|
Activity Number:
|
224
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #304235 |
|
Title:
|
Statistical Models Comparisons for Recurrent Events with Application to Hypoglycemia Data in Diabetes Clinical Trials
|
|
Author(s):
|
Xiaodan Wei*+ and Yujun Wu and Peng-Liang Zhao
|
|
Companies:
|
sanofi-aventis and sanofi-aventis and sanofi-aventis
|
|
Address:
|
200 Crossing Blvd, Bedminster, 08807,
|
|
Keywords:
|
recurrent event ; hypoglycemia ; model comparison ; repeated events
|
|
Abstract:
|
In this paper, we compare various models on fitting recurrent event data. First we compare the existing models including Poisson (with or without dispersion parameter estimated), Negative Binomial model, zero-inflated Poisson and zero-inflated Negative Binomial model. GEE estimation of the variance is also discussed. Semiparametric approaches using cumulative mean function with robust estimation of the variance are also compared via simulation results. Then we discuss the advantages and disadvantages of using each method on event count (or event rate) with application to the hypoglycemia events from diabetes clinical trials. Simulation results based on the recurrent event data will show the performances of each method on controlling the actual type I error. Finally we provide recommendations on the most suitable statistical model and testing method for hypoglycemia data.
|
