Bayesian adjusted R 2 for the meta-analytic evaluation of surrogate time-to-event endpoints in clinical trials

摘要: A two-stage model for evaluating both trial-level and patient-level surrogacy of correlated time-to-event endpoints has been introduced, using data when multiple clinical trials are available. However, the associated maximum likelihood approach often suffers from numerical problems different baseline hazards among imperfect estimation treatment effects assumed. To address this issue, we propose performing second-stage, evaluation potential surrogates within a Bayesian framework, where may naturally borrow information across while maintaining these realistic assumptions. Posterior distributions on measures interest then be used to compare or make decisions regarding candidacy specific endpoint. We perform simulation study investigate differences in performance between traditional new representations common meta-analytic measures, assessing sensitivity characteristics such as number trials, trial size, amount censoring. Furthermore, present frequentist evaluations time recurrence overall survival two meta-analyses adjuvant therapy colon cancer. With results, recommend an attractive numerically stable alternative multitrial assessment surrogate endpoints.