Evidence synthesis with reconstructed survival data

We present a general approach to synthesizing evidence of time-to-event endpoints in meta-analyses of aggregate data (AD). Our work goes beyond most previous metaanalytic 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 modeling and integrating reconstructed survival information with other types of summary data, to estimate the hazard ratio function and survival probabilities. The method attempts to reduce selection bias, and relaxes the presumption of proportional hazards in individual clinical studies from the conventional approaches restricted to hazard ratio estimates. Theoretically, we establish the asymptotic consistency of MARS, and investigate its relative efficiency with respect to the individual participant data (IPD) meta-analysis. In simulation studies, the MARS demonstrated comparable performance to IPD metaanalysis with minor deviation from the true values, suggesting great robustness and efficiency achievable in AD meta-analysis with finite sample. Finally, we applied MARS in a meta-analysis of acute myeloid leukemia to assess the association of minimal residual disease with survival, to help respond to FDA’s emerging concerns on translational use of surrogate biomarker in drug development of hematologic malignancies.