Skip to main navigation Skip to main content

Clin Mol Hepatol : Clinical and Molecular Hepatology

OPEN ACCESS
Search
Advanced Search
ABOUT
BROWSE ARTICLES
FOR CONTRIBUTORS

Articles

Page Path

Review

Predictive modeling and clinical decision tools for risk stratification in steatotic liver disease

Jeanette Girard1, Elliot B. Tapper2, Vincent L. Chen2orcid
Clinical and Molecular Hepatology 2026;32(Suppl):e2026.0204.
Published online: April 24, 2026

1Division of Metabolism, Endocrinology & Diabetes, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA

2Division of Gastroenterology and Hepatology, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA

Corresponding author : Vincent L. Chen Division of Gastroenterology and Hepatology, Department of Internal Medicine, University of Michigan Medical School, 1500 East Medical Center Drive, Ann Arbor, MI 48109, USA Tel: +1-734-647-0785, Fax: +1-734-936-7392, E-mail: vichen@umich.edu

Editor: Seung Up Kim, Yonsei University, Korea

• Received: February 11, 2026   • Revised: April 17, 2026   • Accepted: April 17, 2026

Copyright © 2026 by The Korean Association for the Study of the Liver

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

  • 1,506 Views
  • 76 Download
  • 1 Crossref
  • 1 Scopus

Full Article

Download PDF Download PDF
  • Metabolic dysfunction-associated steatotic liver disease (MASLD) is a leading cause of hepatic decompensation and liver-related death, but most patients with MASLD do not develop these liver-related complications. Risk prediction and care pathways are crucial to identify which patients with MASLD are highest risk and link them to appropriate care. Risk prediction is usually done with blood-based algorithms such as Fibrosis-4, AST-platelet ratio index, and more recent scores such as steatosis-associated fibrosis estimator and LiverRisk Score. Second-line tests include enhanced liver fibrosis and imaging-based tests such as vibration-controlled transient elastography, shear wave elastography, or magnetic resonance elastography. We propose a consensus risk stratification care pathway that can be adapted for different clinical settings. We also discuss key needs to improve upon the state of the art: improving diagnosis/prognostic accuracy, especially of blood-based models, optimizing calibration, and ensuring interpretability of predictive models. Finally, we discuss recent advances in clinical decision support systems including best practice advisories, dashboards, and dynamic guidelines. We highlight factors critical to clinical decision support systems, including integration with existing systems and clinician workflows, minimizing additional burden to clinicians, provision of decision support and recommendations at the time/place of decision-making, and continuous evaluation and local user involvement.
  • Keywords: Care pathways; MASLD; Electronic medical record; Best practice advisory
Metabolic dysfunction-associated steatotic liver disease (MASLD) is common and increasingly morbid [1]. It is associated with an increased risk of cardiovascular disease [2,3], cirrhosis [4,5], liver and extrahepatic cancer [6], poor health-related quality of life [7-9] and healthcare utilization [10,11]. Although it afflicts >30% of people worldwide [12,13], its disease course is highly variable and characterized by distinct, often exclusive phenotypes of cardiovascular risk, liver-related events, and/or neither [14]. Most people with MASLD are at low risk of liver-related events, with the risk of death from non-liver causes vastly exceeding that from liver disease [5,15-17]. However, with progressively more severe liver fibrosis, the risk of liver-related mortality increases greatly [5,16,18]. Compounding the problem, most are unaware of their liver disease status or its risk implications [19,20]. These findings highlight the importance of risk prediction to identify patients at highest risk of liver-related complications, which is central to effective care and linkage to care pathways tailored to the patient’s unique risk profile. Herein, we review the state of the art for these twin aims.
Simple scores
The most important role of a clinical prediction model is to establish the risk of clinical events. We will discuss primarily prediction of major adverse liver outcomes (MALO) – typically defined as cirrhosis decompensations, liver cancer, and liver-related death [21,22]. We acknowledge that overall mortality and specific extrahepatic outcomes such as major adverse cardiovascular events and extrahepatic malignancy are crucial, but predictive scores for these outcomes already exist and a detailed discussion is beyond the scope of this review [23].
The earliest predictive models that are still in wide use today were originally developed for viral hepatitis (Table 1) [24]. Aspartate aminotransferase (AST)-platelet ratio index (APRI) was developed in 2003 to predict advanced fibrosis or cirrhosis in patients with chronic hepatitis C [25]. APRI includes only two predictors: AST level and platelet count. Fibrosis-4 (FIB4) was published shortly after in 2006 and was developed to predict advanced fibrosis/cirrhosis in patients with HIV/hepatitis C coinfection [26]. FIB4 includes four components: age, AST, alanine aminotransferase (ALT), and platelet count. Unlike APRI and FIB4, the nonalcoholic fatty liver disease fibrosis score (NFS) was specifically developed to identify advanced fibrosis or cirrhosis in biopsy-confirmed MASLD and additionally includes body mass index, albumin, and diabetes/prediabetes status [27]. The BMIAST/ALT Ratio-Diabetes (BARD) score was also developed in biopsy-confirmed MASLD to distinguish advanced fibrosis/cirrhosis from earlier fibrosis stages and includes body mass index (BMI), AST/ALT ratio, and diabetes status [28]. Each is fairly accurate for stratifying the risk of advanced fibrosis [29-33]. More importantly, they have also been associated with prognosis [34-37]. In one meta-analysis, FIB4, APRI, and NFS all had high prognostic accuracy for liver-related events (area under the curve 0.69–0.92, depending on the interval examined) while FIB4 and NFS outperformed APRI for prediction of all-cause mortality [38]. Similarly, in a meta-analysis of cohort studies, FIB4 and NFS were both associated with risk of cardiovascular events after adjustment for comorbidities while APRI was not [39]. Higher FIB4 has also been associated with increased risk of extrahepatic malignancy [40,41].
Advanced predictors
Several advances beyond the simple scores characterize the current era of risk prediction. FIB4, APRI, and NFS all have limitations including inadequate sensitivity and specificity. For example, in one study, 9% of incident cases of severe liver disease occurred in individuals with baseline FIB4<1.3 [34]. More recent scores have expanded upon these models. The steatosis-associated fibrosis estimator (SAFE) score was trained to distinguish significant fibrosis (stage 2–4) from earlier fibrosis stages in patients with biopsy-confirmed MASLD [42], which differed from the fibrosis cutoffs used in FIB4, APRI, and NFS (stage 3–4). In the original study, SAFE had higher diagnostic accuracy than the other scores and was also associated with overall mortality [42]. In independent cohorts, SAFE demonstrated higher sensitivity for liver-related events than other blood-based scores (though no difference in area under the curve vs. FIB4) [43] and has also been associated with incidence of hepatocellular carcinoma [44]. Unlike the earlier tests, LiverRisk score (LRS) was trained against liver stiffness measurement (LSM) by vibration-controlled transient elastography (VCTE) [45]. LRS includes AST, ALT, platelets, gamma-glutamyl transferase, total cholesterol, and fasting glucose. LRS outperformed FIB4 and APRI against VCTE LSM cutoffs in two independent validation cohorts and was also associated with liver-related mortality and liver cancer, and weakly with non-liver hospitalizations and extrahepatic cancer [45]. Yet another score generated a regression-based model to predict incident liver-related events defined as hepatic decompensation or hepatocellular carcinoma (nonalcoholic fatty liver disease outcomes score [NOS]) [46]. Unlike other models which were trained against fibrosis or liver stiffness and subsequently assessed performance in predicting liver-related events, NOS was primarily trained on liver-related events. NOS included age, type 2 diabetes, albumin, bilirubin, platelet count, and international normalized ratio and outperformed FIB4 and NFS in predicting risk of liver-related events [46]. Proprietary blood-based tests including enhanced liver fibrosis (ELF) have also been developed and may improve upon the non-proprietary blood-based scores detailed above and offer an intermediate between inexpensive, universally available non-invasive tests (NITs) such as FIB4 and the imaging-based predictors discussed in the next section, especially in the primary care setting where access to imaging-based biomarkers is limited [47-49]. Unfortunately, ELF cutoffs are not widely validated and although this test has value where access to VCTE is limited, additional research is needed.
Imaging based predictors
Beyond blood-based tests, imaging-based NITs including VCTE, shear wave elastography (SWE), and magnetic resonance elastography (MRE) have demonstrated efficacy in risk stratification in SLD. VCTE and MRE typically have higher prognostic accuracy than blood-based NITs but are more expensive with more limited availability [50]j; as such, they are typically positioned as second-line tests in individuals with increased FIB4 scores [51-53]. VCTE or MRE has also been used in conjunction with other predictors such as laboratory tests, cardiometabolic comorbidities, and FIB4 scores to develop combined scores such as Fibroscan-AST (FAST) [54], MRE-AST (MAST) [55,56], Agile 3+/4 [57,58], and MRE-FIB4 (MEFIB) [59,60], which may have higher test performance than imaging or laboratory parameters alone. Newer scores, including the fibrotic metabolic dysfunction-associated steatohepatitis (MASH) index, may also have potential [61].
Generally, non-proprietary blood-based predictors utilize predictors that are already readily available in clinical practice but have lower prognostic accuracy than imaging-based predictors. Imaging-based predictors may more accurately stratify risk of MALO but are less widely available, require specialized equipment and staff training, and are more expensive with the potential for longer wait times for patients. Imaging-based predictors are also susceptible to technical failures such as difficulty finding an adequate acoustic window (for VCTE or SWE) due to body habitus, poor shear wave transmission (for MRE) due to body habitus or poor driver-patient contact, or hepatic congestion resulting in false positives (for all methods) [62,63]. Proprietary blood-based biomarkers (e.g., ELF) may provide a middle ground between non-proprietary blood-based biomarkers (e.g., FIB4) and imaging biomarkers (e.g., VCTE) because they are widely available (at least in the United States, where the authors practice) and likely have superior test performance to FIB4 [47]. We will further discuss care pathways involving combinations of different modalities in the Care Pathways and Clinical Guidelines section of this manuscript.
Future directions in predictive modeling
The future of risk prediction must be responsive to several goals highlighted in Table 2.

