Assessing NTCP Models for Predicting Radiation-Induced Toxicity in Head and Neck Cancer: A Systematic Review

Abstract

Abstract

Background and Purpose: Radiation therapy of the head and neck (HNC RT) is often associated with xerostomia and dysphagia, making it extremely important to predict these toxicities. This review aims to evaluate the Lyman-Kutcher-Burman (LKB) model against alternatives such as relative seriality (RS) for estimating normal tissue complication probability (NTCP) to improve the management of HNC RT-induced side effects.

Methods: An Extensive literature search was carried out in PubMed, Embase, Web of Science, Google Scholar, and Cochrane Library to trace published studies between 1991 to 2023 on the above topic. Studies were included if they presented quantitative NTCP models for xerostomia and dysphagia induced by radiation therapy. The key LKB parameters (TD₅₀, m, n) were extracted, and comparative analyses were then carried out with the RS and clustering methods based on clinical and dosimetric data.

Results: 60-90% of the cases of HNC RT are reported to have xerostomia, which is raised to 100% when salivary glands are involved. Dysphagia occurs in 50-70% of cases, which increases to 90% with chemotherapy. Xerostomia occurs at doses above 40 Gy, and dysphagia at doses above 60 Gy. These are reduced by up to 50% with intensity-modulated radiation therapy (IMRT) and swallowing exercises. The prediction of radiobiological models is at around 75-85% for xerostomia and 70-80% for dysphagia. LKB prediction accuracy is at 70-75%; RS achieves 75-80%. Machine learning models show 85-90% accuracy.

Conclusion: Advanced radiation therapy techniques, in combination with validated toxicity prediction models, can improve treatment personalization and reduce the incidence of radiation-induced complications in HNC patients. Future research should focus on refining predictive models, investigating long-term treatment outcomes, and developing novel strategies to protect normal tissues.