Nonparametric Generated Effect Modifiers for Placebo Response

• Activity Number: 356
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Mental Health Statistics Section
• Author(s): Eva Petkova* and Thaddeus Tarpey and Adam Ciarleglio and Bei Jiang and R. Todd Ogden and Zhe Su
• Companies: New York University School of Medicine and Wright State University and New York University and New York University School of Medicine and Columbia University and New York University
• Keywords: personalized treatment decision ; complex biomarkers ; moderator of treatment effect ; tailoring variables ; placebo response ; treatment effect

Abstract

The problem of identifying baseline covariates that predict placebo response and distinguish placebo response from specific drug response is an on-going problem in clinical research that has evaded satisfactory solutions. Much of the difficulty is due to the fact that typical baseline predictors individually have very little modifying effects in differentiating placebo response from specific drug response. This talk proposes a novel approach to this problem whereby several baseline predictors are combined to form a single powerful moderator ­ a generated effect modifier (GEM). Previous attempts have considered GEMs in linear regression models. In order to add flexibility and increase the power of a modifying effect, this talk investigates GEM models with quadratic effects as well as nonparametric regression models where the outcome is modeled as a function of treatment and a smooth nonparametric function of the GEM predictor. Results will be illustrated via simulation studies and an application to a placebo controlled depression trial.