Preoperative and early postoperative albumin thresholds, rather than ΔAlb, predict severe complications after colorectal cancer surgery: a multifactorial model and nomogram

Background: Postoperative complications of colorectal cancer significantly impact prognosis and healthcare burdens. While the dynamic change from preoperative to the lowest albumin level on postoperative day 2 (ΔAlb) has been linked to complications, their predictive value remains controversial. This study aims to explore the relationship between ΔAlb and postoperative complications, and evaluates a multifactorial predictive model for severe complications. Methods: This retrospective study analyzed 932 patients undergoing colorectal cancer resection (August 2020-April 2022). Data included preoperative/postoperative albumin levels, demographics, surgical/anesthetic details, and 30-day complications. ΔAlb's association with severe complications was assessed via logistic regression, and a nomogram was developed. Results: Severe complications (Clavien-Dindo III-V) occurred in 14.2% of patients. ΔAlb showed no significant association with severe complications in univariate (P = 0.185) or multivariate analysis and discriminated poorly on ROC analysis (AUC 0.536, 95% CI 0.479-0.593). In the final multivariable model using clinically dichotomised predictors, preoperative albumin <35 g/L (OR 2.05, 95% CI 1.29-3.25, P = 0.003) and POD2 nadir albumin <28.5 g/L (OR 1.55, 95% CI 1.01-2.38, P = 0.044) were independent risk factors. Other independent predictors were age >65 years (OR 2.14, P = 0.003), ASA ≥3 (OR 2.14, P = 0.001), preoperative NRS2002 ≥2 (OR 1.97, P = 0.028) and rocuronium dose ≥117.5 mg (OR 4.44, P<0.001). A nomogram integrating these factors showed good discrimination (apparent C-index 0.794; optimism-corrected 0.784 by 1000-sample bootstrapping), good calibration (Hosmer-Lemeshow P = 0.72) and net clinical benefit on decision-curve analysis. Conclusion: ΔAlb alone has limited predictive value, whereas absolute preoperative albumin <35 g/L and POD2 nadir albumin <28.5 g/L are independent predictors of severe complications. A multifactorial model integrating these absolute albumin thresholds with age, ASA score, preoperative NRS2002 and rocuronium dose improves prediction (optimism-corrected C-index 0.78) compared with individual factors (AUC <0.7). Because rocuronium dose may also reflect surgical complexity and perioperative management, these associations require prospective, multicentre validation.