ABSTRACT
This chapter focuses on the statistical aspects of efficacy and futility monitoring using group-sequential boundaries and other tools such as conditional power and stochastic curtailment. When conditional power drops below a fixed threshold like 15% is called stochastic curtailment, although some people erroneously use the terms stochastic curtailment and conditional power interchangeably. Many different testing scenarios, such as comparisons of means, proportions, or survival, can be unified into the same Brownian motion framework. Therefore, the same boundaries can be used in these diverse settings. Popular efficacy boundaries are quite high early in the trial, allowing the final boundary to be close to what it would be with no interim monitoring. Inference following a trial that uses group sequential monitoring is complicated by two factors: the trial may have stopped on a “random high” and the sufficient statistic consists of both the information fraction when the trial was stopped and the Z-score at that time.
