. In a critic3t revicw. Willett" summarized the evidence as being remarkably consistent in suggestiiag an inverse association between carotenoid; sources of vitarnin A and the risk of the disease. Since both exposure to passive smoking; and a diet poorin fruit and vegecable.c. may reflect inadequate health education6 it is conceivable that eachof the two; factors could confound the relation ~ of the other to the risk of lung cancer. The present study suggests that thiss is not the case: the effects of' passive smoking, and' dict appear to be independent. Rcsidual confounding on the basis of a conceivable association between husband's smoking of hig)O tar (rathcr than lbw tar) cigarettes and inadequa'te fruit intake by his wife is unlikely; because high- and low-tar cigarettes,confer similar exposures in the context of passive smoking (tar intake depends primarily on the filter used). Furthermore. che!specix.l effort to exclude ex-smokers from the study of lung cancer among non-smokers provid'rs assurancc that the result s Tabsb 6~ Mvltiplc, IoKistic rc;Rression.derivcd relativc riak (93% confidence inten'aLc)for lung cancer h)' hiirolociccM1'pe amonRpon- smokcrs.according to husband's tobacco smoking status.and to high rr. loww quartile of fruic concumption' Nistoiogr<al tvpcb Husband smoker c's. non.amoket Fruits consumption high v;r. low quanile All lung cancer :,11 (1.09-6'.OS) 0.27 (0!M-0J4'), ildFnoearcinomae $quamous~, small-/ .'.04 (0.98 -.t':241 0.22,(OA'-0.i3) 15rpe-(oll :.58 f0.881- 7.57) 0:2+1 (0.04 - s'.3G) 'CuntrollinR fot:aee. Vears of'schoohngf intervieMer: and total encrgy intake. sFor i!3 of'the 911 cases histolos;ic mpe Mas nou aYailabk. 19

it. , 1Csfirxdidi et al. of passive smoking, do not reflect bias generated from misdassification!of ex.smofters among the non-smokcrs. The present study has advantages and disadvantages. Adl1 women were intervicwc& in person by medically qualified interviewers in the hospital wards; therrwere very few refusal5; and mosr of the lidng cancer cases were histologicalli,v or cytologically confirrned, Howevcr. the study is a liospital.based case•controi irnvestigation. and the possibility of'selection bias cannot be exelhdtd. .t~Jeversheless. the participating hospii:als admit more than one-half of the lttng cancer cases and accident patients hospitalized in the Greater Athens areat cases and controls were very similu with respect to the demographic and' socio•cconomic characteristics examined: the RR linking active smoking to1ung cancer (3.3') in women is very similar to the corresponding cstiinates deriwedi from other, larger studics:'& andl the ai'r-pollutimn findings in the presenc study. suggesting that there is no association beroacen air pollution and lung cancer risk, arc in line with the collective evidence in the intcrnationall lii;erature.'"0'The appropriateness of using patients with fractures and other traumatic conditions as controls m!ayy be~questioncd on the basis of'a postulated assot:i• ation betn+cendierand osteoporosis and thc wcll-knowm link between alcohol intake and'risk of accident: The link between d'iet and osteoporosis is tenuous, however. and there is no evidence thac increased intake of fruit or vegetables can lead to ostcoporosis or increase the probability of a traumatic lesion through another mechanism. Thxrc is also no reason to believe that alcohol intake deserves important consideration in studies of mostly elderly Greek womem and there is no obvious link between high alcohol intake on the one hand. and l ow exposure to passive smoking or high intake of fruits: on the other. The asscxiation bctwccn diet and lung cancer has been exhaustively and critically reviewed by Willett:'_ It appears that physiologic considerations and some results from animal studies have pointed to vicamin A as a possible prosectivc factor against