Improving diagnostic and prognostic accuracy

Two key approaches are needed to improve the prognostic accuracy for MALO of existing models. First, though blood-based tests are the simplest to implement for the largest population, their accuracy must be improved. FIB4, which is recommended for liver fibrosis screening in at-risk populations, may miss up to 10% of incident severe liver disease [34,52,53,64,65]. An additional >40% of cases of significant fibrosis [66] may be missed by the low FIB4 cutoff (<1.3) and up to 13% of patients presenting with hepatocellular carcinoma had FIB4 <1.3 in one recent study [67]. There is also a large indeterminate FIB4 range (1.3–2.67) of >30–40% of adults in the general population for whom second-line would be indicated based on current recommendations, most of whom do not have significant fibrosis [36,47,68]. The sensitivity of LRS and SAFE for prediction of MALO is not well-defined at their lower cutoffs [43,69]. There is not an agreed-upon acceptable threshold for false positive or false negative rates in blood-based screening tests for liver disease, but we believe that missing 10% of incident MALO at a 2-year interval is unacceptably high [65].
Second, we must improve case identification. Models that rely on International Classification of Diseases-10 codes or diagnoses from treating clinicians will miss many at-risk patients [70]. To increase detection rates of significant fibrosis, it is crucial to identify more at-risk patients to bring in for screening. For example, in the emergency department, 1 in 10 abdominal radiology reports document hepatic steatosis [71,72]. Yet, studies suggest only 14–19% of patients have disclosure of these findings, and in one study, 65% of disclosures were merely the inclusion of the radiology report in the discharge paperwork [71-73]. Artificial intelligence (AI) may be of service. Several studies have been performed to improve detection of patients with hepatic steatosis utilizing various AI techniques such as natural language processing, large language models, machine learning, and deep learning algorithms. For example, one study using a natural language processing model screened for hepatic steatosis and additionally reviewed charts for those meeting diagnostic criteria for MASLD. The model was able to identify MASLD with a positive predictive value (PPV) of 93% and appropriately excluded patients with alcohol use with a negative predictive value (NPV) of ~95% [74]. Of these patients found to have MASLD, only 14.6% had a MASLD-related diagnosis code, implying that most patients had been previously undiagnosed [74]. Another algorithm based on natural language processing of imaging reports also identified hepatic steatosis with PPV of 92% [75]. Similar approaches must be applied toward detecting patients at risk for events.

