ABC Score Outperforms Rockall and C-WATCH in Predicting Cirrhosis Bleeding Deaths

When a patient with cirrhosis starts bleeding from ruptured varices in the esophagus or stomach, clinicians face one of the most feared emergencies in hepatology. Acute variceal bleeding is one of the most life-threatening complications of liver cirrhosis, and in this cohort 36 percent of patients died within one year. Several risk assessment scores, including pre- and post-endoscopy evaluations, have been developed to predict outcomes such as rebleeding and death after variceal bleeding. A new prospective study from Sohag University Hospital in Egypt now offers a direct comparison of three widely discussed risk scores in exactly this setting, and the result is a clear but nuanced verdict: the ABC score, built from age, blood tests and comorbidities, predicted one-year mortality better than the older Rockall score or the newer Cologne-watch score, though even its best performance came with a sobering limitation.

The study, published in the Egyptian Liver Journal, enrolled one hundred cirrhotic patients admitted with upper gastrointestinal variceal bleeding between March 2024 and March 2025. Every participant was followed for a full year to determine whether they survived without further bleeding, experienced rebleeding, or died. The cohort was drawn from the Tropical Medicine and Gastroenterology Department, and only patients whose bleeding was confirmed to come from ruptured esophageal or gastric varices were included. Patients bleeding from non-variceal sources such as Mallory-Weiss tears or ulcers were excluded, as were those with portal hypertension unrelated to cirrhosis, portal vein thrombosis without cirrhosis, or anticoagulant therapy. This careful filtering matters, because the study’s central question is whether general bleeding scores, designed for mixed populations of upper gastrointestinal hemorrhage, hold up when every patient has cirrhosis and portal hypertension as the underlying driver.

The headline outcome figures are stark. Over the one-year follow-up, mortality occurred in 36 percent of the patients, rebleeding in 8 percent, and 56 percent enjoyed what the researchers defined as a good prognosis: survival without rebleeding or death during the study period. The mortality rate falls within the range of 15 to 40 percent reported in earlier variceal bleeding literature, which the authors attribute to factors such as advanced portal hypertension, severe underlying cirrhosis, delayed presentation and constrained healthcare resources. The rebleeding figure of 8 percent is notably lower than the 24 percent reported in a prior study that followed cirrhotic patients for twelve weeks after control of a first variceal bleed, and lower still than the roughly 10 to 30 percent rebleeding range cited in the broader literature.

To understand why some patients died and others survived, the team assembled an unusually complete clinical picture. Each patient underwent a full history and examination, laboratory testing including complete blood count, liver and renal function tests, C-reactive protein and ascitic fluid analysis, pelvi-abdominal ultrasound with Doppler, and upper gastrointestinal endoscopy with formal classification of varices by form, color and red color signs. Liver severity was graded with the Child-Pugh system, which combines measures of encephalopathy, ascites, bilirubin, albumin and clotting. Against this backdrop, the three candidate scores were computed: the Rockall score, which blends age, shock, comorbidities and endoscopic findings; the ABC score, introduced by Laursen and colleagues in 2021, which uses age, blood test results and comorbidities on a scale from 0 to 18; and the C-watch score, a pre-endoscopic tool relying solely on laboratory values including CRP, white cell count, ALT, platelets, creatinine and hemoglobin.

The comparison produced an unambiguous winner. The ABC score was the only one of the three that reached statistical significance in predicting mortality, with a P value below 0.001, and patients who died carried significantly higher ABC risk categories than survivors. The optimal cutoff, derived by Youden’s index, was 9.5 points, at which the score showed a specificity of 93.7 percent but a sensitivity of only 44.4 percent. Neither the Rockall score nor the C-watch score achieved statistically significant diagnostic performance for mortality prediction in this cohort.

