[Recommendations for design and statistical analysis of the rat Pig-a gene mutation assay]

To make recommendations on the design and analysis of Piga gene mutation assay on the basis of the historical control data.

All negative historical control data( total: 128) were collected to create the historical control data base, and descriptive statistics was used to determine the distribution of the data. The power of the number of animals per group, number of measured cells per animal and number of groups were discussed through simulation test based on historical data base.

The group mean of mutant red blood cells( RBCs~(CD59-)) and mutant reticulocytes ( RETs~(CD59-)) were 26. 63 × 10~(-6) and 35. 85 × 10~(-6), respectively, and the standard deviation were 27. 71 × 10~(-6) and 31. 06 × 10~(-6), respectively. By mapping the frequencies of data, the variables had skewed distributions and translated to normal distributions after logarithmic transformation. The confidence level for population means were 100% and 92% for RBCs( 1 × 10~6 cells per animal) and RETs( 0. 3 × 10~6 cells per animal), and it increased to 100% when 1 × 10~6 cells was scored for RETs. Group sizes of 5 had low statistical power while the minimal detectable difference was 2 times. The power had increased to > 80% when 4- or 5- fold of minimal detectable difference was employed.

In brief, the recommendations include:① A log( 10) transformation of mutant frequencies often has been found satisfactory for data when parametric method such as an analysis of variance are used. ② Young adult animals( approximately 6 weeks of age rats) are recommended in this assay. ③ The recommended number of RETs and RBCs are > 1 × 10~6 and > 5 × 10~7, respectively. ④ Based on power calculations, 3 treatment groups and 1 negative control group are appropriate when fourfold increase was employed. ⑤ Group sizes of 6 or 7 are recommended but 5 analyzable animals per group may be acceptable when 4 test groups were used.