Cost-Effectiveness of Personalized Policies for Implementing Organ-at-Risk Sparing Adaptive Radiation Therapy in Head and Neck Cancer: A Markov Decision Process Approach

Abstract

Purpose: To develop a clinical decision-making model for implementation of personalized organ-at-risk (OAR)-sparing adaptive radiation therapy (ART) that balances the costs and clinical benefits of radiation plan adaptations, without limiting the number of re-plannings per patient, and derive optimal policies for head and neck cancer (HNC) radiation therapy. Methods and Materials: By leveraging retrospective CT-on-Rails imaging data from 52 HNC patients treated at the University of Texas MD Anderson Cancer Center, a Markov decision process (MDP) model was developed to identify the optimal timing for plan adaptations based on the difference in normal tissue complication probability (∆NTCP) between the planned and delivered dose to OARs. To capture the trade-off between the costs and clinical benefits of plan adaptations, the end-treatment ∆NTCPs were converted to Quality Adjusted Life Years (QALYs) and, subsequently, to equivalent monetary values, by applying a willingness-to-pay per QALY parameter. Results: The optimal policies were derived for 96 combinations of willingness-to-pay per QALY (W) and re-planning cost (RC). The results were validated through a Monte Carlo (MC) simulation analysis for two representative scenarios: (1) W = $200,000 and RC = $1,000; (2) W = $100,000 and RC = $2,000. In Scenario (1), the MDP model's policy was able to reduce the probability of excessive toxicity, characterized by ∆NTCP ≥ 5%, to zero (down from 0.21 when no re-planning was done) at an average cost of $380 per patient. Under Scenario (2), it reduced the probability of excessive toxicity to 0.02 at an average cost of $520 per patient. Conclusions: The MDP model's policies can significantly improve the treatment toxicity outcomes compared to the current fixed-time (one-size-fits-all) approaches, at a fraction of their costs per patient. This work lays the groundwork for developing an evidence-based and resource-aware workflow for the widespread implementation of ART under limited resources.