Assessing replicability and power estimates of behavioral performance of control rats across standardized...

Behavioral assays are critical in evaluating impacts on nervous system function in rodents due to genetic or environmental factors and are frequently incorporated into regulatory decision-making studies. Despite numerous sources of guidance for such studies, results across behavioral assays are reputed to be highly variable with questionable replicability. Behavioral data obtained from control rats within four contract laboratory studies were used to evaluate replicability across studies, calculate the level of statistical power, and estimate the number of animals required for a specific effect size. For the three behaviors evaluated here (motor activity, acoustic startle response, and learning and memory), control rats from all studies showed the expected pattern of behavior, e.g., open field acclimation, startle habituation, % prepulse inhibition (PPI) over pre-pulse intensities, and acquisition and goal quadrant preference in the Morris Water Maze (MWM). For selected representative individual endpoints, power analyses were conducted to evaluate sample size requirements. Across all endpoints, a drop in power occurred as differences between two groups became smaller. Power analysis of multiple representative endpoints suggested that a sample size of 20 may detect a 30 % effect with 80 % power. Sample size requirements changed with the effect size, and achieving 80 % power with a 20 % effect size generally required a sample size of 30 rats. While the behavioral performance was replicated over the Study Cohorts, power analyses suggested a need for moderation of expectations regarding detectable differences if decisions relied on single endpoints or small effect sizes. Reporting results from a low powered study can have significant and wide-ranging impacts, including undermining confidence in data interpretation, misleading future research, and failing to adhere to the ethical framework of the 3 R's.