3P and 5P models of limited value for the detection of clinically significant portal hypertension in patients with hepatitis delta.

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 AccessReply to: “3P and 5P models of limited value for the detection of clinically significant portal hypertension in patients with hepatitis delta”Journal of HepatologyVol. 79Issue 1PreviewThe non-invasive 3P/5P HVPG model requires evaluation in rare diseases Full-Text PDF We read with interest the excellent article by Reiniš et al. published in the Journal of Hepatology.[1]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 Feb; 78: 390-400Abstract Full Text Full Text PDF PubMed Scopus (3) Google Scholar 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]de Franchis R. Bosch J. Garcia-Tsao G. Reiberger T. Ripoll C. Baveno VII - renewing consensus in portal hypertension.J Hepatol. 2022 Apr; 76: 959-974Abstract Full Text Full Text PDF PubMed Scopus (436) Google Scholar The gold standard for diagnosing PH is the invasive measurement of the hepatic venous pressure gradient (HVPG) which is not broadly available. To overcome this limitation, non-invasive parameters to reliably identify CSPH are a major unmet need. Based on an elegant machine-learning approach, the authors developed three- and five-parameter models (3P, 5P) to predict CSPH (HVPG ≥10 mmHg) and sPH (HVPG ≥16 mmHg). The authors suggest that the model works independently from the underlying liver disease. However, we assume that some rare diseases, i.e. chronic hepatitis D virus (HDV) infection, might have been underrepresented in their study. Thus, we decided to evaluate the predictive value of the suggested 3P and 5P models in a cohort of patients with chronic HDV infection and cACLD who recently underwent invasive HVPG measurement as part of the standard diagnostic work up for cirrhosis at our center. We included all 20 HDV-infected patients who received HVPG measurement between March 2021 and September 2022. Laboratory values were measured at the day of HVPG measurement. In our cohort, mean HVPG and viral load were 13.1 mmHg (± 5.7 mmHg) and 4.91 log10 IU/ml, respectively (Fig. 1A). None of the patients received antiviral treatment against HDV infection. HVPG measurement was performed according to Baveno VII recommendations[2]de Franchis R. Bosch J. Garcia-Tsao G. Reiberger T. Ripoll C. Baveno VII - renewing consensus in portal hypertension.J Hepatol. 2022 Apr; 76: 959-974Abstract Full Text Full Text PDF PubMed Scopus (436) Google Scholar without the use of sedative medication during measurement. Treatment with non-selective beta blockers (NSBBs) was continued if the patient had been on treatment prior to measurement (n = 8). In our cohort, all patients (20/20) showed portal hypertension (HVPG ≥6 mmHg). 13 of 20 patients (65%) had CSPH of whom six patients (30%) showed HVPG ≥16 mmHg and were therefore diagnosed with sPH (Fig. 1B). Seven patients (35%) had portal hypertension without clinical significance (HVPG <10 mmHg) of whom one patient received NSBBs. The remaining seven patients on NSBBs had CSPH. When applying the online tool for the 3P model to our cohort, only 38.5% (5/13) of those with CSPH showed a probability of ≥90% for having CSPH according to the model (Fig. 1C). By using the 5P model, the detection rate increased to 58.3%, translating to seven of 12 patients who showed a probability of ≥90% for having CSPH. Six patients in the cohort had sPH with an HVPG ≥16 mmHg. None of the patients showed a probability of ≥90% for having sPH when using the online tool for the 3P model whereas one patient (1/6, 16.7%) would have been identified using the 5P model. When applying the 3P and 5P models to the subgroup of patients not receiving NSBB treatment, detection rates remained unchanged (3P/CSPH: 16.7% [1/6], 3P/sPH: 0% [0/2]; 5P/CSPH: 50% [3/6], 5P/sPH: 0% [0/2]). With a cut-off of 90% for exclusion of CSPH, all patients without CSPH would have been correctly identified by the 3P and 5P models. However, eight (53.3%) and five (41.7%) patients from our cohort would have been misclassified as having no CSPH. Taken together our analyses indicate that the machine learning-derived 3P and 5P models have limited value to reliably identify patients with CSPH and sPH in our cohort of viremic HDV-infected patients. In the analysis by Reiniš et al. patients with viral hepatitis were also included in development and validation. However, subgroups of viral hepatitis were not shown and therefore it is not clear how many patients with chronic HDV infection had been included. Due to the rarity of the disease, the number is expected to be low. Yet, patients with chronic HDV infection may show especially high levels of liver damage. It has been shown that existing non-invasive fibrosis scores have limitations in the context of chronic HDV infection.[3]Lutterkort G.L. Wranke A. Yurdaydin C. Budde E. Westphal M. Lichtinghagen R. et al.Non-invasive fibrosis score for hepatitis delta.Liver Int. 2017 Feb; 37: 196-204Crossref PubMed Scopus (32) Google Scholar Cut-off values for non-invasive tests to assess severity of liver disease have to be adapted for chronic HDV infection and cannot be transferred unchanged from other viral liver diseases.[4]Da B.L. Surana P. Kleiner D.E. Heller T. Koh C. The Delta-4 fibrosis score (D4FS): a novel fibrosis score in chronic hepatitis.D Antivir Res. 2020 Feb; 174104691Google Scholar,[5]Da B.L. Surana P. Takyar V. Kleiner D.E. Heller T. Koh C. Vibration-controlled transient elastography for the detection of cirrhosis in chronic hepatitis D infection.J Viral Hepat. 2020 Apr; 27: 428-436Crossref PubMed Scopus (13) Google Scholar The same reason might play a role when applying non-invasive tools to assess portal hypertension in this special etiology of liver disease. In our cohort, a higher proportion of patients received NSBB treatment than in the study of Reiniš et al. (40% vs. 2.27%). However, detection rates of CSPH and sPH by the non-invasive model did not improve when excluding patients on NSBB treatment from our analysis. Based on the data from our cohort, we think that the proposed model should be used with caution in the setting of replicative, chronic HDV infection. Larger studies are needed to validate the 3P and 5P models for this special patient population. No financial support was received for this study. Lisa Sandmann is partially funded by the German Research Foundation (DFG), through the PRACTIS–Clinician Scientist Program of Hannover Medical School (ME 3696/3-1). Lisa Sandmann reports lecture and personal fees from Falk Pharma e.V., Roche and Gilead and travel support from Abbvie. Tammo L. Tergast has nothing to declare. Heiner Wedemeyer reports grants/research support from AbbVie, Biotest, BMS, Gilead, Merck/MSD, Novartis, Roche; Personal fees from Abbott, AbbVie, Altimmune, Biotest, BMS, BTG, Dicerna, Gilead, Janssen, Merck/MSD, MYR GmbH, Novartis, Roche, Siemens. Katja Deterding received lecture and personal fees from Gilead, Falk Pharma e.V., Abbvie, MSD/Merck and Alnylam Benjamin Maasoumy reports grants/research support from Abbott, Fujirebio, Roche; Personal fees from Abbott, AbbVie, BMS, Janssen, Merck/MSD, Roche, Fujirebio, Astellas. Please refer to the accompanying ICMJE disclosure forms for further details. Lisa Sandmann: Conceptualization, investigation, analysis, data generation, writing of the manuscript (original draft), Tammo L Tergast: Conceptualization, data generation, critical revision of the manuscript for important intellectual content, Heiner Wedemeyer: Conceptualization, critical revision of the manuscript for important intellectual content, Katja Deterding: Conceptualization, critical revision of the manuscript for important intellectual content; Benjamin Maasoumy: Conceptualization, supervision, project administration, writing, critical revision of the manuscript for important intellectual content. The following are the supplementary data to this article: Download .pdf (.61 MB) Help with pdf files Multimedia component 1