|
Activity Number:
|
521
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Learning and Data Mining
|
| Abstract - #305137 |
|
Title:
|
Developing an Adaptive Individualized Therapy Trial for Life-Threatening Chronic Disease
|
|
Author(s):
|
Yiyun Tang*+ and Michael Kosorok
|
|
Companies:
|
The University of North Carolina at Chapel Hill and The University of North Carolina at Chapel Hill
|
|
Address:
|
477 Melanie CT., Chapel Hill, NC, 27514,
|
|
Keywords:
|
Clinical Trial Design ; Adaptive Design ; Individualized Therapy ; Dynamic Treatment Strategy ; Reinforecement Learning ; Cystic Fibrosis
|
|
Abstract:
|
"Adaptive individualized therapy" is the dynamic treatment rule that depends on demographic, biomarker and treatment history at the treatment assignment time. In the sequential multiple assignments randomized trial design setting, we propose to use the statistical learning based on Support Vector Regression and Reinforcement Learning for constructing the treatment assignment rules. The temporal difference reinforcement learning with Markovian assumption called Fitted Q Iteration is utilized to learn from the chronic disease progression. We conduct a simulation study of cystic fibrosis patients with Pa infection and antibiotic therapy based on real studies of CF neonatal screening project. The study results show the great capacity and advantages of adaptive individualized therapy clinical trial for discovering effective treatment rules and improving long term outcomes in chronic disease.
|
