Developing and Understanding Functional Modifiers of Treatment Effect with Applications to Psychiatric Clinical Trials

• Activity Number: 638
• Type: Topic Contributed
• Date/Time: Thursday, August 7, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Adam Ciarleglio*+ and Eva Petkova and R. Todd Ogden and Thaddeus Tarpey
• Companies: New York University School of Medicine and New York University School of Medicine and Columbia University and Wright State University
• Keywords: functional data ; effect modification ; clinical trials ; imaging data ; depression

Abstract

There has been a recent push to move from the analysis of treatment effects on the population level to effects on the individual. This requires that we have a firm understanding of effect modification. Often, the characteristics that we wish to consider as potential modifiers of treatment effect can be classified as functional data. For example, an individual's brain structure or function as measured via imaging modalities such as magnetic resonance imaging or electroencephalography may provide information on how a treatment might differentially affect subjects with the same condition. Unfortunately, it is difficult to detect and measure the effects of functional modifiers in practice. We discuss several approaches for generating "important" functional modifiers from functional data and estimating their effects on outcomes of interest. We use simulations to demonstrate properties of these procedures as well as apply them to real data from clinical trials that compare treatments for major depressive disorder.