Abstract:
|
Personalized medicine has received increasing interest among clinicians and statisticians. Recently, powerful machine learning methods have been proposed to estimate optimal individualized treatment rule (ITR) but are only restricted to the situation with only two treatments. When multiple treatment options are being considered, these methods have to handle complex treatment-treatment interactions by transforming multi-treatment selection into binary treatment selections, which are not straightforward and may lead to inconsistent decision rules. There is a lack of literature on using multicategory learning to estimate optimal ITR. In this article, we fill this gap by proposing a novel and efficient method to generalize outcome weighted learning to multi-treatment settings via sequential weighted support vector machines. We demonstrate the performance of the proposed method with numeric examples. An application to a three-arm randomized trial of treating major depressive disorder shows that an individualized treatment strategy tailored to individual characteristics reduces depressive symptoms more than non-personalized treatment strategies.
|