Tobacco Documents Online

.trMu,. Jouar.a oe Er+oce.1ot.ocr Vel 153, Nb 2 t'opynght C t9M by'lae Joana Hopk,ns usiv.»ay Schml ef Hypeoe aed PuWic H..kk hawtd in US:A A1l rKhu nwr+.ad Letters to the Editor RE: 'PASSIVE SMOKING IN ADULTHOOD AND CANCER RISK' lt is unfortunate that Sandler at aL (1) did not obtain information on, or aceount for„addiuonal pos- sible confounding variables in their study. 'I'he only onu that they repon having included were ap. ses. :m, active tmoking, eduation, hlw colitr vs. white oolinr oavpation, and parental smoking. We have lound that the amount of paaaive smoking wu corr+- latad witM self•rtport.d esposure to oavpational bas• ards, tw of trurijuana, and akrolwl intake (2): 'EZtus" for example, the increased risk of brea.t cancer ob- wrred by Sandl.r et aL may be due to incrtna.d t cohol intaYe (3) rt+ther than to passive smoking. ~erhapa ezpoaws to octvpational! hasardl ezplains he slightly increa.ed risk of hematopoietic cancer (4) ong passive smokers, and husbanda' sesual behav- ar auociat.d with the smoking habit (51,accounu for in inereas.d risk of tirvical eitncrr.'The authon a1- vded to this last possibility, but did not rule it out. Sandler et al; (1) were careful to point out that the essociations that they found'... might rtlate to other fact.ors we have not measured or to d.ficienciss in atudy tirsign.' However. t.bey went on to state that they 'have not been able to identify a poasible con• foundar .:. that could have caused the difference in smoking patterna of spou.cs between cases and con- troia.' Tbe above suRestiona repraent a fw posi'- biliti.a. Saadltwr ar al. also pointed otu that sidestream smoke has higher tonesnustiota of certain eareino- Eens than mainstream straks. N.vertAelew because (a) the dosap of smoke ic so much lower in passive than in actire smoking. and (b) smoken also `pu• ' sively' breathe sidestresm smoke in addition ta inhal- ing mainstream sawke„ it would be surprisin= if pas- sive smoking aetually caused any coocen th.t were not associated with active ueokin=. Rssfltsrcrs L 5andler DP. Everson RB. INilbss AJ. Passi.e seiokin4 is adulthood aad osricer ri.L M J Epulemvol 1liS:121i3?- ea I i Feiedmas CD. PetiW VB tlawd NtD. Pre.+lena and eure.taue .f paesi.e asokiea: Aaw J Public H..ltb Ixi:n;.ors. • i H;istt Rrl, bswe/ 1tD. AStolislic le.erase mnwuaptioa and be.ast ta.wr i.eidrat. Ar J Ep-d.awt 146s:Mi%6•4J1 r 4. D.roufle t. Oaupation. ls: Schottenf.ld D. Frsura.ni JF, Jr. ad. Gncer .p.drv+etop awd' Or.»++uos Pl+iada. Phia WB -,;aund.» Coinpanr. 19e23111-35. f, buckl.y JD. H.rris RWC. Doll R et sL Ca»-conuoi Kudy of the Au.band. ef womes atb dyspia.+. .r oucs. soms of tlr arsa wri I'..r.cet Tl61'2tG10-S. Gary D. Friedman Deanrv,wnt o( .Vedical MerAodi 1Resenrch JKaiaer Permonrrtu Medu ol Care Prodram Oe,kta+d CA 9e677 . RE: 'PASSIVE SMOKING IN ADULTHOOD AND CANCER R1SK' Sandler et al. (1) removed much of the impropriety 3f their demonstrstion of an effect of passive smoking =enerally in raising the risk of cancer by focusing sartitvl.r attention on the i4+stunces in which the study csae was a nonsmoker. But subauntial impro- Priety r+maina. Finu, it gives cause to wonder bow passive smoking could' pve rise tc a doubled rate of cancer generally, including cancer of sites not previously associated with 3'irect smoking of cigarettes. Those specific aites are .only identified in their table giving resulu for non- .otok'ers plus smokers combined, with the added risk ior passive smoking then shown at 60 per cent '17te :n+de addkd riak of lung cancer frorn passive smoking. srnokers plus nonsmokers combined, is shown as 90 -jer cent, but bassd on only 22 ca.ses of lung cancer- nf whom only two were nonsmokers. So only two .sonsmokinR lung cancer cases were available for judt- ag the true effecu of pasa,.e smoking on,lung cancer. Else+rhett 1'21, the study of paasive smoking haa focused specifically on lung cancer and specuically among individuall who themselves were nonsmokera, Le, the nonsmoking wives of