Esophagus. 2026 Jun 22. doi: 10.1007/s10388-026-01223-1. Online ahead of print.
ABSTRACT
BACKGROUND: Risk assessment is essential for planning esophagectomy in patients with esophageal or gastro-esophageal junction (GEJ) cancers. However, previous reports using only preoperative variables (preoperative risk models) have poorly predicted postoperative anastomotic leakage. This study aimed to develop a novel risk model for anastomotic leakage using a combination of preoperative, intraoperative, and postoperative variables (perioperative risk model).
METHODS: Clinical data of 20,113 patients with esophageal or GEJ cancer who underwent esophagectomy followed by reconstruction between 2016 and 2019 were retrieved from the National Clinical Database (NCD), a Japanese web-based nationwide registry. Preoperative and perioperative risk models for anastomotic leakage were developed using only preoperative variable and a combination of preoperative, intraoperative, and postoperative variables within 72 h, respectively. The performance of the perioperative risk model was validated using NCD data of 5,147 esophagectomies registered in 2020.
RESULTS: In the overall population, 11,360 (45.0%) patients were aged ≥ 75 years, and 81.3% were male. Preoperative variables were comparable between the development and external validation cohorts. Anastomotic leakage was observed in 13.7% and 14.4% of the development and validation cohorts, respectively, and in 13.9% of all patients. The optimism-corrected C-statistics was higher in the perioperative risk model (0.610; 95% CI, 0.599-0.621) than in the preoperative risk model (0.565; 95% CI, 0.554-0.577). In the validation analysis, the C-statistics was 0.602 (95% CI, 0.580-0.623) for predicting anastomotic leakage.
CONCLUSION: Postoperative risk assessment using perioperative variables, including operative factors and early postoperative events, may help surgeons predict anastomotic leakage and improve patient management after esophagectomy.
PMID:42329555 | DOI:10.1007/s10388-026-01223-1
