Tobacco Documents Online

COMMENTS ON ENVIRONMENTAL TOBACCO.SMOKE: A Compendium of Technical Information Chapter 4:- Environmental Tobacco Smoke and Cancer Prepared by: Maxwell W. Layard, Ph.3. I am a partner in Layard Associates, a firm of consulting statisticians in-Alameda, California. I received the Ph.D. degree in Statistics from StanfDrd University in 1969. I was formerly an assistant professor of mathematics, University of California, Davis, a staff ::nember in the Biometry Branch of the National Cancer Institute, and a senior biostatistician in the Veterans Administration clinical trials program. My professional activities have involved statistical analyses of epidemiologic data, including data pertaining to studies of exposure to environmental tobacco smoke and disease incidence. My curriculum vitae is attached. I have been asked to review "Environmental Tobacco Smoke and Cancer," by J.M. Samet, which is Chapter Four of an EPA draft compendium of technical literature on environmental tobacco smoke. In this chapter, Dr. Samet cursorily reviews the epidemiologic evidence concerning exposure to environmental tobacco smoke (ETS) and lung cancer in humans, and discusses the conclusions of three review committees on this subject. He also briefly considers epidemiologic studies of ETS exposure and cancer at sites other than the lung. 1

In his introduction, the author states (p. 42) that.- the World Health organization (WHO), the U.S. Surgeon General,- and the National Research Council judged the evidence [i.e., reports of studies of ETS exposure and lung cancer] sufficient to support the conclusion that involuntary inhalation of tobacco smoke by nonsmokers causes cancer„ As regards the conclusions of the WHO IARC Working Group on tobacco smoking, that statement is not strictly accurate. The relevant passage from the report of the IARC Working Group (IARC, 1986, p. 314) reads: "The observations on nonsmokers that have been made so far are compatible with either an increased risk [of cancer] from 'passive'-smoking or an absence of risk. Knowledge of the nature of sidestream and mainstream smoke, of the materials absorbed during 'passive' smoking, and of the, quantitative relationships between dose and effect that are commonly ob:aerved from exposure to carcinogens, however, leads to the conclusion that passive smoking gives rise to some rLsk of cancer." Clearly the Working Group's conclusion was not based on the sufficiency of the epidemiologic evidence, but on considerations of biologic plausibility, ;3pecifically similarities between the composition of m,3instream smoke and ETS, and an a.9sumed absence of a dose threshold below which there is no carcinogenic effect. This point is acknowledged later in Chapter Four (p. 46), and is dis::ussed again-below. 2

) i It is pertinent to note that-in-an earlier review paper Samet- (1985) reached essentially the same.conclusion regarding the epidemiologic evidence.as did the'IARC Wo:rking Group, perhaps not surprisingly since he was a member of that group. In that paper it was stated that the observed association between ETS exposure and lung cancer did not yet meet the criteria for making causal inferences which were applied to smoking in the 1964 Surgeon General's Report, although he did consider that the criterion of biological plausibility l-iad been met. Apparently at the time of writing the earLier review Samet did not consider that biologic plausibility oE itself was - sufficient to warrant a causal inference. In the present review, Samet does not explicitly review the epidemiologic studies of ETS a:zd lung cancer in the light of the usual criteria for making•ca.i-sal judgments based on epidemiologic evidence. Instead, he briefly reviews some of the studies (pp. 44-45), and offers very little critical analysis of the overall evidence. Two taoles are presented (Tables 1 and 2, pp. 52a-52d), containing summary information about 3 cohort studies and 17 case-control studies. The review of the epidemiologic studies begins with a discussion of the Japanese cohort study reported by Hirayama (1981, 1984). The author states that Hirayama satisfactorily answered most'of the published criticisms of the study, but that assertion is debatable. The fact is that there is a remarkable.paucity of information of Hirayama's puk;~-lished ~ `. 3

