Abstract
BACKGROUND: Postoperative pneumonia is a common and serious complication after McKeown esophagectomy for esophageal squamous cell carcinoma (ESCC), particularly in patients receiving neoadjuvant chemo-immunotherapy. The ARISCAT score is widely used for pulmonary risk assessment in general surgery, but its predictive value in this specific oncologic setting remains unclear. METHODS: We retrospectively analyzed 312 patients with resectable ESCC who underwent two cycles of platinum-based chemotherapy plus camrelizumab followed by McKeown esophagectomy between 2018 and 2023. Preoperative ARISCAT scores, neutrophil-to-lymphocyte ratio (NLR), and systemic immune-inflammation index (SII) were recorded. The primary outcome was pneumonia within 30 days postoperatively, diagnosed by standardized criteria. Logistic regression identified independent predictors, and restricted cubic splines (RCS) assessed dose-response patterns. Model discrimination and calibration were evaluated using AUC, Brier score, and calibration plots. RESULTS: Postoperative pneumonia occurred in 86 patients (27.6%). Compared with unaffected patients, those with pneumonia had higher ARISCAT scores (53.9 vs 37.4), NLR (8.3 vs 4.9), and SII (1379.8 vs 917.2) (all p < 0.001). Multivariable analysis confirmed ARISCAT (OR 1.43, 95% CI 1.26-1.62), NLR (OR 1.66, 95% CI 1.31-2.10), and SII (OR 1.09, 95% CI 1.02-1.71) as independent predictors. RCS showed a linear association for ARISCAT (p_non-linearity = 0.794) and threshold effects for NLR (>≈5.0) and SII (>≈1300) (both p_non-linearity < 0.05). The combined model (ARISCAT + NLR + SII) demonstrated superior discrimination (AUC 0.962) and calibration (Brier score 0.152) compared with individual predictors. CONCLUSION: In ESCC patients undergoing McKeown esophagectomy after neoadjuvant chemo-immunotherapy, the ARISCAT score independently predicts postoperative pneumonia risk. Integrating ARISCAT with inflammatory biomarkers enhances predictive performance, enabling refined preoperative risk stratification and potentially guiding targeted preventive strategies. Prospective multicenter validation is warranted.