Determination of dose dependence in repeated dose toxicity studies when mid-dose alone is insignificant

Abstract

Repeated dose toxicity studies with rodents are regulatory requirements for registering chemical substances like drugs and pesticides with the government regulatory agencies. Usually 4 groups of animals, including a control group, are used in repeated dose toxicity studies. Williams' test, Dunnett's test and Jonckheere's trend test are generally used to evaluate the data obtained from these studies. Selection of a statistical tool is relatively easy, when the data obtained from the groups of animals show a dose-dependency. But, occasionally a significance difference, compared to control, is not seen in the mid-dose group alone, thus losing the dose-dependency. We attempted to find the appropriate statistical tool for analyzing the quantitative data obtained from repeated dose toxicity studies, when the data of the mid-dose group alone do not show a significant difference, compared to control. The commonly used Williams' test to analyse such data has a disadvantage as it assigns an estimated mean value for the mid-dose group, rather than the original mean value, for the analysis. Hence, it is likely that Williams' test may misjudge in establishing a dose dependency, when in reality it does not exists. Therefore, to analyse such data we suggest the use of Dunnett's multiple comparison test, to compare each dose group with the control, followed by Jonckheere's trend test for examining dose dependency.