Tobacco Documents Online

l.ettert to tl>e editor - N 0 T 1 C E 149 This material may be pfvtetted by copyright law (Tilr, 1' 111.S Crd-' Pron, G: &.; Burch. J. D.; Ho.e„G.R.; Miller, A.B.'1Le reliability of passive smoking histories reported in a ease-eontrol!study of lung cancer. Amer. I. Epidemioi. 1i27:267-273;,1988. Repace, J. L; Lo.my, A. H. A qaantiiative estimate of nonsmokers' lung cancer risk from passiva smoking. Eaviroa. lat. 11: 3-22; 1985. Rickert, W. S. Some eonaidenttions when estimating exposure to environmental tobacco smoke (BTS) with particular reference to the home eavironmeat. Can. J. Public Health 71eS33-S39c 1988. Schwartz, S. L;, Balter, N. J. ETS-lunB cancer e" 'oloBY: supportability of mirdksei6cant and risk u ons. In: Perry, R.; Kirk, P.W., eds. Indoor aad [ air quality. Loadon: Selper Ltd.;1988: pp. 159-1 . Shimisu, H. et al. A u ntrol study of lung eancer in oonsmok- ia8 .romea T J. Bxp. Med. 154:389+397; 1'988. Tricb ; Kalt<ndidi, A.; Sparros, L; MacMabon, B. 1.ane cer and pasdve smoking. 1at. J. Cancer 27:1-4; 1981. Oberla. K Lung cancer from passive smoking: hypothcn onm- viaeia8 evidmceT lat. Arch. Occup. Envimn. tb 59:421- 437;1987: USSG (U!S. Surgeon Genenl) Tbe [h oonuqueaeer of invol- untary smokin8: a repon e Sureeoo General. DHHS (CDC) 87-8398. Wash' , D.C.: U.S. Public Health Servtce;,1986. VoBt, T. M~8 behavioral factors as prrdicton of rieks. ln: rcb on smoking behavior. NIDA Monograph 17, Nauonal Iastitute of Drug Abuse, U:S. Public Healtb Service; 1977: pp. 98-110. Weisa, S. T. What art: tbe health effects of passive smoking? J. Resp. Dis. 9:46-62; 1988. Wells, A. J! An estimue of adult mortality in the United Statei from passive smokiaB.,Bn.iinn. lnt. 14a249•263; 1988. Wn, A. Hl;, Henderson, B. E.; Pike, M.C;, Ya; M.C. Smoking and otber ritk factors for l®8 eaneer in .omen. J! NkL Cancer Inet. 74:747-751;,1985. 1` e e, (P, N'. F_ , z.~-, AN ESTIMATE OF ADULT MORTALITY IN THE UNITED STATES FROM PASSIVE SMOKING; A RESPONSE Dear Editor:. Wells (1988) estimates that exposure to environ- mental tobacco smoke (ETS) causes 46 000 deaths per year in the U.S.; 3000 from lung cancer, 11 000 from other cancers, and 32 000 from heart disease. Theseasumates are scientifically unjustified. Far too much faith is placed on results from often fragile epidemiological studies, with major sources of bias ignored or totally underestimated.. In contrast, far too little faith is placed on evidence that nonsmokers have very much lower exposure to tobacco smoke ~I6 ~-L~ ry 1-7 19qo constituents than do smokers, and that smokers are much more exposed to ETS than nonsmokers. The evidence that exposure to ETS increases the risk of developing heart disease is extremely uncon- vincing. Of the studies cited by Wells, some are based on unacceptably small numbers of cases, e.g., Gar- land et a1. (1985) where only two deaths occurred in women married to never-smoking• husbands, while the only two studies with substantial numbers of deaths are both open to question. When referencing the Japanese prospective study, Wells uses Hirayama's 1984 report of a statistically significant positive trend in wife's age-adjusted risk according to husband's smoking, but does not com- menron the fact that, in 1981, Hirayama reported no association whatsoever. As shown in Table 1, the Table L Female relative risks for heart disease from passive smoking in Japaaese study. Husband's smokinR habit Total Ex or Fo1Lov-uD oeriod cases Non-smoker <19/dav, 20+/day 1966-79 406 I 0.97 1.03 1980-82t' 88 1 2.85 5.07 11966-82 494 ~ 1 1.10 1.30 © ~ t Estimated from 1966-79'data (Hirayama 1981) and from 1966-82 data (Hirayama 1984). The 1984 paper provided retati've numbers of deaths as 118, 240, aad 136.

