|
Activity Number:
|
307
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 11, 2004 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #300896 |
|
Title:
|
Comparison of Analyses for Repeated Measurements from Ophthalmic Data
|
|
Author(s):
|
Kuolung Hu*+ and Vipin K. Arora and David Manner
|
|
Companies:
|
Eli Lilly and Company and Eli Lilly and Company and Southwestern Medical School
|
|
Address:
|
LCC, Indianapolis, IN, 46285,
|
|
Keywords:
|
generalized estimating equations (GEE) ; survival analysis ; ophthalmology ; clinical studies
|
|
Abstract:
|
This work considers two different model-based approaches for analyzing repeated binary data from longitudinal ophthalmology clinical studies. The two statistical approaches considered are: non-likelihood (GEE) regression and partial-likelihood model built from the simulated survival times. They are compared in an analysis of the binary response outcome (which can reverse during the study). The ophthalmic data are simulated for a two treatment arm (treated and nontreated group) study with persons with diabetes. The primary endpoint is progression to center-involved Diabetic Macular Edema (DME) (Yes/No) and is observed over time (maximum exposure 36 months with 12 time points including baseline). The probability of dropouts and reversal (of DME outcome) of events are considered in several simulation scenarios. By focusing on the efficiency of study design, we will demonstrate that GEE is superior to the partial-likelihood model in detecting the difference between the two treatment arms. Furthermore, we will also demonstrate that GEE performs asymptotically as good as the partial-likelihood model under the assumption of no reversals.
|
