|
Activity Number:
|
430
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #303445 |
|
Title:
|
Sample Size Estimation for Trials with Recurrent Events as the Primary Endpoint
|
|
Author(s):
|
Kuolung Hu*+ and Robert A. Parker
|
|
Companies:
|
Amgen, Inc. and Amgen, Inc.
|
|
Address:
|
One Amgen Center Drive, Thousand Oaks, CA, 91320-1799,
|
|
Keywords:
|
Poisson regression model ; negative binomial model ; recurrent event data ; Anderson-Gill model ; sample size estimation
|
|
Abstract:
|
In some clinical trials, the primary endpoint is repeated occurrences of the same or related types of event. Poisson regression, Anderson-Gill (AG), and the negative binomial (NB) model are commonly used to analyze such data. Signorini (1991) proposed a method to estimate sample size for the Poisson regression, but the adequacy of this estimation for analyses using the AG and NB models have not been reported. The Metcalfe and Thompson (2006) approach to simulate recurrent events data is used. We assess the operating features for detecting treatment differences using these statistical methods with the sample size estimated by Signorini's method. Based on our simulations, Signorini's method is also appropriate for the NB and AG models. We found only small discordant results between these three analyses. The specific analysis method only rarely impacts the conclusion draw from a study.
|
