Statistical analysis of dose-response data from in vitro assays: An illustration using Salmonella mutagenicity data

One of the advantages of in vitro methods is ease of testing multiple concentrations of chemicals for dose responses. A well defined response is used to confirm toxicity, especially for marginal effects, and a significant dose response alone indicates some effect of treatment for further clarification. For sublethal processes (e.g. enzyme induction, organelle damage or mutagenesis), decreases in response can arise at high dose levels that kill cells (cytolethality). This compromises statistical analysis of dose responses using standard approaches, such as ranking, which do not allow for omission of cytolethal dose data. An alternative is the recursive, non-parametric S (M) ( built1 2 )/ Jonckheere-Terpstra test , where omission of results is permitted. Use of this approach is illustrated in conjunction with recommended statistical analyses (Dunnett's 't'-test and Wahrendorf ranking) for non-parametric data from Salmonella mutagenicity assays where the problem is frequently encountered. It is shown that the recursive test can be used for analysing non-parametric dose responses from in vitro assays, where decreases in response are seen at high test chemical concentrations.