|
Abstract:
|
Randomized clinical trials (RCTs) represent the gold standard to assess whether an intervention (drug or biologic) is causally associated with a clinically favorable benefit to risk ratio in subjects. In a RCT, the analysis of covariance (ANCOVA) model, also called a pre-post model, is a common approach that uses the within subject correlation between baseline ("pre") and final ("post") measurements to increase efficiency. Enrichment of "high risk" subjects according to their baseline measurement is also common in prevention trials. In these cases regression to the mean (RTM) may arise and a consequence of this may be reduced within subject correlation, which will affect power in the ANCOVA design. In this talk we present analytic results and quantify the impact of trial enrichment through evaluation of statistical operating characteristics (scientific estimand, bias, precision, power) under RTM and mean-variance relationships when pre-post models are utilized. We discuss the pros and cons of enrichment in prevention trials generally, and in the context of the Nicotinamide in Early Alzheimer’s disease trial (NEAT), an ongoing phase 2 proof of concept trial.
|