IEO 1966-79 and 1980-82 data are totally inconsis- tent and statistical, tests confirm the highly sig- nificant (p<0.001) interaction between relative risk and period of follow-up. A possible expla- nation might be that the 1981 dara, but not the 1984 data, were standardised' additionally for occupation,, but if this was important, why did Hirayama not standardise for occupation in 1984? The Maryland prospective study (Helsing et al. 1988); which reported a 24% increase in heart dis- ease risk in women, based on 988 deaths, andd a 3196 increase in men, based on 370 deaths, in relation to living with a smoker, has a number of features that should be considered when interpreting the data. No attempt was made to follow-up people moving out- side Washington County, thus presumably missing large numbers of deaths. No dose-response relation- ship was reported~ Adjustment for age, marital status, years of school and quality of housing had an enor- mous effect on relative risk, changing estimates from 11. 117 to 1.31 in men and from 0.66 to 1.24 in women. No direct adjustment was made for household' size, despite the fact that the larger the household, the more likely it is to contain a smoker. Furthermore, no direct adjustment was made for the possible correla- tion of household size with various coronary risk factors. Also, data were unavailable on a whole range of factors, such as diet and exercise, which might differ in families with and without smokers. In short, several potential confounders were apparently not controlled for. Wells does not consider the problem of publ'ication bias: This may be particularly acute for heart disease. After all, it is a vastly more common disease than lung cancer in nonsmokers, but the numbers of deaths in Wells' tables are only slightly greater. The possi- bility can surely not be excluded that other researeh- ers, perhaps with much larger and better data bases, have looked at the relationship and found nothing. The data for cancer other than the lung are even less convincing than for heart disease. In view of the much greater passive smoke exposure of smokers than nonsmokers, observations that nonsmokers ex- posed to passive smoking have increases in cancers at sites not increased' in smokers seem to me to suggest that something is wrong with the epidemi- ological studies. And, indeed, the paper showing the strongest association ('Sandler et al. 1985) is open to a number of serious criticisms (Lee 1985). Wells, however, remains content to include all epidemio- logical studies in his meta-analyses, regardless of quality, and attempts to explain obviously spurious relationships by an unsupported, and implausible hy- Leneri to the edieor pothesis, involving an especially, susceptible group of individuals who all die early if they smoke but die later by passive smoking if they do not. Mortality patterns for lung cancer in terms of age, dose, and duration of smoking are in fact weU described by models involving no component for variation in sus- ceptibi4ity at all. Wells' estimate of 3000 lung cancer deaths per year based on the epidemiological data contrasts with that of 12 by Arundel et al. (1987) based on exuapo- lation using relative amounts of particulate matter retained in the lung by nonsmokers and smokers. As I argue at length elsewhere (Lee 1987;,1988a; 1988b; 1989a; 1989b), it is far more plausible to conclude that the associatiom observed between lung cancer an& exposure to ETS arises predominantly because of bias than it arises because of a carcinogenic effect of such low doses of ETS. Misclassification of smokers as nonsmokers is likely to be a major source of bias in most studies and is one for which Wells' correction is totally inadequate. He does not allow at all for the possibility of misclas- sified current typical regular smokers, whereas a re- cent summary of data from~ large studies shows an average rate of about 4% (Lee 1989a), Nor do his calculations take into account recent data (USSG 1989) showing much higher relative risks in active smokers than in older studies. Preliminary calcula- tions based on these data suggest that the total num- ber of lung cancers occurring in self-reported never smokers in the U.S. may have been