Tobacco Documents Online

t er J: C.wrrr (19le). SL 97_ 103 Relationship of passive smoking to risk of lung cancer and other smoking-associated diseases P.N. Lee,' J. Charrlberlain~ & M.R. Aldcrsorlt /nstiturc of Cancer Rrsearch. Clijlon Road. BBeUnont: Surrey. UK: $uraaar7 In the latter t+an of a large hospital osc<ontrol study of the nasuonship of tyf}c of ugarette smoked'to rssk of vanous uaok,ng-assoaattx7 dtsases. patlents answersd questions on the smoltnb hutwts uf thc r first spouse and on the estenn of passtve smoke exposure at home. at work. dunns travel jnJ dbnn; leisure. In an extens,on of this study, an attempt was made to obtain smoking habit data diraetly from the spouses of all lift*lont non-smoling lung ono:ri cases and of two lifelong non-smoking matchcd controls fnr each case. The attempt was made regardlas of whether the pauents had answered passrve smoking qucstrons in hospital or not. Amongst lifelong non-smokers. passive smoking was not assoeiated' with any s+gntfieant increase in nsk' of lung ancer, chronic bronchitis. ischacmic hcart disease or stroke in any analysis. C.irnttatrons of past studies on passrve smoking arc discusst:d and the need for funhen recarch underlined.. From all the avaiiable evidencx. it appears that any efieet of passive smoke on nsk of any of the trulor disa.es that have been associated with aatve smoking is at most srnall, and may not esist at all. Sfudu of hospital in-patienrs In 1977 a large hospital case-control was initiated to study the relationship of the type of eigarettc smoked to risk of lung cncer, chronic bronchitis. ischaemic heart diseasc and stroke. This study was nrried out in 10 hospital regions in England; interviewing ended in January 1982. The original questionnaire did not include questions on passive smoking as it was not considered an important issue in 1977. However, in 1979 it was decided to extend the questionnaire to covcr passive smoking for marrie4 patients for the last four regions to begin interviewing. Subsequently, in 1981. ,'following publication of thc pape-s by Hirayama (1981) and by Tnchopoulos rr aL (1981) claiming, that non- smoking wives of smokers had a si¢niftcantly greater risk of lung cancer than, non-smoking wives of non-smokers, it was decided to incretse the number of interviews of marned lung cancer cases and controls. The eztcnded questionnaire was then administered to thcse patients imaltihospitals where interviewing was still continuing. Follow•up srr+dr of spouses of nnn-smnhinp hospital irs-paN[nts In 1982- after interviewing of hospiutl in-patients had bcrn completed- it was decided to carry out a follow-up study. In this study. an auempt was CorTapondenwz: P.N. Lec. •Ptaent addrac 25 Cedar Road Suuon, Surrey, SM2 SDG: t,Praent addi•as: ofl-ics of Poputauon Cetrsuses and Sur.eys. St. Cathcnnc's House. 10 Kingsway, l.ondon. WC2B WP. made to interview the spouses of all of thc married hospital ip-pauents with 11tng cancer who reported never having smoked, as well as of two marned non-smoking controls for ueh of thcsc index lung canecr cases. The follow-up study was intended partly to compare information on spouses' smoking habits obtained Grst-hand: with that obtained second-hand during the in-patient intcrviews. and panly to obtCtin some ditta on spouses' smokin;; habits for those patients who had not answcrt:d passive smoking questions in hospital. This papcr concentrates solely on the issue of passive smoking in lifelong non-smokcrs. Results rzlating to type of cigarette smoked arc described elsewhere (Aldcrson rr a1:. 1985), whilc a dctailcd' rcport, avaitablc on trquat from f NL, eonsiders the over-all fandings from this clscrontrol study. MrKtiods and response Study of liacpita!'in-paticnts For each of the 4 index diagnoses (lung cnc::r' chronic bronehitis, isehaemic heart disease ar,L: strole), the intention was to interview 200 eascs and~ 2W matched controls in cach of the eight sex/age cells (i.e. malc or femalc, and aged 35-4-a: 45-C4; 55-64 or 65-74):. This gave a target of 12-800 patienu. though for some etcgories (e.g: young femak chronic bronchittcs) this would be unattainable. Paticnts were sclected' from medical (induding chest medicinc): thoracic surgery, and radiothcrapy wards. Controls were patients without one of the four index diagnoses, individually matched to cases on sex. age, hospital region ar.d, 0 The Marrnillan Pras Ud'-. I9M6

