Comparison of Two Treatments in Animal Carcinogenicity Experiments

Abstract

Abstract In animal carcinogenicity experiments, tumors are often undetectable until death, and thus information about the tumor development rate is confounded with both the mortality rate in animals without tumors and the tumor lethality rate. In Malani and Van Ryzin (1986), the problem of estimating the tumor incidence rate and other related functions was discussed, using data from survival/sacrifice experiments. Estimates were derived using a discrete multistate model. Many of the existing methods for comparing the tumor incidence rate of two treatments either require restrictive assumptions about the tumor lethality or assume that treatment primarily affects the tumor development rate and does not alter the death rate in the presence or absence of tumors. In this article tests are derived for (a) checking the tumor nonlethality and (b) comparing two treatments with regard to the death rate in the presence or absence of tumors. Each test involves testing the hypothesis that a certain odds ratio in a 2 × 2 table is equal to a known constant and then combining results over r fourfold tables. In addition, two new tests for comparing the tumor incidence rate of two treatments are derived. The first test requires the assumption that treatment does not affect the mortality rate in animals without tumors but does not require any assumption about tumor lethality. The second test is derived without any such assumptions.