A Multiresolution Model for Multicenter Survival Studies

The Cox proportional-hazards model is a popular method for estimating effects of covariates on survival from possibly censored follow-up data. However, in multicenter studies, there is often a need to make inference about a population survival curve based on multiple, possibly heterogeneous survival data from individual centers. We propose a flexible Bayesian method for estimation of a population survival curve based a semiparametric model that accounts for center heterogeneity. The method provides a smooth estimate of the survival curve for ``multiple resolutions'', i.e.,\ for multiple sets of time points of interest. The Bayesian model used has the capability of accommodating general forms of censoring and {\it a priori} smoothness assumptions. We also develop model checking and diagnostic techniques based on the the posterior predictive distribution. The method is used to analyze data from a large multicenter clinical trial of tamoxifen in the treatment of breast cancer with 110 participating centers. Of particular interest are the estimates of center heterogeneity in the baseline hazard curves and in the treatment effects.