Assessing the impact of extreme values in clinical studies−a latent variable approach

Abstract

Assessment of change in exercise capacity using the 6-minute walk distance (6MWD) tests has been the primary endpoint in many pulmonary arterial hypertension (PAH) and neuromuscular disorder clinical trials. However, large power losses were observed in the primary endpoint of 6MWD analysis in well-powered placebo-controlled studies. One study is a new drug application (NDA) of drisapersen in 2015. The drisapersen NDA included 3 placebo-controlled studies to demonstrate efficacy for Duchenne Muscular Dystrophy, a rare progressive neuromuscular disorder that is ultimately fatal for boys at a young age. Change in 6MWD is the primary endpoint. While the two smaller proof-of-concept pilot studies showed consistent treatment differences, the larger and only well-powered placebo-controlled study failed to detect a treatment difference. The statistical power of the pre-planned primary analysis was reduced from the planned 90% to only 53% as a result of the increased standard deviation from the planned 55 meters to the actual 87 meters, based on the parametric model of mixed model repeated measurement (MMRM) assuming normal data distribution.