For most pharmaceuticals, the assessment of carcinogenic risk to humans can be made from information available from a) genotoxicity studies in vivo, b) 3-6 month toxicology studies in two or more animal species and c) clinical investigations in Phase I and II studies aimed at assessing the existence of risk factors (genotoxicity, immune suppression, hormonal activity and chronic irritation/inflammation) associated with cancer in humans caused by pharmaceuticals. The rodent carcinogenicity bioassay is redundant for compounds with such properties. In considering the utility of a bioassay, one must recognize that the outcome of the bioassay has been shown to be predictable for about half of a random selection of chemicals in the US National Toxicology Program (NTP). The predictability of bioassay outcomes for many pharmaceuticals should be even better, given the availability of extensive knowledge on genotoxic potential in vivo and of pharmacological mechanisms, this adding to the redundancy of the bioassay. Furthermore, the value of the bioassay is itself questionable. The inconsistencies in tumor responses between rodent species and strains, the simultaneous tumor increase and decreases within a study and the susceptibility to tumorigenicity from non-genotoxic chemicals by mechanisms now shown to be of no relevance to humans, together make the use of rodents highly misleading as predictors of human cancer risk. For pharmaceuticals with a novel or poorly-understood pharmacodynamic mechanism, useful information on long-term adverse effects that might presage a carcinogenic hazard to humans may be obtained from a 12 month study, usually in rats, conducted at clinically relevant dose levels.