Quantitative correlations amongst alkaline DNA fragmentation, DNA covalent binding, mutagenicity in the Ames test and carcinogenicity, for 21 compounds.

21 compounds from different chemical classes were quantitatively compared for their carcinogenic potency according to 4 parameters: (1) potency in inducing covalent binding with DNA in vivo; (2) potency in inducing alkaline DNA fragmentation after treatment in vivo; (3) acute toxicity; (4) mutagenic potency in the Ames test. Establishing well-defined conditions for normalization of the different types of data and determination of the set that had to be submitted to statistical analysis appeared to be a difficult task, for which only compromise solutions were possible. A statistical analysis of the data suggested that all parameters considered were correlated with carcinogenic potency. However, we found that there are about 3 chances to 1 that carcinogenicity is better correlated with DNA covalent binding in vivo than it is to mutagenicity in the Ames test. With due precautions, even acute toxicity could be of predictive value. DNA adducts and DNA fragmentation, both in vivo, appeared to be 2 parameters strongly correlated between them. From a multivariate statistical analysis it appeared that: (1) a significant improvement of quantitative predictability is in principle obtainable with a battery of short-term test; and (2) the improvement is obtainable only if the short-term tests considered, while all correlated with carcinogenicity, are relatively independent amongst themselves.