Prospective cohort study to validate esophageal dose constraints and predictive models for esophagitis in patients with breast cancer undergoing hypofractionated regional nodal radiotherapy.

Purpose: Esophageal RV25 < 20 % and AV35 < 0.27 mL were reported as dose constraints predictive of grade ≥ 2 radiation esophagitis (RE) for breast cancer in our previous study. This prospective study aimed to validate the effectiveness of esophageal dose constraints and develop RE prediction models.

Methods: We enrolled 465 patients with breast cancer receiving 43.5 Gy in 15 fractions to the chest wall and nodal regions using IMRT/VMAT between January 2022 and February 2024. The esophagus was contoured from the cricoid cartilage level to the aortic arch's lower margin. RE was assessed weekly during radiotherapy and at weeks 1 and 2 and months 3 and 6 post-RT using CTCAE v3.0. Analyzed esophageal dosimetric parameters: total volume, mean/max dose, the relative and absolute volumes receiving at least 5-45 Gy by 5 Gy increments (RV5-RV45 and AV5-AV45). Predictive models incorporating tumor laterality, internal mammary nodal irradiation (IMNI), and RV25 or AV35 thresholds were developed. Discrimination (AUC) and calibration [Hosmer-Lemeshow (H-L) test] were evaluated, and risk stratification was performed using decision tree analysis.

Results: The grade 2 RE incidence (23.7 %) was considerably lower than in a previous report (40.9 %), and no grade ≥ 3 RE was observed. Both models performed well (RV25 model: AUC, 0.688, H-L, p = 0.974; AV35 model: AUC, 0.651, H-L, p = 0.776). Risk factors for RE included left-side tumor, IMNI, and RV25 ≥ 20 % or AV35 ≥ 0.27 mL. Patients with no risk factors were classified as low risk, those with one risk factor as intermediate risk, and those with ≥ 2 risk factors as high risk. The grade ≥ 2 RE incidence differed significantly across groups (RV25: 14.8 % vs. 24.7 % vs. 48.3 %; AV35: 14.7 % vs. 23.7 % vs. 45.4 %).

Conclusion: Clinical validation confirmed the effectiveness of esophageal dose constraints and the predictive accuracy of the RV25 and AV35 models. Avoiding unnecessary IMNI and maintaining RV25 < 20 % and AV35 < 0.27 mL could reduce the risk for RE.