several tnrmors, in particular lung cancer. Most ofthe epidemiologic studies, however, provide no~ evidence that retinoll plavs an ifnportant~ role in the etiology of lung cancer. whcrca-c, che majioritp of tht:sc studics indicated Ithat a high intake of fruits and vegetables is inversely rclated to the, occurrence of the disease." It has been assumed inicialhy that the effect offruits and vegetables could be r.cplained' throughtheir high content of carotenoids. some of'whichf are eventuallv converted into vitamin A_ However. ca'rotenoids may have othrr actions :that arc not shared by vitamin i A - notably quenching sirogler oxygen andd free radicals that could otherwise initiate harmful biochemical reactions like lipid' peroxidacion.29 Gradu- allv, a conFensus' has begun to devclop that catotcnoids' (and' in particular, beta-catotenc) are important in themselves rather than as precursors of vitarnimA. Althoughi this view is ccrtainln compatible with the cm©irical evidence which,strongll: suggests that fruits and vegetables protect against lung canccr; it is not the onlp- crediblt hypothesis. As Willett' and other authors han•c warned, it is possible that other components of certain fruits and vegetables, such as carotenoids' unrelated to vitamin A. or indole compounds."' are the protective factor(s): The f ndings of the present study^ point towuds : this alternatitro interpretation of the overall evidence. Thev indicate that fnlits. rather than vegctables; ue the important protective factors and that their cffect. if real. is mediated neither through beta-earotcne, nor exclus- iveln• through vitamin C. It is possible. of course: that the results of this stud~•% notR•ittistanding their statistical significance. are due to chance. It is also true that the utilized scmi-quantitative food.frequency . questionnaire was limited to on1v 4, f,ood iterns and that the available nutrient database was not specifically developed to represent the Greek dietars intakes: I+ttvertateless, both issues cannot readily explain the pattern of associations seen in Table 5. Furttiermore. these findings are compatible with the rclativelv loww incidence of lung, cancer in the Greek population - a population with the! highest per capita tobacco con- sumption in the world. but with a vcn• high fruit consumption as well.Z'' The findings of this study with respect to pusivee smoking and histologic ty;pearc compatiblr with findings presiously reported.,It is now accepted that active smok- ing causes alli histologic forms of lung canccr but thzc the association is weakcr for, adenocarcinoma.14 Passivce smoking has becn linked to adcnocarcinorna6 " as a•ell l as to other forms of lung cancer.3•° but the association, has been dcmonstrated less consistently for the former.'4•`'` On the contrarn•. the dietarV association npted in the prescnt:studv dbes not appear to depend on r,istologic tvpe., in line with the evidence emerginF from other:'y though noa ald. _ ! previous studics. References 1.: Trithr,poulos D. Kalandidi A. $,paros: L. MacMahon; B. Lung cancer and pa.ssive smoking. lnrl Carcer 19S'11: 27: 1 -4. 2., Hiravama T tion•smokinR wives of'nea.•.• srnol:;'s have a higher risk of lung canccr: rsaudl- frorn)apan. Br.:Ned J 1981: 2!62: 1183 - 5. 3. Trichopoulos D: Pusive smoking and lung canccr. i nc lpscn Lccture 1987. SrxKdJ'Spc Mcr! 1qw IiG "5-9. 4. $aracci R. Riboli E. Pa.uis•c smoking and lung canccr: currenrevidencc and on¢oing studirs ai the lntcrnational rA'gencv ftir Research on Cancer. Marntaon tt'eleanh' 1'989: 22: 11--27. 5'. Lam TH. N:unR 1TM. Wang CM. et al: Smoking. Fusivc 20

