Association and discriminatory performance of preoperative geriatric nutritional risk index, platelet-to-lymphocyte ratio, and their combined score for anastomotic leakage after McKeown esophagectomy in esophageal squamous cell carcinoma

Anastomotic leakage (AL) remains a serious complication after esophagectomy for esophageal squamous cell carcinoma (ESCC). We investigated the associations and discriminatory performance of preoperative geriatric nutritional risk index (GNRI), platelet-to-lymphocyte ratio (PLR), and an internally derived GNRI-PLR score for AL after McKeown esophagectomy. This retrospective single-center cohort included 434 patients with histologically confirmed ESCC who underwent curative-intent McKeown esophagectomy without neoadjuvant therapy between January 2019 and December 2024. GNRI and PLR were calculated from routine preoperative data, and a combined GNRI-PLR score was derived from a logistic model including both markers. Associations with AL were assessed using parsimonious multivariable logistic regression, with Firth penalized regression as a small-event sensitivity analysis. Discrimination was evaluated using ROC analysis, clinical-model comparison, bootstrap internal validation, calibration assessment, decision curve analysis, and restricted cubic splines. AL occurred in 27 patients (6.2%). In perioperative/pathological-adjusted models, higher GNRI was associated with lower odds of AL (OR = 0.78, 95% CI 0.66–0.93, P = 0.005), whereas higher PLR was associated with higher odds (OR = 1.33 per 10-unit increase, 95% CI 1.17–1.51, P < 0.001). The combined score was also associated with AL (OR = 2.56 per 1-unit increase, 95% CI 1.73–3.79, P < 0.001), with similar Firth results. Apparent AUCs were 0.753 for GNRI, 0.807 for PLR, and 0.854 for the combined score; the optimism-corrected AUC of the combined score was 0.848. Adding the combined score to a preoperative clinical model increased the optimism-corrected AUC from 0.791 to 0.888. Restricted cubic splines showed significant overall associations without nonlinearity. Preoperative GNRI, PLR, and an internally derived GNRI-PLR score were associated with AL after McKeown esophagectomy in ESCC. The combined score improved internally validated discrimination, particularly when added to clinical variables, but remains exploratory and requires external validation before clinical implementation.