Models for Combined Action of Alcohol and Tobacco on Risk of Cancer: What Do We Really Know from Epidemiological Studies?

Abstract

There is a considerable amount of literature on the subject of interaction or combined effect of etiological factors on the risk of cancer^1–7. It is, however, still uncertain which model — multiplicative or additive — is adequate for describing the combined effect of alcohol and tobacco on the risk of esophageal, laryngeal and pharyngeal cancer^8–15, even if there is convincing evidence that each factor increases “independently” the risk of these cancers. Even though many analyses of epidemiological data are based on the implicit assumption of a multiplicative model and a good fit is often obtained, the size of most studies does not permit the testing of interaction in the conventional way^16,17. Therefore, the use of such statistical tools for inferring on the underlying biological process would be misleading. The complete understanding of the effect of alcohol on the risk of cancer is in addition complicated by the lack of carcinogenicity of ethanol or alcoholic beverages in animal experiments. As a consequence, there are divergent opinions on the role played by alcohol in increasing the risk of specific cancers among non-smokers. The latter issue has furthermore been confused with the problem of demonstrating a synergistic effect of the two factors because almost tautologically alcohol and tobacco would have a greater effect in combination than alone if the former had no effect in non-smokers. The present paper will deal only with the problem of distinguishing between additive and multiplicative joint effect of the two factors by reanalyzing the data of several epidemiological studies on oral, esophageal, laryngeal and hypopharyngeal cancers with a method recently proposed by Breslow and Storer⁸.