JP Journal of Biostatistics

In some biomedical cohort studies, recurrent or repeated events can be terminated by a dependent terminal event like death. In this case, the process of recurrent events may lengthen or shorten the survival time which indicates the dependence between time of recurrences and death. Furthermore, some observed or non-observed prognostic factors made some patients more prone to experiencing relapse earlier or more than others. Therefore, the dependence between the occurrences of these events and the potential heterogeneity across subjects should be considered. General joint frailty model can assess the effect of covariates on the risk of recurrent and death events, simultaneously. The two gamma distributed frailties in this model can consider both the inter-recurrences dependence and the dependence between the recurrences and the survival times. When the sample size is small, using maximum likelihood estimation may lead to erroneous results. That being so, we propose a Bayesian joint frailty model that not only estimates the effects of covariate on recurrent and death events in data with small sample size but also deduces the origin of dependences. The performance of Bayesian joint frailty model is compared with Classical approaches. Our proposed estimation is evaluated by employing a simulation study and illustrated using a real dataset on patients with breast cancer who have undergone Mastectomy