A Liver Stiffness-Based Etiology-Independent Machine Learning Algorithm to Predict Hepatocellular Carcinoma.

BACKGROUND & AIMS: The existing hepatocellular carcinoma (HCC) risk scores have modest accuracy, and most are specific to chronic hepatitis B infection. In this study, we developed and validated a liver stiffness-based machine learning algorithm (ML) for prediction and risk stratification of HCC in various chronic liver diseases (CLDs). METHODS: MLs were trained for prediction of HCC in 5155 adult patients with various CLDs in Korea and further tested in 2 prospective cohorts from Hong Kong (HK) (N = 2732) and Europe (N = 2384). Model performance was assessed according to Harrell's C-index and time-dependent receiver operating characteristic (ROC) curve. RESULTS: We developed the SMART-HCC score, a liver stiffness-based ML HCC risk score, with liver stiffness measurement ranked as the most important among 9 clinical features. The Harrell's C-index of the SMART-HCC score in HK and Europe validation cohorts were 0.89 (95% confidence interval, 0.85-0.92) and 0.91 (95% confidence interval, 0.87-0.95), respectively. The area under ROC curves of the SMART-HCC score for HCC in 5 years was ≥0.89 in both validation cohorts. The performance of SMART-HCC score was significantly better than existing HCC risk scores including aMAP score, Toronto HCC risk index, and 7 hepatitis B-related risk scores. Using dual cutoffs of 0.043 and 0.080, the annual HCC incidence was 0.09%-0.11% for low-risk group and 2.54%-4.64% for high-risk group in the HK and Europe validation cohorts. CONCLUSIONS: The SMART-HCC score is a useful machine learning-based tool for clinicians to stratify HCC risk in patients with CLDs.

Adequate versus deep response to ursodeoxycholic acid in primary biliary cholangitis: To what extent and under what conditions is normal alkaline phosphatase level associated with complication-free survival gain?

BACKGROUND AND AIMS: Normal alkaline phosphatase (ALP) levels in ursodeoxycholic acid (UDCA)-treated patients with primary biliary cholangitis (PBC) are associated with better long-term outcome. However, second-line therapies are currently recommended only when ALP levels remain above 1.5 times the upper limit of normal (×ULN) after 12-month UDCA. We assessed whether, in patients considered good responders to UDCA, normal ALP levels were associated with significant survival gains. APPROACH AND RESULTS: We performed a retrospective cohort study of 1047 patients with PBC who attained an adequate response to UDCA according to Paris-2 criteria. Time to liver-related complications, liver transplantation, or death was assessed using adjusted restricted mean survival time (RMST) analysis. The overall incidence rate of events was 17.0 (95% CI: 13.7-21.1) per 1000 out of 4763.2 patient-years. On the whole population, normal serum ALP values (but not normal gamma-glutamyl transpeptidase (GGT), alanine aminotransferase (ALT), or aspartate aminotransferase (AST); or total bilirubin < 0.6 ×ULN) were associated with a significant absolute complication-free survival gain at 10 years (mean 7.6 months, 95% CI: 2.7 - 12.6 mo.; p = 0.003). In subgroup analysis, this association was significant in patients with a liver stiffness measurement ≥ 10 kPa and/or age ≤ 62 years, with a 10-year absolute complication-free survival gain of 52.8 months (95% CI: 45.7-59.9, p < 0.001) when these 2 conditions were met. CONCLUSIONS: PBC patients with an adequate response to UDCA and persistent ALP elevation between 1.1 and 1.5 ×ULN, particularly those with advanced fibrosis and/or who are sufficiently young, remain at risk of poor outcome. Further therapeutic efforts should be considered for these patients.

Diagnosis and assessment of disease severity in patients with nonalcoholic fatty liver disease.

