Statistical analysis of teratologic data: problems and advancements.

A large scale replicated dose-response teratology study of 2,4,5-T was done in mice. Variability and variance of fecundity parameters and fetotoxicity endpoints are discussed. Another study on rats indicated less variation among teratologic endpoints than in mice. Calculations for the number of animals in strains of mice and rats needed to detect a 5 percent and 10 percent reduction in mean fetal weight or increase in resorptions are given. We concluded that at least 3 replicates with appropriate numbers of pregnant animals are needed to estimate variance for comparison among laboratories or among species. The utility of these calculations for standardizing teratologic studies is discussed.