Abstract
Randomized clinical trials are often performed to determine the efficacy of new drug treatments. In recent years, there has been an interest to use group sequential methods in such trials to allow for the possibility of early stopping when there is clear evidence of an established benefit. The purpose of this paper is to develop a new statistic for monitoring the significance of observed results in a K-stage group sequential test. When K is fixed in advance, the statistic can be used with sequential data collected at equal interim stage sample sizes. When K is not fixed, but the information fractions of the past and present analysis stages are known, the statistic can be formulated through the α-spending function for any family of testing procedures.
Acknowledgment
We appreciate the suggestions from the Associate Editor and an anonymous referee which have improved our presentation.