โก Quick Summary
This study developed and validated machine learning models for predicting the risk of osteoradionecrosis (ORN) in patients with head and neck cancer following radiation therapy. The 5-feature Random Survival Forest model demonstrated superior performance, achieving a time-dependent AUC of 0.776 and effectively addressing the overestimation of risk associated with competing mortality.
๐ Key Details
- ๐ Dataset: 2,466 patients treated with curative radiation therapy from 2011 to 2018
- ๐งฉ Features used: Tumor site, D10cc, smoking pack-years, periodontal condition, dental insurance
- โ๏ธ Technologies: Fine-Gray regression, Random Survival Forests, DeepHit
- ๐ Performance: 5-feature RSF model: Time-dependent AUC 0.776, C-index 0.772
๐ Key Takeaways
- ๐ Machine learning models can effectively predict individualized ORN risk in head and neck cancer patients.
- ๐ก The 5-feature RSF model was selected for its simplicity and accuracy.
- ๐ฉโ๐ฌ Competing risks were crucial in accurately estimating ORN risk, avoiding significant overestimation.
- ๐ฅ The study included a large cohort of 2,466 patients, with 183 developing ORN during follow-up.
- ๐ An interactive web application was developed to facilitate clinical implementation of the model.
- ๐ Non-competing risk models overestimated ORN risk, predicting an average cumulative incidence of 8.7% compared to 6.8% with the 5-feature RSF.
- ๐ Feature selection was guided by SHapley Additive exPlanations (SHAP) for better interpretability.
๐ Background
Osteoradionecrosis (ORN) is a serious complication that can arise after radiation therapy for head and neck cancers. It significantly impacts patients’ quality of life and poses challenges in treatment planning. Traditional risk assessment methods often fail to account for competing risks, such as all-cause mortality, leading to potential overestimations of ORN risk. This study aims to leverage machine learning to create more accurate predictive models that consider these complexities.
๐๏ธ Study
Conducted between 2011 and 2018, this prognostic study collected comprehensive sociodemographic, clinical, and dosimetric data from 2,466 patients who underwent curative radiation therapy for head and neck cancer. The researchers aimed to develop predictive models for ORN using advanced machine learning techniques, including Fine-Gray regression and Random Survival Forests, while also addressing the issue of competing risks in their analysis.
๐ Results
The study found that the 5-feature Random Survival Forest model outperformed other models, achieving a time-dependent AUC of 0.776 and a C-index of 0.772. The model’s simplicity, incorporating only five key features, made it particularly appealing for clinical use. In contrast, non-competing risk models significantly overestimated ORN risk, highlighting the importance of considering competing events in risk assessments.
๐ Impact and Implications
The findings from this study have significant implications for clinical practice in oncology. By providing a reliable method for predicting individualized ORN risk, healthcare professionals can make more informed decisions regarding treatment plans and patient management. The development of an interactive web application further enhances the accessibility of these predictive tools, potentially improving patient outcomes in head and neck cancer care.
๐ฎ Conclusion
This study illustrates the transformative potential of machine learning in predicting complications such as osteoradionecrosis in head and neck cancer patients. By accurately estimating individualized risk while accounting for competing mortality, these models can significantly enhance clinical decision-making. Continued research and development in this area are essential for further improving patient care and outcomes.
๐ฌ Your comments
What are your thoughts on the use of machine learning for predicting ORN risk in head and neck cancer patients? We invite you to share your insights and engage in a discussion! ๐ฌ Leave your comments below or connect with us on social media:
Machine Learning Models for Individualized Osteoradionecrosis Risk Prediction in Head and Neck Cancer.
Abstract
To develop and validate predictive models for osteoradionecrosis (ORN) after head and neck radiation therapy (RT) using time-to-event data with death as the competing risk, and to quantify the degree of risk overestimation when the competing risk is ignored. In this prognostic study of patients who underwent curative RT between 2011 and 2018, with ongoing follow-up, sociodemographic, clinical, and dosimetric data were collected. The binary ORN outcome was defined by the ClinRad system (gradeโโฅโ1); all-cause mortality was the competing event. Fine-Gray regression (FGR), Random Survival Forests (RSF) with Gray’s test splitting rule, and DeepHit were implemented using repeated nested stratified cross-validation. Feature selection and interpretation were guided by SHapley Additive exPlanations (SHAP). For comparison, non-competing risk models such as Cox proportional hazards (Cox PH) and standard RSF (S-RSF) with log-rank splitting rule were also trained. Of 2,466 patients, 183 developed ORN during follow-up, and 714 died. Three versions of each model were developed using 20, 10, and 5 features. The 10- and 5-feature RSF models performed best. Considering simplicity, the 5-feature model, which included tumor site, D10cc, smoking pack-years, periodontal condition, and dental insurance, was selected for production. At 60 months, Brier Score was 0.061 (95% CI: 0.060-0.063), Integrated Brier Score 0.038 (95% CI: 0.037-0.040), time-dependent AUC 0.776 (95% CI: 0.762-0.789), and C-index 0.772 (95% CI: 0.757-0.787). FGR closely followed, whereas DeepHit underperformed. Non-competing models, including the S-RSF, overestimated ORN risk, predicting an average 60-month cumulative incidence of 8.7% versus 6.8% with the 5-feature RSF. A parsimonious RSF model reliably estimated individualized ORN risk while avoiding overestimation from ignored competing risks. An interactive web application was developed to support clinical implementation.
Author: [‘Moharrami M’, ‘Watson E’, ‘Huang SH’, ‘Madathil S’, ‘Kim J’, ‘McPartlin A’, ‘Malik NH’, ‘Singhal S’, ‘Hahn E’, ‘Waldron J’, ‘Bratman S’, ‘de Almeida J’, ‘Yao C’, ‘Hope A’, ‘Quinonez C’, ‘Glogauer M’, ‘Hosni A’]
Journal: J Med Syst
Citation: Moharrami M, et al. Machine Learning Models for Individualized Osteoradionecrosis Risk Prediction in Head and Neck Cancer. Machine Learning Models for Individualized Osteoradionecrosis Risk Prediction in Head and Neck Cancer. 2026; 50:(unknown pages). doi: 10.1007/s10916-026-02359-4