Compilation and use of genetic toxicity historical control data

The optimal use of historical control data for the interpretation of genotoxicity results was discussed at the 2009 International Workshop on Genotoxicity Testing (IWGT) in Basel, Switzerland. The historical control working group focused mainly on negative control data although positive control data were also considered to be important. Historical control data are typically used for comparison with the concurrent control data as part of the assay acceptance criteria. Historical control data are also important for providing evidence of the technical competence and familiarization of the assay at any given laboratory. Moreover, historical control data are increasingly being used to aid in the interpretation of genetic toxicity assay results. The objective of the working group was to provide generic advice for historical control data that could be applied to all assays rather than to give assay-specific recommendations. In brief, the recommendations include: 1. The experimental protocol should remain fixed throughout the period during which the historical control data relevant to the current experiment are being built up, unless it can be demonstrated that changes to the protocol have not affected the values. 2. All data (both individual and group mean values) should be accumulated. 3. No negative control values (i.e., vehicle/solvent controls and absolute/culture medium controls, when available) should be eliminated from the data set, even if considered unusual, unless there is a scientifically justified reason, such as when they were obtained by an identified technical error. However, experiments may need to be repeated if disqualified by historical control data. 4. A minimum set of data resulting from at least 10, preferably 20, independent experiments is recommended to create the historical data set, depending upon the complexity of the assay. 5. It is not appropriate to use the simple range (minimum and maximum value observed during the data accumulation period) of the accumulated historical, especially negative, control data for an assessment. Rather, the distribution of the data together with appropriate descriptive statistics should be considered (e.g., confidence intervals, 95–99% percentiles). 6. For an experiment, when statistically significant increases over the concurrent negative controls (i.e., vehicle/solvent controls) are comparable, i.e., within confidence intervals, with the negative historical data, the biological importance needs to be ca