Nutritional risk score drives postoperative morbidity in colorectal cancer: findings from a LASSO-selected multivariable model in 1013 consecutive patients

Abstract

Background Postoperative complications afflict approximately 40% of patients undergoing colorectal cancer (CRC) surgery and adversely affect long-term oncological outcomes. Reliable perioperative risk stratification tools that are both clinically interpretable and actionable remain limited. Methods This retrospective cohort study enrolled 1,013 consecutive patients undergoing elective radical resection for histopathologically confirmed CRC at a single tertiary center. A nine-system composite complication endpoint was defined. Missing data were handled via multiple imputation by chained equations (MICE; m = 5). Candidate predictors were screened by univariable logistic regression and LASSO regularization; final predictors were entered into multivariable logistic regression with Rubin’s rules pooling. Model performance was assessed by AUC, bootstrap internal validation (1,000 iterations), Hosmer–Lemeshow calibration, and decision curve analysis (DCA). A nomogram was constructed for individualized risk estimation. Results Of 1,013 patients, 372 (36.7%) experienced ≥1 postoperative complication. LASSO selected eight independent predictors: NRS-2002 nutritional risk score (aOR 1.502, 95% CI 1.312–1.720), laparoscopic approach (aOR 0.457, 95% CI 0.297–0.702), intraoperative blood loss (aOR 1.002/mL), right colon tumor location (aOR 1.502), NLR (aOR 1.080/unit), operative time (aOR 1.002/min), hemoglobin, and total bilirubin. The model achieved an apparent AUC of 0.689 (bootstrap-corrected 0.680), satisfactory calibration (Hosmer–Lemeshow p = 0.709), and net clinical benefit across threshold probabilities of 10–90% on DCA. Sensitivity analyses confirmed robustness across four pre-specified scenarios (AUC range 0.666–0.715). Conclusion This LASSO-derived nomogram provides transparent, bedside-applicable risk stratification for postoperative composite complications in CRC surgery, identifying nutritional status as the dominant modifiable predictor and supporting targeted perioperative optimization.