A quantitative analysis of statistical power identifies obesity end points for improved in vivo preclinical study design.

Abstract

The design of well-powered in vivo preclinical studies is a key element in building knowledge of disease physiology for the purpose of identifying and effectively testing potential anti-obesity drug targets. However, as a result of the complexity of the obese phenotype, there is limited understanding of the variability within and between study animals of macroscopic endpoints such as food intake and body composition. This, combined with limitations inherent in the measurement of certain endpoints, presents challenges to study design that can have significant consequences for an anti-obesity program. Here, we analyze a large, longitudinal study of mouse food intake and body composition during diet perturbation to quantify the variability and interaction of key metabolic endpoints. To demonstrate how conclusions can change as a function of study size, we show that a simulated pre-clinical study properly powered for one endpoint may lead to false conclusions based on secondary endpoints. We then propose guidelines for endpoint selection and study size estimation under different conditions to facilitate proper power calculation for a more successful in vivo study design.