Cung c,rncer arno»g non•snro4eri smoking and histological tvpes~ in lung cancer in Hong hona Chinese women. BrJ'Canccr 1987: 56:,6'3-8. 6. Koo LC. HoJH•c. Saw D, Ho C:v. Measurements of passivc smokinR and cstimatcs of lung, cancer risk among non- smoking Chinese fcmales: !nr J Cancer 1987. 39: 162 - 9. '. GaoAZ?: Blot \t%): ZhenY W.,et al: Lung cancer among Chinese women. l>rt J Cancer 11987: 40: 6P4 -1): S. Ihouc R. Hirayama T. Pa,csive smoking and' lung cancer in .romen. In: Aoki' Mi H'isamichi S. Tominaga S: eds. S'ntoiargand Hedritli 1987. AAmsterdam: Excerpta Mcdiea. 1988. 2&3 -3. 9. Gen GY. Lang ZJ4. Zhang AY. \17u GL. On thc,relation. .hip brnveemamoking and fomale lung cancer. in: Aoki A4. Hisamichi S. Tominaga S. eds. Srnokrng, and HealtJr 196•'. Amsterdam: Excerpta Medica. 1988: 483'- 6.. 10. Shimizu H. Morishita M. Mizuno K. eral. A case~cwnrrol studM of lung cancer in inon-smoking women i rohohr< J Exp hlyd 1988: 1154: 389 - 97. " Katsouvanni K. Willett W.11cichopoulos.D. e. aL' Risk of breast cancer among Greek women in relation to nutrienr intake_ Cancer 1988: 61: 181 - 5'. Mictcincn OS! Th'ronrtrcalFiDidtnriology. New York:lViley Aedical. 1985!: 21iG - 4a. ! SL Trichopoulos D. Hs;eh:C., MacMahnn B er al. Aoc at anw binh and'breast cancer risk. lntJ Cancer 1983: 31: 701 - 4. 14. tiationzll .Research Council. Committee on • Passive Smoking. Emniionnrenral Tobacco Smoke: Mearurimg ExpoturrtahdAJsrJtinp Healih Ef,/`acrt. Washington DC: National Academy Press, 1986. 5 Sur¢een Gcnetal. The Health Conrequences of lnvofuntary. Snr.okragr e Report. Rocka•ille: Alarylknd: US Department of Health and Human Services. Public Health Scrvice. Centers for Disease ControL Center foeHealth Promotion and Edbcation. Offce on Smoking and Health. 11985: DHHS (CDC) 8"•83'98: 332. 6 World Health Organization. Tobacco Smoking. IARC Monographs on the Evaluation of the{arcinogcnic Risk of Chemicalf to Humans: Vol., 38; Lvon: France: World Health OrRani¢ation IARC. 1986,. ;. A7i11czt,W, Nutrriioncl Epidemiology, New York: Oidord! E,')nisersitt• Press: 1990: 292'- 310. Surgcon General. Th'r Healrli Conrequencet ojSmo+f'rrrg for Womrnr A Ri;porr. Rockville; Mhrviandt CiS'Dcpan- ment of Health and Human Setvicesl Publie Health Serti•ice., Office of' the Assistant Secreran for Health. Office on Smoking and Health, liS: Government Printing Offce. 1980,: 0 - 32G - 003. 19! Ven.)IE. Air pollutio,i as a risk factor in lung cancer. Am J F/tiaemro( 1982; 116: 42 - 56. 20: Haenszel W. Loveiand DB. Sirkcn MG: Lung canccr mortality as related to residence and smoking histories. 1. White males. JrVCI' 1962: 28: 9d"-100'P. 21. Peto R'. The marked difierences between carotenoids and recinoids: mcthodologicall imp6cations for biochemical epidemiology. Cancer Srrrv 1'983: 2: 257 - 41U. 22. Watcenberg LW. Loob WD. Inhibition of polycyclic aromatic hedroearbon-induce& neoplasia by naturally occurring indoles. Cancer Res 1978: 315- 14101- 3. 2'3'. Sasco A. Da1le-VorYia P. Skalkidis Y. Katsouvanni K. Trichopoulos D: An Evaluation of che Fffcctirerreta of Tobaoco controll tguldtis~e Polickt i» EEC Member Statef. 1949 - '1987. a report to the Commission of the European Communities. Athens: Europe Againsr Cancer. 1989: . 1'4'1 - 64. , 24. Surgeon General. Redycing r+5e Healtb Contequencer of Smoking: 25 Years of Progrerr: a Report. Rockvil'de, 4farvlandi US Depart:nent of Health and Human Services, Public Health Scrxiccs: Centcrs for DiseascControl l Center for Chronic Disease Prevention and Health Promotion. Office for Smoking and Healoh. 1989: DHHS Publication (CDC) 8J-8411. 25_ Brownson RC. RcifJSL Keefe Tj. Ferguson SW. Pritxl JA. J' Risk factors for adenocarcinomi of the lung: Am Enidemiol 1987: 125: ~ 25 - 34', 26. Dalagcr NA. Pickic LCW', Mason T1 et aL' The relation of passive smoking to lung, cancer. Cancer Res 1986: I46c 4808'- 111. 2'. Prrshagen G, Hruber2. Svensson C. Paxsivesmoking and lun¢cancerin Swcdi.ch Nomcn. AmJEprdemiol'1987, : 125i r - 24. 28. \Vu AG; Hcndcrson BE. Pikc MC: er aL' Smoking and other risk,facnors for lung clneer in, women, J.N`C1 19~85; 74: 74?-51. 21