The rat bone marrow micronucleus test--study design and statistical power

Abstract

Although the rodent bone marrow micronucleus test has been in routine use for over 20 years, little work has been published to support its experimental design and all this has used the mouse rather than the rat. When it was decided to change the strain of rat routinely used in this laboratory to the Han Wistar, a preliminary study was performed to investigate the possible factors influencing experimental variability and to use statistical tools to examine possible study designs. Subsequently, a historical database comprising of vehicle controls accumulated from 65 studies was used to establish test acceptance criteria and a strategy for analysing equivocal results. The following conclusions were made: (i) no statistically significant differences were observed in experimental variability within or between control animals; although not statistically significant, the majority of experimental variability seen was found to be between separate counts on the same slide, with minimal differences found between duplicate slides from the same rat or between individual rats; (ii) power analyses showed that, if an equivocal result is obtained after scoring 2000 immature erythrocytes (IE), it is appropriate to re-code the slides and score an additional 4000 IE, i.e. analysing a total of 6000 IE; no meaningful increase in statistical power is gained by scoring >6000 IE; this is consistent with the variability observed between separate counts on the same slide; (iii) there was no significant difference between the control micronucleated immature erythrocyte (MIE) values at 24 and 48 h after dosing or between males and females; therefore, if an unusually low control value at either time point results in apparent small increases in MIE in a treated group, it is valid to pool control values from both time points for clarification and (iv) similar statistical power can be achieved by scoring 2000 IE from seven rats or 4000 IE from five rats, respectively. However, this is based only on control animals and does not consider possible differences in responses between animals to treatment with a potential genotoxin. In order to minimize the possible influence of responders and non-responders, the preferred study design in this laboratory is to score 2000 IE from groups of seven rats. Study data obtained over time confirmed observations made in the control study. Also from an ethical viewpoint, clarifying equivocal responses by combining control data from the 24- and 48-h time points and/or increasing the number of IE scored per animal has minimized the numbers of repeat studies necessary to determine the genotoxic status of a novel compound. However, before any laboratory can use these procedures, experimental data must be generated to demonstrate their validity.