Optimized calibration

Most models were developed in medical center-derived or clinical trial cohorts and validated either in other clinical cohorts or a small number of research cohorts (e.g., UK Biobank). Validation in more diverse settings, both geographically (i.e., outside of the US, Europe, and East Asia) and in practice settings (community or population-based settings, primary care, and secondary and tertiary care), will increase confidence in external validity. While power calculations for predictive models are less well-established than some other areas of statistics, it has been estimated that to achieve 80% power to detect a 0.04 difference in area under the curve (which is likely to be clinically meaningful), >200 events are required [76]. Calibration—which measures whether predicted risk matches actual risk—is also critical to avoid over- or under-estimating risk and may need to be done separately depending on the context in which the patient is being seen [77]. Calibration curves (reliability diagrams) plot predicted probability of outcomes vs. actual frequency of outcomes, where deviation from a 45° diagonal indicates poorer calibration [78], and can be adjusted during training using regularization and other methods [79]. Ideally, risk scores would provide an estimated 5 or 10-year risk of MALO, analogous to Pooled Cohort Equations used for prediction of incident atherosclerotic cardiovascular disease [23].

Achieving simpli and 86% of those patients found to have MASLD city, transparency, and interpretability

Existing widely used models are all derived from regression models using a relatively small number (2–7) of readily available parameters. There is increasing interest in using machine learning/AI to derive models using a greater array of features to predict advanced liver disease. Deep learning approaches have shown some promise and may better model the complexity, non-linearity, and interactions between different aspects of disease pathophysiology better than more traditional regression-based models and thus result in superior tests [77,80]. We offer the following cautions. First, while machine learning models may prove to have superior diagnostic or prognostic accuracy compared to simpler regression-based models, this is not guaranteed. For example, the original SAFE and LRS studies also developed and tested random forest and neural network machine learning models but did not find that they outperformed regression models in independent validation cohorts, despite relatively large training cohorts [42,45]. Second, all things being equal, parsimonious models are generally preferable to models with large numbers of predictors. These predictors may not be widely available depending on the clinical setting, and an unexpected positive or negative result may be more difficult for a clinician to interpret or troubleshoot if she must consider dozens of variables that she suspects may be unduly influencing the test result. Finally, the “black box” nature of machine learning models is a major concern. Ultimately, clinicians are responsible for interpreting test results in the appropriate context but if they cannot understand or explain why a model returned a positive or negative result, crucial information may be missing. For example, a patient with immune thrombocytopenic purpura may have a markedly elevated FIB4 solely due to thrombocytopenia; this will be apparent from inspecting the FIB4 equation but may not be obvious if using a black box model. We propose that at a minimum, machine learning models should output the relative contribution of each feature in driving the output so that providers can note major outliers.

Actionable findings

Liver risk prediction models must follow the lead of cardiovascular risk models. In atherosclerotic cardiovascular disease risk prediction models, guidelines surrounding statin therapy are directly tied to estimated risk [81]. In contrast, criteria for use of currently approved drugs in MASLD are based not directly on estimated risk of hard endpoints but on the surrogate endpoint – albeit a strong surrogate endpoint [18] – of severity of liver fibrosis [82,83]. NITs may already be more strongly predictive of MALO than histology [84,85] and we expect they will continue to improve with further development. As predictive models improve, one could envision a direct link between estimated risk—using predictive models that incorporate non-proprietary blood and/or imaging-based markers—and therapy rather than a strictly fibrosis-centric approach. Incorporating the competing risk of extrahepatic endpoints, including mortality from non-liver causes, major adverse cardiovascular events, and extrahepatic malignancy, is also crucial [16].

