Recommendations on the statistical analysis of the Comet assay

Abstract

In 2010, the Statisticians in the Pharmaceutical Industry (PSI) Toxicology Special Interest Group met to discuss the design and analysis of the Comet assay. The Comet assay is one potential component of the package of safety studies required by regulatory bodies. As these studies usually involve a three-way nested experimental design and as the distribution of the measured response is usually either lognormal or lognormal plus a point mass at zero, the analysis is not straightforward. This has led to many different types of analysis being proposed in the literature, with several different methods applied within the pharmaceutical industry itself. This article summarises the PSI Toxicology Group's discussions and recommendations around these issues.

Re-analysis results using medians of the data from the JaCVAM-organized international validation study of the in vivo rat alkaline comet assay

“…TG 489 [3] mentions that statistical analysis can be conducted using various ways [18][19][20][21], and the need to log transform the data and/or add a small number, such as 0.001 to all values to mitigate the effects of zero cell values should be considered [18][19][20][21]. In the individual laboratory analysis, 5 of 12 laboratories used logtransformation values of the medians for statistical analysis, and the others used the non-log-transformation values.…”

Single cell gel electrophoresis (SCGE) and Pig-a mutation assay in vivo-tools for genotoxicity testing from a regulatory perspective: A study of benzo[a]pyrene in Ogg1−/− mice

“…Subsequently, the % tail intensity (TI, i.e. % DNA in tail) per individual was calculated as suggested by Bright and coworkers [23]: 50 comets of one replicate were summarized as median % TI, and three technical replicates were summarized as mean % TI, giving one data point for each animal.…”

Analysis of Nanoparticle-Induced DNA Damage by the Comet Assay

“…In addition, linear regression analysis is also tested to fi nd out a linear dose-response effect. As recommended by Bright et al [ 33 ], the positive control group is Scoring omitted from the linear model (because the variability in the animal summary statistic is usually markedly smaller in this group than in the other groups) and compared separately versus the negative control group to ensure the reliability of the assay. Another recommendation from the same authors is that the focus should be on the confi dence intervals for the treatment effects and not on the p-values because the former conveys the sizes of effects consistent with the study data.…”

“…In vivo, we perform only one experiment, with a minimum of 3 doses of nanomaterials plus corresponding controls, minimum of six mice (or rats) per dose, and two slides per animal, scoring 50 cells per slide. Measurements made on 50 cells per slide have appeared as a recommendation following extensive realistic simulations of both in vitro and in vivo studies that compared measurements on different number of cells per slide [ 33 ].…”