Evaluating statistical analyses and reproducibility of microbial mutagenicity assays.

The NCI/NTP has completed the first phase of a 4-laboratory study on the reproducibility of testing chemicals for mutagenicity in the Salmonella/microsome assay. This paper is a report of the statistical analysis of some of that data. This analysis involved (1) identifying and removing spurious data; (2) determining the adequacy of the remaining data in making a decision on the mutagenicity of the test chemical; (3) performing the statistical tests; and (4) interpreting the results. Using this procedure, 7 approaches were used to determine the mutagenicity of a test. These approaches were the (1) 2-fold rule, (20 modified 2-fold rule, (3) one-way analysis of variance (homogeneity test), (4) test for linear trend, (5) combination of 3 and 4, (6) 97.5th percentile threshold rule and (7) confidence interval threshold rule. The conclusions drawn by each rule were compared to the microbiologists' interpretation, and the results of these comparisons were presented. In addition, the strengths and weakness of each rule were discussed. The reproducibility of the assay in this study was examined, and a discussion of the significance of these results was presented.