Tobacco Documents Online

I I s Ex-oiwr.a lnW-Uaul. `/cl. 16. Pp. 173-197. 1990 P:feswd ia d.1J.S.A AU non I ~ V LETTERS TO THE EDITOR AN ESTIMATE OF ADULT MORTALITY IN THE UNITED STATES FROM PASSIVE SMOKING: A RESPONSE Dear Editor. The health implications of environmental tobacco smoke (ETS) remain controversial. Neither the pub- lished reports nor statements from public health of- ficials and agencies have resolved the question of ETS health effects, nor are they likely to in the near future. A. Judson Wells' paper. "Estimate of Adult Mor- tality in the United States from Passive Smoktng" (1988) is yet another effort to draw sctentific verity from a reassessment of published epidemiological data. But this new look does not change the quality or meaning of the existing evidence, which remains equivocal. Neither does it substantively support the author's statement that exposure to ETS "can have adverse long term health effects that are more serious than previously thought'. The conclusions of nonsmokers' increased risk of lung cancer from ETS exposure found in the reports of the National Academy of Sciences (NRC 1986) and of the Surgeon General (USSG 1986) were based on epidemiological studies that produced a wide range of findings. The relative risk (RR) values summa- rized in Table 12.4 of the NAS report ranged from 0.50 to 3.25, with 17 out of 20 risk estimates (for subgroups by sex) lacking statistical significance. In seven additional reports since the NAS docu- ment was published, relative risk values ranged from "<1.00' to 1.65, with only the latter being stausu- cally significant. The RR values from all 29 sub- groups in the 20 studies included in the NAS report plus those published later are summarized in Table I herein. All of the epidemiological studies that comprise the data base for estimating nonsmokers' risk of lung cancer in relation to ETS ue actually estimates of association based on spousal smoking. In not a single study was either exposure to ETS or retained dosage determined. A few studies have attempted to estimate 0160-412.0f90 53.00 r.00 C.opyn{pt -:0 1990 Per{.mm ?seu pue the degree of exposure to spousal smoking in terms of hours per day or total years of exposure, but none of the studies measured ETS exposure in objective and quantitative terms or even estimated ETS expo- sure with any degree of reliability. Proximity to a smoker sitting across the dining table does not permit an estimate of the nonsmoker's exposure to ETS, which will vary according to room volume, ventila- tion rates, the physical and chemical changes in ETS as it ages, and other factors that influence the con- centrations and duration of ETS exposure. A spouse's smoking in another room or in another building can have even less or no significance at all in assessing the possible role of passive smoking on a subject's health. It should be recognized, also, that association can never establish causality. At best, association can only suggest the possibility of causality Feinstein (1988), discussing public alarms based on epsdem,o- logical studies, recently pointed out that "a causal suspicion is supported if an impressive statistical association appears in the 2 by 2 tabulation for sub- groups of people reported as being exposed or non- exposed, diseased or nondtseased". There are many ways to look at data and try to draw meaning from the aggregation of values. After deciding that the 13 studies which survived critical assessment did not, individually or collectively, sup- port a definitive conclusion on the risk of lung cancer in relation to spousal smoking, the NAS Committee performed a meta analysis on the aggregated data, leading to an esttmated risk increase of about 34% for nonsmokers married to smokers. This estimate has been questioned on a variety of grounds by a number of investigators (Letzel et al. 1988). It can be argued that even if a first order relation- ship does not exist between disease and passive smok- ing in the epidemiological studies, the data used by Wells are the best evidence available. And it can be argued that even the array of values shown in Table 1 is not impressive ia the sense that Feinstein specifies, there are other ways of testing the data, as has been done by Wells. 173 TI DN 001335

