Machine learning model for predicting recurrence following intensity-modulated radiation therapy in nasopharyngeal carcinoma
Abstract Background Nasopharyngeal carcinoma (NPC) exhibits unique histopathological characteristics compared to other head and neck cancers. The prognosis of NPC patients after intensity-modulated radiation therapy (IMRT) has not been fully studied, and there remains a high risk of recurrence. This...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
BMC
2025-06-01
|
| Series: | World Journal of Surgical Oncology |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s12957-025-03860-9 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Background Nasopharyngeal carcinoma (NPC) exhibits unique histopathological characteristics compared to other head and neck cancers. The prognosis of NPC patients after intensity-modulated radiation therapy (IMRT) has not been fully studied, and there remains a high risk of recurrence. This study aims to construct a reliable model for predicting post-treatment recurrence by integrating high-accuracy machine learning (ML) models. Methods A total of 859 NPC patients who underwent IMRT treatment at the First Affiliated Hospital of Zhejiang University School of Medicine from January 2013 to December 2020 were included in this study. The average follow-up period was 6.42 years, from July 2020 to March 2021. Candidate predictive factors were selected from demographic, clinical characteristics, medical examinations, and test results. We developed Cox proportional hazards regression, competing risk, and extreme gradient boosting (XGBoost) models. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves assessed model calibration, and decision curves evaluated the net benefit of the models in practical decision-making. Results EBV DNA, TNM stage, T stage, N stage, HGB, Tumor volume and LDH were the primary predictive factors for early recurrence, with the first three factors contributing the most to the XGBoost model. The average age of all participants was 49.39 ± 11.29 years. In both the training and validation sets, the XGBoost model demonstrated the best predictive performance, with the highest AUC compared to the Cox proportional hazards model (3-year AUC in the training set: 0.84, 95% CI: 0.81, 0.87; 3-year AUC in the validation set: 0.84, 95% CI: 0.79, 0.89) and the competing risk model (3-year AUC in the training set: 0.74, 95% CI: 0.70, 0.78; 3-year AUC in the validation set: 0.71, 95% CI: 0.64, 0.77). The XGBoost model achieved the highest AUC (3-year AUC in the training set: 0.93, 95% CI: 0.91, 0.96; 3-year AUC in the validation set: 0.94, 95% CI: 0.91, 0.97). Conclusion The XGBoost model is a simple and accurate tool for predicting the prognosis of newly diagnosed NPC patients undergoing IMRT treatment, which may aid in preventing NPC recurrence and postoperative follow-up. |
|---|---|
| ISSN: | 1477-7819 |