A1741
Title: A flexible joint modeling approach for recurrent events, death, and cure fraction with application to breast cancer data
Authors: Pingyu Lu - National Yang Ming Chiao Tung University (Taiwan) [presenting]
Abstract: In clinical research, patients are typically followed from baseline until the occurrence of a failure event or censoring. However, individuals may experience recurrent events repeatedly, and such events are often associated with, and ultimately terminated by, the failure event. Moreover, with advances in medical treatment, a subgroup of patients may remain free of both recurrent events and the terminal event over a sufficiently long follow-up period, suggesting the presence of a cure fraction. Cured individuals are assumed to be not susceptible to either recurrent events or the terminal event, whereas uncured individuals may experience recurrent events until the occurrence of the terminal event. To appropriately account for the cessation of recurrent events following the terminal event, a conditional additive mean function defined among survivors is proposed. A three-stage estimation procedure is developed: Stages 1, 2, and 3 estimate the cure fraction, the terminal-event model, and the recurrent-event process, respectively. Without specifying the association among these processes, estimating equations for all model parameters are derived. Large-sample properties of the estimators are established and their performance is evaluated through extensive simulation studies. Finally, the proposed method is illustrated using breast cancer data.