Lauen Eo the e,0iW Tabt. 1. Statisciesl asnificases of nsk valuws for lana canear 'a eel.uoa to spoesal emokina. I I , I i Investi,cator •CAan and Funq (1982) •Buftler et al (1984) Dalaqer et al. (1986) •Kabat and ttynder (1984) Gao at al. (1987) *Gillis at al. (1984) *Lee at al. (1986) Gao at al. (1987) Shiaizu et al. (1988) *Oarfinkel (1981) *Pershaqen at al. (1987) Wu at al. (1985) *Lee at al. (1986) *Cariinkel at al. (1985) •Akiba at al. (1986) *Koo at al. (1984) 9rownson at a1. (1987) Huable at al. (1987) •Correa at al. (1983) *8irayaaa (1981) Laa at al. (1987) •Trlchopouios et ai. (1981) *Cillis et al. (1984) 3.25 Stattstically Sic:•;ant "_alE Fesfale 2.215 1.63 1.65 :.11 • Rish values from Table 114, Naaoaat Academy of Sciences Repoct (1986) a Exposnte in adult life. b Exposure in childhood. c Satisaeaily si8mtieaac asads in one or more data subseu Widua the study. There rematns, however, the fundamental question of the quality of the individual underlying studies whose data are under consideration. Many of the epidemsological studies assessing the risk of lung cancer from spousal smoking have been criticized for a variety of methodological flaws and weaknesses, especially with regard to the potential for mtsclasst- fication (Uberia 1987; Balter et al. 1986; Lebowitz 1986; OTA 1986). Misclassification of subjects is a source of error where patients claiming to be never smokers are in fact current or exsmokers. Wells conceded the likeh- hood of 5% misclassification. But misclassification of smoker status has been found at levels from 10% to 40% (Schwartz et al. 1988; Weiss 1988). NAS noted the likelihood of misclassification and lowered its estimate of the elevated risk to 25% from 3496, but it failed to indicate whether the lower value was statisttcally stgnificant. (NAS found the combined risk from American studies a 14% increase, which was not statistically significant.) Misclassification of disease can also be a source of error. There was a marked potential for misclassi- fied disease in the studies having staustically signif- icant risk ratios in the NAS and Surgeon General's reports. In Hirayama's study of Japanese women, his 1984 report suggests that only 21 of the 200 lung cancer cases (10.S9b) were histologically confirmed, while the Surgeon General's report states that "none" were verified. Akiba et al. (1986) studying survivors of the Hiroshima and Nagasaki atom bombings. noted 43% of the lung cancer cases had not been histolog- ically confirmed. Weiss (1988) notes that "thtrteen percent of the cases (in Garfinkel's study] proved on review not to involve lung cancer". Not Statisti,cally Sicniicant :lale Female 0.75 0.50 0.78 <1.00 1.00 0.79 0.9 • 1.00 1.00 1.1 e 1.1 1.17e 1.Z0 1.20 1.30 1.31e 1.80 1.50 1.64 1.63 >1.80 1.80 2.00 2.0?e T= DN 001336

, . sysa j .os .ouo. Misciassificacion of exposure can be a source of uncertainty in studies that attempt to find exposure- response relattonshtps. There is little basis for con- sidering estimates of spouses' smoking to be reliable. Pron et al. (1988) concluded that "test-retest esu- mates of reliability [over a six-month time span) would suggest that miselassificatuon of such expo- sures may be extenstve". Vogt (1977) found "twenty- two percent of persons gave different answers on the two questionnaires (on the number of cigarettes smoked per day] given about an hour apart". Among the variety of flaws and weaknesses found in the various epidemtologicai studies on ETS and lung cancer, it is worth noting the age bias found by Ahlborn and Oberla (1988) in Hirayama's study and their conclusion that "ft rtsit increase ... disappears completely when one removes selection bias by age" (Jberla (1987), highlighting the weaknesses of the epidemiological studies comprising the NAS data base. had earlier