Activity Number:
|
590
- Missing Data
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 1, 2018 : 2:00 PM to 3:50 PM
|
Sponsor:
|
Biopharmaceutical Section
|
Abstract #329304
|
|
Title:
|
Approaches to Tipping Point Analyzes for a Binary Endpoint in Longitudinal Clinical Trials
|
Author(s):
|
Joseph Wu* and Huaming Tan and Neal Thomas and Cunshan Wang
|
Companies:
|
Pfizer and Pfizer, Inc. and Pfizer and Pfizer, Inc.
|
Keywords:
|
Binary Endpoints;
Missing Data;
Tipping Point Analysis;
Longitudinal Data;
Multiple Imputation;
Clinical Trials
|
Abstract:
|
Missing data are common in randomized longitudinal clinical trials. Sensitivity analyses based on missing not at random (MNAR) assumptions are routinely performed to evaluate the robustness of trial conclusions. Some statistical methodologies for missing data of a continuous endpoint were proposed. These methods are based on multiple imputations, such as Jump-to-Reference (JTR) and Tipping Point analyses. As binary endpoints are common in clinical trials, methods are also needed. A sequential logistic regression model as the imputation model has been used but it can be computationally intensive as the number of timepoints increases. We propose a tipping point analysis for binary endpoints using a generalized linear mixed effect model with logit link function on the binary outcome as the imputation model. Non-informative prior densities are used for model parameters to generate multiple posterior samples used in the multiple imputations. Penalty on treatment difference can be applied on the logit or probability scale with one scenario corresponding to JTR and another one to missing at random (MAR). We illustrate and discuss this method using simulated correlated binary trial data.
|
Authors who are presenting talks have a * after their name.