Early stopping in clinical trials and epidemiologic studies for “futility”: Conditional power versus sequential analysis

Abstract 

Early stopping of a clinical trial is well accepted when there is enough evidence for a significant effect. However, during the course of a trial, there can be reasons to consider early termination for “futility.” In epidemiologic studies, costly or destructive laboratory tests or slow case accrual can make it desirable to stop a study early for reasons of efficiency. Estimation of the conditional power (CP) is proposed as a decision tool to stop a study early or to continue it. We consider the disadvantages of this method. We propose (group) sequential continuation of the trial or study as a less arbitrary strategy. We re-analyzed two data sets from the literature to illustrate the advantages of a sequential approach. We conclude that (group) sequential analyses have several advantages over CP. More studies should consider a sequential design and analysis to enable early stopping when enough evidence has accumulated to conclude a lack of the expected effect. Such a strategy can save valuable resources for more promising hypotheses.

Keywords:  Sequential tests, Conditional power, Early stopping, Randomized clinical trials, Epidemiologic studies