concluded, "False plus false does not equal true." In addition, most of the eptdemiological studies have failed to take into account significant confound- ing factors in assessing lung cancer risk in relation to ETS. Many risk factors for lung cancer have been tdentified, including exposure to heavy metals, or- gamc chemicals, combustion by-products, natural and man-made radiation, diet, and nutritional status, per- sonal health history, emotional, and psychological factors. Hoist et al. (1988) recently reported stgnifi• cantly increased risk in relation to keeping pet birds and to reduced vitamin C intake. Gao et al. (1987) found no significant increased risk for Chinese women in relation to passive smoking or type of employment but did find significantly increased risk in relation to previous lung disease, cooking practices, and shorter menstrual cycles, reflecting hormonal factors. Some of these factors may act independently, but many may tnteract. Any attempt to assess the role of one factor must take into account all other relevant factors. None of the epidemtological studies on spousal smoking took into account confounding factors other than attempting to mateh cases with controls by age, residence. and general socio-ecooomic status. Of the 20 eptdeeniologieal studies, those by Hirayama and by Lam et al. (1987) have the two largest number of tung cancer cases. with the increased risk in both being statistically stgnificant. Both studies are of Oriental populatuons, which suggests that many fac- tors like cooking practices and fuels for cooking and heating should have been controlled. All of the studies included in Wells' Table 4, on which he based his estimate of heart disease deaths t» related to passive smoking, similarly fail to constder the confounding effect of the many cardiovascular disease risk factors that have already been estab- lished for that disease. Some observers have commented that increased risk of lung cancer from ETS exposure seems impiau- sible because the ETS components are so dilute in ambient air compared to the concentrations of these substances in mainstream smoke. In addition, it has been found that nonsmokers retain far less of inhaled ETS than active smokers retaut of mainstream smoke. Wells noted that "smoke retention by a passive smoker is only about 1/400 that retained by a direct smoker in a 16 hour day". This is more than one order of magni- tude greater than Rickert's calculation (1988) that non- smokers exposed to ETS retatn about 1/8000 the amount of particulate matter retained by the active smoker. Lee (1988) cited estimates of the same range: 1/3000 for males, 1/10 000 for females. All of these estimates are probably on the high side, since none of the studies appears to have considered the chemi- cal and physical changes that occur as ETS ages and the losses of ETS through evaporation, fallout, and deposition over time. Other observers have commented on the implausi- bility that lung cancer in nonsmokers might be caused by ETS. Aviado (1988) noted that none of 17 constit- uents of ETS "designated as suspect carcinogens ... [hasl been adequately shown to cause pulmonary cancer via inhalation in animals" Crawford (1988) noted that "no atypical cellular changes have been found in the lungs of nonsmokers". Lee (1987) con- cluded "that exposure to smoke constituents of non- smokers is too low to explain the moderate increase in risk of lung cancer seen in epidemiological studies in self-reported never smokers married to smokers. This increase tn risk is much more plausibly ex- plained by misclassification of smokers as nonsmok- ers than by a direct effect of passive smoking". Wells has attempted to make his calculation of annual deaths from exposure to ETS appear more reasonable by comparing it to the larger estimate of Repace and Lowrey, but their estimate has been se- verely criticized because the controls were Seventh Day Adventists (SDA) whose life style is so radically different from that of the non-SDAs married to smok- ers that the comparison is considered inappropriate (OTA 1983; Balter et al. 1986; Cberla 1987). Taking these and other factors into account, Gostomzyk (1986) concluded, following the Interna- tional Experimental Toxicology Symposium on Pas- sive Smoking in Essen. FRO, that "even toxicology has not been able to ascertain with any greater degree TI DN 001337

