|
Activity Number:
|
322
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #304730 |
|
Title:
|
Reinforcement Learning Treatment Strategies Based on Support Vector Regressions in a Non-Small Cell Lung Cancer Trial
|
|
Author(s):
|
Yufan Zhao*+ and Michael Kosorok
|
|
Companies:
|
Amgen, Inc. and The University of North Carolina at Chapel Hill
|
|
Address:
|
709 Audubon Lake Drive, Durham, NC, 27713,
|
|
Keywords:
|
Adaptive design ; Individualized therapy ; Non-small cell lung cancer ; Q-learning ; Reinforcement learning ; Support vector regression
|
|
Abstract:
|
We present a reinforcement learning design to discover optimal individualized treatment regimens for a non-small cell lung cancer trial. In addition to the complexity of the problem of selecting optimal compounds for first and second-line treatments based on prognostic factors, another primary scientific goal is to determine the optimal time to initiate second-line therapy, either immediately or delayed after induction therapy, yielding the longest overall survival time. $Q$-learning is utilized and approximating the $Q$-function with time-indexed parameters can be achieved by using support vector regressions. A simulation study shows that the procedure not only successfully identifies optimal strategies of two lines treatment from clinical data, but also reliably selects the best time to initial second-line therapy while taking into account heterogeneities of NSCLC across patients.
|
