Liver cirrhosis is a common consequence of prolonged stress on the liver, for example due to alcohol abuse or because of hepatitis. In close collaboration, scientists from Thomas Reiberger’s research group at CeMM, the Medical University of Vienna and the Ludwig Boltzmann Institute for Rare and Undiagnosed Diseases (LBI-RUD) have now developed an algorithm that allows the risk of severe complications in patients with liver cirrhosis to be estimated easily and without invasive procedures.
Liver cirrhosis represents an advanced stage of virtually all types of liver disease. It develops in response to repeated or severe damage to the liver, for example due to fatty liver disease or viral hepatitis. Doctors distinguish between two clinical stages of cirrhosis: an early-asymptomatic (compensated) stage, in which patients usually have only few and unspecific symptoms such as fatigue, and an advanced-symptomatic (decompensated) stage, in which severe complications such as internal bleeding (variceal hemorrhage) or accumulation of water in the abdominal cavity (ascites) can occur, which can even lead to death. To determine the risk for such complications, it is currently necessary to perform invasive measurement of the hepatovenous pressure gradient (HVPG). Above an HVPG value of ≥10 mm Hg, there is in principle a risk of liver-associated complications (hepatic decompensation), and above an HVPG value of ≥16 mm Hg, the likelihood of hepatic decompensation increases significantly. To date, this risk in patients with compensated cirrhosis could only be determined by invasive HVPG measurement, which is not widely available. Scientists from Thomas Reiberger’s research group at the CeMM Research Center for Molecular Medicine of the Austrian Academy of Sciences, LBI-RUD and at MedUni Vienna, as well as from Stefan Kubicek’s research group at CeMM, have now developed an algorithm that enables the early detection of the risk of hepatic decompensation quickly and easily. In their study, first authors Jiří Reiniš (CeMM) and Oleksandr Petrenko (CeMM, LBI-RUD, MedUni Vienna) trained machine learning models based on blood test parameters from patients with compensated cirrhosis to detect HVPG values ≥10 or ≥16 mm Hg to identify those who are in principle at risk or at particularly high risk for developing hepatic decompensation.
Identification of the most important clinical parameters for risk prognosis
The main source of data for the project was the ongoing Vienna Cirrhosis Study (VICIS), which is being conducted at the Department of Gastroenterology and Hepatology at MedUni Vienna and Vienna General Hospital. From the set of clinical variables, three and five optimal parameters for the detection of high-risk patients were computationally determined, and based on them, a 3-parameter and a 5-parameter model for the simple determination of the risk of hepatic decompensation were developed. Study leader Thomas Reiberger explains: “The results show that with the help of our developed algorithm, a more precise risk assessment than with other available methods can be achieved very simply and quickly and without additional burden for patients by means of a simple blood sample.
Method easy to perform, online calculator developed
To evaluate the diagnostic performance of the noninvasive 3P as well as the 5P model for predicting an HVPG ≥10 mm Hg and ≥16 mm Hg, respectively, the two models were validated using a multicenter data set of 1,232 patients with compensated cirrhosis from eight European clinical centers. “The novel approach proved to be of excellent diagnostic value. Since the new test is based on only three or five commonly available parameters, and does not require any special and expensive equipment, as would be necessary for measuring liver stiffness, for example, and – in contrast to the invasive HVPG measurement – also poses no risk to patients, it is easy to use anywhere,” says Reiberger. “While HVPG measurement remains necessary for accurate follow-up of clinically significant or severe portal hypertension, the new, noninvasive approach may serve primarily for early risk detection and thus prevention of cirrhotic decompensation, or for selection of patients who may particularly benefit from participation in clinical therapy trials.”
Due to its ease of application, the proposed method can be used in routine examinations without additional costs. For ease and widespread use, an online calculator was developed for clinicians to calculate HVPG levels and associated risk for affected patients with compensated cirrhosis.
Original publication:
Jiří Reiniš, Oleksandr Petrenko, et al. Assessment of portal hypertension severity using machine learning models in patients with compensated cirrhosis, Journal of Hepatology,2022, DOI: https://doi.org/10.1016/j.jhep.2022.09.012