Lettsrs to Wa souor of probability than did epidemiology that there exists a link between damage to health and passlve smok- ing". Perhaps it is the weight of these facts, interpreta- tions, and opinions that caused no less an authority than the American Cancer Society to assert last year that "the currently available evidence is not suffi- clent to conclude that passive or involuntary smoking causes lung cancer in noasmoketrs..." (ACS 1988). A final comment: both the title and the content of the editorial that accompanied the Wells paper sug- gests that the paper provides stronger evidence of rtsk of cardiovascular disease (CVD) for nonsmokers married to smokers than the paper in fact offers. In 1986. both the NAS and USSG reports noted the lack of convincing evidence of significant CVD risk from ETS exposure. More recently, Fielding and Phenow (1988) commented on papers reporting an assocla- tion between ETS exposure and CVD risk, conclud- Lng that 'no f"trm conclusion that a causal relation exists is yet warranted'. Wells' calculations with respect to CVD are based on data from epidemiologlcal studies that have the same weaknesses as the lung cancer studies. There is, thus, no basis for greater confidence in his esti- mate of hean disease deaths in relation to ETS than hls estimate of lung cancer deaths. It is commendable that those who are not satisfied continue to seek more meaning from the data. But in an issue as serious as t6is. it is important to note when the data fail to meet the standards for sclentific in- ference. Alan W. Katxensteln Katzenstetn Associates Larchmont. NY 10538 REFERENCES Ahlboen, W.: Qbsris, K. Passive smoking and lung ea.e.r' nanalyass of Hitryw'a data. L: Parry, R.: Ki:k, P. W., ads. Indoor and ambtent atr qaslity. London: Salpar Ltd.: 1981: p. 169-i7E. Akiba, S.. Kato. H., Blot, W. J. Pasaivs smoking aad 1na9 eaao.er tmoss Japenas wamss. Cancer Rss. 46:4i04-+407; 1986. ACS (Ameriw Canear Soeiaty). Gsaeral faets oa smoking sad tiaaltb: 1988 (p. 3). Anado. D. M. Suapsetsd paimosary earataobaas m.avuanmrW tobacco anoks. In: Psrry. R; Kirk, P. W., ada. London: Salpsr, Ltd; 1981: pp. 141-146. Baltsr, N. J.; Schwarts, S. L. Kl7paaick. S.J.; Witoneb, P. Caoasl niatioaabip between sn.iroamotal tobscco rmoks and ltta{ csaear in noa-+makers: a critieai ravisw of tha literature. Proe. Air Poilnt. Cantrol A.aoe. t6-i0.9: 1986. Browmaa. R. C„ Raif. J. S.; at al. Risk facton for adsaocaretaoma of tha lung. Amsc L 8ptdaanoL 125: 2J-34; 1987 Bnfflsr, F. A.; Ptckls, L. W.t btaaon. T.J.; Contant. C.. The causes of lna6 eanesr ln Taxaa. Ia: Misall, bi.; Corread, P., eda. Lun6 cancer eanaea aad preventson. Nsw York: Verial Chemie Ia- taraatioaal lac.,1964: p. 83-99. Chan. W. C., Fua=, S. C. Luni eaaesr in nonsmokers in Hong Koss. In: Grundmana, E.. ad. Cancer epidemaology, vol. 6. Nsw York: Gustav Fisohar Varlaf; 1912: pp. 199-202 Corras. P.. Plckls. L. W. Fontham, E.. Lin. Y.. Hasnazsl. W. Passive smoking and lung cancer. Lancet 11 2:395-397; 1983 Crawford, 1V.A. Health effects of pssatve smoking in tha work- piacs. In: Perry, R.; Kirk. ?.W., edt. Indoor and ambient ur qwlity, Londoa: Saiper Lid., 19Ei:p. 203-210. Daiapr, N. A. at al. The rslauon of paaivs smoking to lung caacec Cancer Rsa. 46:4ioi•4l11; 1986. Faiaatata, A. R. Scientific standards in epidsmioloqie ttadia of tbs maaaes of dally Life. Sclencs 242:12J7- t 263. Fiddiag, l. B., Pbaaow, K. J. Health affseta of lnvolaotary tmok- laf. N. Sag. I. btad. 319:1432-1460: 1988. Gao, Y T. at aL Loag osaosr among Chinese womsn. Iat. I Cancer 40:6W-609; 1967. Gerfidtal, L 13ss trands in lang canesr mortality among oon• smokers aad a aats on passive smoking. I. Nat. Cancer lan. 66:1061-1066 1981. Garfiaksi, L, Atwrb.cb, 0., loabert, L Involuntary rmokini and lung caaaar: a eaas-eontrol study. 