Tobacco Documents Online

1 REVIEW 1136 CONFIDENTIAL Subject ref 8b "Environmental tobacco smoking, mutagens sensitivity, and head and neck squamous cell carcinoma" Z-F Zhang et al Cancer Epidemiology, Biomarkers & Prevention (2000), 9 1043-1049 Evidence relating head and neck cancer to ETS exposure is extremely limited. Early reports by Hirayama" and by Sandler et a13 did not find any association with ETS, and a more recent study by Tan et al", claiming implausibly strong associations with ETS exposure at home and at work had major weaknesses, as discussed in review 936. The present paper describes the results of a case-control study conducted at the Memorial Sloan-Kettering Cancer Center between 1992 and 1994 involving 173 cases of histologically confirmed squamous carcinoma of the head and neck and 176 blood donors without a history of cancer acting as controls. Data were collected on active smoking, passive smoking (at home, at work, spouse), alcohol use, and a variety of other variables. In the abstract the authors note that their study "suggests that ETS exposure may increase the risk of head and neck cancer with a dose-response pattern," although their "results need to be interpreted with caution because of potential residual confounding effects of active tobacco smoking and other methodological limitations." Their conclusion of a suggestive relationship appears to be based mainly on the following odds ratios reported in the abstract: 1) Crude odds ratios for ETS exposure in smokers and nonsmokers combined: 2.8 (95% CI 1.3-6.0), 2) Adjusted' odds ratio for ETS exposure in smokers and nonsmokers combined: 2.4 (0.9-6.8),

4 7) No reference is made to the possibility of misclassification bias, occurring if some of the self-reported never smokers had actually smoked. Clearly far more and better evidence is needed before one can start to believe that ETS exposure might cause head and neck cancer. P N Lee 16.1.01. Reference List 1. Hirayama T. Passive smoking and cancer: an epidemiological review. GANN Monograph on Cancer Research 1987;33:127-35. 2. Hirayama T. Passive smoking and cancer: The association between husbands smoking and cancer in the lung of non-smoking wives. In: Kasuga H, editor. Indoor air quality International Conference on Indoor Air Quality, Tokyo, November 4-6, 1987. Berlin Heidelberg: Springer-Verlag, 1990;299-311. 3. Sandler DP, Everson RB, Wilcox AL Passive smoking in adulthood and cancer risk. Am J Epidemiol 1985;121:37-48. 4. Tan E-H, Adelstein DJ, Droughton MLT, van Kirk MA, Lavertu P. Squamous cell head and neck cancer in nonsmokers. Am J Clin Oncol 1997;29:146-50.

3 4) It seems remarkable that the authors do not present in the abstract estimates of risk associated with ETS exposure in nonsmokers that are adjusted for covariates. After all, we know that adjustment reduced ETS risk estimates for smokers and nonsmokers combined, and it seems quite plausible that it would also do so when attention is restricted to nonsmokers. At first I thought that Table 4 gave us the relevant results, with the crude estimates of 2.2 (0.6-8.4) falling to 1.5 (0.3-6.5) after adjustment for age, sex, race, education, alcohol consumption, marijuana use and pack-years of smoking. But the inclusion of pack-years of smoking in this analysis as an adjustment factor makes no sense, as can be seen from the fact that it virtually eliminates the smoking relative risks. In any case the results seem to be based on one logistic regression analysis with all the patients included, and not on an analysis restricted to nonsmokers. 5) The results for ETS at home and for ETS at work in Table 3 are interesting in that both show a higher odds ratio for occasional than for regular ETS exposure. Subjects were asked "Have you ever been regularly exposed to other people's cigarette smoke at home?" (and similarly for work), with three choices for each question, "never," "occasionally" and "regularly." In principle this makes no sense -the only legitimate answers to the question are clearly yes, no or don't know. Furthermore, the questions are open to recall bias, most obviously as regards the answer "occasionally," with cases being less likely to regard themselves as having no ETS exposure at all. If one were to take only "regularly" answers as the exposed group, taking others with little or no exposure as the denominator, the crude odds ratios then become 0.96 for at home ETS and 1.01 for at work ETS, i.e. no indication of an association (as was also observed for spousal smoking with an odds ratio of 0.9). One must have concern when the conclusions depend so markedly on the definition of exposure. 6) I also find the results as presented for degree of ETS exposure to be deceptive. "Heavy ETS exposure" is merely defined as having reported some ETS exposure both at home and at work, while "Moderate ETS exposure" includes those subjects with some ETS exposure in one place but not the other. This is hardly an adequate way of determining degree of total ETS exposure, and again is subject to recall bias.

2 3) Adjusted* ETS dose-response in smokers and nonsmokers combined: 2.1 for moderate exposure, 3.6 for heavy exposure, trend p=0.025, 4) Crude odds ratio for ETS exposure in nonsmokers: 2.2 (0.6-8.4), 5) Crude ETS dose-response in nonsmokers: 1.8 for moderate exposure, 4.3 for heavy exposure, trend p=0.008. (*Adjusted analyses are adjusted for age, sex, race, education, alcohol consumption, pack-years of cigarette smoking, and marijuana use.) There are a considerable number of important limitations with the study and the way the results have been analysed and presented that render its tentative conclusion that ETS may affect the risk of head and neck cancer most uncertain: 1) The number of lifelong nonsmoking cases, 23, is extremely small. These are the important group for evaluating ETS effects. The number of nonsmoking controls, 46, is also small. 2) The cases represent a mixture of tumour sites, about a third larynx, a third tongue and the rest various other sites. Numbers are too small to separate results by site, but it must be bome in mind that "head and neck cancer" is a heterogeneous mixture which is composed of tumours of varying sites with possibly differing relationships with smoking and ETS. 3) Use of blood donors as controls seems highly unsatisfactory as they are unlikely to be representative of the population at large. There is discussion of this issue on p1047, where it is pointed out that, because the majority of the blood donors were relatives or friends of cancer patients at the Center, they are likely to have characteristics more similar to the cases in this study than one would expect, so tending to understate associations with ETS. However, this is all rather speculative and it would clearly have been better to chose a most appropriate control group in the first place.