Development and internal validation of the CNA-Score (CRP–NLR–Albumin): a simplified model to predict postoperative complications after gastrectomy for gastric cancer

Postoperative complications after gastrectomy for gastric cancer remain common. We aimed to develop and internally validate a simplified model based on routine inflammatory and nutritional biomarkers to estimate the risk of any postoperative complication within 30 days. This prospective cohort included 270 adults undergoing curative-intent total or subtotal gastrectomy for gastric adenocarcinoma at two Colombian tertiary centers. The outcome was any postoperative complication within 30 days, irrespective of Clavien–Dindo grade. Candidate predictors included C-reactive protein (CRP), neutrophil-to-lymphocyte ratio (NLR), and albumin. Receiver operating characteristic-derived thresholds and multivariable logistic regression were used to develop a points-based score. Discrimination, calibration, and internal validation were assessed. A post hoc secondary analysis evaluated major morbidity, defined as Clavien–Dindo grade III or higher. In this analysis, the existing 0–4-point CNA-Score and the prespecified cutoff of at least 2 points were applied unchanged; no coefficients, thresholds, or point weights were re-estimated. Complications occurred in 68 patients (25.2%). In the multivariable association analysis, NLR ≥ 8 on postoperative day 3 (adjusted odds ratio 2.91, 95% confidence interval 1.49–5.67) and NLR ≥ 5 on postoperative day 5 (adjusted odds ratio 1.99, 95% confidence interval 1.05–3.74) were independently associated with the outcome. The early CNA-Score, designed for assessment by day 3, assigned 2 points to NLR ≥ 8 and 1 point each to CRP ≥ 22 mg/L and albumin ≤ 3.5 g/dL; the day-5 NLR measurement was not incorporated into the score. Complication rates increased from 12.5% with a score of 0 to 60.0% with a score ≥ 3. The multivariable logistic model had an apparent area under the curve of 0.689 (95% confidence interval 0.642–0.732), while the points-based CNA-Score had an area under the curve of 0.665. The optimism-corrected area under the curve was 0.673. Major complications occurred in 45 patients (16.7%). When the unchanged CNA-Score was applied to this secondary endpoint, the area under the curve was 0.700 (95% bootstrap confidence interval 0.613–0.783). At the prespecified cutoff of at least 2 points, sensitivity was 60.0%, specificity was 71.6%, and negative predictive value was 89.9%. The CNA-Score showed moderate discrimination and may support preliminary early risk stratification for 30-day complications after gastrectomy. Its exploratory performance for major complications was similar, but external validation is required before clinical implementation.