|
Activity Number:
|
424
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #306583 |
|
Title:
|
Variability in Visit Times in Clinical Trials and the Implications for a Common Mixed Model for Repeated Measures
|
|
Author(s):
|
Tristan Massie*+
|
|
Companies:
|
U.S. Food and Drug Administration
|
|
Address:
|
10903 New Hampshire Ave., Silver Spring, MD, 20993-0002,
|
|
Keywords:
|
repeated measures ; visit windows
|
|
Abstract:
|
A popular mixed model for repeated measures in a clinical trial fits time as a class variable. When there is variability around planned visit times this assumes the overall response over time is a step function, constant within each visit window. This may be suboptimal for detecting treatment differences if the response is changing moderately fast around a particular visit and there is variability in the actual time of visit. In such a case one can also fit a local slope for each visit window at the price of only k (or a multiple of k) more parameters, where k is the number of post-baseline visits. Like the step function approach this obviates the need of specifying the functional form of the response over time and when the slopes are of moderate size it can also have better power than the step function approach for testing for a treatment difference at the planned visit time.
|
