|
Activity Number:
|
149
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #305673 |
|
Title:
|
Comparison of Methods for Handling Dropouts in Longitudinal Ordered Categorical Data
|
|
Author(s):
|
Mohamed Alosh*+
|
|
Companies:
|
FDA
|
|
Address:
|
10903 New Hampshire Ave, Silver Spring, MD, 20814,
|
|
Keywords:
|
Longitudinal categorical data ; missing data ; threshold model
|
|
Abstract:
|
Missing data is a common problem in clinical trials and several methods are proposed in the literature for handling dropouts. Yet, it is commonly recognized that no universal imputation method is expected to outperform other methods, as the analysis goal and underlying mechanism for dropouts are expected to vary from trial to trial. There is an extensive literature on comparing different approaches for handling missing data for normally distributed endpoints. In this presentation we compare the performance of several methods for handling dropouts for ordered categorical response data arising from longitudinal clinical trials. For modeling this type of data we consider threshold models with random components and investigate robustness of findings through a simulation experiment under different dropout mechanisms. Also we present an application related to a dermatological indication.
|
