Early prediction of clinically relevant anastomotic leakage after colorectal surgery: development and internal validation of a simplified risk score

Early identification of anastomotic leakage after colorectal surgery remains challenging. This study aimed to develop and internally validate a simplified risk score using early postoperative clinical and biochemical parameters to predict subsequently diagnosed clinically relevant anastomotic leakage. This retrospective cohort study included 256 patients undergoing colorectal resection between 2016 and 2025. Risk assessment was performed at the end of postoperative day (POD) 4 using parameters recorded during POD2–4. Clinically relevant leakage was defined as grade B or C according to the International Study Group of Rectal Cancer classification within 90 days. The primary logistic regression model included peak C-reactive protein (CRP), persistent tachycardia, and peak lactate. Weighted and simplified scores were developed, and additional baseline-adjusted models were evaluated. Missing predictor values were addressed using multiple imputation, and internal validation was performed using 1000 bootstrap resamples. Clinically relevant anastomotic leakage occurred in 37 patients (14.5%). Peak CRP (adjusted odds ratio [aOR] 1.12 per 10 mg/L), persistefignt tachycardia (aOR 2.94), and peak lactate (aOR 1.65 per mmol/L) were associated with leakage. Optimism-corrected C-indices were 0.82 for the primary model, 0.80 for the weighted score, and 0.80 for the simplified score. The extended model achieved a corrected C-index of 0.83. At a simplified-score threshold of ≥ 2, sensitivity was 70.3% and specificity was 76.7%. Early postoperative CRP, persistent tachycardia, and lactate demonstrated potential for predicting subsequently diagnosed clinically relevant anastomotic leakage. The simplified score retained comparable discrimination, although its clinical applicability remains uncertain. Prospective multicenter external validation is required before clinical implementation.