reports about the design, conduct, and re:cults-of the study. Most seriously, no explicit information h<<s been published about the number of subjects lost to follow-up. Detailed analysis of the published results on ETS and lung cancer reveals serious internal and external inconsistencies in those results which raise questions about their validity. (See Layard and Viren, 1989.) The author next briefly describes the.Greek case-control study of Trichopolous et al. (1981, 1983), but does not mention the serious criticisms which have been made of that study, although those criticisms were discussed at some length in his own 1985 review. In the following paragraph (p.44), Samet states that the results of subsequently reported case-control studie:; also demonstrated a significantly elevated risk of lung cancer in nonsmokers exposed to ETS, and asserts that the more recent studies "greatly strengthen" the evidence from the earlier studies. It is not clear how he arrived at that conclusion, since he offers no explanation beyond a reference i.o Table 2. In fact, reports of 19 case-control studies have been published since the initial report of the Trichopolous et al. study. Most of those studies used spousal smoking as the index of ETS exposure, and based on comparisons between "non-exposed" and all "exposed"Isubjects, only three of the 19 reported a significant association between ETS exposure and lung cancer. (See the reviews by Layard, 1989,- and Lee, 1989.) The author 4

goes on to say that several of the newer studies- included relatively large numbers of nonsmokers. However, two of the five studies he lists following that remark, namely Akiba et al. (1986) and Gao et al. (1988) did not report a significant elevation in lung cancer risk. (The resu7.t given in Table 2 for one of the listed studies, that of Da7.ager et al., 1986, was based on an analysis of combined data, with 48 lung cancer cases of both sexes, from the Correa et aL., 1983, and Ziegler et al., 1984, studies.) Samet next refers to studies by Knoth et al. (1983) and Gillis et al. (1984) as having been interpreted as- showing an association between ETS and lung cancer, and remarks that both studies have limitations. The Knoth et al. study did not incorporate a control group, and for that reason was discounted by the NRC and Surgeon General's committees; it was not mentioned by the IARC Working Group. The Gillis et-al. cohort study'reported a- spousal smoking relative risk of 3.25 for males which was based on only six lunq cancer cases, and was not statistically significant; the re:lative risk for females was 1.0. These results can hardly be interpreted as showing an ETS-lung cancer association. In the next paragraph, the author notes that other investigations indicate lesser or no effects of ETS exposure Gn Cn on lung cancef risk. He goes on to say that these studies ~ ~ have relatively few subjects and large statistical ?J ~ uncertainty, so their. apparently_.negative-findings are m 0. 5

statistically compatible,with those-of the_"studies-judged-as- positive." It is not entirely clear whicll-studies-Samet is placing in the negative and positive categories. Following these remarks, he discusses, as having negative or weak findings, the case-control studies of Chan and Fung (1982), Koo (1987), Lee et al. (1986), Shimizu et al. (1988), Kabat and Wynder (1984), Wu et al. (1985), and 13rownson et al. (1987), as well as the American Cancer Society cohort study (Garfinkel, 1981). (Another negative study which is not listed in Table 2 is that of Buffler et a:L., 1984.) However, six of the other studies listed in Tables 1 and 2, which by implication Samet categorizes as "positivi~," reported nonsignificant relative risks. The implication that "negative/weak" studies had few lung cancer cases relative to "positive" ;studies is rather I U misleading. Four of the "negative/weak" studies listed above had more than 80 cases, and the Gao et al. (1988) study, surely "weak" with a nonsignificant relative risk of 1.19, had 226 cases. On the other hand, four of the remaining, implicitly "positive," studies had fewer than 30 cases. Apart from remarks, in connection with two of the studies, to the effect that exposure misclassification may have biased the results, Samet in this review offers no further comment on the quality of the epi3emiologic data or on potential biasing factors. (Later in the review he does mention the--NRC committee's analysis of-pDssible_sources_=of--- 6

