A Comparison of Statistical Models for Recurrent Events with Applications to Hospitalization and Death Events

• Activity Number: 524 - Statistical Issues in Drug Development - 2
• Type: Contributed
• Date/Time: Wednesday, August 1, 2018 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Jihong Chen* and Tim Lin and Matthew T. White and Christine Powell and Andrew Strahs
• Companies: Alnylam Pharmaceuticals and Alnylam Pharmaceuticals and Alnylam Pharmaceuticals and Alnylam Pharmaceuticals and Alnylam Pharmaceuticals
• Keywords: recurrent events; efficacy endpoint; hospitalization; death; simulation; dropout

Abstract

In clinical studies, a composite outcome consisting of events such as hospitalizations and death events are often assessed as either a safety or efficacy endpoint. In the literature, there are multiple methods for analyzing recurrent events including Poisson and negative binomial regression, Andersen-Gill, joint frailty modelling, and the Win-Ratio method. Each of these methods focuses on addressing different underlying scientific questions. Furthermore, in practice, an imbalance in the dropout rates between the groups being compared, particularly informative dropouts, will introduce bias in the estimates. These methods are illustrated and compared by extensive simulation studies and are applied to the analysis of hospitalizations and death events from a randomized phase 3 study conducted in patients with hATTR amyloidosis.