Reflections on the declining ability of the Salmonella assay to detect rodent carcinogens as positive.

It is suggested that urgent attempts should be made to define and gain general agreement for the existence of two classes of animal carcinogen, those which are genotoxic and those which are not. In the absence of such a step, attempts to validate in vivo genotoxicity assays, and to derive a meaningful structure-activity database for chemical carcinogenesis, will be frustrated. These suggestions are supported by the preliminary findings of a detailed analysis of the carcinogen database accrued by the United States National Toxicology Program. The possibility that many non-genotoxic carcinogens should be regarded as tumour-promoting agents is considered.