Carcinogenicity prediction and battery selection procedure: an in-depth analysis of cyclamate and its major metabolite cyclohexylamine.

Abstract

The carcinogenicity prediction and battery selection (CPBS) method can be used to predict the probable carcinogenicity of a chemical based on the results of a battery of short-term assays. The method uses Bayesian statistics and the estimated performance characteristics of the assays (i.e., sensitivity and specificity). For routine use, the prior probability of carcinogenicity (or of noncarcinogenicity) is assumed to be unknown and is assigned a nondiscriminatory value of 0.5, i.e., the chemical is assumed to have an equal probability of being a carcinogen or a noncarcinogen, which implies that the expert's intuition regarding the chemical's potential as a carcinogen, based on structural features, known metabolic transformation, or potential electrophilicity, is not taken into consideration. In the present study, it is shown with cyclamate and its metabolite cyclohexlamine that when a battery of assays is used, assigning values to the prior probability between 0.1 and 0.9 has no significant effect on the predicted carcinogenicity. In the view of the fact that the performance of short-term tests is calibrated against known carcinogens and noncarcinogens, and since in the available data bases there is a preponderance of carcinogens, it may be argued that the estimation of sensitivities may be biased toward elevated values. It is shown, however, that when a battery of assays is used, assigning decreased values to the sensitivity does not result in significant effects on the predicted activity of the two test chemicals. Frequently, in the compilation of short-term results within a single assay, different laboratories may report varying results. Heretofore the "consensus result" was derived by majority rule. Because the variability in results may have biological significance, and in view of the fact that for a widely used sweetner one might be even more risk-adverse, a modification of Bayes's formula was derived to take these differences into consideration when calculating the probable carcinogenicity of cyclamate and its major metabolite.