EcoSta 2019: Start Registration
View Submission - EcoSta2019
A0161
Title: A mixture cure model with multilevel frailties for analysing recurrent event data Authors:  Richard Tawiah - Griffith University (Australia)
Shu-Kay Ng - Griffith University (Australia) [presenting]
Abstract: In the field of medical and health sciences, researchers frequently encounter multivariate survival data consisting of multiple failure outcomes of (recurrent) events (e.g. tumour relapses) from multi-centre studies. Data collection in these studies often exhibits a multilevel structure, inducing strong methodological challenges in modelling intra-subject and between-subject correlation. The complexity in modelling increases further by the presence of long-term survivors who respond favourably to treatment and are thus insusceptible to tumour relapses (a characterised feature is the marginal survival being levelled off to non-zero probability). A new mixture frailty model is developed to analyse multilevel recurrent event data within the context of cure models, in which multivariate random effects with an AR(1) structure are used to impose serial dependence between gap times to recurrent events and another set of random effects are used to model main centre effects and treatment-by-centre interaction effects for uncured patients. The cure probability is modelled by a logistic mixed model with random (patient, centre, and treatment-by-centre) effects. Estimation inference is developed via an EM-type algorithm based on residual maximum likelihood (REML) through the generalised linear mixed model (GLMM) methodology. The method is illustrated using simulated data and a publicly-available dataset concerning a randomised multi-centre trial of rhDNase for treating cystic fibrosis.