1t1 P.1V. ILF£ rr al whcn possiblt„ hospttal I waJd and time of interview. Subscqucntly, whcn final discharge diagnosa bccamc available, they were used to reallocate cases and eontrols as neccsury. Patients without a final diagnosis kept their provisional diagnosis. Where changes in casc{ontrol status occurrrd, patients were regrouped into new ease-eontrol pairs as appropriatc. With the assistance of Str Richard Doll and Mr Rtuhard Pcto; non-indcx diagnoses wcre classified as follows: class IA 'dcGnitcly not smok'ing,associated' class 1B 'probably not smoking associated' class 2A 'probably smoking,associated' class 2B 'd'efinitely smoking associated' Controls with no final diagnosis were considered class 18 Overall, there were 12,693 interviews carricd out which resultcd in +,950 pairs with class I controls and 7?t* pairs with class 2 controls. Thcrc were 3.832' intcrvicws of married eases and' controls whcre the passive smoking questionnaire was completed In order to avoid substantial loss of data, duc to one mcmbcr of a pair not being marricd or not eomplcting the passive smoking questionnairc, it was decided to ignore matching when analysing the passive smoking data and' to eompare each indcx group with the combined controls. Numbers by sex and casc-control status art given in Table I'. Tabte I Numhers cf nsMed hospital in-patients completing passive smoking questionnaircs Mdr Frrwalr Tord Lung umzr 347 245 792 Chronic bronchitis 182 94 266 Ischacmic hcart disease 286' 221 507 Stroke 161 137 298 t: ontrols l'lass I'A and 1 B' 239 713 U32 Class 2A and 2B' 269 149 417 Total 2-283 13A 9 3,132 'Othcr di.ean were elassifted by degree of smoking aisociat,on - class IA: dcfinitcly not, class IB. probably not„clras 2A probabty, class 2B. dcfinitclX. _ ln the passrvc smoang part of the qucstionruirc, paticnts were asked when the marriage started, if and when it had ended; the number of manufactured cigarettes per day smoked by, the spouse both during the last 12 months of marriage and also al the period of maximum smoking during the marriagc;,and whether the spouse ever regularly smoked hand-rolled QgarTttes, cigars or a pipe during the marriagc. For ser.ond or subsequentt marriages, questions related to the first marriage to give thc longest latcnt intcrvali bctwccn exposure and dts,casc onscn The paticnts wcrr also askc& to quantify, according to a fourrpoint scak (a lot,. avcragc, a little, not at all), the extent to which they were rrgularly exposed to tobacco smoke from other pcoplc prior to coming into hospital in 4 situations: at homc; at work; dunng daily travel; during leisure time. In thc main questionnaire, detailcd' questions wcrc askcd' on smoking habits and on a whole range of possibk confounding variabks. Follow-srp study ojsporeses oJnon-smoking hospital in-porienrs From the hospital study there were 56 lung cancer cases who rcporncd bcing lift:long non-smokers, who were married at the timc of intcrvinm and who were not known to have been married previously. In a follow-up to the main study, an, attempt was made to interview the spouses of thcsc 56 cases and also tihc spouses of t,wo life-long non-smoking controls for each casc, individually matchcd for sex, marital status and 100.ycar agc group and! as far as possible, hospital. Where multiple potential controls in the same hospital were availablc, those interviewed nearest in timc to the case were sclccted Where suitable controls in the same hospital wcre not available, those in the nearesu hospital wcrc choscn. Bccausc namcs and addresscs of the patients were not rccordcd in the hospital study, it was neerssary, to go back to the hospital both to obtain this information and also to get pcrmission to interview their spouses. Following some rxfusals both by the hospital and by, the spouscs„ sucecssful interviews wcrc obtained from spouscs of 34 cases (10 wives an&2d' husbands) and 80 controls (26 wivcs and 54 husbands) whose condition was dcfinitcly or probably not related to smoking.... Interviewing was carricd out betwcen July 1982 and! August 1983; The spouses were asked about their consumption oG manufacturcd eigarettcs, cigars and pipcs (a), nowadays. (b) during the year of admission of the psticnt or (c)i maximum during the whole of the marriaFc: The spouses were not asked about thc smoking habits of the index patient. The sl+ouscs wcrc also askcd qucsuons on agc, occuputron, social class and a range of other potential confounding faetors. Srarisriral 'ntrrhods The statistical methods art based on classical procedures for analysis of grouped data dcrived from ease{ontrol studies (Breslow & Day. 1980). In general, the material has bccn examined as a 2 x A" x S tabfc, with A' representing the kvels of the t t

