Risk prediction model for radiation pneumonitis in breast cancer radiotherapy based on dose–volume parameters combined with the neutrophil-to-lymphocyte ratio

Purpose To develop and validate a risk prediction model for radiation pneumonitis (RP) and radiation-induced pulmonary fibrosis (RIPF) following breast cancer radiotherapy by integrating the V40 dose–volume parameter with the neutrophil-to-lymphocyte ratio (NLR), providing guidance for individualized treatment strategies. Methods This retrospective cohort study analyzed clinical data from 164 patients with breast cancer who underwent postoperative radiotherapy between May 2018 and August 2020. Clinical–pathological characteristics, radiotherapy dosimetric parameters and NLR values were collected at three time points: pre-surgery, 1 week before radiotherapy and 1 month after radiotherapy. Radiation pneumonitis (0–6 months) and RIPF (≥6 months) were assessed according to the Common Terminology Criteria for Adverse Events (version 5.0). Receiver operating characteristic (ROC) curves were used to determine the optimal predictive indicators. Variable selection was performed using least absolute shrinkage and selection operator regression followed by multivariate logistic regression to construct the prediction model. Internal validation was conducted using 1,000 bootstrap resampling iterations. Results Of the 164 patients, 107 (65.2%) developed varying degrees of RP (grade 1: n = 103, 62.8%; grade 2: n = 4, 2.4%), and 118 (72.0%) developed RIPF (all grade 1). The ROC analysis revealed that ipsilateral lung V40 had superior predictive performance for RIPF (area under the curve AUC = 0.728, 95% confidence interval CI: 0.651–0.805, cutoff value: 10.45%). The pre-radiotherapy NLR showed significant predictive value for RP (AUC = 0.685, 95% CI: 0.605–0.765, cutoff value: 2.82). Multivariate analysis identified independent risk factors for RP: V40 ≥ 10.45% (odds ratio OR = 3.24, 95% CI: 1.78–5.89, P 0.001), pre-radiotherapy NLR ≥ 2.82 (OR = 2.56, 95% CI: 1.42–4.61, P = 0.002) and regional nodal irradiation (OR = 2.13, 95% CI: 1.18–3.84, P = 0.012). The combined prediction model achieved an AUC of 0.816 (95% CI: 0.748–0.884), significantly outperforming single indicators (ΔAUC = 0.088–0.131, P 0.05). Bootstrap internal validation demonstrated robust model stability (C-index = 0.803). Conclusions The integrated prediction model combining V40 and the NLR effectively identifies patients a high risk of RP following breast cancer radiotherapy, facilitating personalized treatment planning and early intervention strategies.