Recurrent Events Methods in the Analysis of Clinical Trials for Cardiac-resynchronization Therapy

• Activity Number: 509
• Type: Topic Contributed
• Date/Time: Thursday, August 11, 2005 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Rui Song*+ and Michael Kosorok and Susan Anderson and David Breiter and David L. DeMets and Elizabeth Galle and Michael Gruber
• Companies: University of Wisconsin, Madison and University of Wisconsin, Madison and University of Wisconsin, Madison and Guidant Corporation and University of Wisconsin, Madison and Guidant Corporation and University of Wisconsin, Madison
• Keywords: Recurrent events ; Informative censoring ; Model misspecification ; Cardiac morbidity ; Frailty

Abstract

Recurrent events data with informative censoring are encountered frequently in clinical trials. Consider the COMPANION trial of Cardiac-Resynchronization Therapy. In addition to the primary endpoint of time to first hospitalization or death, an important secondary endpoint consisted of possibly recurrent cardiac morbidity events, such as hospitalization for nonfatal events or resuscitated cardiac arrest. An important issue is the dependent censoring of these recurrent events due to death. Frailty models and inverse probability of censoring weighted (IPCW) methods are two recent techniques dealing with informative censoring. Because the correctness of the estimated censoring mechanism is important in the dependent censoring setting, results of these methods may differ and it may be unclear which method is most appropriate. We implement mismatch simulations to study the behavior of these two methods under model misspecification. The robustness of the methods are discussed and some inferential tools are illustrated. We also present practical examples from the COMPANION trial to illustrate the proposed data summary and model diagnostics.