|
Activity Number:
|
424
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #306248 |
|
Title:
|
Mixed Model: an Alternative to LOCF as Primary Analysis
|
|
Author(s):
|
Cunshan Wang*+ and Naitee Ting and Greg C. G. Wei
|
|
Companies:
|
Pfizer Inc. and Pfizer Inc. and Pfizer Inc.
|
|
Address:
|
50 Pequot Ave., New London, CT, 06320,
|
|
Keywords:
|
mixed model ; likelihood ; primary analysis ; Macugen ; LOCF
|
|
Abstract:
|
Last observation carried forward (LOCF) has been a common statistical tool for handling missing data in clinical trials. Numerous drugs have been approved based on LOCF analyses. Recently, newer statistical methods are available to better deal with missing data. In this article, we recommend a likelihood-based linear mixed effects model as the primary analysis for continuous longitudinal clinical data. This model uses all available data. Treatment and time are considered discrete to allow for different means over time without parametric modeling. Covariances among repeated measures are modeled through using subject specific random effects that can be expressed as functions of time. We propose to use it as an alternative to LOCF as primary analysis in protocols. Efficacy data from the Phase 3 Macugen studies are used as an example to illustrate its usage.
|