Capacity to implement and act on screening

Widespread uptake of MASLD fibrosis screening guidelines has not occurred yet, likely due to lack of awareness among at-risk patients and primary care providers alike [86-89]. There is also limited capacity for hepatologists, endocrinologists, and obesity specialists to manage the 3–6% of the general population across the world with severe/advanced fibrosis based on VCTE [90-93]. These challenges can be addressed by outreach to primary care providers and health systems to implement fibrosis screening, as detailed in the next section. Hiring advanced practice providers may also help with risk stratification and treating patients found to be at increased risk.
Care pathways and clinical guidelines
Pipelines to integrate automated detection into care pathways may improve detection. With the increasing prevalence of MASLD worldwide, several professional societies including the American Gastroenterology Association (AGA), American Association for the Study of Liver Diseases (AASLD), American Diabetes Association (ADA), Asian Pacific Association for the Study of the Liver (APASL), European Association for the Study of the Liver in conjunction with the European Associations for the Study of Diabetes and the Study of Obesity (EASL), and the Korean Association for the Study of the Liver (KASL) have offered recommendations for fibrosis risk stratification (Fig. 1) [52,53,64,94-96].
The target of these care pathways differs by society between the general population (considered low risk) and high-risk individuals. All societies emphasize the use of a 2-step pathway with FIB4 as the primary assessment, though alternative NITs are listed by AGA and APASL [52,53,64,94-96]. Compared to most other non-proprietary blood-based scores, FIB4 has high diagnostic and prognostic accuracy and has been well-validated across a range of practice settings. Except for APASL, the remaining societies recommend secondary assessment in those with FIB4 score >1.3 (or >2 in individuals over age 65 per AGA, KASL, and EASL) but below 2.67 with a secondary NIT, typically VCTE, an ultrasound-based NIT, though ELF and/or MRE are options offered by AASLD, AGA, and EASL. EASL also includes the option of lifestyle management with close 1 year repeat instead. If the second line of testing is consistent with low risk, the societies recommend reevaluation in 1–3 years though specific societies may suggest a shorter range for follow up [52,53,64,94-96]. Otherwise, intermediate and high-risk individuals are recommended for referral to GI in all but AGA who recommends that intermediate risk individuals can be followed in 1 year with serial non-invasive image monitoring. Biopsy in high-risk individuals is still recommended or considered by KASL, APASL, AGA, and AASLD where clinical guidance is specifically recommended to hepatology/GI care rather than primary care providers or other non-GI specialists.
To support the use of these care pathways, several studies have evaluated both the effectiveness of FIB4, ELF, and VCTE as well as the two-step approach recommended by these professional organizations. Kjaergaard et al. [47] in 2023 evaluated the effectiveness in risk stratification of the general population and at-risk individuals using VCTE as a reference standard comparing FIB4 alone, ELF alone, NFS alone, and FIB4 to ELF strategies. In at-risk populations—defined as excess alcohol intake, metabolic syndrome, type 2 diabetes, or BMI >30 kg/m2—the current guidance of FIB4 alone would require VCTE in 37% of the total population, but of those 20% had VCTE LSM ≥8 kPa. The FIB4 alone strategy also missed 36% of individuals with VCTE LSM ≥8 kPa, with even higher false negative rates with age-adjusted FIB4 cutoffs. An alternative strategy of using FIB4 followed by ELF in indeterminate cases (FIB4 1.30–2.67) reduced the rate of false positives albeit at the cost of more false negatives.
In 2025, Yip et al. [97] also evaluated the two-step referral pathway utilizing FIB4 and VCTE looking at several different pathways of non-invasive testing and determining the sensitivity, specificity, and positive and negative predictive values for these pathways in individuals with MASLD. Reviewing the AGA pathway specifically, it was found that the sensitivity and NPV of two-step testing for liver-related events at 5 years were 79% and 99.5%, respectively, for low-risk patients, and specificity and PPV were 91% and 12.3%, respectively, in high-risk patients. However, this study population consisted of those who have already undergone second-line testing with VCTE and is not necessarily representative of the at-risk population. Another randomized trial testing a best practice advisory (BPA) related to FIB4 found that the number needed to screen to identify one case of advanced chronic liver disease (LSM ≥10 kPa) was 54 in the overall population, though among those with elevated FIB4 the prevalence of advanced chronic liver disease was 6.7% [98].
These papers together show that two-step algorithms with FIB4 or APRI followed by VCTE result in a population enriched for advanced liver disease. How frequently to apply these tests should be based on our best understanding of the natural history, where it is estimated that 7–14 years are required to progress one fibrosis stage depending on the presence or absence of MASH [99] and diabetes status [100-102]. The frequency of repeat testing varies between society recommendations, but typically a 1–3 year interval is recommended, with a shorter interval for those with type 2 diabetes mellitus (T2DM) or more advanced liver disease. We propose a consensus screening algorithm in Figure 2, though we acknowledge areas of uncertainty. First, the optimal screening interval is not well established. Second, which testing modality to use in patients with abnormal first-line testing (e.g., FIB4) but normal second-line testing (e.g., VCTE) remains unclear. In our practice, we typically repeat second-line testing at least one more time, and if persistently normal return to first-line testing with a patient-specific cutoff. Finally, false negative rates for ruling out fibrosis or liver-related events are still substantial, and false positive rates are still relatively high for significant fibrosis. Implementing such algorithms into clinical practice is also challenging. Improving the status quo requires the following: improved predictive models, identifying more patients who may benefit from risk assessment, and implementation of clinical decision tools.
The role of automation and AI
There is a clear role for AI in case identification and linkage to care [103,104]. Applying AI-based algorithms, Fox et al. [105] utilized natural language processing techniques to evaluate ultrasound reports for patients over the age of 18 with a diagnosis of active diabetes mellitus in a primary care clinic to identify studies over a 20-year period containing positive identification of fatty liver or steatosis. With additional chart review and calculation of FIB4 and NFS scores, the team identified those at highest risk for advanced fibrosis and notified all PCPs with targeted messaging suggesting hepatology referral, leading to 73% of primary care physicians (PCPs) placing referrals, 76% of patients completing hepatology visits, 86% of those patients were found to have MASLD, of whom 36% had F3 to F4 disease [105]. The study required significant effort to directly contact PCPs, but automated approaches may be more readily scalable.
Despite several professional society clinical guidelines and care pathways, it is evident from automated detection studies that care pathways and guidelines alone are not enough [74,105,106]. While automated detection is useful in finding the appropriate patients, translating that into clinical management requires additional steps. This is where clinical decision support (CDS) systems can be valuable.
The role of CDS systems
CDS systems are systems that provide clinicians, staff, patients, or other individuals with knowledge and person specific information that is filtered and presented at appropriate times to enhance health and healthcare. CDS systems can be broken down into 6 subtypes including alerts/reminders, order sets, dashboards, tailored forms and flowsheets, dynamic guidelines, and infobuttons and reference guides.
