Abstract:
|
There has been increasing interest in testing the interaction between the treatment group and biomarker to identify patient subgroup with enhanced treatment benefit. Statistical power is the main challenge in testing the interaction effect. Meta-analysis of treatment-biomarker interaction across trials can be performed to enhance power, but it can also involve subtle statistical issues with challenges in data harmonization from various data sources. In this talk, we review two approaches of fixed-effect meta-analyses: standard inverse-variance weighted meta-analysis and meta-regression. We showed the relative efficiency of the two methods depends on the ratio of within- versus between- trial variability of the biomarker. We then propose to use an adaptively weighted estimator (AWE), between meta-analysis and meta-regression, for the interaction parameter. The AWE used only summary statistics from each trial without sacrificing efficiency as compared to joint analysis using individual level data, and bypasses issues with sharing individual level data. The accompanying Wald-type test, based on AWE and its asymptotic variance, is shown to have enhanced power for detecting treatment-bi
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.