Activity Number:
|
260
|
Type:
|
Contributed
|
Date/Time:
|
Tuesday, August 13, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Biometrics Section*
|
Abstract - #301111 |
Title:
|
Extension of a Shared Parameter Model Applied to a QOL Analysis
|
Author(s):
|
Yunrong Ye*+ and Thomas Ten Have and Stephanie Land and H. Wieand
|
Affiliation(s):
|
University of Pittsburgh and University of Pennsylvania and University of Pittsburgh and University of Pittsburgh
|
Address:
|
Suite 350 - 201 North Craig Street, Pittsburgh, Pennsylvania, 15213, USA
|
Keywords:
|
mixed effects ; qualify of life ; informative missing data
|
Abstract:
|
In an analysis of depression, hot flashes and other quality-of-life data from the Breast Cancer Prevention Trial, conducted by the NSABP (which compared tamoxifen to placebo), investigators were confronted with potentially informative missing data. One tool they used to analyze the data was a model described by Ten Have and colleagues that assumed there was a shared parameter between the probability of dropout and the QOL outcome. A restriction on the particular model was that it assumed the relationship between missing data and QOL outcome was in the same direction (positively related or negatively related) in both the placebo and tamoxifen groups and over time. This assumption did not hold for the hot flash data. We investigate ways to remove this restriction. In addition, we consider other variations to the model and perform sensitivity analyses to determine its utility in addressing practical problems.
|