Group Sequential Tests of Futility: The Relationship Between Type II Error Spending and Conditional Power Approaches

• Activity Number: 517
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Thomas W. Dobbins*+
• Companies: Merck & Co., Inc.
• Keywords: Adaptive Design ; Conditional Power ; Futility Boundaries ; Group Sequential Test ; Type II Error Spending ; Stochastic Curtailment

Abstract

In a clinical trial incorporating a group sequential test that allows early stopping for efficacy or futility (GSTEF), or early stopping for futility only (GSTF), two common approaches to testing for futility are Type II error spending and conditional power. As their names imply, these approaches differ in how the probability of incorrectly stopping for futility is specified; consequently, the statistical properties of the two approaches are not easily compared. We utilize a partitioning of the Type II error derived by Dobbins (2011) to obtain the exact relationship between conditional power and Type II error spend for any K-group GSTEF. In particular, the conditional power approach can be completely specified as a GSTEF using Type II spending and non-binding futility boundaries. We describe a simple procedure to obtain the Type II error spending sequence for the interim tests of futility directly from the futility boundary corresponding to a specified conditional power, and vice versa, and we consider an example that illustrates the correspondence.