smokers. and, even so,. such studies hove besn subject to criticism (3). In Sandler et al. t11, the authors have rven oaten- ab{e coruern for puuenuel coniounding factors. Per- haps othern would agtee that confounding facton have been taken into account, but not in my opinion. The i.sue of what is a confounding f.etor bas become so confuaed'that I have hesitated to get into the fray. I juat took to aee what is sensible. (lacidentally. the saa+e is.ue of the lournot does carry an itum M on eion(oundin=). What the authora (1),treat as confounding factors are are (in broad aie caurones)„ ses, race, smoking (not relevant for the analysis on nonsmokers). three broad education gtoupa, two broed occupation groups. and wFietAer either parent stauked Presumably, the factors adjusted for simultaneously in the analysu of passive smoking among nonsmokers ue age• race. sez. and eduution. But, somehow. I do not think of ap as a confound- ing facwr. Rathetn it is a lorical'fsctor which must be taken into aceount in any neasonabli analysis if one wanu to come up with rRa.onuble resulta. 'Much the satne is trtx of race and ses. though I am leu certain about level of educstion. A truly eonfounding factor to my mind would be something like smoking and altrohol eonsumption. I'f, heavy dnnken are also heavy smukers, wninK out their separate eifecu could be a problem. Any-ay. if we an concentrating on passive smuk,nr among nun• 3r

36$ 1.bTTE(ZS TO 7?tE [DITpR amoken, then we art sontrutin6 familin with a sin6it smoker aYainst families wherw neither spouse smokes.. And differences could e:ist between them. Thus we might aak: 1) Art their alcohol eonsumption habits the satru' ?) W hat an the differences in their dietary dabiu' 31 How might they difTar in age at first inur- eourse, frequency o( intrrrourse„eircumcikion stwtus of husband, etc.? 4) Are numbers of children or prtg- nancies and a6es at fint birth or pr.gnancy the satne for sinEle-smoktr and nonsmoker families? 5)Wbat true difTenrues may e:ist in occupations ltading to cancer? These po.sible corralates of smoking practice might arcount for the observed'difQreneas reported in an- cern of a variety of sites„ and these an the true potential confounding faeton, whether or not we are in a pwition to get at.them. What Sandler etal. have uken u potential confounding facton artr just the straightforward fatton requiring adjustment that any nasonable epidemiologist would employ. One partitvlarr thing strikes me about the Sandler at al. study. Passive smoking has been shown to double the risk of' all ancen among nonsmokers, albeit true confounding facton have not been taken into account But stud.es focusing on lung cancer aaiong nonarnok- an consequent on passive amokint have also eome up with about the same doublin6 of risk. To other enu- eianu m.de of such other studies may be added that tbey may be overlooking true confounding factors. The trut criticism of passive smoking nemain.-it is an unpleasant burden to impose on the nonsmoker. R=scttvress l. Sandkr DP;,Everson RB'. M'ikoa AJ. Pauive smokmr in adulrhood and cancer nak Ae J Epdeauul lttdl.t2(x .-4d 2 Nir.yarea T. Non-anwkang wives af heavy seaoken have a luther risk ot lunW cancer. a study (rom Japan. Br Mnd J' 1i6t2>l2I63-S 1 Mantel N. Epdemioloiric invfuiptiona-esn in toedun. nn in analysis and care in nryantn6 J Cancer Rn C6n Onco1 1M:103:1 t3-16. l: Boi.in,J~F, M'.cholder S Conditan& for tunfoundinr o( the nak eario and of tlie o6E& ratio Aae J Epidee+.ol tf6S.121 a 52 -6. Nrthan'Mantel AfatAnrtoticr, S'tatutio ond Corrtputer Science The Amtricon IJnitxrsily BctJrrdo, AID?OBl1 RE: 'AASSIVE SMOKING IN' ADULTHOOD AND CANCER RJSK' It is ironic that whereas a few of us atonise over the question of whether active smoking is a major cause of lung cancer othen seem willing to believe that passive smoking causes an appreciable incidence of lung and other cancers. Sandler et al. (1) purpon, to have studied arxer risk in relation to 'pauive smoking in adulthood_* They did rwt. They