Repeated dose administration toxicity studies - Use of t-tests in multiplicity data analysis

Abstract

In conducting repeated dose administration toxicity (RDAT) studies with rats and mice, a minimum of three dose groups and one control group are normally set for determing NOEL/NOAEL (no-observed effect level/no-observed adverse effect level) of the test item. For comparison of data among the groups, initially, the data are analysed by analysis of variance (ANOVA). If ANOVA shows a significant difference, then groups means are compared by a multiple comparison range test (MCRT). However, in RDAT studies, at the end of the long duration of the test substance administration, the distribution of the data obtained varies considerably among the groups and the number of animals decreases due to mortality/morbidity, especially in the high-dose groups. Increased variance in the distribution of the data and decreased animals in one or more groups may result in an insignificant ANOVA, though the low-dose group may show a marked difference compared to the control. Dunnett's multiple comparison test is commonly used to compare each treatment group with the control group. However, Dunnett's test has a lower ability to detect significant differences than the t-test, and its detection power decreases with the increase in the number of groups. Therefore, we recommend the t-test, by-passing ANOVA, which has a high detectable significant difference in the two-group test. In addition, the application of the t-test eliminates the need to select an MCRT. However, the final judgment of the adverse effects may be made based on the toxicological relevance in consideration of the statistical analysis results.