Prediction of risk for human developmental toxicity: how important are animal studies for hazard identification?

To determine how developmental toxicity studies in animals can be used in human risk assessment, a data base was assembled from the literature. It included probable, suspected, unknown, and probably negative teratogenic or embryotoxic drugs and chemical compounds. For each of 175 substances, we recorded the results of any developmental toxicity testing in up to 14 animal species and any reports of mutagenicity or carcinogenicity. Logistic regression and discriminant analysis were used to predict a compound's effect in humans. A measure of the number of positive animal studies and bacterial mutagenicity were important predictors in both models, as were the specific results in hamsters and subhuman primates. The fact that a compound had been tested at all in subhuman primates was more important in predicting human risk than was the result itself. The methods used correctly classified the study compounds 63-91% of the time, depending upon how the suspicious and unknown compounds were treated. The models had a sensitivity of 62-75%, a positive predictive value of 75-100%, and a negative predictive value of 64-91%. These findings imply that studies in laboratory animals carry weight in predicting human developmental toxicity. The more animal species in which a compound is positive, the more likely it is to have a human effect, albeit not the same effect as in the animals. Clinicians should not ignore developmental toxicity tests in animals "because laboratory rats are not like humans."