ABSTRACT
The principles of design and evaluation are discussed in the following by means of selected items in short-term repeated toxicity studies. The aim of repeated dose toxicity studies is to characterize the toxicological profile of a compound usually administered between 4 weeks [283] and 3 months [284]. Although baseline values or data of a recovery period are available in some studies, the typical data are multiple endpoints at the end of the administration period, such as continuous endpoints (e.g., hemoglobin), rates (e.g., proportions of histopathological findings), or ordered categorical data (e.g., graded histopathological findings). The standard design uses a negative control (C) and several doses D1, ..., Dk where k = 3 is common. A one-way layout will be considered during evaluation, i.e., both sexes are analyzed independently (see their joint analysis in Section 2.4.1). The comparisons of doses versus control are performed by unadjusted two-sample tests (see Section 1.3), or the Dunnett procedure [102] without order restriction (see Section 2.1.1) or the Williams procedure [400] with order restriction (see Section 2.1.2). The usual sample sizes per group i, ni = 10 in rodent studies allow the use of standard tests, including asymptotic ones, though only under particular assumptions, or only for selected endpoints. For all other endpoints and particularly small sample sizes like in dog studies, specific tests are needed. This small sample size restriction is the first feature of statistics in toxicology; it is repeatedly discussed in the book.
