(Vienna, 07 November 2025) Researchers at MedUni Vienna have tested the prognostic significance of their previously developed artificial intelligence (AI) model based on simple blood tests. The Vienna 3P/5P models, which are using just three or five routine laboratory parameters, can predict the clinical course of patients with advanced chronic liver disease – without invasive procedures or the need for special equipment. Originally developed at MedUni Vienna and first published in the Journal of Hepatology in 2023, the models are used for the non-invasive detection of clinically significant portal hypertension (CSPH). In the follow-up study now available, the AI models were tested for the first time in independent patient cohorts and were able to accurately predict the occurrence of severe liver-related complications – such as abdominal fluid (ascites), internal bleeding (variceal bleeding) or liver-related consciousness disorders (hepatic encephalopathy). The study was conducted in collaboration with Hannover Medical School (MHH) and was recently published in the renowned journal "JHEP Reports".
The study focused on patients with compensated advanced chronic liver disease (cACLD) – a stage in which cirrhosis is already present but no hepatic decompensation events have yet occurred. Early identification of patients at high risk of disease progression is crucial as it allows a timely initiation of preventive measures.
The research team led by Georg Kramer and Thomas Reiberger from the Clinical Division of Gastroenterology and Hepatology at MedUni Vienna's Department of Medicine III analysed data of 266 patients who were examined at the Vienna Hepatic Haemodynamic Laboratory and validated the results in an independent external cohort at Hannover Medical School, which included 215 additional patients.
Portal hypertension – elevated blood pressure in the portal vein system of the liver – is a key driver of complications in cACLD. The hepatic venous pressure gradient (HVPG) is considered the gold standard for assessing this pressure, but requires an invasive, catheter-based procedure that is only performed in specialised centres.
Elastography-based methods such as liver stiffness measurement (LSM) offer a non-invasive alternative, but require expensive equipment and expertise from trained personnel, which limits their routine use in many healthcare facilities. To overcome these limitations, the Vienna-based research team developed the Vienna 3P and 5P models. These are completely blood-based models developed using AI that estimate the severity of portal hypertension and the risk of future complications based solely on standard laboratory values. In the study, their prognostic accuracy was comparable to that of HVPG and exceeded that of the imaging method elastography. "Our models offer a simple, cost-effective and easily repeatable way to determine the severity of portal hypertension and – as we have now been able to show – also to predict the further course of the disease," explains first author Georg Kramer. "This enables individual risk monitoring, even outside specialised centres."
A step towards individualised treatment of patients with chronic liver diseases
The ability to assess the risk for progression of liver disease exclusively through blood tests represents an important step towards individualised care for patients with chronic liver disease. Since the Vienna 3P/5P models can be easily repeated as part of routine check-ups, they enable continuous tracking of the disease progression and dynamic adjustment of therapy and monitoring strategies.
This approach could help doctors identify high-risk patients, initiate preventive therapies at an early stage and provide closer care, while reducing stressful and costly examinations for other patients. In the long term, these models could thus contribute to a more efficient use of medical resources – especially in regions where access to specialised diagnostic procedures such as invasive HVPG measurement or elastography is limited.
Publication: JHEP Reports
Blood-based Vienna 3P/5P risk models accurately predict first hepatic decompensation in compensated advanced chronic liver disease.
Georg Kramer, Benedikt Simbrunner, Mathias Jachs, Lorenz Balcar, Benedikt Silvester Hofer, Nina Dominik, Lukas Hartl, Michael Schwarz, Georg Semmler, Christian Sebesta, Paul Thöne, Sophia Geisselbrecht, Benjamin Maasoumy, Eduardo Alvarez, Martin Sebastian McCoy, Oleksandr Petrenko, Jiří Reiniš, Philipp Schwabl, Albert F. Stättermayer, Michael Trauner, Mattias Mandorfer, Thomas Reiberger.
DOI: 10.1016/j.jhepr.2025.101642