studied cancer cases in actave and' twnacuve smokers in rolation, mainly, to•the smoking habits of the spouse. Two bto.d aspects were cotuid- ered whether or not the spouse 'smoked n;ularly ao any timt during their mar3ia6e*; and, for t3oaee-rr aponae relationa, 'the number of years of marriage durin6 which a spouse srnoked'c' or 'tlie average amount smoked by sDouat.' Also„ the 'nonerposed group consisted of persoru married to non.mokerrand p.rsona who never married.' Tb. smoking habita of the spotse ars, however, a very poor surrogate for passive e:po.ure to cigarette saake. Thus, Repace and T.owrey (2) have estimated that the typiul' workplace eaposure ut the United States is about four times the tyyialie:posure in the home. Mat.ukun et al. (3) found in Japan that the mean (t SE), urinary, eotinine eacretion for 200 non- sa+okrrs living with no smoken in the home was 0.51 s 0.09 rt/in[ of creatininr for 272 nonsmokers withh smokers in the home it was sli`htly hi6her, at 0.79 s 0.101(same uniu), butnotsi6nifieantly so. The assumption by Sandltret aL thaithe smoking statusof the apou.e is a uaefulitneasure of esposure to ambient cigarette smoke u, therefore, highly suapett. This doubt is borne out by their own study of lung ancercases in which'. .. there was no apparent do.e- response using either yean married to a smoker or average amount smoked by spouse u the measure of do.e ...' (1). The observed cancer aasociationa were with the status of the spouse and not with e:posure to cigarette smoke. That uncorTected confoundinr fac- ton might have been present is further suuestrd by tbeir finding of'sutietically significunt' assocutions 'with several specific tumor sites ... including wnx which are not ordinarily ttprdedau srnoking•rel.ted' Asaorvtive mating is a wellleswbli.hed phenome• non that.e:tends to the smokinll habit (1) and henre we hr<ve to considbrthatthe underlying factors leadinrc to tlie choice of amokin6/nonamukin6 apouae tnight asaociatt with the riak of certain cancers. la principle, randomized trials can of course distin guish between auul„constitutronal, and cauuLplus eonatitutional hypotheses of aa.ociatiuns but they have yet to be conducted in eonnection with paurve smokint. However, two itrtpnrtant studits have been carried out; with rttndotni=ation„ in the conte><t of active smoking (5. 6):and, because active srnoken are inevitably pas.ive satokers u..ell, the rewlu of these trials are of immense imponanee not only to the Undinp of Sandler et ali but also to general the.es about the arcinogenicity of mainstream and side• uream tobacco smoke. We obtain the beso available direct epidemioloer test of causal hypotheses by combining the re.ults of the MRF1T in the Unitrd Sut.es (5)wiih those of the Whitehall Study in London (6). In the combined, low- amoking intervention groups some 56 cases of lung ancer/deaths in MRFIT, deaths and rtgi~trations in the Whitehall Study) were recorded in the twnllentry group of 7.142 men„a frequency of 0.7A per cent F,.r the combined, relatively, hi=h-amokinr 'uau.l curr' groups the corresponding numbers were 53 in 7.169 w 0.74 per cent,. Findings for all cancers other than lunr ancer„aevenl.ot which associate atrunt<ly' wrth utn- rettt srnokrnK, sre auunishinl;. Sume tiR caN+. ur,. 1'._a

Iarr7:RS TO THE EDrran per cent, were r.cord.d in the combined low-amoking intervention groups but only 60, or 0Bi per cent, in the high-amokin f usual are froups. 