Nonalcoholic fatty liver disease (NAFLD) includes simple steatosis, nonalcoholic steatohepatitis (NASH), fibrosis, and eventually cirrhosis and hepatocellular carcinoma (HCC). The diagnosis of NAFLD is based on the detection of excess fat disposition in the liver, which is the first step to trigger further evaluation of NAFLD, including necroinflammation and fibrosis. In this review, we discuss non-invasive biomarkers and imaging tools that are currently and potentially available for different features (steatosis, necroinflammation and fibrosis) and disease severity assessment of NAFLD. In the past 2 decades, advances in non-invasive tests of fibrosis have transformed the management of NAFLD. Blood and imaging biomarkers have already been evaluated in multiple studies for the diagnosis of fibrosis and cirrhosis. Among the various histological features of NAFLD, the degree of fibrosis has the strongest correlation with liver-related morbidity and mortality. Non-invasive tests of fibrosis have been shown to predict liver-related outcomes, both in the general population and among patients with NAFLD. What is lacking, however, is good data to support the use of non-invasive tests as monitoring and response biomarkers. With the conclusion of several large phase 3 studies in the next few years, the availability of paired liver biopsy, non-invasive test and clinical outcome data will likely advance the field and shed light on new biomarkers and the way to use various non-invasive tests in a longitudinal manner.

Risk of liver-related events in metabolic dysfunction-associated steatohepatitis (MASH) patients with fibrosis: A comparative analysis of various risk stratification criteria.

BACKGROUND AND AIMS: International regulatory agencies recommend testing drug therapy for patients with noncirrhotic high-risk metabolic dysfunction-associated steatohepatitis (MASH) because they are at risk of liver-related events (LRE). We aimed to compare the risk of LRE in patients with MASLD stratified for F2-F4 fibrosis and MASH. APPROACH AND RESULTS: Overall, 1938 consecutive patients with biopsy-proven MASLD were enrolled. High-risk MASH was defined as MASH with F2-F4 fibrosis. LSM was measured by transient elastography. LRE were recorded during follow-up. Cox multivariate models were used to assess the association between high-risk MASH or F2-F4 fibrosis without MASH, of LSM (≥8 or ≥10 Kpa), and of AGILE 3+ with LRE. The diagnostic performance for the prediction of LRE was assessed using the area under the receiver operating characteristic curves. The observed 5-year actuarial rate of LRE was 0.4%, 0.2%, 5.1%, and 6.6% in patients with F0-F1 fibrosis without MASH, F0-F1 fibrosis with MASH, F2-F4 fibrosis without MASH, and high-risk MASH, respectively. At multivariate Cox regression analysis using F0-F1 fibrosis without MASH as a reference, both F2-F4 fibrosis without MASH [adjusted HR (aHR) 9.96] and high-risk MASH (aHR 10.14) were associated with LRE. In the 1074 patients with available LSM, LSM ≥ 10 kPa (aHR 6.31) or AGILE 3+ > 0.67 (aHR 27.45) independently predicted the development of LRE and had similarly acceptable 5-year area under the receiver operating characteristic to high-risk MASH and F2-F4 fibrosis (0.772, 0.818, 0.739, and 0.780, respectively). CONCLUSIONS: The risk of LRE is similar in patients with high-risk MASH and with F2-F4 fibrosis without MASH. The use of LSM ≥ 10 kPa or AGILE 3+ > 0.67 could be an accurate option to identify patients with MASLD worthy to be included in clinical trials.

Point-of-Care Noninvasive Prediction of Liver-Related Events in Patients With Nonalcoholic Fatty Liver Disease.

BACKGROUND & AIMS: Individual risk prediction of liver-related events (LRE) is needed for clinical assessment of nonalcoholic fatty liver disease (NAFLD)/nonalcoholic steatohepatitis (NASH) patients. We aimed to provide point-of-care validated liver stiffness measurement (LSM)-based risk prediction models for the development of LRE in patients with NAFLD, focusing on selecting patients for clinical trials at risk of clinical events. METHODS: Two large multicenter cohorts were evaluated, 2638 NAFLD patients covering all LSM values as the derivation cohort and 679 more advanced patients as the validation cohort. We used Cox regression to develop and validate risk prediction models based on LSM alone, and the ANTICIPATE and ANTICIPATE-NASH models for clinically significant portal hypertension. The main outcome of the study was the rate of LRE in the first 3 years after initial assessment. RESULTS: The 3 predictive models had similar performance in the derivation cohort with a very high discriminative value (c-statistic, 0.87-0.91). In the validation cohort, the LSM-LRE alone model had a significant inferior discrimination (c-statistic, 0.75) compared with the other 2 models, whereas the ANTICIPATE-NASH-LRE model (0.81) was significantly better than the ANTICIPATE-LRE model (0.79). In addition, the ANTICIPATE-NASH-LRE model presented very good calibration in the validation cohort (integrated calibration index, 0.016), and was better than the ANTICIPATE-LRE model. CONCLUSIONS: The ANTICIPATE-LRE models, and especially the ANTICIPATE-NASH-LRE model, could be valuable validated clinical tools to individually assess the risk of LRE at 3 years in patients with NAFLD/NASH.

