Optimal treatment regimes for the net benefit of a treatment

We develop a mathematical framework to define an optimal individualized treatment rule (ITR) within the context of prioritized outcomes in a randomized controlled trial. Our optimality criterion is based on the framework of generalized pairwise comparisons. We propose two approaches for estimating optimal ITRs on a pairwise basis. The first approach is a variant of the k-nearest neighbors algorithm. The second approach is a meta-learning method based on a randomized bagging scheme, which enables the use of any classification algorithm to construct an ITR. We investigate the theoretical properties of these estimation procedures, evaluate their performance through Monte Carlo simulations, and demonstrate their application to clinical trial data.