A common method for CDS tool implementation is the use of alerts or reminders, often called BPA. These can be active alerts that pop up and limit clinician engagement with other areas of the electronic health record (EHR) until managed or passive alerts that sit on the dashboard for clinicians to review. Among the simplest of examples include quality improvement projects piloted with internal medicine programs utilizing a “Smart Phrase,” in which a FIB4 score is calculated as part of a clinical note template allowing for easy integration into the clinician workflow [107,108].
Mohanty et al. [109] developed a BPA alert for a weight management clinic that utilized an automatic calculation for FIB4 score from existing clinical data and provided recommendations on follow-up per clinical guidelines. For scores of 1.3–2.67, the BPA recommended referral for VCTE while a FIB4 score of >2.67 or liver stiffness >8 kPa prompted a BPA recommendation for referral to hepatology. This particular case provides great context into the successes and limitations of BPAs. While there was an increase in patients with automated FIB4 scores that were calculable (22.4% to 54.2%), there was virtually no change in clinical practice amongst providers with average VCTE referrals over a 6 month period changing from 0 to 3 and hepatology referrals remaining the same at 2 pre- and post-activation respectively [109]. Mohanty et al. [109] notes that there were several challenges with their study citing clinician skepticism over the FIB4 score, clinician view of responsibility, and workflow disruption. This highlights the importance of meeting the 5 rights of CDS: “The right information, to the right person, in the right format, through the right channel, at the right time in the workflow” [110].
Another recent example of a dynamic workflow was reported by Spann et al.111 in 2023, which initially tested the use of EHR systems to develop a dynamic workflow to identify care gaps in those with NAFLD/MASLD [111]. This used a CDS tool to identify patients with diagnosed MASLD who lacked liver or platelet studies to calculate a FIB4 score, and found that >50% of patients did not have a calculable FIB4 score within the past year. A subsequent follow up study in 2025 evaluated the use of this with a linked BPA focusing on providing support to PCPs in the management for patients with existing MASLD or NAFLD diagnosis codes [112]. The active alert included information on an automated FIB4 calculation directing providers to open and use a corresponding order set with patient-specific recommendations. The outcome evaluated was the linkage to liver care at the conclusion of the visit. Ultimately, the study found an increase in provider adherence to guidelines from 29.7% to 45.8% post-implementation [112]. Some key pieces of this program that was developed included collaboration with PCPs in development as well as end-user feedback. The program was highly targeted with very specific, actionable steps and provided in-line decision support for management.
Zhang et al. [98] reported on a partially blinded randomized controlled trial to investigate the use of electronic reminders in improving linkage to appropriate care for patients with elevated FIB4 scores in several clinics across Hong Kong and Malaysia. Adults with type 2 diabetes were randomized to a control arm vs. an interventional arm in which their medical providers received a BPA if the patient had elevated FIB4 (≥1.3 or ≥2.0 if over 65 years) or APRI (>0.5). The alert was a simple active alert stating the elevated FIB4 score and recommending referral to hepatology or VCTE. Ultimately, the simple alert led to 33% of those in the intervention group with high fibrosis scores with linkage to care within 1 year, while only 3.1% occurred in the control group [98]. It was importantly noted that despite the intervention, 2/3 of patients still did not receive care in the intervention group. A limitation of this CDS tool is the lack of feedback built into the program to determine the reasons for dismissing without acting on the information.
Continuous feedback between CDS tool developers and users is crucial to maximize the impact of CDS. In the absence of such feedback, CDS use in a clinical context is likely to be limited, with one systematic review noting only a 5.8% increase in the proportion of patients receiving desired care [113]. Evaluation of CDS systems can take two forms: Summative and Formative. Summative evaluations assess the effects or outcomes of CDS systems, most commonly the process or outcome measure that led to CDS tool development or, less commonly cost-benefit analysis [114]. Formative evaluations are equally important and should be a part of the pre-implementation and early implementation process [114]. These include brainstorming, focus groups, structured interviews, concept mapping, and Delphi polls. These help with the development of the CDS tool and are crucial in success [114]. Ge et al. [115] reported on the use of CDS systems in hepatology and emphasized the importance of human-centered design as part of planning. When developing a CDS tool, there are several checklists to assist with success. Use of the GUIDES checklist in development helps ensure that all steps of the process are reviewed. This checklist includes 4 domains of evaluation, including context, content, system, and implementation, with a total of 16 factors associated with the success of CDS [110]. Other studies have been conducted, and Table 3 lists at least 9 features related to these 4 domains that are associated with successful CDS implementation.
As described above, FIB4 may be a good target for CDS systems because it is simple, calculable, requires little to no additional testing beyond basic evaluation, and is already part of national society recommendations. Incorporating other scores such as SAFE and LRS into CDS could be another future direction. Regardless, the development of a successful CDS tool for SLD screening requires significant effort to ensure clinician engagement. Additionally, adapting to changes in clinical guidance requires that CDS systems be reviewed and refined regularly to ensure the use is appropriate as research evidence progresses. Another important consideration is what defines the success of a CDS tool in screening for liver fibrosis. In previous studies on care pathways, more than half of alerts did not result in medical action [98,112]. Currently, there are no benchmarks for success in the hepatology space [96]. Adherence to screening for other extrahepatic complications of T2DM may provide guidance on this topic, particularly disease of the kidney, which is another transplantable solid organ. One national US study of over half a million primary care patients found the adherence to screening for chronic kidney disease by both annual urine creatinine/albumin ratio and eGFR varied from a median of 52% to a 90th percentile of 73% [116]. Based on this, we propose that improving MASLD fibrosis screening rates to around 50% is a reasonable initial target for liver fibrosis screening in patients with T2DM, with approximately 75% being an ideal target. Modestly lower adherence levels could potentially be acceptable for those without T2DM who are on average at lower risk of liver fibrosis.
In order to meet the needs of the growing population with MASLD, new strategies must be developed and existing strategies must be optimized with deliberate care for effective implementation. The path forward involves implementing care pathways, which are robust and capable of updating their inputs as novel prediction models are developed. It is also crucial to move beyond the hepatology clinic and integrate with other clinical settings enriched for advanced liver disease (endocrinology, obesity medicine, and bariatric surgery) and primary care. Finally, expanding the clinical capacity to see patients at increased risk for advanced liver disease—likely with advanced practice providers—is essential.

