Online Program Home
My Program

Abstract Details

Activity Number: 612
Type: Contributed
Date/Time: Wednesday, August 3, 2016 : 2:00 PM to 3:50 PM
Sponsor: Biopharmaceutical Section
Abstract #320261 View Presentation
Title: Optimizing Dynamic Treatment Regimes via Quality-Adjusted Q-Learning and Threshold Utility Analysis for Subgroup Analysis in Clinical Trials
Author(s): Geoffrey Johnson* and Andrew Topp and Abdus S. Wahed
Companies: and University of Pittsburgh and University of Pittsburgh
Keywords: Dynamic treatment regime ; Inverse probability weighting ; m-out-of-n bootstrap ; Quality adjusted lifetime ; Q-learning ; Threshold utility analysis
Abstract:

We provide a new method of Q-learning, which we call quality adjusted Q-learning. The methods are developed in a specific 2-stage SMART design using quality adjusted lifetime as the outcome, but are easily generalized to any continuous outcome in SMARTs with an arbitrary number of stages. Quality adjusted Q-learning adds an interesting wrinkle to the field of dynamic treatment regimes (DTRs), in that the optimal regime will not only depend on patient information (including treatments taken, intermediate outcomes, and other patient covariates), but it will also depend on information on the treatments themselves, e.g. monetary cost or toxicity. The focus of this paper is to investigate a form of Q-learning using estimating equations for the quality adjusted survival outcome. We use m-out-of-n bootstrap for inference, and threshold utility analysis to show how the patient-specific optimal regime varies according to the treatment characteristics (e.g. cost, side effects). Methodologies investigated are demonstrated to construct optimal treatment regimes for the treatment of children's neuroblastoma.


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

Back to the full JSM 2016 program

 
 
Copyright © American Statistical Association