Rodent mutation assay data presentation and statistical assessment

摘要

Carr and Gorelick (1995) and Piegorsch et al. (1995) have discussed statistical aspects of study design and data interpretation for transgenic mouse mutation assays. In order to derive reliable data from these or any other assay it is obviously important to study known sources of variability and thereby to diminish or eliminate incorrect conclusions caused by chance or low assay power. It is perhaps equally important for somebody to admit that equations such as:taken from Carr and Gorelick (1995) mean little to most investigators or regulators. On the surface that probably does not mutter so long as the overall message regarding animal group sizes/plaque numbers etc. is accepted. However, it is not that simple.The elegance and complexity of the statistical equations shown in these two papers implies that the transgenic assays have well studied and regular sources of variability that have been precisely modelled. Our experience of these assays is not like that. We cannot pin down so convincingly the causes of the differences between apparently identical repeat experiments. Because of this we have adopted the conservative approach of representing our group data, as bar charts with standard deviations. When itseems as if an increase in mutant frequency has occurred in a test group, we conduct a repeat experiment using the same sample of DNA. If necessary, we also re-isolate DNA from the frozen tissue, or repeat the whole experiment. This means that at the present state of development of these assays in our laboratory we incorporate the study design advice of Piegorsch et al. (1995) and Carr and Gorelick (1995) but we look at the test data and require to be convinced by appropriate repeat experiments - we donot conduct statistical assessments of the data. What we are attempting to avoid is handing over responsibility for the assay outcome to a standard statistical package. This was not advised in either of the above two papers, but the trend to it was implicit, and it will happen when such assays are eventually accepted for regulatory use.