Personalized Medicine with Multiple Treatments

Personalized medicine is an intriguing topic in academia and the pharmaceutical industry. A major branch of research focuses on estimating individualized treatment rules that map patient pretreatment covariates to the space of available treatments to maximize the expected clinical outcome. Most existing research focused on scenarios where only two treatment options are applicable. Nevertheless, it is not uncommon to have more than two treatment options available to patients in practice. Learning the optimal treatment rules among multiple candidate treatments given the patient’s pretreatment characteristics is thus of great interest and impact in practice, which can be a much more challenging problem than the problem with only binary treatments. In this chapter, we provide a selective review of the methods proposed in the recent literature for estimating the optimal treatment rules with multiple treatments.