1s smoking pro- phylsctic' Does quitting smoking result in changes of Gfe-styk Ihat cause cancer7'Or is this finding a.iki but accepuble, random sscunion! We may not be entitled to estimate confidence IimitJ from this post hoc scrutiny but the rtsulta of theu metFiodolopcally reputable nndoreiked triaL cast grave doubts on the validity of orthodox cLints about the hazards of smoking. These claims derive, in the main, from caaes.ontrol and prospective studies of aelf,selected smokers, si-amokrrs and oonsmoken which, by their very nature, can tell us much~about association but nothing about cause. The study ot Sandler et at (1) does noceven tall us about aaaocia• tiona with pasaive amokinr it does, however, s.had some inttresuns lijhton cancerasaociations with the sa+okint/nonsmoking wtus of tl+e apouse. ltwsu"csa L Sandler DP. Everson RB.,Wilmr AJ Parive wokint; in adultAood and cancer ri.J< Am J F-0~demiol 1965:1Z137- k THE AUTH'ORS REPLY Friedman (1) and Mantet (2) have expressed some concerns about our pauivaamoking findinp that ar* concerns for us as well! (3). We agree that it would have been useful to have information, on a greater number of potent,al confounding variabllra Uhfortu• nately; when conducting a study of many cancers, it is diff;cult to do justice to individual sites because risk factors vary widely by site. While the possibility of confounding variables es- iata, some of those suuested by Friedman and Mancel ar: not likely to explain our findinp.. We collected information on alcohol consumption, and adjustment foralcohollintake did not alur our results. Msrijuana use is also not likely to be a factor, since most of our cases were older than t0 yeara of age and from largely poorand rural,areaa Occupational exposures may be importa nt, although it is difficult to see how occvpa- tion of the subject would be aasociatsd'with spouse's smoking, other than as an indicator of social claas. Simple tantiftcation into white and blue co(lar occu- pations did not change our results (3h and thert were too few individuals with the same job for a more detailed analyaia to be meaningful. Information on reproductive factors would have been useful and should certainly be included in subsequent studies given the observed aarociauoaa with cancen of the breast and cervix. Since many of the factors thought to increase the risk of cervical cancer are thought to decrease the risk of breast caneer, however, it is not clear that thoae factors could explain our finding of elevated risk at both sites. Sexual behavior of hus- bands may also be important in cervical cancer ruk, but reports suggest that this does not entirely explain tAe observed asaociations with spouse smoking (4„ 5). Burch points out that the smoking habits of the spouse mav be a poor surrogate for passive smoke etpo.ure (6)i,Certatnly toul paasive smoke eaposurc 0 all be under.stiroated as Friedman's data suggest (7): However, Friedman's data and others suggest that individualswho live with smokers are likely to be more exposed to other people'a cipretta atnoke-either be- cause they ars more toler.nt of other peoplt'. smoking or beau,e t.tiey a{e often e:pwed' to other smokers when in the company of their smoking spouse (7, S): Thus while smoking by apcuse may only roughly re- flrct the amount of tobacco smoke e:posure. it appears to diitinpish the most exposed from the least e:- posed Misc.asaificauon of e:posura u likely to dilute any asaociation nther, than create a spurious asaoci- ation, unless only controls who are married to non- smoken sre also exposed at work.,Furthermore, one fourth of our cases and controls (3.S per cent of the females) wen either housewives or unemployed, and thus likely to receive most of their smoke exposure at bome. Forthese subjects alone, the odds ratio aaaoci, .ted with spouse's smoking was unchangedat 1.6. As Msntel' has pointed out, past atudies of cancer risk from passive smoke esposure have focused on _ lung cancer risk among nonsmoking women married' to smoken. Our study was not designed:to focus onn this iuue, and with only 22 lung cancers, and only two among nonsmok'en, we can add little to clarify results from thoae studies. While it may be surprising tkat strong relative risks were seen for sites not generally aasociated with ciga- eette smoking, we believe that such effects are pouible, Studies indicate that passive smoking has a pervasive biologic effect (8-121, making the boundaries between 'amoking rilated' cancer sites and other sites unclean Some sites may be considered unrelated'to cigarette smoke because risks are less dramatic