While the primary objective of multicenter clinical trials is to compare treatments for a specific disease, they also contain a lot of relevant additional information. For example, it may be interesting to investigate heterogeneity in outcome and in treatment effect over centers. Considering a frailty model including a random center effect and a random treatment by center interaction term, we demonstrate how to quantify this heterogeneity, to interpret it based on medically relevant quantities, and to identify center and patient specific factors explaining it. Based on the same idea, we also propose to study heterogeneity in prognostic index effect over centers, and to use this information to get a new insight in the validation of prognostic indices. Various approaches are discussed to fit such a frailty model, and a Bayesian approach based on the Laplace approximation is detailled. We illustrate this research using data from an EORTC multicenter clinical trial for breast cancer and from a pooled database of seven EORTC clinical trials in bladder cancer patients.