Abstract:
|
Biomarkers play an important role in identifying subgroup of patients who are most likely to receive benefit from investigational treatments. In early-stage clinical trials, it is challenging to identify biomarkers due to limited sample size and short study duration. In oncology, multiple surrogate endpoints have been considered to evaluate treatment effects and detect biomarkers, for example, treatment response and progression-free survival (PFS). In this work, we propose to use weighted sum of test statistics for multiple endpoints to improve the power for detecting biomarker-treatment interactions. We derived the optimal weight for combining these statistics. With simulations of tumor growth data, we showed that the proposed method outperforms statistical methods using single endpoint. In addition, we showed that power is robust to misspecifications of parameters, such as the correlation between test statistics for different endpoints.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.