Carcinogenicity of chemicals: the weight of evidence.

1. The evidence discussed here is derived from epidemiology, long-term bioassays in laboratory animals, and predictive short-term tests. 2. Epidemiological data are obtained directly from human studies and are most compelling when they demonstrate a large relative risk and a clear dose-response in association with a distinctive tumour type. Exposure to a suspected carcinogen and the doses involved are, however, often difficult to determine, and the most sophisticated epidemiological methods are relatively insensitive. There are no epidemiological data for most occupational/environmental chemicals. 3. Long-term bioassays can present major problems in design, interpretation and extrapolation. Particular difficulties are associated with the planning of appropriate dose levels and the occurrence of certain tumours at high incidence in both control and test groups. Results from animal bioassays set priorities for concern and action but they cannot be reliably used for quantitative assessment of human risk. 4. Evidence of potential carcinogenicity derived from short-term predictive tests, involving a wide variety of systems with diverse end-points, is increasingly important. Emphasis is placed on the need for more in vivo procedures with a broadening of the scope of somatic cell targets.