substantially overestimated. Rather than 12,000 the figure may be nearer 8000. If reasonable corrections are made for misclassification, the figure of lung cancer deaths among actual never smokers may be less than 6000. Wells considers his overall estimate of 46 000 deaths conservative. I disagree, When better data are avail- able, it may prove to be about 46 000 too high. Peter N. Lee P. N. Lee Statistics and Computing Ltd. Surrey. United Kingdom REFERENCES Arandei, A.; Starting, T.; weinkam, J. Never smoker 1nng canoer rieks from eiposors to particnlate tobacco rmoi{e. Baviron. Int 11:4Q9-426; 1997. Garland, C.; Banau-Connor, B.; Saarea, L Cnqai, M. H.; Winaard, D. L. Hftecu of paaeive emoldng on iiehemic hean diYeaes monality in nonrmoken lirina with emoken. Am. J. BpidemioL 121:643-63Q; 1995. 2023511.SS9

Letters to tde editor M" T t lhis ^'3'= " protecteC 07 C-tw (f,de 17 U..) Helsih8, K. J.; Sandler, D. P.; Comstock, G. W.; Chee, E. Heart disease mortality ia nonsmokers living with smokers., Am. J. Epidcmiol. 127:915-922; 1988. Hirayama, T. Non-smokin8 wives of heavy smokers have a hi8her, risk of lung caneer. a study from Japan. Br. Med. J. 282:183- 185; 1981,. Hirayama, T. Lung cancer in Japan c effects of nutrition and passive smokin8. In: Wtizelll M: Corres„ P., eda. Lung cancer: causes and prevention. New York: VerlaB Chemie International. 1984':175-195. Lee, P. N. Lifetime passive smoking and eaneer risk. Lancet 1:1444; 1985, Lee, P. N. Passive smoking and lun8 cancer. Association a result of' biasl Human Tozicol! 6:317-524; 1987. Lee, P. N. Misclltuification of smoking habits and passive smok- ing. A review of the evidence., Inc International Archives of Occupationali and Health Supplemeaa Heidelberg: Springer- Verlag; 1968a. 1S EMIC HEART DISEASE; RES ONSETO LEE Dr. P. Lee questiolitEd of my reports in 19 the reasons for a discrepancy and in 1984 on husbands' M disease risk in nonsmok- smoring and ischemic he ing wives. Table 1. Iscbernic heart disease mortality Lee, P. N. An,alternacive explanauon fortheincreased risk of 14ng cancer in non-smokers marned to smokern. In: Perry, R.; Kirk, P. W., ed1. Indoor and ambient air quality. London: Selper,, 1988b:149-151. Lee, P. N. Passive smoking Fact or fietion7 Paper presented at Conference on Present and Future of Indoor AirQuality.,Brus- sels, February 14-16, 1989; 1989a. Lee„ P. N; Problems in interpreting epidemiological data. Paper presented at Conference on Assessment of Inhalation Hazardt.. Hmover„February 19-24„ 1989; 1989b. Sandi'er, D. P. et al. Passive smoking in adulthood and cancer nak., Am. J. Epidemioll 121:37-43; 1985. USSG (U.S. Surgeon General) Reducing the healtb consequences of smoking. 25 yean of pro8ress. A report of the Surgeon General. Rockville, MD: U.S. Public Health Service; 1989. Welli, A. J. An estimate of adult mortality in the United States from passive smoking. Ei+viron. Iat. 11:249-265; 1988. The 1981 report was based on a 14 year follow-up (p=400) and the 1984 report was based' on a 16 year follow-up (n•494) of nonsmoking wives. The rela- tive risks of ischemic heart, disease when husbands were nonsmokers, exsmokers, or daily smokers of 1-19 cigarettes and'20 or more cigarettes were 1.00, 1.06, and' 1.18 (trend p : 0.061 not significant),in~the 141 year foilow-up; and 1.00, 1.10, and 1.31 (trend p: 0;019'significant) in the 1984 report. women by age group, by occupation„and by huebandi' smoking habit (patient herself a nonemoker): Musband's accupation Nusband's age group Nonsmoker Exsmoker 1-19/day 20•/day Total Agricultural 40-49 8 2.502 Z\ 5.941 17 3.636 50~ 12.079 worker 50-59 1S 3.497 Z7 812 2l 3,514 69 13,823 60-69 36 4„084 79 6, 27 2,152 142 13,081 70- 5 323 11 446 2 89 1B 858 TotaT 64 10,406 142 20.044 9,391 279 39,841 Other 40-49 5 3,,727 15 9,093 1S 128 35 19,948 50-59 11 4„294 29 8,830 23 6, 6 63 19,430 60-69 29 3.036 46 5,598 20 2,499 95 11,133 N 10- 9 432 B 619 5 137 2 1,188' 4= Total 54 11,489 98 24,140 63 16,070 21 51,699 .N C..) ' The weighted point eatieu te of rate ratio 1.00 1. 11 1.33 ~ 1.36 .11.68 Mant: el eNtension. and'test-based 90S 0.92 \ 1.09 I-A 2 conftdence 1'imits cn i ~,539 A I- One tail p value Mantel-MA.enS2e1 Chi One-tail p value U:882 2.331 0.00916 0.18889 0.00988