|
Activity Number:
|
224
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #303731 |
|
Title:
|
Missing Data Imputation for Estimating Time-to-Event from Longitudinal Continuous Data
|
|
Author(s):
|
Lei Xu*+ and Kaifeng Lu and Bret Musser and Devan V. Mehrotra
|
|
Companies:
|
Merck & Co., Inc. and Merck & Co., Inc. and Merck Research Laboratories and Merck Research Laboratories
|
|
Address:
|
126 E. Lincoln Ave, Rahway, NJ, 07065,
|
|
Keywords:
|
Missing Data ; Multiple Imputation ; Repeated Measures ; Cox Model
|
|
Abstract:
|
Diabetes is a chronic disease characterized by abnormally high glucose levels in the blood. The American Diabetes Association recommends a target A1C level < 7% to prevent microvascular complications of diabetes. As existing anti-diabetic drugs lose efficacy over time, treatments in diabetes clinical trials are often compared in terms of subsequent durability, e.g., time to A1C >7%. In a longitudinal clinical trial, A1C is measured periodically over time. The prevalence of missing data is a major issue and complicates the data analysis. We discuss the usefulness of imputing the missing A1C data in order to derive the event time for the downstream Cox regression analysis. Simulation results will be presented to evaluate the performance of various imputation strategies, in terms of bias and efficiency for the Cox regression analysis.
|
