ABSTRACT
Competing risks data are commonly encountered in clinical trials and observational studies, when subjects are subject to failure from one of distinct causes (Kalbfleisch and Prentice, 1980). In several applications, competing risks data cannot be considered as independent because of a clustered design, for instance in multicentre clinical trials and family studies. Two challenges arise naturally in the analysis of clustered competing risks data. First, methodologies for standard survival data cannot be applied to competing risks data directly without modifications. Secondly, appropriate procedures are needed to account for the correlations of event times among subjects from the same cluster.
