Online Program Home
My Program

Abstract Details

Activity Number: 653 - Machine Learning and Other Statistical Methods in Clinical Trials
Type: Contributed
Date/Time: Thursday, August 1, 2019 : 10:30 AM to 12:20 PM
Sponsor: Biopharmaceutical Section
Abstract #304340 Presentation
Title: Alternatives to Logistic Regression for Detecting Treatment by Covariate Interactions with Binary Endpoints
Author(s): Radha Railkar* and Devan Mehrotra
Companies: Merck & Co., Inc. and Merck & Co., Inc
Keywords: clinical trial; logistic regression; interaction; covariate; scale

In clinical trials there is often interest in evaluating whether certain patient characteristics affect response to treatment. This is particularly important for optimizing the drug development process and advancing the goal of personalized medicine. It is therefore important to use appropriate methods to detect a treatment by covariate interaction. When the endpoint of interest is binary, logistic regression is commonly used to test for a treatment by covariate interaction. Logistic regression implicitly tests for an interaction on the odds ratio or logit scale. In practice an interaction may exist on one or more scales (e.g., original or log scales) and it is important to be able to detect these interactions. We propose alternative methods to test the null hypothesis of no treatment by covariate interaction on any scale. The advantages of the new methods will be demonstrated through simulation studies and real examples.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program