Assessment of portal hypertension severity using machine learning models in patients with compensated cirrhosis

BACKGROUND & AIMS

In 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.

METHODS

A detailed laboratory workup of individuals with cACLD recruited from the Vienna cohort (NCT03267615) was utilised to predict clinically significant portal hypertension (CSPH, i.e., HVPG ≥10 mmHg) and severe PH (i.e., HVPG ≥16 mmHg). The MLMs were then evaluated in individual external datasets and optimised in the merged cohort.

RESULTS

Among 1,232 participants with cACLD, the prevalence of CSPH/severe PH was similar in the Vienna (n = 163, 67.4%/35.0%) and validation (n = 1,069, 70.3%/34.7%) cohorts. The MLMs were based on 3 (3P: platelet count, bilirubin, international normalised ratio) or 5 (5P: +cholinesterase, +gamma-glutamyl transferase, +activated partial thromboplastin time replacing international normalised ratio) laboratory parameters. The MLMs performed robustly in the Vienna cohort. 5P-MLM had the best AUCs for CSPH (0.813) and severe PH (0.887) and compared favourably to liver stiffness measurement (AUC: 0.808). Their performance in external validation datasets was heterogeneous (AUCs: 0.589-0.887). Training on the merged cohort optimised model performance for CSPH (AUCs for 3P and 5P: 0.775 and 0.789, respectively) and severe PH (0.737 and 0.828, respectively).

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 individuals with CSPH or severe PH, who are thus at risk of hepatic decompensation.

IMPACT AND IMPLICATIONS

We used machine learning models based on widely available laboratory parameters to develop a non-invasive model to predict the severity of portal hypertension in individuals with compensated cirrhosis, who currently require invasive measurement of hepatic venous pressure gradient. We validated our findings in a large multicentre cohort of individuals with advanced chronic liver disease (cACLD) of any cause. Finally, we provide a readily available online calculator, based on 3 (platelet count, bilirubin, international normalised ratio) or 5 (platelet count, bilirubin, activated partial thromboplastin time, gamma-glutamyltransferase, choline-esterase) widely available laboratory parameters, that clinicians can use to predict the likelihood of their patients with cACLD having clinically significant or severe portal hypertension.

Decompensation in Advanced Nonalcoholic Fatty Liver Disease May Occur at Lower Hepatic Venous Pressure Gradient Levels Than in Patients With Viral Disease

BACKGROUND & AIMS

Portal hypertension is the strongest predictor of hepatic decompensation and death in patients with cirrhosis. However, its discriminatory accuracy in patients with nonalcoholic fatty liver disease (NAFLD) has been challenged because hepatic vein catheterization may not reflect the real portal vein pressure as accurately as in patients with other etiologies. We aimed to evaluate the relationship between hepatic venous pressure gradient (HVPG) and presence of portal hypertension-related decompensation in patients with advanced NAFLD (aNAFLD).

METHODS

Multicenter cross-sectional study included 548 patients with aNAFLD and 444 with advanced RNA-positive hepatitis C (aHCV) who had detailed portal hypertension evaluation (HVPG measurement, gastroscopy, and abdominal imaging). We examined the relationship between etiology, HVPG, and decompensation by logistic regression models. We also compared the proportions of compensated/decompensated patients at different HVPG levels.

RESULTS

Both cohorts, aNAFLD and aHVC, had similar baseline age, gender, Child-Pugh score, and Model for End-Stage Liver Disease score. Median HVPG was lower in the aNAFLD cohort (13 vs 15 mmHg) despite similar liver function and higher rates of decompensation in aNAFLD group (32% vs 25%; P = .019) than in the aHCV group. For any of the HVPG cutoff analyzed (<10, 10-12, or 12 mmHg) the prevalence of decompensation was higher in the aNAFLD group than in the aHCV group.

CONCLUSIONS

Patients with aNAFLD have higher prevalence of portal hypertension-related decompensation at any value of HVPG as compared with aHCV patients. Longitudinal studies aiming to identify HVPG thresholds able to predict decompensation and long-term outcomes in aNAFLD population are strongly needed.

Acute kidney injury in patients with cirrhosis: Prospective longitudinal study in 405 patients

BACKGROUND

Acute kidney injury (AKI) is common in patients with cirrhosis. In 2015, the International Club of Ascites (ICA) proposed new definitions of AKI in order to improve the prediction of outcomes. Our aim was to assess the prevalence and prognostic value of ICA 2015 - AKI criteria in hospitalised patients with cirrhosis.

