Prediction of carcinogenic potency by short-term genotoxicity tests

Past attempts to correlate carcinogenic potencies of chemicals with potencies in causing mutations or other endpoints in short-term genotoxicity tests have found quantitative correlations to be weak except for restricted classes of chemicals. We have classified 379 chemicals on the basis of the rodent carcinogenic potencies calculated by Gold and co-workers as strong carcinogens, moderate carcinogens, weak carcinogens, or non-carcinogens, and have compared the qualitative results of short-term tests (positive or negative) for each of the four classes of chemicals. For most of the short-term tests analyzed, the proportion of positive results increased with carcinogenic potency, being highest for strong carcinogens, lower for moderate carcinogens, lower still for weak carcinogens, and least for non-carcinogens. This differential sensitivity based on carcinogenic potency implies that short-term test responses could be used to predict the carcinogenic potency of unknown chemicals. We have quantified these predictions using Bayesian analysis so that positive or negative results in short-term tests can be interpreted as predicting that a chemical may be a strong carcinogen, a moderate carcinogen, a weak carcinogen, or a non-carcinogen. Possible screening strategies based on these predictions are discussed.