Authors’ contribution

JG: drafting and critical review of manuscript. EBT: drafting and critical review of manuscript. VLC: drafting and critical review of manuscript.

Conflicts of Interest

VLC received grants from KOWA, AstraZeneca, and Ipsen, paid to University of Michigan. JG: No conflicts of interest. EBT has received grants from Madrigal and consults for Ipsen, Madrigal, Satellite Bio, Tortugas, Kineticos.

Figure 1.
Liver risk stratification algorithms recommended by major professional societies. Society recommendation statements are from the Korean Association for the Study of the Liver (KASL) [94], Asian Pacific Association for the Study of the Liver (APASL) [95], American Association for the Study of Liver Diseases (AASLD) [53], European Association for the Study of the Liver (EASL) [52], American Gastroenterological Association (AGA) [64], and American Diabetes Association (ADA) [96]. Note that APASL recommends that patients with type 2 diabetes, overweight/obesity, or at least two metabolic syndrome components should be screened for MASLD and that those with MASLD should undergo FIB4 screening; FIB4 screening in those not known to have MASLD is not explicitly recommended. Other societies do not explicitly require hepatic steatosis to undergo screening. Unless otherwise noted, recommendations are common to all societies; discrepancies between recommendations are shown in parentheses. ELF, enhanced liver fibrosis score; FIB4, Fibrosis-4 score; LSM, liver stiffness measurement; MASLD, metabolic dysfunction-associated steatotic liver disease; MRE, magnetic resonance elastography; NFS, nonalcoholic fatty liver disease fibrosis score; NIT, non-invasive test; SWE, shear wave elastography; VCTE, vibration-controlled transient elastography. “Consensus” is defined as three or more of the non-APASL recommendation statements recommending screening. *AASLD and ADA: every 1–2 years if type 2 diabetes/prediabetes or at least two metabolic syndrome components and every 2–3 years otherwise; AGA: every 2 years; APASL: every 2–3 years; EASL and KASL: every 1–3 years. **Only AASLD provides specific ELF cutoffs, but we show cutoffs here for clarity.
cmh-2026-0204f1.jpg
Figure 2.
Proposed consensus risk stratification algorithm for liver fibrosis screening. Screening is recommended in individuals with cardiometabolic risk factors, most notably type 2 diabetes or prediabetes, metabolic syndrome, or obesity complicated by at least one cardiometabolic risk factor. Fibrosis assessment is also recommended in those with known hepatic steatosis or elevated liver enzymes. If not already done, universal screening for other key risk factors for liver disease such as hepatitis B and C, alcohol use disorder, diabetes, hypertension, and dyslipidemia should be performed. In at-risk individuals, we recommend Fibrosis-4 (FIB4) for initial risk stratification, though aspartate aminotransferase-platelet ratio index (APRI) and steatosis-associated fibrosis estimator (SAFE) are reasonable alternatives depending on clinical context. In patients with low FIB4 (<1.3 if <65 years old or <2 if 65+ years old), no additional risk stratification is routinely required and the NIT should be repeated. In those with intermediate FIB4 (1.3–2.67 if <65 or 2–2.67 if ≥65 years), we suggest initial risk stratification with the enhanced liver fibrosis (ELF) test. However, it may be reasonable depending on availability, local expertise, and patient characteristics, to proceed directly to imaging-based noninvasive tests preferably vibration-controlled transient elastography (VCTE), though shear wave elastography (SWE) and even magnetic resonance elastography (MRE) are alternatives. In individuals with low ELF, the noninvasive test (NIT) should be repeated, while those with elevated ELF should be referred to subspecialty care. Those with high FIB4 (>2.67) should proceed to imaging-based NITs for risk stratification though direct referral to subspecialty care can be considered. In those with normal liver stiffness measurement (LSM) <8 kPa, the non-invasive test should be repeated; if LSM is elevated, referral to subspecialty care should be pursued. We suggest that non-invasive tests be repeated annually if FIB4 >2.67, every 1–2 years if there is diabetes, metabolic syndrome, or markedly elevated liver enzymes, and every 2–3 years for other patients. NIT, non-invasive test.
cmh-2026-0204f2.jpg
Table 1.
Risk prediction models
Table 1.
Model Components Outcomes predicted
Major adverse liver outcomes Major adverse cardiovascular events Overall mortality
Blood-based predictors
 AST-platelet ratio index (APRI) AST, platelets ↑ [37] ↑/= [39] ↑/= [38]
 Fibrosis-4 (FIB4) AST, platelets, ALT, age ↑ [34,37] ↑ [39] ↑ [38]
 Nonalcoholic fatty liver disease fibrosis score (NFS) AST, platelets, ALT, age, albumin, body mass index, diabetes/pre-diabetes ↑ [37] ↑ [39] ↑ [38]
 BMI-AST/ALT Ratio-Diabetes (BARD) AST, ALT, body mass index, diabetes ↑ [117] ↑ [118] ↑ [118]
 Steatosis-associated fibrosis estimator (SAFE) AST, platelets, ALT, age, globulin, body mass index, diabetes ↑ [65] ↑ [42,119]
 LiverRisk score (LRS) AST, ALT, platelets, gamma glutamyl transferase, cholesterol, fasting glucose ↑ [45] ↑ [45]
 Nonalcoholic fatty liver disease outcomes score (NOS) Age, diabetes, albumin, bilirubin, platelet count, international normalized ratio ↑ [46]
Imaging-based predictors
 Vibration-controlled transient elastography (VCTE) Liver stiffness measurement (LSM), controlled attenuation parameter (CAP) ↑ [84] ↑ [120] ↑ [121]
 Shear wave elastography (SWE) LSM ↑ [122] ↑ [122]
 Magnetic resonance elastography (MRE) LSM, liver fat content (LFC) ↑ [123] ↑ [124]
Imaging- and blood-based
 FAST VCTE LSM, CAP, AST ↑ [125]
 MAST MRE LSM, LFC, AST ↑ [126]
 Agile 3+ and Agile 4 VCTE LSM, multiple laboratory values, comorbidities ↑ [84,125]
 MEFIB MRE LSM, FIB4 ↑ [126]

ALT, alanine aminotransferase; AST, aspartate aminotransferase; FAST, Fibroscan-AST; MAST, MRE-AST; MEFIB, MRE-FIB4.

↑ implies that higher values of a score are associated with increased risk of the outcome.

