1. Assessment of portal hypertension severity using machine learning models in patients with compensated cirrhosis
- Author
-
Jiří Reiniš, Oleksandr Petrenko, Benedikt Simbrunner, Benedikt S. Hofer, Filippo Schepis, Marco Scoppettuolo, Dario Saltini, Federica Indulti, Tomas Guasconi, Agustin Albillos, Luis Téllez, Càndid Villanueva, Anna Brujats, Juan Carlos Garcia-Pagan, Valeria Perez-Campuzano, Virginia Hernández-Gea, Pierre-Emmanuel Rautou, Lucile Moga, Thomas Vanwolleghem, Wilhelmus J. Kwanten, Sven Francque, Jonel Trebicka, Wenyi Gu, Philip G. Ferstl, Lise Lotte Gluud, Flemming Bendtsen, Søren Møller, Stefan Kubicek, Mattias Mandorfer, and Thomas Reiberger
- Subjects
hepatic venous pressure gradient ,non-invasive testing ,machine learning ,Hepatology ,Human medicine - Abstract
BACKGROUND & AIMS In patients with compensated advanced chronic liver disease (cACLD), the severity of portal hypertension (PH) determines the risk of decompensation. Invasive measurement of the hepatic venous pressure gradient (HVPG) is the diagnostic gold standard for PH. We evaluated the utility of machine learning models (MLMs) based on standard laboratory parameters to predict the severity of PH in cACLD patients. METHODS A detailed laboratory workup of cACLD patients recruited from the VIENNA cohort (NCT03267615) was utilised to predict clinically significant portal hypertension (CSPH, i.e., HVPG≥10mmHg) and severe PH (i.e. HVPG≥16mmHg). The MLMs were then evaluated in individual external datasets and optimised in the merged cohort. RESULTS Among 1232 cACLD patients, the CSPH and severe PH prevalence in VIENNA (n = 163, 67.4%/35.0%) and the validation cohort (n = 1069, 70.3%/34.7%) were similar. The MLMs were based on 3 (3P; platelets, bilirubin, INR) or 5 (5P; +cholinesterase, +gamma-glutamyl transferase, +aPTT replacing INR) laboratory parameters. The MLMs performed robustly in VIENNA with best AUROCs for CSPH by 5P-MLM: 0.813 and for severe PH by 5P-MLM: 0.887 and compared favourably to liver stiffness measurement (AUROC: 0.808). Their performance in external validation datasets was heterogeneous (AUROCs 0.589-0.887). Training on the merged cohort optimized the MLM performance for CSPH (AUROCs: 3P: 0.775, 5P: 0.789) and for severe PH (AUROCs 3P: 0.737, 5P: 0.828). CONCLUSIONS Internally-trained MLMs reliably predicted PH severity in the VIENNA cACLD cohort but exhibited heterogeneous results on external validation. The proposed 3P/5P online tool can reliably identify patients with CSPH or severe PH, thus, at risk for hepatic decompensation. LAY SUMMARY The gold standard for diagnosing portal hypertension is the invasive measurement of the hepatic venous pressure gradient. In this work, we selected the most suitable, widely available laboratory parameters for machine learning models to predict the likelihood and severity of portal hypertension in patients with compensated cirrhosis. This will aid in the identification of patients who are at the highest risk for subsequent hepatic decompensation.
- Published
- 2023
- Full Text
- View/download PDF