Tobacco Documents Online

Potential biases also exist in the Sole and Gillis study. For example, those exposed to ETS within the home may have had higher exposures to ETS outside of the. Kome compared with controls. A second potential bias is misclassification of women as non-smokers when they may in fact be former smokers or current smokers. Although the Hole and Gillis study suggests an association between ETS and cardiovascular mortality in non-smokers, the data lack any statistical significance. Also, the study reports so(he confusing and similar relative risks for active and passive smokers, and is vulnerable to several important methodological biases. This study would have to be replicated in a much larger study population to achieve adequate statistical power. D. Svendsen With respect to the Svendsen report, one problem is possible misclassification of the husband's smoking status either at entry or subsequently. A second problem is that the wife's smoking status was based on interviews with the husband, and not on direct questioning of the wife. There is also an alcohol-related bias in this study, as MRFIT ETS-exposed husbands had on average two drinks more per week more than non-ETS exposed husbands, and this alcohol effect

could explain the observed statistical significance in dose Finally, by combining ex-smoking husbands with never smokers, the Svendsen paper confounds any.possible past effects of active smoking by the husband with exposure to ETS. The XRFIT study serves in general as an exemplary prospective trial for its design and conduct. 8owever, lack of statistical significance, failure to control for several confounding variables such as alcohol consumption, misclassification, and misgrouping make it difficult to draw any conclusions relating to ETS'exposure from the study. The last prospective study is the work of 8elsing, Sandler and co-workers. The first problem I see here is that the only smoking data that were collected on every person were from 1963. Hence, no changes in smoking habits over the 12-year period were ascertained. In addition, no data were collected on acknowledged risk factors for heart disease such as diet, exercise, blood pressure, and cholesterol. Also, no ETS exposure outside the home was measured. And finally, death certificates were obtained only for those 1963 participants who still resided in Washington County, Maryland, at the time of their deaths. .__.__....~..,.,.,~:..~..~~,Y..,.........~.w,

The only retrospective case-control study in English thus far is the work by Lee. Although this study is one of the few to attempt to examine non-spousal ETS exposure, it raises the methodological difficulties that plague retrospective case control studies in general. In its finding of non-statistical significance for any trends of association between ETS and cardiovascular illness, the Lee paper confirms the need for execution of better controlled prospective trials. In 1989, He reported a retrospective study of 34 women. The odds ratio reported was 3.0 and a significant dose response was reported in the English language abstract of this study, which was itself published in Chinese. Methodological problems of the study include a small sample size and retrospective recall of number of cigarettes smoked. It is impossible to comment further on the study without an adequate translation of the paper from Chinese into English, which is not feasible given the short time period permitted for comments on this chapter. As noted in my critique of the seven epidemiological studies, misclassification is a pivotal variable in accurately - 12 -

determining exposure to ETS. It is critical that a mechanism for accurate measurement of ETS exposure, including exposure outside the home, and adequate follow-up of exposure status be determined. Glantz and Parmley unjustifiably minimize the potential effects of misclassification and assert, without proof, that the effect of misclassification errors leads to an underestimation of the effects of ETS. Without an accurate measure of ETS exposure, their assertion cannot be justified. The Surgeon General's 1986 Report examines the studies of Hirayama, Hole and Gillis, and Garland, and concludes that "further studies on the relationship between involuntary smoking and cardiovascular disease are needed in order to determine whether involuntary smoking increases the risk of cardiovascular disease." Likewise, in 1986 the National Research Council reviewed the prospective studies of Garland, Bole and Gillis, Hirayama, and Svendsen, as well as several experimental designs examining the biological plausibility of the association of ETS and cardiovascular disease, and concluded that: "With respect to chronic cardiovascular morbidity and mortality, although biologically plausible, there is no evidence of statistically significant effects due to ETS exposure, apart from the study by Hirayama in Japan."

In my opinion, contrary to the position taken by Glantz and Parmley, none of the more recent studies in this area provides a basis for altering the Surgeon General's and NAS's conclusions concerning ETS and.cardiovascular disease. Although Glantz and Parmley contend that cultural factors do not confound the interpretation of the ETS and CHD data, I would argue that they do, as illustrated in the previous critiques. Likewise, Glantz and Parmley assert that several studies showed a dose-response relationship between increasing amounts of ETS exposure and the risk of heart disease, and that this indicates a "real effect" of ETS on heart disease. I believe this conclusion is unjustified and is contradicted by a careful evaluation of the epidemiological studies. The few studies that reported statistically significant dose response relationships were invalidated by serious methodological flaws. I agree with the statement of Glantz and Parmley that most of the studies cited do not yield statistically significant relative risks. In addition, most of the studies have low to moderate power and are not capable of supporting a conclusion that ETS elevates the risk of heart disease. The authors assert that the two studies (Helsing 1988; Hole 1989) that have power above the desirable level of 80% "identified significant increases of heart disease with ETS exposure," and imply that these studies should be given greater - 14 -

