Confidence Intervals and Point Estimates for Adaptive Group Sequential Trials

• Activity Number: 416
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Lingyun Liu*+ and Cyrus Mehta and Ping Gao and Pralay Senchaudhuri and Pranab Ghosh
• Companies: Cytel Inc. and Cytel Inc. and The Medicines Company and Cytel Inc. and Cytel Inc.
• Keywords: Adaptive design ; median unbiased estimate ; confidence interval ; stagewise ordering ; flexible design ; conditional type I error

Abstract

Adaptive sequential designs have been intensively investigated in the literature. It is well known that type I error can be preserved by preserving the conditional type I error. The inference problem was addressed by Mehta et al (2007). This approach (RCI), however, is only guaranteed to provide conservative coverage of the treatment effect. In addition, this method cannot produce an unbiased point estimate. Brannath et al (2009) generalizes the stage wise adjusted confidence intervals (SWACI) of Tsiatis et al (1984) to adaptive setting. This method provides nearly exact coverage. Both of these two methods are implemented in East®. The SWACI method is limited to one-sided test and is only applicable when there is a single adaptive change through the whole trial. For one-sided test, the SWACI method can only provide either lower or upper confidence limits but not both at the same time.