~.,..~ ..... ~-~ ..r.w~r,..om. ,....aw...~-~........,.,,.. . -., .,rn~... .,... s.,,,,. .... PASSIVE SMOKING AND SM'OKING-RI:LATEf) DISFaiSfS !9 t risk' factor of interest and S the numbcrr of strata used to t'akc account of potential confoundcrs. Results presented are for the combined strata and show the relative risk (Mantcl'.-Hacnsu) estimate) together with the significancc of its dificrrncc from a base level (risk 1.0), andtor the dosc-rclatcd trend. In analyscs "of the data eollectcd in hospital, comparisons arc made bct..rcn oces with, a particular index disease and all thc controls with, discases definitely or probably not related to smoking Six simple indices of passive smokc exposure were eonsidered in these lattcr analyscs, (i)-(tv) exposure at home. at work, during travcl, during leisure. (v)', spouse smoking manufamured cigarettes in the l5st 12 months- and' (vi) spousc smoking manufactured eigarcttcs in thc wholc of the marriage. Bases for (it) are reduced as not all patients worked': In addition, a combined index of passive smoke exposure was nkulatcd' by the unweighted sum of the four individual' exposure indices (i}-(iv), counting 'not au all' as 0, 'little' as 1, 'average' as 2 and 'a lot' as 3. Resvlts. LLung concrr The follow-up study concerned 56 lung eanecr, eascs and 112 matched eontrols who reported never having smoked in their hospiul'intcrvicw. Of thcsc. there were 47 eascs (1S madc and 32 fcmalc) and 96 controls (30 male and 66 fcmalc) for whom some information on smoking habits of their spouscs was availabit. Of these 643 patients, information on spouse smoking was available both from the sf+ouse and from the patient for 59 (011'/.), from the spouse only for 55 (:;R9.) and from the paticnt onl), fur 29' (20'/.). Table II shows the estimated agc-adjusted relative risk of lung cancer in rclatinn; to sriousc smoking during the whole of the marriage„ hy scx, source of data, and period of smoking. None of the 9 relative risks shown in the table arc stnuztically significant. Whcn data fion both sexes and both sources arc considered, the cstimatcd relattvo risks in relation to spouse smoking arc closc to I( I1. 11). For individual sexes or sources, whcrc numbcrs of cases and controls are smaller, relative risks vary more from unity„ but no eonsistcnt~ pattern is evident. Similar conclusions were reached; when analyses were based omsmoking during the year of hospital interview. Here, the overalll relative risk was again close to I(0.93 with limits 0.4,1-2.09). Table lI1 summarises concordance between spovsr s manufactured eiFarette smoking habits as reported directly and indirectly for the 59' patients with, data from both sources. Discrepancies were seen for 9 spouses (IS'/.) in respect' of smoking at some time during marriage and in the crst of 2 T.Wr 11 Relationship between spousr's manufactured cigarette smokinF dunng the whok ofi the marriage and risk of lung cancer among lifelong non-smokers Isiandardised for age) Spr>ru did' a ot' s"r Spnu.sr s+no4rd Ses of Rtlutrtr ri.%L ppcrirnr Casn Conrrnfs• Cavs Conrrois' (93'. IJwws) Basrd on intrrr.rws oJthr slwm.v in fnlln.-up srrd)• (lYl prsretrt)', Malc S 13 S 13 1.0110.23J41) Fcmal6 5 16 19 38 1.6010"-5.78) Combined 10 29 24 51 IJ3(0.50-34h1 Based oe intrrnrws of the indr: /sotinu'in bspital (M prirers) Male 7 IS S 7 1.S3(0:37-6:31), Female 9 17 b 20 0 75t0 R4-2 40) ', Combined 16 32 13 27 1.001041-2Wll based on h.>rh' sovrcr.% nf rronnotinn,(IlJ patirrtts)• Male 7 16 a 14 1.30(0.3l1:1.391 Female 110 21 22 45 1.00(0.37-2.7)) Combined 17 37 - 30 59 1.11',10.51-2.39) '(!nt} controls rncluded in follor.-up study eonsidered: aIn ibis analysis the spouse was countc& u a smoker if nrponed to bc so either diraaly, by the spouse during l01/ow-up intervicw, or, indir>zt1y, by the patient in hospital. Notc thar the 59 patients for whom information on spouse smoking was availabic from both sources are included in a1113 ana1yw& ~

Idt P.N t_r-.f ri al Tabk /1I Cnncnrdannc hn.reen st+nuse's manuracturtd eigarette smoking hahits as repnrted a dtrenly and rndirectiy Sex of patrrnrlcasr consrol saatms A4 a/t Frrnalr Casrs Ca.urois Cnsrs ConrroLs Tord Stwusc a smr+ker somcumc in marri.ge acxnrding tn Subjen and spouse 2 6 5 13 26 Only suh)ect I 0 0 3 4 Only spouse 1 1 3 0 5 Neither 3 11 1 9 24 `/, subrct/spouse agreement 71'/,. 