1. Nat. Ctnesr laut. 7s:463- 469a 1915. Gi11is, C. R.; Hais, 0. J., Hawtbataa, P. Boyle. P. Ths affect of sae7roa~stttal tobaoeo amoks in two urban commanitias is t6i wast o(Soaals:wL Eat: l. Rap. Dia. (Supp. 133) 65:121-126: 19i4. Goataussyk, G. 1. Paasivrasebaa - Bancht tiber eia intsrnatio- salsa Sympoaam (23-2? Oktober 1916 Esan) Public HaaltL 49: 212-213: 19t7. Holat, P. A., Ktasbast. D.; 8rmd. R. For debate: pet birda as an iadapsadot ruk factor for Lung cancer. Brit. Med. I 297:1319-1321; 19ii. Ptombl., C G.: Sasat, J. M.; Patbak. D. R. Marriage to a smoker aM laag caassr rtalt. Amer. J. Public Health 77:J9i-602, 1987 Hirayaau,l: Notr-emokiab wives of heavy tmoksri bavs s higher risk of Iaa6 caacar. a study fromlapan. Brit. b/sd. J. 2aZ:1t3 1ut 1981. Kabst, G. C.; Wyadsr, E. L Lunb cancer m nonsmokers. Cancer 13:1214-1221; 1984. Koo, LC.: Ho, LH-C„ Saw, D. Is passive smoking an added risk fsator far ltri cancer to C3iasaa women? J. E:p. CIin. Cane.r Ra. 3:277; 1944. Las. T. H. at al. at at. Smoking, pasaive smoking and hiuological types in lnn; cancer in Hong Kong Clnnses womsn. Biit. 1. Caaar 16:673-67i; 1917. Labowits, M. D. The potential assoaatlos of lang cancer with passive aaaksaS. Ea.itoa. Iat. t2:3-9: 1986. La, P. N.: Chamberlain. l.. Alderson. hf. lt. Rslauoashtp of passive smoking to risk of lna= cancer and other tmoksni-.s- toaiusd diseases. Brit. l. Cancer J4:97-10J; 1966. Laa, P. N. Pasaivs smoking sad lung cancer aa.oclatioo: a result of baaal Hataar ToticoL 6:117-J24; 1997, I.es. P. N. Aa aitseaativs asplanauoa for the increased risk of laab cancer in saa-emokats marrisd to smokers. In: Perry. R.. Kirk, P. W., eda. Indoor sad ambient air quality. l.aados- Saip.r Ltd.: 1988: p. 149-13t. NRC (National Rassareb Comeii) Environmental tobacco tmoke: maasariai s:posora and assessing health effects. WaabiaStoa, D.C.: National Acadamy Press; 1916. OTA (OQios of Tsebaoloby Assessment). Passive smoking in tbs workplace: ulseed issues. Ws.ltin6toa, D.C.: Offios of Tech- notoW Aasstassat, U.S. Cosgrsss; 19{6. Psrahasaa, G.: Hnbso, Z; Swoaaa, C. Pauivs smoking and lung eaaosritt Swdiab women. Amaa J. Eptdemioi. 121:17-24; 19t7 TI DN 001338

t I I Y l.sten to the editor Pro.. G.1.; S.eob, I. D.: Ho... G.R.: b1iWr. A.S.13. rsiiabiltty of passiw smotiwg bisaeri•a npaet.r om a aas.-eourd st.dp a( lua3 oa.eas: A..r. I. EpidsioL 127:267•273; 19Si. R.p.es, J L.t lowny. A. 8. A qarosati.a.K iow at sa..mokam' lasg ea.om tisk trao psasito s.obs.g. Ss.itao. lat. 11: 3-2Zt 1995. Riok.rt. W. S. Sw ootwsnnou wYr ssdsati.3 aspoare to !O~liOMtataa fO1MCC0 tROw (1rs) with pptlCmar Mfmsa :o :ba ba.e s.irorr.a. Ca.. 1. Pabiis H.tlti 71:539-399; 19ti. Sehw.ess. S. L. 8dtse: N. 1. BTS-Io.g a.a.r apii..ioleq: sappenabiHsy of miselasdamt aod riak a.aspuaau IL: P.rry. R.; Kiel:. P.W., .dt. Isioor atl a.6inc av qsality. L,a.Oaa: Sdp.r U4.:19it: pp. 119-16i. Sh:miaa, HL as aL A a.so•ao.ceoi seNy a( lo.1 sa.o.r ia samssdo- mi wa.., Toioin I&p. l1d. 1S4:3i9-397; 19ii. Tadummaia, D» Kala.Oidi. A.; Spsnos. L; blaelL.ia.. 9. Lrses ca.a.r sd pnss.m s+.oiQ.p Lc J. Ca.m 27:1-4; 1981. 179 Obda, K. Lnsg erar from pwi.m c~matsog: bYpabosu or eaa- viM=g adaoa7 lat. AneB. Goeep. 6aniw. Health 39:421- 437; 19i7. USSG (U.S. Sae{eaa Cimanl) T1u bsaltlt eoawqasaess of iavol- auary snoiti.g: a npoat ai tbs Surs.o. G.avaL DHHS (CDC) 37-439t. Waabntpoo, D.C: U.S. Poblie H.altb Samcs; 1986. Vogt, T. 3L Ssatiy b.dsiriotai faotors u pe.dieson of riska. Ia: Rasaeet aw saskiaS brYwior. NIDA blomogrsp6 17. Nanoaa! lsiaw of Dr.{ Ab.... U.S. Pabiie Hniti S.r.:as: 1977: pp. 9t-110. Wasi, 3. T. Whac aro the bsalcK effects ot pssa.o smoitiaf7 J. Rap. Db. 9: 4f-62: 19Si. WiIL. A. 1. Ao aus:w at addt moeuJity u:b. Uaiud Stats: f~ pasaw ssaio.g. 8a.um. Ist 14:249-265; 1988. iV., A. IH.; Hasr.nor, a. 8.: Pl1n.1+t.C: Yo. lif.C Satokial aea other risk faamn torlr{ erar m wa.ua. l. Nu Gmcar 1a:4 74:747•791: 1915. TI DN 001339