|
Activity Number:
|
203
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2003 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
ENAR
|
| Abstract - #301839 |
|
Title:
|
Indeterminacy in Analysis of Longitudinal Change in the Presence of High Mortality
|
|
Author(s):
|
Tom Greene*+ and Xuelei Wang and Brett Larive and Guofen Yan
|
|
Companies:
|
Cleveland Clinic Flundation and University of New Hampshire and Cleveland Clinic Foundation and Case Western Reserve University
|
|
Address:
|
Dept. of Biostatistics & Epidemiology, Cleveland, OH, 44195-0001,
|
|
Keywords:
|
longitudinal analysis ; informative censoring ; counterfactuals ; quality of life
|
|
Abstract:
|
Informative censoring models are often used to analyze change in longitudinal outcomes to avoid bias from dropout when the process leading to dropout may be related to the outcome. When attrition from death is high, this approach has been criticized because the analyses estimate parameters referring to hypothetical populations with complete data. For example, an estimated treatment effect at 3 years refers to the hypothetical difference in the mean response at 3 years between the treatment and control groups in all patients, including those who died prior to 3 years. We apply a counterfactual modeling approach to suggest a modification of pattern-mixture informative censoring models in which the estimated parameters refer to subsets of patients who would have survived sufficiently long for the parameters to be meaningful if the patients had been randomized to the control group. We contrast the uncertainties associated with this counterfactual approach to uncertainties of extrapolation of longitudinal outcomes following death in conventional informative censoring analyses. Concepts are illustrated with data from the HEMO Study, a randomized trial in dialysis patients.
|