94% 6T:; 88;; >!S% Spouse a smoker during ycar or hospital interview accordrng los Sub1ect and spouse 1 6 2 1 13' Only sub1cct. 0 0 0 1 1'. Only spnusc 1 0 0 0 1 Neither 5 12 7 20 4.4 '/,'suhyw/spousc agreement 96. ; g6'; 100„ 100;; 96i 97•,; spouses (3•/.) in respect of smoking during the year of hospital intcrvicw. Thcre was no eonsistcnt pattern in the direction of dtxrepancy: Table IV' summardscs the results of analyses earried out rclitting 7 indices of passivc smoke exposure recordcd in the hospital intcrviews to risk of lung canccr among lifcir+ng non-smokcrs. Here the controls used for eomprrison arc all never smoking paticnts with discases classified as definitcly or probably not associated with smoking who completed the passive smoking questionnaire. Overall the results showed no evid'cncc of an effect or passive smoking on lung cancer incidcnec among lifelong non-smokcrs. In male patients, relative risks were increased for some of the indiaxs but numbers of cases were small and none of the differences approached statisticat significanee. in femalcs, where numbers of cases were larger, such trends as existed tcndcd to be negative and indccd were marginally significantly negative (P<0.05) for passive smoking d'uring trrvel and during Icisurc. For the combined sexes no difTcrcnccs'or trends were statistically significant at thc 95'/% confidence kvcl; such trends as existed lending to be slightly negative. The relative risk in rclation to the spouse smoking during the whole of the marriagc was estimated to be 0.80' for the sexes combined, with 95•/% confidence limits of 0.43 to 1.50. Standardisation for working in a dusty job; the variabk apart from smoking found to have the strongest asvociation with lung cancer risk in the analyses dcscribcd' in Alderson rt a/. (1985), did not affect the conclusion that passive smoking was not associated with risk of lung cancer among never smokers in our study.. Chronic bronchiris, ischarmic lrrorl disrasr and strokr Analyses similar 1o that shown in Table IV'for lungg cancer were also carried out for chronic bronchitis, ischaernic heart disease and stroke Illustrative results for two of the indices ara presented in Table V. No significant relationship of any index of passive smoking to risk of the 3 discases was seen. For the sexes eombinod, the relative risk in rclation to the spouse smoking during the whole of the marriage was 0.83 for chronic bronchitis (95% confidence limits 0.31-2.20). 1.03 3 for ischacmic heart disease (limits 0.65-1.62) and 0.90 for stroke (limits 0.53-1.52). For stroke there was, in both sexes, an approximate 2-fold increase in risk for patients with a combined passive smoke index that was high (scorr of 5 to 12) compared with those where it was low (score of 0 or 1). Howevcr„ numbers of cases with a high scorc were low (14 tnales and 7 femaks) and even for the sexes oombine.d, the relative risk estimate of 2.18 was not uatistiially signifinnt (limits 0.86-5.t8): In interprcting this fsnding, it should be noted that active smoking was not, found to be ekarly related to stroke in the main study (Alderson rr a1., 1985),, rendering a two-fold inerusc in rclation to passive smoking a priori unlikely.

PASSIVE SMOKING ANt) SMOKiNG-RfLJhTF.n DIS[ASES 1 T.Wr IV RelatMnshrp hn.esn various indi¢% of yssive smoke erctKxurt and nsk of lung eaneer among lifclbna non- smokors Isundardisc.l for agr and, for sf+ousc smoking. .'hcsFicr thc marnagr .a% ongoing or endatl Passnr srw,ir Nfolr parrnes r:fnsvrr mdre//nr! Casri Ci+wrrnls R s Frnn/r parwwts Sr:rs rrawAurd Ccs, Cnetrnts R Cavs Crwttrolt R At home Not at all 9 1101 1 21 192 II 30: 293 .1 Little 2' 21 1.22 6 65 0.92 >< 96 09K Avcragc/a lot 1 I I 1.11 5 61 0.81 6 72 0.86 At .ork Not,at all 3 40 1 12 113 1 IS 153 1 s . Lrnlc 6 29 3.24 3 26 ' 1.19 9 55 1.R2 Avcragc.'a lot 1' 29 046 0 19 0.0 1 4R 0.19 , Dunng travel Not,at all E 101 ' I 28 239 1 36 339 1 : Littlc 3 16 2.06 2 51 0.33 5 67 064 Avctage/a lot 0 13 0.00, 0 13 000 0 26 000 Trend (negative) P<0.05 During kisurs Norat all 3 45 / IS 116 11 l8 161 1 Little 4 49 1.12 14 107 1.05 18 155 1.06 Averagc'a lot 5 39 3.1'8 2 95 018 7 134 0.59 Trend (ncgatrve) , P < 0.05 Combined index' Score 0~ 1 1 27 1 10 - 75 1 11 102 1 Score 2J' 7 55 4.34 5 61 0.63 12 116 1.08 ' Score 5-12 2 15 310: 0 21 0.00 2 36 0.50 Spouse smoked man algs. in last 12 months. No 10 105 1 20 193 1. 30; 298 1 Yes 2 29 0.96 11 122 0.76 13 151 0 79 Spouse smoked man. aEs in whok ofirnarrugc No 7 93 1 13 89 1 - 20 182 1 Yes 5 40 2.47 19 229 0.55 24 269 0.80 'Basod on sum of 0 - not at all. 1'=litllc. 2=averaEe, 3- a lot for at,homct at rork, during travol, dunnE li:isurc. r . Discrssion Over the past 4 years there has been considerable research intrrest in the relationship between passive smoking and, risk of lung cancer in nonsmokers. V1'hilc some studies ha.r claimed a positive effect (Hirayama- 1981. Trichopoulos rr al.. 1981. Correa ri al.. 198?r Garfinkcli rr al:- 1995. Giflis et ol:. 1984, 'Knoth er ol., t98?). others (Buffler rt aL, 1984:. Chan, 1982; Garfinkel, 1981; Kabat and' Wyndcr, 1984; Koo rt at. 1984), have found no signifi,::..;; .,._.., ...~i;. °:!