Improving the Analysis of Randomized Controlled Trials: a Posterior Simulation Approach

The randomized controlled trial (RCT) is the standard for establishing efficacy and tolerability of treatments. However, the statistical evaluation of treatment effects in RCTs has remained largely unchanged for several decades. A new approach to Bayesian hypothesis testing for RCTs that leverages posterior simulation methods is developed. This approach (1) employs Monte Carlo simulation to obtain exact posterior distributions with fewer restrictive assumptions than required by current standard methods, allowing for a relatively simple procedure for inference with analytically intractable models, and (2) utilizes a novel approach to Bayesian hypothesis testing.