weight than the other studies. This analysis fails to account for the fact that there were only 84 cases in the Hole study, and Helsing failed to control for important confounding variables such as diet, exerciser blood pressure, and cholesterol. Glantz and Parmley "pool the results of these studies to increase the power of the tests and so produce estimates of risk that are significantly elevated." They cite a pooled relative risk of 1.3 for men, and in a second analysis they cite a pooled relative risk of 1.2 for women. Even assuming that this pooling technique is justified, both pooled relative risks fall significantly below 2.0, which is generally regarded as necessary to support causality in epidemiology. In addition, for many of the reasons already cited in the individual critiques, pooling studies that are fraught with existing biases and confounders only leads to a pooled relative risk which combines and strengthens the effects of the biases and confounders. Such a meta-analysis is intrinsically unreliable. I disagree with Glantz and Parmley's implication that scientists currently understand the true exposure of ETS in the home, and in the workplace. Without a mechanism for accurate measurement of ETS exposure, including exposure outside the home, and adequate follow-up of exposure status, it is impossible to come to any definite conclusion as to actual ETS exposure levels. This conclusion is supported by the findings of Schieveblein and Richter (1984). They report that, under real-life conditions,

persons exposed to ETS inhale only one to two percent of the amount of particulate matter taken„up by active smokers. Also, level, and the increase in carboxyhemoglobin rarely~exceeds 1%. within a=ange that is barely distinguishable:..fr,om the,background nicotine concentration in serum of ETS-exposed individuals is The authors thus conclude that "passive smoking is not likely to have an effect on the development and progression of CBD."' At a November, 1989 international symposium on ETS'at McGill University in Montreal, in which I participated,• the participants expressed the view that the available epidemiological studies examining the relationship between ETS and coronary heart disease report at most a small statistical association, and that ETS is unlikely to contribute significantly to the incidence of coronary heart disease. This view is in sharp disagreement with Glantz and Parmley's conclusions, and provides further evidence of the weaknesses of Chapter 11 of the draft EPA ETS compendium. .~......,,,.~.~.~:a,w...~..r,..~.~_.....,.,.~..~..:~:.:,..:........~......~..~..,,,.~:~,..~.....~,..

References Ecobichon, D.J.; Wu, J.M. En'vironmental Tobacco Smoke: Proceedings of the International Symposium at McGill University. Lexington Books 1990. Environmental Tobacco Smoke - Measuring Exposures and Assessing Health Effects, National Research Council, National Academy Press, Washington, D.C., 1986. Garland, C. et al. (1985). Effects of passive smoking on ischemic Heart disease mortality of non-smokers. American Journal of Epidemiology 121(5). Helsing, K.J., D.P. Sandler, G.W. Constock, and E. Chee, (1988). Heart disease mortality in non-smokers living with smokers. American Journal of Epidemiology 127(5). Hirayama, T. (1984). Lung cancer in Japan: Effects of nutrition and passive smoking. Lung Cancer: Causes and Prevention. Verlag Chemie International, Inc. Hirayama, T. (1981). Non-Smoking wives of heavy smokers have a higher risk of lung cancer: a study from Japan. British Medical Journal 282:183-185. Hole, D.J., C.R. Gillis et al. (1989). Passive smoking and cardiorespiratory health in a general population in the west of Scotland. British Medical Journal 299:423-427. - 17 -

Lee, P.N., J. Chamberlain, and M.R. Alderson (1986). Relationship of passive smoking to risk of lung cancer and other.r smoking-associated diseases. British Journal of Cancer 54:97-105. Schievelbein, H. and F. Richter (1984). The influence of passive smoking on the cardiovascular system. Prevention Medicine 13:626-644. Seltzer, C. (1989). Framingham study data and established wisdom about cigarette smoking and coronary disease. Journal of Clinical Epidemiology 42(8):743-750. Surgeon General's Report (1986). The Health Consequences of Involuntary Smoking. U.S. Department of Health and Human Services. Svendsen, K.H., Lewis H. Kuller et al. (1987). Effects of passive smoking in the multiple risk factor intervention trial. American Journal of Epidemiology 126(5). Wynder, E.L. (1987). Workshop on guidelines to the epidemiology of weak associations. Prevention Medicine 16:139- 141.