# Post-Bleed Prognosis Beyond Hospitalization: The CLIF Consortium Acute Decompensation Score (CLIF-C AD) Predicts Six-Week Mortality After Upper Gastrointestinal Bleeding in Patients With Cirrhosis

## TL;DR

This study shows that the CLIF-C AD score can predict six-week mortality after gastrointestinal bleeding in cirrhosis patients, even after hospital discharge.

## Contribution

The study demonstrates the CLIF-C AD score's effectiveness in predicting post-discharge mortality in cirrhosis patients after upper gastrointestinal bleeding.

## Key findings

- CLIF-C AD had an AUROC of 0.891 for six-week mortality prediction.
- Six-week mortality was 33.9%, with 23 deaths occurring after discharge.
- An endoscopy-augmented model improved five-day rebleeding prediction (AUROC 0.720).

## Abstract

Background

In cirrhosis, outcomes after upper gastrointestinal bleeding (UGIB) extend beyond the index hospitalization. Six-week mortality is a standard endpoint for portal hypertension studies because it captures early deaths related to recurrent bleeding and post-bleed complications (infection, renal dysfunction, organ failure).

Objective

Our objective was to evaluate the CLIF Consortium Acute Decompensation score (CLIF-C AD) for post-bleed risk stratification, using six-week mortality as the prespecified primary endpoint and five-day rebleeding as a prespecified secondary endpoint, in comparison with MELD-Na (Model for End-Stage Liver Disease with sodium) and AIMS65 (albumin, international normalized ratio (INR), mental status, systolic blood pressure, age ≥65 years).

Methodology

We analyzed a retrospective cohort of 224 consecutive adults admitted with cirrhosis-associated UGIB (January 1, 2024, to December 31, 2025). No patients met the European Association for the Study of the Liver-Chronic Liver Failure (EASL-CLIF) criteria for acute on chronic liver failure (ACLF) at presentation. Risk scores were available as computed columns in the database using published definitions. Six-week vital status was obtained from regional registries and supplemented by follow-up phone calls when needed. Rebleeding at five days required recurrent hematemesis and/or melena with or without hemoglobin drop. All suspected rebleeding events underwent repeat endoscopy, and endoscopic confirmation was mandatory for classification. Discrimination was assessed with the area under the receiver operating characteristic curve (AUROC) and bootstrap 95% confidence intervals (CIs). For the secondary endpoint, we additionally evaluated an endoscopy-augmented model incorporating baseline scores, bleeding stigmata, and hemostasis modality.

Results

Six-week mortality occurred in 76 of 224 patients (33.9%), including 53 deaths during the index hospitalization and 23 additional deaths after discharge within six weeks. Five-day rebleeding occurred in 33 of 224 patients (14.7%). Among patients with esophageal varices, large varices (Paquet grade ≥3) were present in 79 of 163 (48.5%). Median CLIF-C AD was higher among non-survivors than survivors (67.7 vs. 50.5). CLIF-C AD discriminated six-week mortality with an AUROC of 0.891, comparable to MELD-Na (0.866) and AIMS65 (0.886). Optimism-corrected calibration for CLIF-C AD was near-ideal (slope 0.99, intercept -0.02), with an optimism-corrected Brier score of 0.119. Decision-curve analysis demonstrated a net benefit for CLIF-C AD over treat-all and treat-none strategies across thresholds of approximately 0.07-0.60. For five-day rebleeding, baseline score discrimination was modest, whereas an endoscopy-augmented model improved discrimination (AUROC 0.720, 95% CI 0.621-0.819). After bootstrap optimism correction, the AUROC decreased to 0.671, underscoring the need for external validation.

Conclusions

CLIF-C AD provides clinically useful post-bleed prognostication in cirrhotic UGIB when the goal is to anticipate deterioration beyond discharge. Early rebleeding prediction improves when lesion-level stigmata and endoscopic therapy are incorporated.