METHODS

We prospectively collected data from 405 consecutive cirrhotic patients admitted to the hospital between November 2016 and November 2017. AKI was diagnosed at inclusion according to ICA 2015 criteria, and was assessed to predict 30-day and 90-day in-hospital mortality.

RESULTS

AKI was diagnosed in 78 (19.3%) patients. AKI was independently associated with 90-day death (HR 7.61; 95% CI 4.75-12.19; p < 0.001). In hospital, 30-day and 90-day survival was lower in the group of patients with AKI compared to the group with no AKI (72% vs. 98%, p < 0.001; 64% vs. 96%, p < 0.001; and 49% vs. 81%, p < 0.001, respectively). Patients with stage 1a AKI had a lower 30-day and 90-day survival compared to the group of patients who did not develop AKI (71% vs. 96%, p < 0.001, and 71% vs. 91%, p < 0.01, respectively) and better survival than patients with more severe AKI (71% vs. 40%, p < 0.01).

CONCLUSIONS

AKI was independently associated with mortality in patients with cirrhosis, even at the very early 1a stage. Response to treatment improved survival, and was inversely proportional to the stage of AKI, which suggests that treatment should be started at the earliest stage of AKI.

Non-invasive diagnosis and follow-up of benign liver tumours

Hepatocellular benign liver tumours are mainly developed on normal liver and include hepatic hemangioma, focal nodular hyperplasia and hepatocellular adenoma from the most frequent to the less frequent. The diagnosis of hepatic hemangioma and of simple hepatic biliary cysts can be performed using non-invasive criteria using liver ultrasonography or contrast enhanced MRI. Most of the time the diagnosis of focal nodular hyperplasia can be achieved using contrast-enhanced ultrasonography or contrast enhanced MRI with an additional value of hepatobiliary contrast-agent in this setting. Rarely, if a doubt persists, a tumour and non-tumour liver biopsy can be required in order to establish the diagnosis. As hepatic hemangioma, simple hepatic biliary cysts and focal nodular hyperplasia are not associated with complications, they don't require any treatments or follow-up. Hepatocellular adenomas are mainly diagnosed at histology on surgical samples or liver biopsy even if some radiological features are highly suggestive of several subtypes of hepatocellular adenomas. Finally, the management of hepatocellular adenomas should be guided according to the tumour size, gender but also to the molecular subtypes.

Non-invasive diagnosis and follow-up of primary malignant liver tumours

Among a wide range of malignant liver tumours, hepatocellular carcinoma (HCC) developed on a background of cirrhosis represents the most frequent clinical situation. In this setting, HCC is one of the rare solid tumours for which histological confirmation is not mandatory. The convergence of multiple arguments obtained by non-invasive parameters using radiological findings allows to avoid liver biopsy in a large proportion of patients when a diagnosis of underlying cirrhosis is ascertained. Conversely, in case of atypical presentation or in order to exclude other rare malignant tumours mostly developed in the absence of cirrhosis, liver biopsy will then be essential. Based on typical radiological patterns described by contrast-enhanced imaging, numerous clinical guidelines have endorsed non-invasive diagnosis, staging and monitoring of HCC patients under treatment since 20 years. These algorithms have evolved over the years, taking into account progress in radiological technology and advances in curative or palliative procedures. Large cohort studies have also helped to refine diagnostic criteria and prognostication in the setting of complex therapeutic strategy. Unsupervised multi-analysis approaches both at the biological and radiological levels will in the future enrich the panel of non-invasive markers useful in clinical practice to manage HCC and other malignant tumours.

Non-invasive diagnosis and follow-up in liver transplantation

The field of liver transplantation directly or indirectly embodies all liver diseases, in addition to specific ones related to organ rejection (cellular and humoral). The recommended non-invasive methods for determining the indication for liver transplantation are the Model for End-stage Liver Disease score, and the alpha-foetoprotein score in case of hepatocellular carcinoma. Radiological methods are the cornerstones for the diagnosis of vascular and biliary complications after liver transplantation. The possible diseases of the liver graft after transplantation are multiple and often intertwined. Non-invasive diagnostic methods have been poorly evaluated in this context, apart from the recurrence of hepatitis C. Liver biopsy remains the gold standard for evaluating graft lesions in the majority of cases, especially graft rejection.

