FDA perspective on trials with interim efficacy evaluations.

Abstract

Over the years FDA has seen a variety of 'adaptive' approaches, including repeated calculation of p-values. More recently there has been interest in adjusting sample sizes based on event rates, easy if the treatment group rates remain blinded but possible even with unblinded analysis. In large trials, group sequential analytic approaches are routine. Early stopping however, can reduce available data and in cardiac trials we have generally urged use of only survival as the basis for stopping, even if the primary endpoint is a composite (death, new AMI, etc). Other possibilities include adaptive randomization and starting additional dose groups. Enrichment designs focus on populations with a higher event rate or on populations more likely to respond. They can reduce sample sizes and target therapy toward people most likely to benefit.