The use and utility of multistage models of carcinogenesis in cohort studies

Multistage models have been widely employed in recent years and are thought to be useful for the interpretation of biological bases of epidemiological results. We have employed two general models, both interpretable under the multistage theory, using GLIM package for the simultaneous analysis of the effect on mortality of several factors, including age at first exposure, duration of exposure, time interval since exposure ended. In the multiplicative model (relative risk, RR) each variable is supposed to act multiplicatively on the number of expected cases. In the additive model (absolute risk, AR; excess of risk, ER) each variable is thought to act multiplicatively on the excess risk; the result is then added to the expected number of cases. These models were applied to data regarding exposure to aromatic amines and bladder cancer mortality. The results suggested that aromatic amines act on more than one stage of the carcinogenic process, most probably on both early and late stages. In addition, hints for prevention measures were obtained. In fact, the effect on late stages implies a decreasing risk after cessation of exposure, and in this case discontinuing exposure can be useful; an effect on early stages, instead, indicates the necessity of intensive medical surveillance even after cessation of exposure. However, it seems that the possibility of obtaining results interpretable under the multistage theory is at present confined to the rare cases in which there is a substantial excess mortality from specific cancers.