Combination of Model for End-Stage Liver Disease and Lactate Predicts Death in Patients Treated With Salvage Transjugular Intrahepatic Portosystemic Shunt for Refractory Variceal Bleeding

BACKGROUND AND AIMS

Data about the prognosis of salvage transjugular intrahepatic portosystemic shunt (TIPS) using covered stents for refractory variceal bleeding caused by portal hypertension are scarce. We aimed to assess survival and to identify predictors of mortality in these patients.

APPROACH AND RESULTS

One hundred sixty-four patients with cirrhosis from five centers treated with salvage TIPS between 2007 and 2017 were retrospectively divided into a derivation cohort (83 patients) and a validation cohort (81 patients). Comparisons were performed using the Mann-Whitney and Fischer's exact test. Six-week overall survival (OS) was correlated with variables on the day of the TIPS using Kaplan-Meier curves with log-rank test and univariate/multivariate analyses using the Cox model. Eighty-three patients were included in the derivation cohort (male, 78%; age, 55 years, alcohol-associated cirrhosis, 88%; Model for End-Stage Liver Disease [MELD], 19 [15-27]; arterial lactate, 3.7 mmol/L [2.0-8.3]). Six-week OS rate was 58%. At multivariate analysis, the MELD score (OR, 1.064; 95% CI, 1.005-1.126; P = 0.028) and arterial lactate (OR, 1.063; 95% CI, 1.013-1.114; P = 0.032) were associated with 6-week OS. Six-week OS rates were 100% in patients with arterial lactate ≤2.5 mmol/L and MELD score ≤ 15 and 5% in patients with lactate ≥12 mmol/L and/or MELD score ≥ 30. The 81 patients of the validation cohort had similar MELD and arterial lactate level but lower creatinine level (94 vs 106 µmol/L, P = 0.008); 6-week OS was 67%. Six-week OS rates were 86% in patients with arterial lactate ≤2.5 mmol/L and MELD score ≤ 15 and 10% for patients with lactate ≥12 mmol/L and/or MELD score ≥ 30. In the overall cohort, rebleeding rate was 15.8% at 6 weeks, and the acute-on-chronic liver failure grade (OR, 1.699; 95% CI, 1.056-1.663; P = 0.040) was independently associated with rebleeding.

CONCLUSIONS

After salvage TIPS, 6-week mortality remains high and can be predicted by MELD score and lactate. Survival rate at 6 weeks was >85% in patients with arterial lactate ≤2.5 mmol/L and MELD score ≤ 15, while mortality was >90% for lactate ≥12 mmol/L and/or MELD score ≥ 30.

Liver Stiffness by Transient Elastography to Detect Porto-Sinusoidal Vascular Liver Disease With Portal Hypertension

BACKGROUND AND AIMS

Porto-sinusoidal vascular liver disease (PSVD) is a rare cause of portal hypertension. PSVD is still often misdiagnosed as cirrhosis, emphasizing the need to improve PSVD diagnosis strategies. Data on liver stiffness measurement using transient elastography (TE-LSM) in PSVD are limited. The aim of this study was to evaluate the accuracy of TE-LSM to discriminate PSVD from cirrhosis in patients with signs of portal hypertension.

APPROACH AND RESULTS

Retrospective multicenter study comparing TE-LSM in patients with PSVD, according to Vascular Liver Disease Interest Group criteria, with patients with compensated biopsy-proven cirrhosis associated with alcohol (n = 117), HCV infection (n = 110), or NAFLD (n = 46). All patients had at least one sign of portal hypertension among gastroesophageal varices, splenomegaly, portosystemic collaterals, history of ascites, or platelet count < 150 × 10⁹ /L. The 77 patients with PSVD included in the test cohort had lower median TE-LSM (7.9 kPa) than the patients with alcohol-associated, HCV-related, and NAFLD-related cirrhosis (33.8, 18.2, and 33.6 kPa, respectively; P < 0.001). When compared with cirrhosis, a cutoff value of 10 kPa had a specificity of 97% for the diagnosis of PSVD with a 85% positive predictive value. A cutoff value of 20 kPa had a sensitivity of 94% for ruling out PSVD with a 97% negative predictive value. Of the patients, 94% were well-classified. Even better results were obtained in a validation cohort including 78 patients with PSVD.

