Carcinogenicity testing of antitumor agents in animal bioassays has been proposed because of the potential for carcinogenicity of this class of agents and the expectation that such testing may indicate prospectively the target organs of any related human oncogenesis. The literature reveals the anticipated confirmations in animals of the carcinogenicity of many antitumor agents. Furthermore, these agents have been associated with human tumors in numerous case reports. Review of the literature also indicates the inability of animal studies to predict the sites of carcinogen-induced tumors in man. The carcinogenic risk assessment of antitumor agents should begin with the determination of the ability of the agent to interact with DNA. Those agents which are capable of alkylating or binding DNA should be tested for mutagenic and teratogenic potential. The presumption of carcinogenicity should be made for DNA-reactive, mutagenic/teratogenic antitumor agents without requiring confirmation in long-term carcinogenicity bioassays in large numbers of animals. The inability of carcinogenicity studies in animals to accurately predict potential human tumor sites must also be recognized.