A finite-horizon Markov decision process model for cancer chemotherapy treatment planning: an application to sequential treatment decision making in clinical trials

Abstract

Cancer is one of the major diseases that seriously threaten the human life. Increasing interest in cancer treatment strategies for chemotherapy treatment planning and optimal drug administration has created new applications for mathematical modeling. In this paper, we develop a finite-horizon Markov decision process (MDP) model for cancer chemotherapy treatment planning that could advise selection of the optimal policy for the chemotherapy regimen according to the patient’s condition. The proposed model uses a finite action space of optimal cancer chemotherapy regimens for gastric and gastroesophageal cancers resulted from the proposed optimization model and a finite state space of patients’ toxicity levels. Results show that the proposed approach yields the optimal sequence of gastric and gastroesophageal cancer chemotherapy treatment regimens for a period of chemotherapy treatment which makes possible designing clinical trials for sequential treatments.