Statistical power in the analyses of brain weight measures in pesticide neurotoxicity testing and the relationship between brain and body weight.

摘要

Few investigations have examined statistical power in studies of pesticide exposure effects on rat brain weight. Using data from developmental neurotoxicity studies conducted by four different laboratories, we evaluated statistical power to detect changes in rat brain weight and examined the relationship between brain and body weight in several age groups. All power calculations assumed an alpha value of 0.05, equal variance between groups, and a sample size of 10/group. Statistical power often varied substantially both within and between laboratories, and for a 5% change in brain weight average between-laboratory differences in power were as large as 34%. Power estimates for individual studies often spanned a range of more than 50% within a given laboratory. The likelihood of detecting a "significant" change smaller than 5% was generally low for all laboratories. Brain weight increased linearly with body weight within age groups but the strength of this relationship decreased with age. Therefore, confounding by body weight may be more likely in brain weight analyses for younger animals. Our findings suggest that caution is required when weighing the importance of statistical "significance" in studies of pesticide exposure effects on brain weight as the power to detect these differences may be low.