=a.r risks of lung cancer for non-smoking women mamed to smokers comparrd to non-smoking women married to non- smokers range from somewhat over 2 in the Trichopoulos and Correa studies to around 0.75 in the BufTlcr and Chan studies. The wcightcd' relative risk from~ thcsc studies has been estimated by us as approiimataly 1.3. Whilc thcrc is, therefore, a lendcncy for a small positive assneiarion between passive smoking and lung unccr, tmcnt reviews of thcsc dutu (l.cc. 19X4. Lchnert rr al:,, 1984) hhave concluded that overall Iherc is no rel'iablc seicntific evidence or a cJusal' relationship between passive smoking and, lung, r.nccr.. In these rcvxws a numbcr of general points have becn made. First. dosimctric studies have show•n that, in eigarettc-cquiv•.lcnt tcrma, passive smoking only results in a relatively small exposure to the non- smoker. Hugod rt al. (1I978), for example, showed that even under quite extreme conditions the time taken for a non-smoker to inhale the equivaknt of

1!02 P.N LfF rr,at Tahle V Rclatrnnchsp hctrccn, two indrezs or r,+accivc smnke eit+osurc and nsk or chronic brnnchius, ischaemic hcan dtseasc and stroke among lifelong non-smoken (standardised for age and, lor spouse smoksng- whether, the marruge was ongoing or endedl Pa.ccar tww.ie - Ma1F parrrnrs Frnwlr panrnu Srsrs co.ehrnrd [ipncyrr uv/ri/lrtr! Casrs Canrr.dsR Casss Contrnls R Casrs Conrrols R Clrr..nrr hrnnr/trrrs. Com bi ned' mdcx •' Score 0-I 1 27 1' 7 75 1 8 102 1 Score 2J' 2 55 0.83 4 61 1.05, 6 116 1.00 Scorc 5-12 I 1'S 1.90 1 21 1.03 2 36 1.30 Spouse smoked man np- in, whok of marnage No 8 93 I 4 89 II 12 182 I Ycs 1 40 0.34 13 229 IL22 14 269 0.83' ffchorrnlr hrarf dlY4Y Combined mdcs• Scorc 0-a' 13 27' 1 23 75 1 38 102 1 Score 2-4 12 55 043 9 61 0.1.4 21 116 0.52' Score 5-12 3 15 043' 4 21 081 7 36 0.61 Spouse smoked nsan eigs in whnk ofimarriage No 26 93 I 22 89 1 48 182 1' Yes 15 40 1.24 55 229 0.93 70 269 I'_03 Strolr Combined indcs• Score 0-1 5 27 I 19 75 1'. 24 102 1 Score 2-4 10 55 1?4 10 61 0 86 20 116 0.97 Score 5-12 4 15 1.77 7 21 2.44 11 36 218 Spouse smoked man eiFs: in whok of marriage No W 93' 1 19 89 I 37 182 1 Yes 6 40 0.84 49 229 0.92 55 269 0.90 •tiased on sum of 0- not'at all. I - little, 2=avera8c, 3- a lot for at homc, at work, during travcli,during kisure. one eigarctte would be 11 hours as regards particulatc matter and 50 hours as regards nicotine. Similarly. Jarvis rt al:(1985) have shown that the increase in salivary cotininc in relation to passive smoke exposure is less than 1'.e of that in relation to active smoke exposure. Extrapolating linearly from the lafold, relative risk of lung cancer in relation to active smoking would therefore predict a relative risk in relation to passive smoking less than 1.I, while a quadratic extrapolation, as suggested by Doll and Pcto (1978), would predict a lower risk still. The conflict bctwren the dose and the claimed response is parnicuiarly clear for the results of Hirayama (1981) who found a similar effect on lung cancer for passive smoking as for active srnoking of 5 cigarettes a day. Second, all the studies suffer from weak exposure data, most studies only obtaining information on the spousc's smoking habits and, none obtaining objectivc data by mcasurement of ambient levels of smoke constituents in the air of the home or workplace and/or of concentrations of constituents in body fluids. Third. no studies adequately take into aceounv the possibility that misclassifscation of active smokers as non-smokers may have consistently biascd relative risk estimates upward. Active smokers have a high relative risk of lung cancer and spouses' smoking habits are positively, correlated. Because of this, it can be shown that if a relatively small proportion of smokers deny smoking, this results in an apparent elevation in risk of lung cancer in 'non-smokers' married to smokers compared to 'non-smokers' married to non.smokers, even when no rrtar effect of passive smoking exists. A demonstration that this source of bias is of rul~imporvnoe can be found in the study of Garfinkel st a/: (1985). Based on unvalidated smoking data taken from hospital notes, a relative risk of lung cancer in relation to husband's smoking at home of 1.66 was okuiated, with relative risks of at least 1'_3 seen in rclauon to each

