|
Abstract:
|
Enrichment design is gaining popularity by focusing on biomarker-positive subpopulations that may be more responsive to treatments. On the other hand, the uncertainty on biomarker-negative subpopulations is also of interested, and is suggested in 2012 FDA draft guidance on enrichment design. Among many design and analysis approaches that enable evaluation of both biomarker-positive and negative subpopulations, a weighted statistic approach was proposed to effectively and efficiently obtain the overall treatment effect for a general patient population under an enriched trial setting (Yang et al, 2015). This weighted statistic utilizes a linear combination of the treatment effect in the positive and negative subpopulations. However, this approach does not apply to the commonly used hazard ratio analysis for time-to-event data. To address this challenge, we apply the weighted-statistic principle to an alternative estimand for time-to-event data, namely, restricted mean survival time (Uno et al, 2014). Treatment difference in this alternative estimand can then be combined in a linear fashion across both subpopulations, enabling an appropriate estimate for the overall treatment effect.
|
