Repeated Confidence Intervals for Adaptive Group Sequential Trials
by
Cyrus R. Mehta
Cytel Inc.
Coauthors: Peter Bauer, Martin Posch and Werner Brannath

This paper proposes a method for computing conservative confidence intervals following a group sequential test in which an adaptive design change is made one or more times over the course of the trial. The key idea, due to Muller and Schafer (2001), is that by preserving the null conditional rejection probability of the remainder of the trial at the time of each adaptive change, the overall type I error, taken unconditionally over all possible design modifications, is also preserved. We show how this principle may be extended to construct one-sided confidence intervals by applying the idea to a sequence of dual tests derived from the repeated confidence intervals proposed by Jennison and Turnbull (1989) for the interim monitoring of a group sequential clinical trial. The extent of the conservatism is studied by simulation and shown to be fairly modest if the O'Brien-Fleming error spending function (Lan and DeMets, 1984) is adopted for early stopping. Furthermore, the conservatism of the confidence bounds can be eliminated completely if the only purpose of the interim monitoring is to modify the trial design, without any possibility of early stopping. In that case the method also provides a median unbiased point estimate of the treatment effect. The statistical methodology is illustrated by application to a clinical trial of deep brain stimulation for Parkinson's disease.

Date received: April 3, 2007