Risk and predictors of hepatic decompensation in grey zone patients by the Baveno VII criteria: A competing risk analysis.

BACKGROUND: Baveno VII was proposed for non-invasive identification of clinically significant portal hypertension. However, a substantial proportion of patients is classified in the grey zone (i.e., liver stiffness 15-24.9 kPa and/or platelet count <150 × 109 /L). AIMS: To evaluate the risk and predictors of hepatic decompensation in patients in the grey zone, and to determine the prognostic role of spleen stiffness measurement. METHODS: We included prospective cohorts (from Hong Kong, Korea and France) of patients who had undergone transient elastography examination for chronic liver disease. We estimated risk of hepatic decompensation using competing risk regression with hepatocellular carcinoma and non-liver-related death as competing events. RESULTS: We identified 2763 patients with compensated advanced chronic liver disease (cACLD). There were 1243 (44.9%) and 536 (19.4%) patients in the Baveno VII grey zone and high-risk groups, respectively. The cumulative incidence of decompensation at 5 years was significantly different among low-risk (0.6% [95% CI: 0.2%-1.3%]), grey zone 4.2% (95% CI: 3.1%-5.4%) and high-risk groups (11.4% [95% CI: 8.7%-14.6%]). By competing risk analysis, aetiology of liver disease (alcohol-related liver disease), albumin-bilirubin score and alkaline phosphatase level were independently associated with decompensation among patients in the grey zone. The combination of Baveno VII and spleen stiffness significantly reduced patients classified into grey zone (12.8% in cACLD patients), while maintaining high discrimination of decompensation in low- and high-risk groups. CONCLUSIONS: Patients in grey zone of Baveno VII criteria remain at high risk of hepatic decompensation. Clinical risk factors and spleen stiffness can further stratify the risk in such patients.

Prognostic value of non-invasive scores based on liver stiffness measurement, spleen diameter and platelets in HIV-infected patients.

BACKGROUND AND AIMS: People living with HIV (PLWH) are at high risk for advanced chronic liver disease and related adverse outcomes. We aimed to validate the prognostic value of non-invasive scores based on liver stiffness measurement (LSM) and on markers of portal hypertension (PH), namely platelets and spleen diameter, in PLWH. METHODS: We combined data from eight international cohorts of PLWH with available non-invasive scores, including LSM and the composite biomarkers liver stiffness-spleen size-to-platelet ratio score (LSPS), LSM-to-Platelet ratio (LPR) and PH risk score. Incidence and predictors of all-cause mortality, any liver-related event and classical hepatic decompensation were determined by survival analysis, controlling for competing risks for the latter two. Non-invasive scores were assessed and compared using area under the receiver operating curve (AUROC). RESULTS: We included 1695 PLWH (66.8% coinfected with hepatitis C virus). During a median follow-up of 4.7 (interquartile range 2.8-7.7) years, the incidence rates of any liver-related event, all-cause mortality and hepatic decompensation were 13.7 per 1000 persons-year (PY) (95% confidence interval [CI], 11.4-16.3), 13.8 per 1000 PY (95% CI, 11.6-16.4) and 9.9 per 1000 PY (95% CI, 8.1-12.2), respectively. The AUROC of LSM was similar to that of the composite biomarkers, ranging between 0.83 and 0.86 for any liver-related event, 0.79-0.85 for all-cause mortality and 0.87-0.88 for classical hepatic decompensation. All individual non-invasive scores remained independent predictors of clinical outcomes in multivariable analysis. CONCLUSIONS: Non-invasive scores based on LSM, spleen diameter and platelets predict clinical outcomes in PLWH. Composite biomarkers do not achieve higher prognostic performance compared to LSM alone.

