Pharmacodynamic Models for Cancer Risk Assessment

Abstract

Dose-response models for cancer risk assessment can be classified broadly into two categories. The first class of models to be considered included statistical models that were constructed as empirical descriptions of the dose-response curve in the observed range of data. While these models have good statistical properties and are generally easy to use, they are not based on knowledge of cancer mechanisms and are, thus, unsuited to addressing two of the fundamental problems of cancer risk assessment, namely the problems of low-dose and inter-species extrapolation. It appears reasonable to assume that incorporation of knowledge of cancer mechanisms into dose-response models will lead to more meaningful extrapolations. To that end, various attempts have been made to develop biologically-based dose-response models for cancer risk assessment. Success has been modest, at best. The problem is that, even if the broad features of the model of carcinogenesis are correct, slight mis-specification of the model can lead to large errors in the extrapolation. Thus, in my opinion, models of carcinogenesis have been more useful from the scientific point of view, i.e., in providing a framework within which the process of carcinogenesis can be viewed, than for risk assessment. Nevertheless, their use in risk assessment must be explored thoroughly because only through incorporation of sound biological principles into dose-response modelling is there any hope of a rational approach to the extrapolation problems.