Statistical evaluation of mutagenicity test data: recommendations of the U.K. Environmental Mutagen Society

Most of the many published guidelines on how to conduct mutagenicity tests do not give advice or references on statistical analysis of data. The U.K. Environmental Mutagen Society decided to address this omission, and in 1985 established 8 working groups comprising genetic toxicologists from all sectors of the science, plus at least 2 statisticians per group, to produce statistics guidelines on 10 different test systems. Each group gave advice on how to determine the suitability of data for distribution fitting, when data are unsuitable, when and how data should be transformed, which statistical tests are suitable for a given set of data, which factors govern the choice of statistical test, an order of preference, and some worked examples using real data. In addition, groups gave advice on statistical issues in the design of experiments. Strong recommendations were made that for in vitro tests, sufficient cells be treated and sampled to provide meaningful values of spontaneous mutant/aberration frequencies, for genuine, independent replicate treatments to be used, and that the acceptability of an experiment should be based on homogeneity between replicates as well as comparison of negative and positive control responses with historical ranges. It was recommended that most in vitro studies should include independent repeat experiments, and advice was given on how to check for consistency between experiments and then combine data for further significance testing.(ABSTRACT TRUNCATED AT 250 WORDS)