Validation of the new 2021 EASL algorithm for the noninvasive diagnosis of advanced fibrosis in NAFLD.

BACKGROUND AND AIMS: The European Association for the Study of the Liver (EASL) has recently proposed an algorithm for the diagnosis of advanced liver fibrosis. We aimed to evaluate the diagnostic accuracy of this algorithm in nonalcoholic fatty liver disease (NAFLD). APPROACH AND RESULTS: One thousand fifty-one patients with NAFLD, liver biopsy, and four noninvasive tests (NITs; Fibrosis-4 [FIB4], vibration controlled transient elastography [VCTE], FibroMeter, Fibrotest) were included. The enhanced liver fibrosis (ELF) score was available in 396 patients. A cohort of 230 patients from primary care/diabetes clinics had FIB4, VCTE, and ELF. Compared with the performance of single NITs, agreement between two NITs (FIB4 and VCTE, VCTE and patented serum tests) increased specificity and positive predictive value by 20%, thus justifying the sequential use proposed in the EASL algorithm. The FIB4/VCTE/FibroMeter and FIB4/VCTE/Fibrotest algorithms performed similarly, providing 85% diagnostic accuracy and a liver biopsy requirement rate of only 10%. The FIB4/VCTE/ELF algorithm performed similarly in the subgroup where ELF was available. Simulations of algorithm accuracies at different prevalence showed that positive predictive values rapidly increased, reaching a plateau above 75% starting at 15% prevalence. Negative predictive values remained higher than 90% up to 25% prevalence. The rate of liver biopsy requirement remained stable, increasing by only 5% between low and high prevalence settings. When the EASL algorithm was applied in the primary care/diabetes clinic cohort, liver biopsy requirement was only 3%, and the agreement among the three steps provided 75% positive predictive value. CONCLUSIONS: Our study validates the algorithm proposed by the EASL in its latest 2021 guidelines for the diagnosis of advanced fibrosis in the setting of NAFLD.

Impact of Type 2 Diabetes on the Accuracy of Noninvasive Tests of Liver Fibrosis With Resulting Clinical Implications.

BACKGROUND & AIMS: Noninvasive tests (NITs) of liver fibrosis have been suggested to be less accurate in type 2 diabetes mellitus (T2DM). We aimed to compare the accuracy of 6 NITs between patients with or without T2DM, explain any differences, and adapt diagnostic algorithms for clinical practice accordingly. METHODS: We included 1051 patients with nonalcoholic fatty liver disease with liver biopsy, blood fibrosis tests (Nonalcoholic Fatty Liver Disease Fibrosis Score, FIB4, Fibrotest, FibroMeter), vibration-controlled transient elastography (VCTE), and the combinatory elasto-blood test FibroMeterVCTE. The study endpoint was advanced fibrosis on liver biopsy. RESULTS: NIT areas under the receiver operating characteristic curve were significantly lower in patients with T2DM, mostly because of a decrease in specificity. For FIB4, this decrease in specificity was only related to the higher age of patients with T2DM enrolled. For Fibrotest, FibroMeter, and FibroMeterVCTE, the decrease in specificity was related to age but also to higher alpha2-macroglobulin level, which is known to increase in T2DM. Sensitivity was unaffected by T2DM, but it masked a doubled raw number of false negatives because of the 2-fold higher prevalence of advanced fibrosis in that setting. The sequential algorithm FIB4-vibration-controlled transient elastography had 90.3% accuracy in patients without T2DM vs 79.0% in those with (P < .001). Algorithms using first-line specialized tests maintained a low rate of false negatives and false positives in T2DM. CONCLUSIONS: The decrease in NIT accuracy observed in T2DM is partly biased by the different characteristics of the groups studied, but also linked to T2DM itself through modification of the levels of some NIT biomarkers. Specialized tests should be used first-line to diagnose advanced liver fibrosis in T2DM.

Liver stiffness measurement by vibration-controlled transient elastography improves outcome prediction in primary biliary cholangitis.

