|
Activity Number:
|
198
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 30, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #308970 |
|
Title:
|
Properties of Missing Data Imputation Methods: Baseline or Worst Observation Carry Forward
|
|
Author(s):
|
Jun Shao*+ and David Jordan and Yannis Jemiai and Yili Pritchett
|
|
Companies:
|
University of Wisconsin-Madison and Abbott Laboratories and Cytel Inc. and Abbott Laboratories
|
|
Address:
|
1300 University Ave, Madison, WI, 53706,
|
|
Keywords:
|
baseline carry forward ; bias ; worst observations ; clinical trials ; multiple visits ; dropout
|
|
Abstract:
|
Baseline-observation-carry-forward (BOCF) or worst-observation-carry-forward (WOCF) is sometimes applied to handle patient dropout in clinical trials with multiple scheduled visits. To many practical users, it is not clear what kind of effects we can estimate using data after applying BOCF or WOCF, and when and how statistical inference can be made using BOCF or WOCF data. We will address these issues in this presentation. After demonstrating the impact of BOCF or WOCF on study power via simulation results, we point out that sample means based on BOCF or WOCF data estimate a fracture of completer's sample mean where the multiplication factors are related to dropout rate, or the dropout rate and the percentage of worsening cases. We also demonstrate how to assess the variability of the sample means based on BOCF or WOCF data, which will ensure the accuracy of the statistical inferences.
|
