|
Abstract:
|
We present a novel approach for meta-analyses of aggregate data (AD) to synthesize evidence of time-to-event endpoints. Our work goes beyond most previous meta-analytic research by using reconstructed survival data as a source of information. A Bayesian multilevel regression model, called the “meta-analysis of reconstructed survival data” (MARS), is introduced, by combining three major types of survival information, to estimate the hazard ratio function and survival probabilities. The method attempts to reduce the selection bias, and relaxes the presumption of proportional hazards in individual clinical studies from the conventional meta-analyses using hazard ratio estimates. Theoretically, we establish the large sample properties, including consistency and relative efficiency, and provide the conditions under which the MARS is as efficient as the individual participant data analysis. Simulation studies demonstrate that our method has excellent performance in a moderate sample size. Finally, we apply the MARS in a meta-analysis of acute myeloid leukemia to assess the association of measurable residual disease with survival.
|