BACKGROUND & AIMS: Liver stiffness measurement (LSM) by vibration-controlled transient elastography (VCTE) has been shown to predict outcomes of patients with primary biliary cholangitis (PBC) in small-size studies. We aimed to validate the prognostic value of LSM in a large cohort study. METHODS: We performed an international, multicentre, retrospective follow-up study of 3,985 patients with PBC seen at 23 centres in 12 countries. Eligibility criteria included at least 1 reliable LSM by VCTE and a follow-up ≥ 1 year. Independent derivation (n = 2,740) and validation (n = 568) cohorts were built. The primary endpoint was time to poor clinical outcomes defined as liver-related complications, liver transplantation, or death. Hazard ratios (HRs) with CIs were determined using a time-dependent multivariable Cox regression analysis. RESULTS: LSM was independently associated with poor clinical outcomes in the derivation (5,324 LSMs, mean follow-up 5.0 ± 3.1 years) and validation (1,470 LSMs, mean follow-up 5.0 ± 2.8 years) cohorts: adjusted HRs (95% CI) per additional kPa were 1.040 (1.026-1.054) and 1.042 (1.029-1.056), respectively (p <0.0001 for both). Adjusted C-statistics (95% CI) at baseline were 0.83 (0.79-0.87) and 0.92 (0.89-0.95), respectively. Between 5 and 30 kPa, the log-HR increased as a monotonic function of LSM. The predictive value of LSM was stable in time. LSM improved the prognostic ability of biochemical response criteria, fibrosis scores, and prognostic scores. The 8 kPa and 15 kPa cut-offs optimally separated low-, medium-, and high-risk groups. Forty percent of patients were at medium to high risk according to LSM. CONCLUSIONS: LSM by VCTE is a major, independent, validated predictor of PBC outcome. Its value as a surrogate endpoint for clinical benefit in PBC should be considered. LAY SUMMARY: Primary biliary cholangitis (PBC) is a chronic autoimmune disease, wherein the body's immune system mistakenly attacks the bile ducts. PBC progresses gradually, so surrogate markers (markers that predict clinically relevant outcomes like the need for a transplant or death long before the event occurs) are often needed to expedite the drug development and approval process. Herein, we show that liver stiffness measurement is a strong predictor of clinical outcomes and could be a useful surrogate endpoint in PBC trials.

Long-term prognosis of patients with alcohol-related liver disease or non-alcoholic fatty liver disease according to metabolic syndrome or alcohol use.

BACKGROUND & AIMS: The boundary between non-alcoholic (NAFLD) and alcohol-related liver disease (ALD) is based on alcohol consumption. However, metabolic syndrome and alcohol use frequently co-exist. The aim of this study was to determine prognostic factors of long-term morbidity and mortality in patients with NAFLD or ALD. METHODS: From 2003 to 2016, all consecutive NAFLD or ALD patients were prospectively included in this cohort study. We evaluated overall survival, specific cause of mortality and occurrence of any complication. The primary endpoint was analysed by the Kaplan Meier method, secondary endpoints were estimated by Gray test method or logistic regressions. Factors independently associated with overall mortality and morbidity were identified by a multivariate Cox model. RESULTS: A total of 3365 patients (1667 with ALD and 1698 with NAFLD) were included. Median follow-up was 54 months (range: 30-86) and 563 subjects died. In the overall population, overall mortality was higher in patients with ALD (HR: 10.1 [7.57-13.3]), and with weekly alcohol consumption >7 units (HR:1.66 [1.41-1.96]). Liver-related mortality was higher in patients with ALD (HR: 11 [7.27-16.5]). In the NAFLD group, weekly alcohol consumption >1 unit was associated with higher overall mortality (HR: 1.9 [1.1-3.4]), and weekly alcohol consumption >7 units was associated with higher overall morbidity (OR: 1.89 [1.61-2.21]). In the ALD group, the presence of metabolic syndrome was associated with higher overall (HR:1.27 [1.02-1.57]), and liver (HR: 1.47 [1.1-1.96]) mortalities, and overall (OR: 1.46 [1.14-1.88]), liver (OR: 1.46 [1.14-1.88]) morbidities. CONCLUSION: In fatty liver diseases, light alcohol consumption and metabolic syndrome are prognosis cofactors.