S l1 I bias.) In fact the studies of ETS and lung cancer suffer from serious methodologic-flaws in their-design-and execution which- could well introduce bias into the results. There is good reason to suppose that biases and confounding factors may have inflated the observed relative risks in many of these studies. Detailed discussions of these points are presented in the recent reviews of Layard (1990) and Lee (1989), and are briefly summarized-here: The study results are weak and inconsistent, and thus do not offer convincing evidence that any observed association is not an artefact-produced by bias or confounding factors. Dose-response relationships are nonexistent or negative in some studies, and none of the studies demonstrates a significant dose-response when attention is restricted to exposed subjects. Some studies display.contradictory results with respect to the lung cancer cell type for which risk elevation with ETS exposure was observed. The results of epidemiologic studies are subject to distortion by various types of bias, and case-control studies in particular are susceptible to selective recall bias, due to the propensity of cases to recall exposure - more completely. None of the studies of ETS and lung cancer used objective ETS exposure measurements such as biologic or environmental markers. One important source of bias in these-studies is under-reporting of current or 7

i r past smoking by professed "never-smokers." This would result in over-estimation of relative risk, since the smoking habits of spouses, as well a:a smoking-and lung cancer incidence, are positively cor:-elated. This point is taken up again below. A number of studies have suggested a,3sociations between lung cancer and factors such as occupation, nutrition, and alcohol consumption. There is evidence that such factors are also correlated with ETS exposure, and they are therefore confounders which could well give rise to spurious associations between ETS and lung cancer. Few of the studies of ETS and lung cancer have controlled for potential confounding factors. Estimates based on epidemiologic data of the relative risk of lung cancer of nonsmokers married to smokers, such as the NRC committee's estimate of 1.25 referred to on page. 46 of Chapter Four, are much higher than would be expected from comparisons of biologi::al markers of smoke exposure between ETS-exposed persons and active smokers. In the case of respirable suspended :particulates, for example, such dosimetric comparisons, coupled with low-dose extrapolation from data on active smokers, lead to risk estimates which are two or mcre orders of magnitude smaller than estimates based on the epidemiologic data. Such huge discrepancies cast doubt on the__validity of-the observed association-between~ETS- 8

j ; r9 t and lung.cancer, and suggest that.the epidemiologic_- results are more_.likely_explained by-bias and.confounding than by an effect of ETS exposure.- On page 46 of Chapter Four, the author discusses the conclusions of the three review committees referred to earlier, namely the IARC, the NRC, and the Surgeon General's committees. He notes that the IARC committee reached its conclusion largely on the basis of biologic plausibility, and states that one factor cited by the committee was the nature of dose-response relationships for carcinogenesis, "which project some risk for any level of exposure." The quoted language, which was not used by the IARC committee, obscures the fact that a no-threshold carcinogenic dose-response is an assumption, not an experimentally verified principle. Samet next discusses the NRC committee's report, noting that after careful consideration of possible sources of bias the committee concluded that the asscciation observed in the epidemiologic studies could not be attributed solely to bias. The NRC committee did not attempt to estimate the possible effects of uncontrolled confounding factors, but did consider the possible effect of misclassification of current- or ex-smokers as nonsmokers, concluding tt.at it was likely to be relatively small. Lee (1988a, 1989) took issue with the NRC committee''s analysis, pointing out tha,t there was an error in their calculations and claiming that their.assumptions about smoking risk and the extent of-.misc]assification_.were- 9

unrealistic. Lee suggested that- miscl.ass:Lf-icat.ion--of smoking== habits could produce an apparent.ETS-1ung_,cancer-association. . similar in magnitude to the average relatdve risk reported in the epidemiologic studies. In view of the dosimetric evidence referred to above, Lee concluded that misclassification bias is a more likely explanation of the observed association than is an effect of ETS. Samet briefly describes the reasoning behind the conclusion of the Surgeon General's Commi-:tee that ETS exposure is a cause of lung cancer, but makes no further comment on it. The Surgeon General's report discusses- dosimetric comparisons of ETS exposure and smoking, and the comparability of low dose extrapolation from smoking data with the epidemiologic results. It reaches quite different conclusions from those of Lee (1988a) and others, on the basis of estimates suggesting that the epidemioLogic data are similar to what would be expected from smoking dose-response models. However, those estimates used cigarette equivalents of ETS exposure, derived from urinary cotinine data, which were higher than those calculated by others (see the NRC report, for example). Moreover, dosimetric comparisons using respirable suspended particulates suggest cigarette equivalents which are far lower than those based on cotinine data (Lee, 19$8a, Robins et al., 1989, McAughey et al., 1989). T W The Surgeon General's report also briefly considers the ~ N questions- of misclassif-ication of smoking status, but ~ N GW 10