ABSTRACT
The purpose of a long-term carcinogenicity study is to evaluate the tumorigenic potency of a compound when it is administered to rodents for most of their life span [109, 270]. An FDA guideline describes study design and statistical analysis [14]. But, more in detail the US National Toxicology Program proposed a balanced design with a control and three dose groups with sample sizes of 50 animals for each sex and related statistical tests [9]. The primary objective is to identify a dose-dependent trend. Commonly the Cochran-Armitage test and its modifications are used. However, because of its limitations (see Section 2.1.6 for details), Williams-type tests (or just Dunnett-type tests for unrestricted comparisons against control) are used here. The primary endpoint in a carcinogenicity study is the number of tumors (more precisely the number of animals developing tumors). However, longer living animals develop tumors more likely than those dying earlier. Therefore the evaluation of crude tumor proportions could be biased, and thus both their mortality-adjusted analysis and the analysis of their survival as secondary endpoint are suggested. Commonly, the joint testing of age-adjusted tumor lethality (fatal tumors) and age-adjusted tumor prevalence (incidental tumors) is performed [300]. However, this classification based on the availability of the valid cause-of-death information for each individual tumor in each individual animal. In some studies this classification of a particular tumor into incidental, fatal, or mortality-independent is difficult or uncertain. As an alternative, the poly-k approach [34] was introduced where cause of death information is not necessary.