than thcne for other sites, or hecause studies have not been donc. While active smoken ant also pasxive smokers, if an effect were due to sume specific property of the stdr- 369 i Repac. JL Isr.eey AH. Madelins e:powre of;nonrmok- en to ambient tobacro aswke Proce+d+np ef the 'itA Annual'Meeury of tl+e Air PoUuuoe Control Asaoasuoea Atlanta. June 1f61: 7 M.taukws S. Taminato T. Kitano N. at al EJiecu of MvieonaNntallWbK'rD.IWIVOke On Ynn.ary COtin-ne eKrf- t,oe m,non.a+oken. N fasl J Med 1%4 ,31112A-JZ a. Srtton CC. Pr.i.e enokinc and kutt canet. Br M.d J 196 1Z92733. l MRFIT Research Gmup MuhipM risk factor intervent,oet trial: risk factor, cSanRs and .o+uLty re.ults. JAMA 1'962:24t26"12 t Rosr G, Hamdton PJS;,Colwtl L trt at. A ra+domi.ed conuotl.d trial af s+ui-smokins ad.rr. Idyear rs.ultt. J Fy-demiol Cemmunity Healt! 1%2.]l.1Ct-& Philip R J. Burth Deaonment of Medicof'Physica llnu+rniq o( l:reds The GenernllnJtrmary Leed, L4FJEX U:K

370 LE'T'i'ERS TO THE EDITDR stream smoke (e.g., increased concentration of chem• icala or decrsased particle aise).it would be difficu)t to detect by comparing smokers to nonsmokers when tAa nonsmokint group also includes pasaive smokers. AJurttstively, one eould speculate that smoken, d• though at increased overall riak. may paradoxically be protected against rome properties of cigarntte smoke,. possibly through auirnulation of'en:yme aystema that raeuboliu cartinoEens into lesa harmful eompounda 113, 14). Theae same tnechanisms may not be stimu- Irted in nonamoken who have different apos+us to ctjaretta sDfOke.. Wt would not rantto arsue tbat a biologic pathway is established, but rather to say that a plausiblt path- way cannot be ruled out Evidence wgTesu that pas# sive smoking is rwt simply a lower dosa of actiw stnokinE. and thua may need to be eonsidend in a dif[erent tight Our (indinp are preliminary and ne.d' to be cronfirmed by other studies. Studies of individual ancersites cartewluau potential confounding fanors in more detail, and may be able to clarity the role of pasaiv' smokirt=. Raattrusr 1. Fri.dmaa GD: Rr. 'Pauibt smokinc in aduhLood and tancrr nak:' A. J Epidrmiol 1l8S.123.M'7. 2. 14rotd iN. Rr 'Pa.sl•e amoking in adulthood and cannr risk.' Am J' Eaid.miol 1986.12J:3b7-i.. i Sandler DP. E*.noo RB. 1ti ",ko: AJ. Passiw amokuq in adulthood and unctar risk. Ape J Epid.maol l9dS:12L37- ~ . 4. BWeklty JD. Harris RWC. Doil R, et aL C.rKOnvol etsrdy o( t1r bu.b.ndl af .omea .ita dyspl.ai. w tiro. anma of tse or.u auri Lat,cst 19d121010-15 i Bro.n DC, Pv.:rs. L Gasn.r JB. Caxv o( tLe t.rvv and the saok.iet il.b.nd. Caa Fas Pby.ciaa 11192_M.99-502 G. Burch PRJ. Rr 'Pa.a,.e stnoking in aduJtliaod and carrer n.k.' Ar. J Ep~d.mal 1996.1:7:36d-9 . 7. Fn.dman GD. Petau DB. Ba-ot RD. Pr"sitere srr torr.laus of pasain seakins Aa J PubaK H'..+Ik 19W:72:.01 -i l W.ld N. Ritthi. C. Validation of studis on II rahi in non•neokers tn.rri.d to aewkers 4ncx 1lt1i.I tui'. 11. Wald NJ. Bor.ham J. Ra1l.y A- et ai. Urlnar) cotln,M al arkeraf brsthing oNtr p.opl.•a lob.cro nawkt. l.erw 19K4:l:~i. 10. Maaukura S, Taminato T. Kiuno N. .t al. Etr.cu ,r enrllOftoltnW.tOhafto anloke on 4rlnarylOllniM.IrT tion in norumokert rid.ner for p.aaw smoking \Eitd J w1.d 116tJ 111i25-]Z Il: Ba RP. TTav-a JLG. Hendemon PT. Eaernion ef.w tarns in hursan utir alsat paaurs lsok ing, Crrcer leu 1lKl:1ti6-l0. It MancMater DK. Jacoby EH. S.mitivity of huetan plar.w u! monoolyt.naw an/.ity to t..urnal soukine (1a Pharmaicol'T)rr 19A1:]0.tid:-tL IS Wilson AGE Kunal H•C„Borou)erd Sl, et aa. Inh.h.tw is vi" a! the formation ol adductu ba.wn weiaWl,ui af b.nsa4i/pyrmt.nd DNA by,aryl hydrxartion h.vvl. yla.e induaers Cancer Ra 19d1:41:2JS3-40. 14. Ho.l DG. ICaplao NL Andernon MN'' Implical... .4 aonlintar kinI on risk esumation in esrcinus.n.... Science 1t16711r1032-7. Daltr P. SandJkr Richard B. Everson A.l1tn J. W ilcos Fpid.miolup Brnnch Biom,ttry and RisA .{autssrnrnt !'nvro,n Nordo+to! Inatit+ru of Enuuonnvnro! Heolth Sciencs: Restaxh Tmnry/4 Park lYC'!T7lrJ ' 9