Table 2.
Current state of the art and eventual goals in risk prediction algorithm development
Table 2.
Current state Goal
Diagnostic accuracy High false negative and false positive rates Low false negative and false positive rates
Calibration in diverse settings Most models have been studied in tertiary care and a small number of community cohorts (e.g., UK Biobank) Validation in community and primary care settings, and in South America and Africa
Quantitative estimates “Low” vs. “intermediate” vs. “high” risk Estimated risk of an event occurring within 5–10 years
Interpretability, simplicity, and transparency Acceptable for most widely-used models, but at risk with increased adoption of machine learning Maintain in the era of machine learning models
Actionable findings Fibrosis (measured directly or indirectly) as the key determinant of pharmacotherapy Predicted risk as the primary determinant of pharmacotherapy
Capacity to implement and act on findings Limited screening (likely due to low disease awareness) and limited specialist capacity to act on findings Increased outreach to primary care providers and health systems, and management of high-risk patients by advanced practice providers
Table 3.
Keys to successful clinical decision support (CDS) tool implementation
Table 3.
Keys to successful CDS tool implementation
Integration with charting or order entry system
Automatic provision of decision support as part of clinician workflow
No need for additional clinician data entry
Provision of decision support at the time and location of decision making
Provision of a recommendation, not just an assessment
Justification of decision support via provision of research evidence
Promotion of action rather than inaction
Local user involvement in the development process
Provision of decision support results to patients as well as providers

