Should We Assess Tumorigenicity With the Peto or Poly-k Test?

Abstract

The IARC guidance document on testing for tumorigenicity in long-term animal experiments (Peto et al. 1980) recommended a statistical test for dose-related trend that stratifies animals dying with tumors of a specific type according to whether the tumor was observed because it killed the animal (fatal context) or because of a competing cause of death (incidental context). The test has become known as Peto’s cause-of-death (COD) test or, simply, the Peto test, in recognition of Richard Peto’s lead authorship of the 1980 IARC document. In 1988 John Bailer and Chris Portier developed the Poly-k test for tumorigenicity which does not require a COD determination (Bailer and Portier 1988). Instead it relies on the assumption that the distribution of time to tumor onset can be expressed as a function of the kth power of time. Animals dying without tumors during the course of an experiment contribute to the test statistic according to weights determined by this kth power. Bailer and Portier recommended using k=3 based on an empirical study of historical bioassay data. This article presents construction and characteristics of the Peto and Poly-k tests, a limited comparison of size and power, a view of their regulatory standing, and some modifications.