This media is currently not available.
Predictors of 30-day mortality in upper gastrointestinal bleeding
Poster Abstract

Aims

Upper gastrointestinal bleeding (UGIB) is a heterogeneous condition with variceal (V), high-risk non-variceal (HRNV) and low-risk non-variceal (LRNV) subgroups. We aimed to identify clinical and laboratory parameters associated with 30-day mortality in these three subgroups in a tertiary care center.

Methods

We retrospectively analyzed all patients undergoing emergency upper endoscopy for UGIB at the Department of Emergency Medicine, Semmelweis University, between 1 January 2022 and 31 December 2023. Within each subgroup we performed univariable tests and univariable and multivariable logistic regression. Associations with 30-day mortality were described using odds ratios (OR) and p-values.

Results

The data of 654 patients (V=126; HRNV=99; LRNV=429) were analyzed.

In the variceal group, non-survivors had lower albumin (p<0.0001), higher creatinine and urea levels, lower GFR (p<0.0001) and more frequently had cardiovascular disease (p=0.003). Univariable analyses showed that GFR 30-60 and 15-30 ml/min markedly increased mortality risk (OR=10.8 and 7.05), as did cardiovascular disease (OR=5.04) and vasopressor requirement (OR=3.67), whereas each 1 g/L increase in albumin reduced the odds of death by 22% (OR=0.78). In multivariable models, liver disease (OR=2506), lower albumin (OR=0.62) and rising urea (OR=5.6) remained independent predictors.

In the HRNV group, non-survivors presented with lower blood pressure (all components p≤0.006), older age and more frequent noradrenaline use (p=0.026). Univariable analyses indicated that higher systolic/diastolic blood pressure and MAP were protective (OR=0.95-0.96 per mmHg), whereas noradrenaline use (OR=3.75) and elevated INR (OR=4.5) were risk factors. In multivariable analysis, increasing urea remained an independent predictor (OR=1.47).

In the LRNV group, non-survivors showed more severe hemodynamic instability (p<0.0001), worse renal function (creatinine p<0.0001), lower albumin (p=0.0010) and more frequent chronic high-dose PPI use (p=0.0215). Univariable analyses demonstrated that lower GFR values (CKD III-V. stages; OR=5.7-7.2) and rising urea values (OR=1.17) increased, whereas albumin (OR=0.90 per g/L) decreased the mortality risk. In multivariable models, albumin (OR=0.87) and rising urea (OR=1.15) remained independent predictors.

Conclusions

Across all three etiologic subgroups, hemodynamic instability, hypoalbuminaemia and urea dynamics were closely linked to 30-day mortality and may serve as a basis for etiology-specific risk stratification in UGIB.