Adjustment of Alpha for Final Analysis per Actual Number of Events in Event-Driven Clinical Trials Using Group Sequential Testing Procedure

• Activity Number: 441 - Bayesian (and Other) Clinical Trials Designs
• Type: Contributed
• Date/Time: Thursday, August 12, 2021 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Li Wei* and Stephen Lane
• Companies: Bristol Myers Squibb and Bristol Myers Squibb
• Keywords: Adjustment of alpha; group sequential testing procedure; alpha spending function; final analysis; number of events; stopping boundaries

Abstract

Group sequential testing procedure along with various types of alpha spending functions are widely used in clinical trials, providing us great flexibility to take interim looks to allow for early trial termination. The alphas for interim analyses are allocated by the spending function based on the corresponding information fraction. However, in event driven trials, the actual number of events at final analysis is very likely to deviate from the target which is pre-specified in the protocol and has already been used to calculate the information fraction at interim. In that case the alpha for final analysis might need to be adjusted. We compared several possible stopping boundaries, evaluated their impact on overall alpha control and corresponding power under different settings in terms of magnitude of deviation, types of spending functions and information fractions at interim, and identified the one that we think most appropriate.