|
Activity Number:
|
235
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #310277 |
|
Title:
|
Assessing Poolability of Data for a Clinical Trial: Why the Data Should Be Poolable and Why the Data Should Not Be Poolable
|
|
Author(s):
|
David Naftel*+
|
|
Companies:
|
The University of Alabama at Birmingham
|
|
Address:
|
LHRB 790, Birmingham, AL, 35294-0007,
|
|
Keywords:
|
Interactions in a randomized clinical trial ; Pooling of data ; Subgroup analysis
|
|
Abstract:
|
Clinical trials require pooling data from multiple sites to obtain adequate sample size. The FDA requires that such pooling be justified. Components of pooling are discussed in this paper with examples. Was the proposed study design and conduct the same at each site? These components are non-negotiable. Are the sites similar as to the baseline variables? The procedure is to compare baseline variables across sites. Lack of differences may indicate a homogenous sample that does not represent the target patients. Multiple tests almost guarantee that sites will be found different. We recommend assessing overlap among sites to understand similarities rather than differences. Do differences exist in outcomes? We recommend identifying the statistical explanation of the differences. Nonpoolable data allows investigation of interactions and may provide a representative study.
|