PASSIVE SMOKING AND SMOKING-RELATED DISEASES 101 kvel of husband's cigarette smoking and in relation to hushand's agar and pipe smoking When additional sources of information on smoking habits were used, the overall relative risk was reduced to a marginally, significant 1.31 with an elevated risk only rcafl) discerniblc in relation to. heavy eigarette smoking by the husband. Even here, it is notable that the eltvatuon in risk was not evident when smoking data were obtained from the subject or her spouse directiy, but was only evident when the data were obtained from the daughter or son or another informant! i.r. from those people who~ were ksc hkeh.• to have known the fulll smoking history. The IoMrr rtlativc risk may still: have arisen wholly or partly asa bias resulting from misclassification of smoking habits. Fourth, many of the studies are open to specific eriticisms. For examplc, the conclusion ofl Gillis rr al. (I'984), that male lung cancer deathsin, non- smokers rose from 4 per 10,000' in those not exposed to passive smoke to 13 per 10,000; in, those who were exposed was based on a total of only 60) deaths and was not statistically significant. Also the claim by Knoth it at. (1983) of a relationship between passive smoking and lung cancer in non- smoking women was based stmply on the observation that the proportion of female non- smoking lung cancer patients living together with a smoker exoeeded the proportion of male smokers as rcported in the previous microcensus, ignoring irrtrr aha the faa that in manv families women live with more than just their husbands. lmthc present study no significant relationship of passive smoking to lung cancer i,ncidencc in lifelong non-smokers was seen, either in the analyses based on the information collected in hospital or in subsequent inquiry of the spouses or both: It must be pointe6 out. howevcr, tha; the number of lung cancer patients who had never smoked was rather small so that, though our findings arc consistent with passive smoking having no effea on lung cancer risk at all, they do not excludc the possibility of a small increase in risk, though the upper 95% confidence limit or, 1_50 for the estimatc of 0.80 (Table IV), in~ relation to the spouse smoking during the whole of the marriage is not consistent with,some or the larger increases elaimed~ by Hirayama (1981. 1984) Tnchol+oulos tt al: (1981i, 1983) an&Corrca et al: (1983). Though the number of lung cancer patients who had never smoked is small, varying around 30=50 depending om. thc analysis. this number is not very diffcrent~ from that reported in a number of other studies, e.g the findinea of Cortea it al. (1983) were based on only 30, whik those of Trichopoulos it al. (1981). even when~ updated~ (Lrichopoulos tt aL. 1983) were tu~ed on only 77, The difTieulty of obtaining an adequate sampk size is underiined when one considers that in our study the 44 never smoking lung, cancer patients who eompkted passive smoking questionnaires in hospital were extracted from a total of 792 lung cancer patients., It would need a very largc research efTort to iixrrasc precision substantially, and even thcn~ one wouid have to take care that the magnitude or any biases did not excz•ed the magnitude of the efLcct one was looking for. The two major prospective studies which have so far rtportcd findings on passive smoking (Hirayama„ 1981i; Garfinkcl! 1981) ' were not actually designcd to investigate this issue and, as a result, could only use spousc's smoking as an, index of exposurc: Our study, on the other hand, though not able to monitor exposure objectively, as woul& have been preferable, was abk to look ao passive smoking in a wider context, by asking about the extent of exposure at homc, at work, during travel and at kisurs. Although the answers to these questions were subjcctivc, and could have exhibited some bias, their inclusion perhaps allows greater confidence in the conclusions. It was interesting that, of the 59 patients for whom sf+ouse-s cigarette smoking habits were obtained from, both the spousc and the patients, there were 9' patients for whom, there was disagrccmcnt as to whether the spouse had been a smoker at some time during the marriage. It seems reasonable to suppose that some of these were in fact smokers and may have been erroneously classified as non-smokers had only one sourec of information been used. It was also noteworthy that there wasquitc a strong correlation in our study between active and passive smoking As illustrated in Table Vi., current smokers were considerably more likely to be exposed to passive smoke exposure at home (from sources other than their own cigarettes) than were never or es-smokers. As noted above, this conreUion, coupled with some misclassification of smokers as non-smokcrs, may spuriously inflate the estimatc ofl risk related: to passive smoking. It is important to carry out further, studies to obtuin more accurate information, on reliability of sstatcments about,. smoking habits because of this possibility of bias. Little other evidence is availLbic concerning the relationship betwcen passive smoking and risk or the other smoking-assonrtcd diseases in (adult) non-smokcrs and much of this is open to criticism. In his original paf+cr, Hirayama (1981) prescnte& relative risks or death for various diseases for non- smoking women according to the husband's smoking habits. Based on a total of 6& dcaths, a slight positive trend for emphyscma and asthma was not signifieant, whilc, based on a total of 406 dcaths, no indication, of a tren& at all was seen for ischacrnic heart discase., Ima later paper, based on

