A robust statistical procedure for evaluating genotoxicity data

摘要

The statistical evaluation of in vitro genotoxicity data is commonly performed by dose-response analysis. Common trend tests cannot be recommended because non-monotonicity may occur due to the use of high dose levels. Several decision tree approaches using preliminary comparisons versus control and conditional eliminating of high doses were proposed. Their heuristic and conservative behavior can be overcome by multiple contrast test, which test on trend and on comparisons versus control simultaneously. Other versions of protected trend test were investigated—each sensitive to any shape of the dose-response. A simple bootstrap approach is proposed using SAS PROC MULTTEST for the analysis of real genotoxicity data.