Reply to: “3P and 5P models of limited value for the detection of clinically significant portal hypertension in patients with hepatitis delta”

3P and 5P models of limited value for the detection of clinically significant portal hypertension in patients with hepatitis deltaJournal of HepatologyVol. 79Issue 1PreviewWe read with interest the excellent article by Reiniš et al. published in the Journal of Hepatology.1 The diagnosis of clinically significant portal hypertension (CSPH) and significant portal hypertension (sPH) in patients with compensated advanced chronic liver disease (cACLD) is of clinical importance as it identifies patients at risk for future decompensation and indicates the need for treatment.2 The gold standard for diagnosing PH is the invasive measurement of the hepatic venous pressure gradient (HVPG) which is not broadly available. Full-Text PDF Assessment of portal hypertension severity using machine learning models in patients with compensated cirrhosisJournal of HepatologyVol. 78Issue 2PreviewIn individuals 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 individuals with cACLD. Full-Text PDF Open Access The non-invasive 3P/5P HVPG model requires evaluation in rare diseases To the Editors: We thank Dr. Sandmann and colleagues of the Hannover Medical School for sharing their results[1]Sandmann L. Tergast T.L. Wedemeyer H. Deterding K. Maasoumy B. 3P and 5P model of limited value for the detection of clinically significant portal hypertension in patients with hepatitis delta.J Hepatol. 2023; https://doi.org/10.1016/j.jhep.2023.03.012Abstract Full Text Full Text PDF PubMed Scopus (0) Google Scholar on the application of our machine-learning models based on three (3P) or five (5P) simple laboratory parameters[2]Reiniš J. Petrenko O. Simbrunner B. Hofer B.S. Schepis F. Scoppettuolo M. et al.Assessment of portal hypertension severity using machine learning models in patients with compensated cirrhosis.J Hepatol. 2023; 78: 390-400Abstract Full Text Full Text PDF PubMed Scopus (3) Google Scholar in a cohort of 20 patients with viremic hepatitis D virus (HDV) infection, and for discussing the limitations of this approach. Specifically, in their cohort of hemodynamically characterized patients with HDV, only 38.5% and 58.3% of patients with clinically significant portal hypertension (CSPH) had a predicted probability of CSPH ≥90% according to the 3P and the 5P models, respectively. Further, none and only one (16.7%) of the six patients with real HVPG ≥16mmHg had a predicted probability ≥90% with the 3P and 5P models, respectively. The authors conclude that our proposed models have to be used with caution in the setting of viremic HDV infection. We agree with this conclusion for several reasons. First, HDV is a rare disease, and as pointed out by the authors, was underrepresented in the training and validation cohorts of our study owing to the overall low prevalence of HDV in the participating countries.[3]Jachs M. Binter T. Schmidbauer C. Hartl L. Strasser M. Laferl H. et al.Hepatitis D virus (HDV) prevalence in Austria is low but causes considerable morbidity due to fast progression to cirrhosis.United Eur Gastroenterol J. 2021; 9: 1119-1127Crossref PubMed Scopus (14) Google Scholar This impacts on the model’s applicability in hepatitis D, as for other non-invasive tests (NITs) developed in patients with viral hepatitis B or C.[4]Takyar V. Surana P. Kleiner D.E. Wilkins K. Hoofnagle J.H. Liang T.J. et al.Noninvasive markers for staging fibrosis in chronic delta hepatitis.Aliment Pharmacol Ther. 2017; 45: 127-138Crossref PubMed Scopus (35) Google Scholar Importantly, in our article we already underlined the importance of evaluating the discrimination and calibration[5]Van Calster B. McLernon D.J. van Smeden M. Wynants L. Steyerberg E.W. Topic Group ‘Evaluating diagnostic t, et al. Calibration: the Achilles heel of predictive analytics.BMC Med. 2019; 17: 230Crossref PubMed Scopus (448) Google Scholar of our 3P/5P models in larger and different patients cohorts,[2]Reiniš J. Petrenko O. Simbrunner B. Hofer B.S. Schepis F. Scoppettuolo M. et al.Assessment of portal hypertension severity using machine learning models in patients with compensated cirrhosis.J Hepatol. 2023; 78: 390-400Abstract Full Text Full Text PDF PubMed Scopus (3) Google Scholar the latter including rare liver diseases. Furthermore, and perhaps even more relevant, the model’s prognostic utility – compared to invasive HVPG-based risk assessment - needs to be assessed in the future. Moreover, the analytical approach chosen by Dr. Sandmann highlights another important limitation of our model, which is inherent to NIT-based strategies. Choosing a high predicted probability or high positive predictive value of a certain threshold (e.g., ≥90%) to rule-in a certain condition (in this case CSPH) will inevitably result in a high proportion of patients who have the condition being missed. This also applies to the current standard of care, i.e. a combination of liver stiffness measurement by vibration-controlled transient elastography and platelet count. Accordingly, lower predicted risk thresholds may be applied, especially if the consequence would be the initiation of an effective treatment, such as carvedilol for CSPH. Alternatively, additional NITs such as the von Willebrand factor-to platelet ratio (VITRO) or spleen stiffness measurement may be added. All things considered, the data presented in the letter by Dr. Sandmann et al. clearly highlights the unmet need for the development and validation of accurate NITs or models for CSPH risk assessment in larger cohorts of hemodynamically characterized patients with HDV, which will evidently require collaborative studies to achieve a sufficient sample size. No financial support was received for this article. The authors declare no conflicts of interest that pertain to this work. Please refer to the accompanying ICMJE disclosure forms for further details. Study concept and design: all authors; interpretation of data: all authors; drafting of the manuscript: all authors; critical revision of the manuscript for important intellectual content: all authors; final approval of the version to be published: all authors; agreement to be accountable for all aspects of the work: all authors. The following are the supplementary data to this article: Download .pdf (.29 MB) Help with pdf files Multimedia component 1