Development and Validation of a Predictive Model for Severe Radiation-Induced Esophagitis in Lung Cancer Patients Undergoing Moderate Hypofractionated Radiotherapy

Anqi Zhang¹Mengjie Quan¹Weiqian Li¹Jing Cao²Yuee Liu¹Bo Zhu¹Xiaocang Ren¹Yueliang Qin¹Qiang Lin^1*

• ¹Oncology Department, Huabei Petroleum Administration Bureau General Hospital, Renqiu, China
• ²Anesthesiology Department, Huabei Petroleum Administration Bureau General Hospital, Renqiu, China, renqiu, China

Background and Objectives: Moderate hypofractionated radiotherapy (MHRT) offers shorter courses and improved local control in lung cancer but increases the risk of severe radiation-induced esophagitis (RIE; grade ≥3), which impairs quality of life and may interrupt treatment. This study aimed to build a prediction model integrating clinical and dosimetric factors to identify high-risk patients and guide individualized strategies. Methods: Lung cancer patients treated with MHRT were retrospectively analyzed, with grade ≥3 RIE as the endpoint. Model development employed fully nested bootstrap resampling (B = 1000). In each outer training set, predictors were imputed (median for continuous, mode for categorical) and standardized, then selected via elastic-net regression (α = 0.5). The final model was constructed using Firth-penalized logistic regression, with performance reported as optimism-corrected estimates and 95% confidence intervals. Discrimination was assessed by ROC/AUC, calibration by calibration curves and Hosmer–Lemeshow test, and net benefit by decision curve analysis (DCA). The Youden index defined the optimal cutoff, and a nomogram was generated for clinical application. Results: Among 105 patients, 17 (16.2%) developed grade ≥3 RIE. Five predictors entered the final model: mean tumor volume, V5, D2cc, circumferential 2.6-Gy length, and circumferential 3.0-Gy length. Apparent performance was AUC = 0.771, Brier score = 0.114, slope = 1.16, intercept = 0.13. After optimism correction, AUC declined to 0.608 (95% CI, 0.464–0.761) with Brier score 0.176 (95% CI, 0.114–0.247). The Hosmer–Lemeshow test indicated adequate fit (χ² = 7.84, p = 0.449). The cutoff of 0.13 stratified patients into high- and low-risk groups. DCA showed net benefit compared with treat-all or treat-none across thresholds 0–0.8. Calibration was less stable due to limited events. Conclusion: Using elastic-net feature selection and Firth logistic regression, we developed a model to predict severe (grade ≥3) RIE in lung cancer patients undergoing MHRT. The model exhibited moderate discriminatory ability with generally acceptable calibration, enables risk stratification and identification of high-risk patients, and is presented as a nomogram to support clinical application. It holds promise for guiding individualized radiotherapy decisions and the prevention of treatment-related complications.