|
Abstract:
|
This talk investigates the use of a general multi-arm multi-stage (MAMS) approach for time-to-event outcomes that would streamline simultaneous comparison of a large number of promising therapies, thus significantly reducing the time and the number of patients needed to test the treatment. Controlling type I error in this setting is harder as the multiple comparison and multi staged approach differs from standard techniques. Historically, the pairwise (PWER) and familywise (FWER) type I error rates are mainly used for control. The focus will be on construction of efficacy and futility boundaries for a MAMS trial in two scenarios. When the same outcome is used throughout, the design is based on the Generalized Dunnett procedure with control of FWER, whereas, when using an intermediate outcome at initial stages, the available methods control for PWER. In the latter, we propose modifications to control for the FWER. Both designs are based on a proportional hazards assumption which is violated in the presence of a delayed treatment effect which results in a loss of power. We suggest an alternative test statistic that can help circumvent this problem to maintain the desired power.
|
