The performance of short-term tests in identifying potential germ cell mutagens: a qualitative and quantitative analysis

In risk estimation, the results of rodent carcinogenesis experiments are often used to quantitatively predict effects in man. The justification for this approach has in large part been dependent upon the good correlation of carcinogenic potency found between mice and rats over large numbers of test chemicals. Using the data base of chemicals tested by the NCI Bioassay Program, we observe that there is a very high correlation of the maximum doses tested (max-d) for rats and mice on a milligram per kilogram body weight per day basis. Next we show that the calculated carcinogenic potency (b-defined in the paper) is restricted to an approximately 30-fold range surrounding log(2)/max-d, which has a biological as well as a statistical basis. Since the max-d's for the set of NCI test chemicals vary over many orders of magnitude, it necessarily follows statistically that the carcinogenic potencies will be highly correlated. This "artifact" of potency estimation does not imply that there is no basis for extrapolating animal results to man. It does suggest, however, that the interpretation of correlation studies of carcinogenic potency needs much further thought.