Statistical Methods to Assess Heterogeneity Effects in a Randomized Clinical Trial

• Activity Number: 643
• Type: Contributed
• Date/Time: Thursday, August 8, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Alok Dwivedi*+ and Luis A. Alvarado and Patrick Tarwater and Rakesh Shukla and Sada Nand Dwivedi and Rebecca A. Pasillas
• Companies: Texas Tech University Health Sciences Center and Texas Tech University health Sciences Center and Texas Tech University health Sciences Center and University of Cincinnati and All India Institute of Medical Sciences and Texas Tech University health Sciences Center
• Keywords: Heterogeneity effect ; Cluster analysis ; Clinical trial ; Interaction effect

Abstract

Randomized clinical trials play a significant role in evidence-based medicine. However, due to distinct characteristics of individuals, we often encounter heterogeneity in study patients. This may cause unreliable treatment effects for individual patients in randomized clinical trials. Exploring heterogeneity in treatment effects (HTE) in clinical trials is necessary for data analysis and reporting. A multivariable prediction model may be preferred in examining HTE. However, it may have feasibility, reliability, and over-fitting issues. We propose a cluster analysis approach to create homogeneous groups of study patients and subsequently a multivariable regression model to examine HTE. We have used published data on motivational interviewing to improve treatment engagement and outcome in the subjects seeking treatment for substance abuse. A Two-Step cluster analysis is used to create different homogenous groups using baseline characteristics. Interaction effect of treatment and cluster group variable on outcomes has also been obtained. Data analysis reflects a significant HTE on each outcome. Such an approach may be preferred as an alternative for assessing HTE in clinical trials.