CONCLUSIONS

This study including a total of 155 patients with PSVD and 273 patients with cirrhosis demonstrates that TE-LSM < 10 kPa strongly suggests PSVD in patients with signs of portal hypertension. Conversely, when TE-LSM is >20 kPa, PSVD is highly unlikely.

Characterization of Blood Immune Cells in Patients With Decompensated Cirrhosis Including ACLF

Background and Aims

Patients with cirrhosis and acute-on-chronic liver failure (ACLF) have immunosuppression, indicated by an increase in circulating immune-deficient monocytes. The aim of this study was to investigate simultaneously the major blood-immune cell subsets in these patients.

Material and Methods

Blood taken from 67 patients with decompensated cirrhosis (including 35 critically ill with ACLF in the intensive care unit), and 12 healthy subjects, was assigned to either measurements of clinical blood counts and microarray (genomewide) analysis of RNA expression in whole-blood; microarray (genomewide) analysis of RNA expression in blood neutrophils; or assessment of neutrophil antimicrobial functions.

Results

Several features were found in patients with ACLF and not in those without ACLF. Indeed, clinical blood count measurements showed that patients with ACLF were characterized by leukocytosis, neutrophilia, and lymphopenia. Using the CIBERSORT method to deconvolute the whole-blood RNA-expression data, revealed that the hallmark of ACLF was the association of neutrophilia with increased proportions of macrophages M0-like monocytes and decreased proportions of memory lymphocytes (of B-cell, CD4 T-cell lineages), CD8 T cells and natural killer cells. Microarray analysis of neutrophil RNA expression revealed that neutrophils from patients with ACLF had a unique phenotype including induction of glycolysis and granule genes, and downregulation of cell-migration and cell-cycle genes. Moreover, neutrophils from these patients had defective production of the antimicrobial superoxide anion.

Conclusions

Genomic analysis revealed that, among patients with decompensated cirrhosis, those with ACLF were characterized by dysregulation of blood immune cells, including increases in neutrophils (that had a unique phenotype) and macrophages M0-like monocytes, and depletion of several lymphocyte subsets (including memory lymphocytes). All these lymphocyte alterations, along with defective neutrophil superoxide anion production, may contribute to immunosuppression in ACLF, suggesting targets for future therapies.

Abdominal Surgery in Patients With Idiopathic Noncirrhotic Portal Hypertension: A Multicenter Retrospective Study

In patients with idiopathic noncirrhotic portal hypertension (INCPH), data on morbidity and mortality of abdominal surgery are scarce. We retrospectively analyzed the charts of patients with INCPH undergoing abdominal surgery within the Vascular Liver Disease Interest Group network. Forty-four patients with biopsy-proven INCPH were included. Twenty-five (57%) patients had one or more extrahepatic conditions related to INCPH, and 16 (36%) had a history of ascites. Forty-five procedures were performed, including 30 that were minor and 15 major. Nine (20%) patients had one or more Dindo-Clavien grade ≥ 3 complication within 1 month after surgery. Sixteen (33%) patients had one or more portal hypertension-related complication within 3 months after surgery. Extrahepatic conditions related to INCPH (P = 0.03) and history of ascites (P = 0.02) were associated with portal hypertension-related complications within 3 months after surgery. Splenectomy was associated with development of portal vein thrombosis after surgery (P = 0.01). Four (9%) patients died within 6 months after surgery. Six-month cumulative risk of death was higher in patients with serum creatinine ≥ 100 μmol/L at surgery (33% versus 0%, P < 0.001). An unfavorable outcome (i.e., either liver or surgical complication or death) occurred in 22 (50%) patients and was associated with the presence of extrahepatic conditions related to INCPH, history of ascites, and serum creatinine ≥ 100 μmol/L: 5% of the patients with none of these features had an unfavorable outcome versus 32% and 64% when one or two or more features were present, respectively. Portal decompression procedures prior to surgery (n = 10) were not associated with postoperative outcome. Conclusion: Patients with INCPH are at high risk of major surgical and portal hypertension-related complications when they harbor extrahepatic conditions related to INCPH, history of ascites, or increased serum creatinine.