Normal Tissue Complication Probability Modeling for Proton Therapy.

Proton therapy requires accurate, reliable normal tissue complication probability (NTCP) models for treatment planning and evidence-based patient selection. This review examines 2 challenges in proton therapy: developing NTCP models that accurately reflect proton-specific toxicity and managing NTCP prediction uncertainty to ensure reliable clinical decision-making. Regarding model accuracy, photon-derived NTCP models work for some endpoints but fail for others when proton therapy exhibits distinct dose-response characteristics, indicating that the need for proton-specific models is endpoint-dependent. Clinical evidence that supports updating NTCP models by incorporating variable relative biological effectiveness is statistically weak due to interpatient heterogeneity and inconsistent findings. Despite ongoing efforts in improving NTCP prediction accuracy, the impact of prediction uncertainty on patient selection decisions has received limited attention. Further study is required to establish a robust, model-uncertainty-aware NTCP-guided patient selection framework for proton therapy that addresses both model accuracy and prediction confidence.