Development and external validation of an inflammation-based nomogram for predicting short-term mortality in cirrhotic patients with esophagogastric variceal bleeding

Abstract

Abstract Background Early and accurate risk stratification is critical for managing cirrhotic patients with esophagogastric variceal bleeding (EGVB). Inflammatory biomarkers derived from complete blood counts may offer prognostic value, yet their comparative performance remains underexplored. Methods The research adopted a dual-cohort design. The training cohort was derived from the MIMIC-IV database, while the validation cohort comprised patients treated at the Second Affiliated Hospital of Xi’an Jiao-tong University. We systematically evaluated six inflammatory markers to predict 28-day all-cause mortality employing a training cohort of 324 cirrhotic patients with EGVB and an external validation cohort of 509 patients. The optimal biomarker was identified through time-dependent receiver operating characteristic (ROC) analysis and further screened against conventional clinical variables, ultimately leading to the prediction model being constructed. Results The natural logarithm of the neutrophil-to-lymphocyte ratio (lnNLR) demonstrated the highest predictive accuracy and exhibited an independent, linear association with 28-day mortality in both training (HR = 1.31, 95% CI 1.00–1.71) and validation (HR = 2.10, 95% CI 1.26–3.51) cohorts. The final nomogram integrated lnNLR with international normalized ratio, serum potassium, mean arterial pressure, and respiratory rate. The model showed excellent discrimination, good calibration, and superior net benefit as compared to the model for end-stage liver disease (MELD) and cirrhosis acute gastrointestinal bleeding (CAGIB) scores, especially for ultra-early (48-h) risk stratification. Conclusion We presented a novel, inflammation-informed nomogram that enabled rapid, and individualized short-term mortality prediction in cirrhotic patients with EGVB. Its reliance on routinely available, pre-treatment variables enhanced feasibility in emergency settings and supports early clinical decision-making.