Value Function Guided Subgroup Identification via Gradient Tree Boosting: A Framework to Handle Multiple Outcomes for Optimal Treatment Recommendation

Abstract

In randomized clinical trials, there has been an increasing interest in identifying subgroups with heterogeneous responses to study treatment based on baseline patient characteristics. Even though the benefit risk assessment of any patient population or subgroups is almost always a multi-facet consideration, the statistical literature of subgroup identification has largely been limited to a single clinical outcome. In the article, we propose a nonparametric method that searches for subgroup membership scores by maximizing a value function that directly reflects the subgroup-treatment interaction effect considering clinical priorities of multiple outcomes based on win difference. A gradient tree boosting algorithm is proposed to search for the individual subgroup membership scores. We conduct simulation studies to evaluate the performance of the proposed method and an application to a colon cancer adjuvant clinical trial is performed for illustration.

Keywords:

• Gradient tree boosting
• Nonparametric
• Subgroup identification
• Win difference

Funding

The author(s) reported there is no funding associated with the work featured in this article.