Statistical Endpoint Selection for Recurrent Events in Clinical Trials

• Activity Number: 31
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Xiaohai Wan*+ and Qianying Liu and Nathalie Ezzet and Paul Gallo
• Companies: and University of Chicago and Novartis and Novartis
• Keywords: recurrent events ; negative binomial process ; time-to-event ; sample size calculation ; power analysis

Abstract

Recurrent clinical events are common and play an important role in assessing both the efficacy and safety of therapeutic molecules in clinical trials. Recurrent events data often comprises both event counts and timing of individual events, therefore both count data methods and time-to-event methods can be applied as the primary analysis for such data, mostly depending on whether event counts or time-to-event is the endpoint of interests. In practice, both types of statistical methods have been used for recurrent events data. In this talk, popular statistical methods such as negative binomial regression, Cox proportional hazards regression, log-rank test and Andersen-Gill model are compared to analyze the recurrent events data generated from a negative binomial process. The relationship between hazard ratio and event rate ratio is presented and its implication to the sample size and power analysis is discussed. Explicit formulas are also developed to calculate sample size with log-rank test and Andersen-Gill model based on the negative binomial process. Performance of the various methods is assessed using simulations and an example to plan a phase 3 exacerbation study is presented.