|
Activity Number:
|
33
- Sensitivity Analysis with Nonignorable Missing Data: Recent Work from Academia, Industry, and Regulatory Agencies
|
|
Type:
|
Invited
|
|
Date/Time:
|
Sunday, August 8, 2021 : 3:30 PM to 5:20 PM
|
|
Sponsor:
|
ENAR
|
|
Abstract #316703
|
|
|
Title:
|
Control-Based Imputation in Longitudinal Clinical Trials for Continuous Outcomes with Outliers
|
|
Author(s):
|
G. Frank Liu* and Yilong Zhang and Gregory Golm
|
|
Companies:
|
Merck Sharp & Dohme Corp. and Merck & Co., Inc. and Merck & Co., Inc.
|
|
Keywords:
|
control-based imputation;
non-normal data;
outlier;
robust regression;
rank-based analysis
|
|
Abstract:
|
In longitudinal clinical trials with continuous endpoints, a primary objective is often to assess treatment effect in terms of mean difference (or difference in mean change from baseline) at the last time point. Control-based imputation for missing outcome data provides a conservative approach to evaluate the treatment effect for a hypothetical estimand in which patients who drop out of the trial are assumed to have a response profile similar to that of the control group. The analysis is commonly implemented using a multiple imputation approach. When there are potential outliers in the outcome measurements, the conventional multiple imputation followed by ANCOVA or mixed model analysis may be compromised. In this talk, we investigate several alternative methods including rank-based analysis and robust regression to handle the outliers. Simulations are conducted to evaluate the statistical properties such as bias, power and type-I error and compare the performance for these methods.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2021 program
|
