Comparison and evaluation of some experimental designs for use in carcinogen screening.

The development and evaluation of experimental designs for routine in vivo screening of chemicals for potential carcinogenic activity were considered. Such designs have played an important role in the Carcinogenesis Bloassay Program of the National Cancer Institute (NCI). In particular, the current one-stage 50-animal/group screen used by the NCI was considered. A specific two-stage alternative was proposed in which 35 animals/group were used; this alternative allowed for retesting of equivocal compounds. The proposed designs were evaluated in terms of sensitivity, specificity, and throughout. Despite the large number of tests made for each compound, the false-positive rate was found to be less than 0.07 for the current screen and less than 0.05 for the proposed two-stage alternative. The power of the one-stage and two-stage screens was comparable. The two-stage screen was shown to make about 30% more decisions per test period with a savings of around 28% in the expected number of animals needed per compound tested.