Activity Number:
|
506
- Statisticians' Approaches to the Realities of Clinical Trials
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 2, 2017 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Biopharmaceutical Section
|
Abstract #322502
|
View Presentation
|
Title:
|
Analysis of Dichotomous Binary Data with Missing Data Imputed Under Jump to Reference and Copy Reference Models
|
Author(s):
|
Zachary Zimmer* and Lei Xu and Frank Liu
|
Companies:
|
Merck and Merck and Merck & Co. Inc.
|
Keywords:
|
Missing Data ;
Dichotomous Data ;
Relative Risk
|
Abstract:
|
In clinical trials, a binary endpoint may be constructed from an underlying continuous measurement. For example, the percentage of patients with hemoglobin A1C < the goal of 7% is a clinically important endpoint for diabetes trials. In this situation, it can be more efficient to use multiple imputation to handle missing data on the underlying continuous responses than directly on the dichotomized binary endpoint. This presentation reviews the analysis of dichotomous binary endpoint with missing data imputed under both Jump to Reference and Copy Reference models. As the standard approach based on Rubin's formula inflates the standard errors of the treatment and control group response rate point estimates, a bootstrap approach is considered for correcting the variance estimates. Simulation studies are carried out for longitudinal trials with data simulated from a joint continuous multivariate model under both missing at random and missing not at random conditions. The improvement in the type I error rate and power for the assessment of risk difference is demonstrated using the bootstrap method. An example of the analysis of A1C < 7% is used as an application of this method.
|
Authors who are presenting talks have a * after their name.