Statistical design of toxicity assays: role of genetic structure of test animal population.

This paper concerns certain statistical aspects of the problem of among-strain differences in cancer susceptibility and how these differences may affect the design of toxicity assays. First, the data of Innes et al. (1969) were examined to investigate the magnitude of within-study, between-strain differences in tumor induction. Although there was a very high overall association between mouse strains with respect to the induction of hepatomas, evidence of strain-to-strain variability was found for several compounds. Next, a number of long-term carcinogenicity studies with DDT were considered, and among-strain differences in cancer susceptibility for this compound were noted. Finally, it was shown that if susceptible subgroups do exist, and certain simplifying assumptions are made, then in many cases tumor increases can be detected more readily by studying several inbred mouse strains rather than a single outbred stock.