Tobacco Documents Online

Review of the April 1990 Internal Draft Document "Lung Cancer Hazards and Other Respiratory Effects Due to Exposure to Environmental Tobacco Smoke". Reviewer: Larry Glass J/fi ^~ T i l ~t i co ox og s Chemical Mixti,Yres Adses)sment Branch This document represents an impressive effort to summarize the data regarding the potential health effects associated with Environmental Tobacco Smoke (ETS). Given the scientific evidence on the health effects associated with "direct" exposure to tobacco products, it is intuitively appealing to assume that "indirect" exposure may also be associated with untoward effects. However, while the authors are to be commended for this fine effort, I have several concerns about using this review as the basis for the classification of ETS as a Group A carcinogen. It is suggested that ETS be classified as a Group B1 carcinogen. SPECIFIC ISSUES 1) The page titled "Preface" is blank as is the page titled "Authors, Contributors, and Reviewers. Is there a reason for this? 2) Figures 2-1, 2-2, 2-3, and 2-4 (pages 2-6, 2-20, 2-23, and 2-25 respectively) are missing. 3) The document specifically addresses two potential health effects associated with exposure to ETS -- i.e. lung cancer in women and respiratory disorders in children. It is suggested that the title be changed to more accurately reflect that the review is not about all respiratory disorders, but is limited to children. GENERAL ISSUES 1) As clearly cited in the document, these issues nave already been extensively reviewed by both the U.S Surgeon General (SG) and the National• Research Council (NRC) . The purpose of this report was to extend the analyses of those committees. It is curious that while both of the other reports were done in the same time frame, and presumably reviewed the same data, that the NRC concluded that ETS was associated with lung cancer but that the SG concluded that ETS was causally related to lung cancer. The authors of this report also conclude that ETS is causa ly related to lung cancer. Since the relationships being discussed in ':his report are obviously complex, it would be useful to dedicate a few paragraphs

for additional discussion of the other reports and their relationship to this one. In particular, it is suggested that the authors include supplementary text reviewing the interpretation of the NRC and how this report offers additional information that allows for a different position. 2) There were numerous limitations in the methods associated with the exposure assessment that the authors attempted to correct using statistical modeling techniques (issues associated with these models are beyond the scope of this review). In turn, the difficulties in addressing the exposure in the different studies produces severe restrictions on the final exposure-response analysis obtained by pooling the results of the different studies. Representative issues include: a) The exposure classification schemes usec to characterize a person as a smoker vs. a nonsmoker were very complicated and varied between studies. opportunities for misclassification were attributed to reasons as diverse as questions asked in the individual investigations to how the individual investigators classified smokers vs. non-smokers. For example, in at least one study (Lam et al., 1987) never-smoker status was defined as having not smoked for at least one year. In other studies the methods for assigning ETS exposure are not discussed. Additionally, in some societies, smoking among women is seen as a very negative trait. Female participants in these cultures may not reveal their true status (p. 2-36). b) Some investigators (e.g. Hirayama, 1984 and Lehnert, 1984) also classified exposure to include distance from the smoker during the average exposure. This was categorized as direct passive smoking (1-1.5 meters from source) and indirect passive smoking (> 1.5 meters). c) There are substantive differences between Mainstream and Sidestream smoke (ETS) and consequently it is difficult to generalize about the properties of one from studies on the other. These include differences in their relative compositions and physicochemical properties. Additionally, measures of ETS are also complicated by distance of receptor from source (see above), and "aging" (which changes the composition and properties) of the ETS. Exposure to ETS is clearly not described by a simple binary distribution. d) Age related susceptibility (a surrogate measure of exposure) was shown to be a poorly understood variable as demonstrated by the discrepancies in the observations of at t".) least three studies. Lehnert (1984) was described as having p demonstrated that the relationship between ETS and probability ~ ~ ~ ~ c~o O ~ C~

of death from lung cancer declined with age. This was substantiated by citing the work of Wells (1988) who postulated that there may be susceptible subpopulation, who for various reasons, have a reduced rate of survival to old age. In contrast, Vutuc (1984) is cited as having observed a marked relationship between higher age grcups and ETS and having concluded that the carcinogenic effect of tobacco smoke is strongly influenced by the duration of exposure. If possible, these diametrically opposed differences should be addressed -- one supports an age-dependent variable associated with duration of exposure, while the other indicates the possibility for early death of the susceptible population. " 3) Usually the designation of a "causal relationship" has been reserved for results obtained from analytical studies (case-control and cohort) in which individual risk could be shown to be a function of exposure. In general, the demonstration of an exposure and effect relationship is obviously limited by the quality of the measures of exposure, the assessment of the exposed population, and having the necessary controls for extraneous factors. Since these types of studies are extremely expensive and time consuming, most investigations tend to be limited to the provision of supportive evidence. To compensate for theses types of limitations, investigators have been applying increasingly sophisticated statistical tools to large data sets. For example, as noted by the authors, meta-analysis can be used to surmount the limitations of any given study by statistically pooling the results of numerous investigations as a means of increasing the statistical power to detect a small increase in disease. As noted above, there were numerous limitations in the data set which may be partially abated by pooling the results of the individual studies. Thus, as demonstrated by the authors, this approach may provide a defensible means to develop an interpretation that would not have been possible using any single study. Nonetheless, it is my understanding that this type of an approach is still comparatively new and controversial. Finally, there are tremendous scientific, regulatory, and political ramifications of categorizing a substance as a Group A carcinogen., With all due respect to the epidemiologists who produced this report, given the inherent limitations of the data, and the comparative novelty of the approach used to interpret the data I~ would recommend that this approach not be used as the basis of a p Group A classification. ~ o'a In line with the "Workshop Report on EPA Guidelines for Carcinogen W" Risk Assessment: • Use of Human Evidence" (Sept. 1989) it is CR GO O C.r C~

recommended that the evidence be summarized as being limited (i.e. "...a causal association is more credible than nonrandom error, although [confounding] cannot be entirely ruled out as an explanation..."). This would classify ETS into a weight-of-the- evidence Group Bi. cc: R. Hertzberg W. R. B. B. Peirano Schoeny Sonawane