What is an appropriate measure of exposure when testing drugs for carcinogenicity in rodents?

This discussion paper argues that in the carcinogenicity testing of drugs, biological measures of drug exposure may often be more relevant than the classical pharmacokinetic approaches applicable to reversible pharmacodynamic phenomena. Chemicals that produce tumors in rodents may do so (either directly or after bioactivation) by mechanisms involving inter alia a mutagenic, cytotoxic, or hormone-like effect. Such mechanisms may involve the formation of reactive metabolites of fleeting existence, and these are subject to the principles of irreversible pharmacokinetics. Examples are given of genotoxic and nongenotoxic substances for which the species and target site for tumor formation correlates not with plasma concentration, but with the amount metabolized and/or the rate of metabolism. Other compounds produce tumors in rodents, often in only one sex or species, in association with an exaggerated pharmacodynamic effect, with an increase in liver weight (by one of a diversity of mechanisms) or with hormonal or hormonal-like effects. In such cases the determinant of tumor formation is the degree of disturbance of homeostasis, not the plasma concentration of the parent substance. For drugs, the disturbance in question may have no relevance to the clinical use of the drug. Plasma concentrations of the parent substance are useful in assessing the proportion of an oral dose that is absorbed and the linearity of kinetics over the full dose range and for exploring the differences between dietary and gavage administration. They do not usually, however, provide information of direct relevance to assessment of "exposure" for the purposes of carcinogenic risk assessment.