That low sensitivity deserves emphasis, and the authors themselves flag it. A sensitivity of 44.4 percent means the score, at its chosen cutoff, missed more than half of the patients who ultimately died, even though it rarely mislabeled survivors as high risk. In a life-threatening condition, the authors note in their limitations section, such a sensitivity is extremely low for a predictive score. The study’s conclusion therefore stops short of endorsing the ABC score as a stand-alone tool, recommending instead that it be combined with laboratory and ultrasonographic parameters, and that further validation in larger multicenter cohorts is warranted before clinical reliance. This is a measured reading of the data: the ABC score is the best of the three tested, but best among imperfect options.

Beyond the score comparison, the study mapped which clinical features tracked with death. Older age, diabetes mellitus, pallor combined with jaundice, ascites, low serum albumin, elevated INR, Child-Pugh class C, elevated creatinine, elevated CRP, spontaneous bacterial peritonitis, and the presence of hepatic focal lesions on ultrasound were all significantly more common in the mortality group. In univariate regression, these factors along with dilated and thrombosed portal veins and the ABC score itself predicted mortality. But when the variables were pushed through multivariate logistic regression with backward elimination, only three survived as independent predictors: increased INR, the presence of a dilated and thrombosed portal vein, and an increased ABC score. The portal vein finding is mechanistically interesting, since thrombosis and dilation in cirrhosis may signal more advanced portal hypertension and impaired hepatic perfusion, raising variceal pressure and the risk of further bleeding.

The findings also speak to a broader debate about how well general gastrointestinal bleeding scores transfer to variceal hemorrhage. The Rockall score was developed for acute upper gastrointestinal bleeding broadly, and prior studies have been split on its usefulness in the variceal subgroup, with some reporting acceptable prognostic performance and others finding limited discrimination. The present result, in which Rockall failed to predict mortality, aligns with work by Stanley and colleagues who found the score of limited utility for this purpose, and with a meta-analysis in which no bleeding score reached an area under the curve above 0.70 for in-hospital outcomes. The C-watch score, designed specifically to incorporate inflammatory and laboratory markers, performed better in non-variceal bleeding in earlier work by Allo and colleagues, and the Sohag team interprets their own null result as evidence that the score’s applicability may narrow when portal hypertension dominates the picture.

What emerges is a picture of prognostication in variceal bleeding that is improving but still incomplete. The ABC score's advantage is plausible on its face: it captures age, which was significantly higher in the mortality group in univariate analysis, along with blood test results and comorbidities. Its 72.6 percent area under the curve is respectable but far from definitive, and the authors note it echoes the 0.72 value Liu and colleagues reported for 30-day mortality prediction in a broader upper gastrointestinal bleeding population. For clinicians, the practical message is that no single number yet suffices. Mortality in this study was best predicted by combining an elevated ABC score with laboratory findings such as INR and ultrasonographic signs such as portal vein thrombosis and hepatic focal lesions. For researchers, the message is that variceal bleeding in cirrhosis remains a distinct prognostic problem, one in which the severity of liver disease, the state of the portal circulation and systemic inflammation all conspire, and in which the scores inherited from general bleeding medicine must be tested, as this study has done, rather than assumed to work.

Subject of Research: Comparing the ABC, C-watch and Rockall risk scores for predicting one-year mortality after variceal bleeding in cirrhotic patients

Article Title: Predictive value of the age, blood tests and comorbidities (ABC) score, Cologne-watch (C-watch) score and Rockall score for risk of mortality following variceal bleeding among cirrhotic patients

Article References: Meghezel, E.-Z. M., Sayed, A., & Makbol, R. M. (2026). Predictive value of the age, blood tests and comorbidities (ABC) score, Cologne-watch (C-watch) score and Rockall score for risk of mortality following variceal bleeding among cirrhotic patients. Egyptian Liver Journal, 16(1), Article 80. https://doi.org/10.1186/s43066-026-00564-x

Image Credits: AI Generated

DOI: 10.1186/s43066-026-00564-x

Keywords: cirrhosis, variceal bleeding, ABC score, Rockall score, C-watch score, mortality prediction, portal hypertension, Child-Pugh, INR, portal vein thrombosis, prognostic scores, Egyptian Liver Journal