104 P.N LEf rr a1 Tabk Vl Rclativc nddk of havinF pacvve smoke eiposurc at home awordinE to paucnt s orn manufaoturcd s:prcttc smnk'mg hahMts /9landrrdiscd~ for aEc base - IF comhmed cl.s liand 2'cumroki RrJburn odds 19.R', rnwfiJrnrr bmtul', O.'n.vnw.lurt} hoh f.% . Af a/n Frwralr Never F. I t.2SIO R& 1.111 I 1-2610 A(- 1.1t51 Current •.0012.67- 5.9x 1 2 510 .7a'-3.62I . Chr-squared for trend (2dr) . 57:81 23.34 Y <00011 <0.001 only a further 88' ischacmic hcart' discasc dcaths, Hirayama (1984) rreported a slight: positive trend in risk, but, this was not statisticallj siEnificant. Garland rr a/: (1985), in a small prospective study. rcportcd a 15-foldi highcr risk or ischacmtc heart disease in non-smukrng Califurnian women whosc husbands werc current or formcr smokers compared with those whose husbands were never smok'ars. bunthis enormous and implausihic relative risk was only, significant at the 90`o confidcncc kvcl and had vcry wide confidence limits, being based on only 2 deaths in women whose husbands were current smokers. Sandlcr rr al. (1985): in a casc-eontrol study carried out in North Carohna, reported a strong rclationship~ between risk of cancer or all sites and passive smoking,. This studyy has been criticised by Lee (1985)~.vho notes that it is basicsily ii•nplausib)c that passive smoking should increase risk of cancers not associated with active smoking. Lcc also criticiscd the method of analysis, showing that no association with cancer risk would bc found if a more standard method of analysis was used. Vandcrbrouckc er al: (1984); based on a 25 year follow-up of 1.070 Amsterdam married couplcs, recently reportcd that passive smoking was associated with somc decrease in total'. mortality: There is evidenac indicating that young children whose parents smoke have an excess incidence off respiratory symptoms and some reduction in pulmonary function. Reviewing this evidence. Lee (1984) noted that the interpretation of the association is fraught with difficultics and that other possible explanations, including social e)ass related factors, parental negelct, nutrition, cross- infection and s+....:i::p ,....:.g p:rgnaney, had not been takcn into aeraunt adcquately„ so that a causal effect of passive smoking could not be infcrred. The relevance of these findings to chronic bronchitis or other diseases in adults is iny any case not clear. Our analyscs showed no significant effect of passive smoking on lifclong non-smokers as regards risk of chronic bronchitis, ischaemic heart disease or strokc. ln, all: the analyses relating thc various indices of passive smoke exposure to thcse discases, no significant diRcrenees were seen and slight decreases in nsk were as common as slight increases. Whill more data would be desirable for these discases, lung cancer continues to be the major smoking associated disease for which passive smoking comes under suspicion.. Since all the difTicultles of carrying out good research have eltarly stillinot yet been overcome, furthcr research is certainly needcd. Our findings appear consistent withthc general view, based on all the available evidence, thati any efTect' of passive smoking on risk of lung cancer or other smoking-associated diseases is at most quite small! if it exists at all. The marked increases in risk noted in some studies are more likcly to be a result of bias in the study design than of a true effect of passive smoking. Any views aprrssed' in t'his paper are those or the authors and not of any other person or company. This study was funded by thc Tobacco Research Council (now Tob.ao Advisory Council), to whom we aro most Enteful I> Abcrson was the hober of' the Cancer Rese.rch Campaiftn endowed Chair ofi Eridcmiology at the Institute of Cancer Research dunng the period of the study, dcsiFn and ficld work. Mr. I. Marks from Rrscareh Surveys of Gror Bruain provided adriec in the planninr phase and was rcponsibic for, the iniervic-wers' vital conunbution to the study. We ttunk the many clinicians at the 46 particifuunF hospitals who permitted us to eontact thcir patients and all the patients and spouses who answered the quations. Dr R. Wan6. who hdd a British Council award for the period 1 960-119 8 3; as well,as a number orothcr colleagues provided useful advice at various stages or the study. Mrs BJt Forcy provided invaluable assistance in rarrying out the statutial analysc.

PASSIVE SMOKfNG AND SMOKING-RF.LATFf2 DISFv\Sf,S 105 s . ( Refen.ca ALDF:RSON; M R.-,LEI_ P N A K'ANG. R(1'9R5) Risks of lung anocr, chronic bronchttis: tschaemic hcarf discasc and sirokc in rel6tton to typc of ca6arettc smoked. J., f4+drm Can+n, H1th .,39, 296 FiRESLOW, N.C i DAY. ItL (1990) Srarirnra! Mrrhodi w CanArr Rrsrarrh 1'ol 1- The Analrsu of Cav-rnnrral Sn.drrs. I'nttrnational AEcncy for Resnrch on Canczr;, Lyon BUFFLER, P A- PICKLE. L W.. MASON. T1. & CONTANT. C. (1984), Thc ouses ofi lung onotr in Tesas In [.ac Canrrr Cm.vs and Prr.rnta.n, Mtrtll. M. & Cotrra„ P. YcrlaE Chcmic Intcrnationa Inc CHAN WC (1982). Zahlen aus Hongkong Alrarh. Med. H och.. /24, I6, CORRih- P.. PICKLE. L W., FONTHAM, E.. LIN. Y: & HALNSZEL, W(1983). P.ssvve smoking and lung an¢r. Lanrrt„ki, 595. DOLL R A PCTO. R(1978) Cigarette smoking and bronchial arnnoma dose and time netationshipss among regular smokers and lifelong non-smokers. J. EPrdrm Cmron. H/rh'. 3Z 303. GARFINKEL. L (1981) Time trends in lung cancer monaltty among non-smokers and a note on passi.c smoking J. Narl'Cancrr. /RSr.. i6, 1061. GARFINKEL L, AUERBACH. O[ 1OUBERT, L(1985) Involuntary smoking and' lung an¢r: A aseeontrol study. J. Nar! Cancer /n.u.. 75, 463. GARLAND. C.. BARRLTf{'ONNOR, E.. SUAREZ. L.-. CRIQUI: M H. & WINGARD. D.L (1985). Efkcis of passive smoking on ischemic hean disease monaliiy of non-smokcrs: A prospxtivc study. Amrr. J. Epidrnr., 121. 615: GILLIS.,C.R „HOLE, DJ.- HAK'THORNC. V.M & lOYLE. P. (1994) TThe effect of environmenul tobacco smoke in Iwo urbam communities in the west of Scotland. E+vop: J. Rrsp Dat.,,s5„(Suln+l 133)- 121. HIRAYAMA- L(1981) Non-smokmp wives of heavy smokers have a higher risk of lung on¢r:, a study from lapan, Br. Mrd J.. 282, 1,93 HIRAYAMA, T, (1986), Lung cancer in Japan. effects of nutrition and passive smoking In Lynx Canccr, Cauva and Prr.rnrun M,1sll, M& Corrca-, P. (ads) VcrLg CAcmtc International 1nc HUGOD. C.. HAM'K1NS. Lit & AST~RttP, *(,19711) Ea)+usurc ofi passive smnktn In Inh.ccn, smoke oonstrtuents lnt. Arch. Orny!. iEr+riron 17Irh, 42, 21. 1ARVIS, M.1.. RUSSCI.L. M A H.. F[iYCRARf.ND: C& 4 others (1985)~ Passive aalwsurc to tobaczo smoke saliva catininc mncrntrations in a nerrc.entatrvc population sampk of non-smoking schoolchildrcn. Br Alyd J'.. 291'.927. KABAT. GC A WYNDER. E.L. (191H). Lunl; tJryorr in non,smokcrs fbnrrr. 53. 1214. KNOTH, A, BOIiN, H t SCHMIDT„ F. (1983) Passive smoking as uusc oflung cancer in fcmak non, tQnokcrs. Nrd Alrn , 711. 54 KOO. L.C., HO: 1H:C. A SAW. D(191(4). Is p,scivc smoking an added risk factnr for lung cancer in Chrncsr womcn? J. Exp. Clm. Cancer Rrc, 3, 277. LEE, P N' (191it) Passive Smnking 6n SmolCuee and lJ+r Larx Cumming. G& Bonsignore. G. (eds) Pknum Publishing Corporaunn~ LEE. P.N (1985'), Lifeumc p.ssivc smoking and! ornorr risk. Lt+nrrt, k, 1'4.t LEHNfeRT, G.. GARFINK'EL., L.- HIRAYAMA. T. • 4 others. (19R4). Round tabk discussion. Prrv. Afrd.., 13, 730. SANDLER. D P.- WILCOX. A 1 t[VCRSON; R.B (19115) CumuL.tix eReas of lifetime smoking on cancer risk (onrrr. L 312. TR.ICHOPOULOS. D.. K.ALANDI1711 A.,. SPARROS. L A - M.cMAHON; B(1981). Lung cancer ard' passive amoking Anr. J. Cancer. 27. 1. TRICHOPOULOS. D.. KALANDIDI, A & SPARRI'1S: L (1983). Lung cancer and' passive smoking Conclusion of'Crreek study. Lantrtt ii, 677:. VANDERBROUCKL 3 P.. VERHCCSEN„1.H H.. DC BRUIN, A.. MAURIT2. BJ. VAN DCR HCIDLW[SSCL. C A VAN OER HEIDC- R.M (1994) : AActive and passrve smoking in marticd cnuplcs results of! 25 year follow up. Br. A1rd. l.. 21tR, 10111.

10