AASLD

American Association for the Study of Liver Diseases

ADA

American Diabetes Association

AGA

American Gastroenterology Association

AI

artificial intelligence

ALT

alanine aminotransferase

APASL

Asian Pacific Association for the Study of the Liver

APRI

AST-platelet ratio index

AST

aspartate aminotransferase

BMI

body mass index

BPA

best practice advisory

CDS

clinical decision support

EASL

European Association for the Study of the Liver

EHR

electronic health record

ELF

enhanced liver fibrosis

FIB4

Fibrosis-4

KASL

Korean Association for the Study of the Liver

LRS

LiverRisk score

LSM

liver stiffness measurement

MALO

major adverse liver outcomes

MASH

metabolic dysfunction-associated steatohepatitis

MASLD

metabolic dysfunction-associated steatotic liver disease

MRE

magnetic resonance elastography

NFS

nonalcoholic fatty liver disease fibrosis score

NIT

non-invasive tests

NOS

nonalcoholic fatty liver disease outcomes score

PCP

primary care physician

SAFE

steatosis-associated fibrosis estimator

SWE

shear wave elastography

T2DM

type 2 diabetes mellitus

VCTE

vibration-controlled transient elastography
  • 1. Li D, Zhang X. From mechanisms to management: early detection and improved treatment of MASLD and its related hepatocellular carcinoma. Hepatol Commun 2025;9:e0846.
  • 2. Mantovani A, Csermely A, Petracca G, Beatrice G, Corey KE, Simon TG, et al. Non-alcoholic fatty liver disease and risk of fatal and non-fatal cardiovascular events: an updated systematic review and meta-analysis. Lancet Gastroenterol Hepatol 2021;6:903-913.
  • 3. Mantovani A, Petracca G, Csermely A, Beatrice G, Bonapace S, Rossi A, et al. Non-alcoholic fatty liver disease and risk of new-onset heart failure: an updated meta-analysis of about 11 million individuals. Gut 2023;72:372-380.
  • 4. Owrangi S, Paik JM, Golabi P, de Avila L, Hashida R, Nader A, et al. Meta-analysis: global prevalence and mortality of cirrhosis in metabolic dysfunction-associated steatotic liver disease. Aliment Pharmacol Ther 2025;61:433-443.
  • 5. Simon TG, Roelstraete B, Khalili H, Hagström H, Ludvigsson JF. Mortality in biopsy-confirmed nonalcoholic fatty liver disease: results from a nationwide cohort. Gut 2021;70:1375-1382.
  • 6. Mantovani A, Petracca G, Beatrice G, Csermely A, Tilg H, Byrne CD, et al. Non-alcoholic fatty liver disease and increased risk of incident extrahepatic cancers: a meta-analysis of observational cohort studies. Gut 2022;71:778-788.
  • 7. Czapla BC, Dalvi A, Hu J, Moran IJ, Wijarnpreecha K, Chen VL. Physical activity, diet, and social determinants of health associate with health related quality of life and fibrosis in MASLD. Sci Rep 2025;15:7976.
  • 8. Younossi ZM, Stepanova M, Lawitz EJ, Reddy KR, Wai-Sun Wong V, Mangia A, et al. Patients with nonalcoholic steatohepatitis experience severe impairment of health-related quality of life. Am J Gastroenterol 2019;114:1636-1641.
  • 9. McSweeney L, Breckons M, Fattakhova G, Oluboyede Y, Vale L, Ternent L, et al. Health-related quality of life and patient-reported outcome measures in NASH-related cirrhosis. JHEP Rep 2020;2:100099.
  • 10. Allen AM, Van Houten HK, Sangaralingham LR, Talwalkar JA, McCoy RG. Healthcare cost and utilization in nonalcoholic fatty liver disease: real-world data from a large U.S. claims database. Hepatology 2018;68:2230-2238.
  • 11. Gedallovich SM, Stephen J, Kang R, Ackermann RT, Ladner DP, VanWagner LB. Geographic variation in NAFLD prevalence and subspecialty care utilization among insured adults in the United States. Clin Gastroenterol Hepatol 2023;21:3440-3443.e2.
  • 12. Le MH, Yeo YH, Li X, Li J, Zou B, Wu Y, et al. 2019 global NAFLD prevalence: a systematic review and meta-analysis. Clin Gastroenterol Hepatol 2022;20:2809-2817.e28.
  • 13. Paik JM, Hobbs K, Gupta A, Alkalbani RJ, Reyes MA, Younossi ZM. Prevalence of MASLD, Met-ALD, and ALD and associated fibrosis among US adults: insights from NHANES 2017 to 2023. J Clin Gastroenterol 2025 May 28. doi: 10.1097/MCG.0000000000002202.
  • 14. Stefan N, Yki-Järvinen H, Neuschwander-Tetri BA. Metabolic dysfunction-associated steatotic liver disease: heterogeneous pathomechanisms and effectiveness of metabolism-based treatment. Lancet Diabetes Endocrinol 2025;13:134-148.
  • 15. Männistö VT, Salomaa V, Färkkilä M, Jula A, Männistö S, Erlund I, et al. Incidence of liver-related morbidity and mortality in a population cohort of non-alcoholic fatty liver disease. Liver Int 2021;41:2590-2600.
  • 16. Ochoa-Allemant P, Hubbard RA, Kaplan DE, Serper M. Cause-specific mortality in patients with steatotic liver disease in the United States. J Hepatol 2025;83:860-869.
  • 17. Stine JG, Bradley D, McCall-Hosenfeld J, Motz-Patel V, Tondt J, Batra S, et al. Multidisciplinary clinic model enhances liver and metabolic health outcomes in adults with MASH. Hepatol Commun 2025;9:e0649.
  • 18. Ng CH, Lim WH, Hui Lim GE, Hao Tan DJ, Syn N, Muthiah MD, et al. Mortality outcomes by fibrosis stage in nonalcoholic fatty liver disease: a systematic review and meta-analysis. Clin Gastroenterol Hepatol 2023;21:931-939.e5.
  • 19. Alp J, Sripongpun P, Roldan GA, Fletcher JA, Davie T, Udompap P. Low awareness of MASLD among U.S. adults: trends from NHANES 2017-2023. Hepatol Commun 2025;9:e0829.
  • 20. Lazarus JV, White TM, Allen AM, Pannain S, Alkhouri N, Bansal MB, et al. Awareness of metabolic dysfunction-associated steatotic liver disease (MASLD) in 4 major cities in the United States. Hepatol Commun 2025;9:e0704.
  • 21. Diaz LA, Thiele M, Louvet A, Lee BP, Ajmera V, Tavaglione F, et al. Clinical trial design, biomarkers and end points in metabolic and alcohol-related liver disease. Nat Rev Gastroenterol Hepatol 2025;22:866-884.
  • 22. Brown RS, Brown KA, Flamm S, Bejarano RE, Rahimi RS, Singal AK, et al. Screening and management of portal hypertension and varices in cirrhosis: expert perspectives. Hepatol Commun 2025;9:e0682.
  • 23. Muntner P, Colantonio LD, Cushman M, Goff DC, Howard G, Howard VJ, et al. Validation of the atherosclerotic cardiovascular disease Pooled Cohort risk equations. JAMA 2014;311:1406-1415.
  • 24. Grady J, Song M, Townsend W, Mahmud N, Tapper EB, Parikh ND. A systematic review of noninvasive laboratory indices and elastography to predict hepatic decompensation. Hepatol Commun 2025;9:e0675.
  • 25. Wai CT, Greenson JK, Fontana RJ, Kalbfleisch JD, Marrero JA, Conjeevaram HS, et al. A simple noninvasive index can predict both significant fibrosis and cirrhosis in patients with chronic hepatitis C. Hepatology 2003;38:518-526.
  • 26. Sterling RK, Lissen E, Clumeck N, Sola R, Correa MC, Montaner J, et al. Development of a simple noninvasive index to predict significant fibrosis in patients with HIV/HCV coinfection. Hepatology 2006;43:1317-1325.
  • 27. Angulo P, Hui JM, Marchesini G, Bugianesi E, George J, Farrell GC, et al. The NAFLD fibrosis score: a noninvasive system that identifies liver fibrosis in patients with NAFLD. Hepatology 2007;45:846-854.
  • 28. Harrison SA, Oliver D, Arnold HL, Gogia S, Neuschwander-Tetri BA. Development and validation of a simple NAFLD clinical scoring system for identifying patients without advanced disease. Gut 2008;57:1441-1447.
  • 29. Han JW, Kim HY, Yu JH, Kim MN, Chon YE, An JH, et al. Diagnostic accuracy of the Fibrosis-4 index for advanced liver fibrosis in nonalcoholic fatty liver disease with type 2 diabetes: a systematic review and meta-analysis. Clin Mol Hepatol 2024;30:S147-S158.
  • 30. Drolz A, Wolter S, Wehmeyer MH, Piecha F, Horvatits T, Schulze Zur Wiesch J, et al. Performance of non-invasive fibrosis scores in non-alcoholic fatty liver disease with and without morbid obesity. Int J Obes (Lond) 2021;45:2197-2204.
  • 31. Mózes FE, Lee JA, Selvaraj EA, Jayaswal ANA, Trauner M, Boursier J, et al. Diagnostic accuracy of non-invasive tests for advanced fibrosis in patients with NAFLD: an individual patient data meta-analysis. Gut 2022;71:1006-1019.
  • 32. Ooi GJ, Mgaieth S, Eslick GD, Burton PR, Kemp WW, Roberts SK, et al. Systematic review and meta-analysis: noninvasive detection of non-alcoholic fatty liver disease related fibrosis in the obese. Obes Rev 2018;19:281-294.
  • 33. Sun W, Cui H, Li N, Wei Y, Lai S, Yang Y, et al. Comparison of FIB-4 index, NAFLD fibrosis score and BARD score for prediction of advanced fibrosis in adult patients with nonalcoholic fatty liver disease: a meta-analysis study. Hepatol Res 2016;46:862-870.
  • 34. Hagström H, Talbäck M, Andreasson A, Walldius G, Hammar N. Repeated FIB-4 measurements can help identify individuals at risk of severe liver disease. J Hepatol 2020;73:1023-1029.
  • 35. Hagström H, Nasr P, Ekstedt M, Stål P, Hultcrantz R, Kechagias S. Accuracy of noninvasive scoring systems in assessing risk of death and liver-related endpoints in patients with nonalcoholic fatty liver disease. Clin Gastroenterol Hepatol 2019;17:1148-1156.e4.
  • 36. Chen VL, Oliveri A, Miller MJ, Wijarnpreecha K, Du X, Chen Y, et al. PNPLA3 genotype and diabetes identify patients with nonalcoholic fatty liver disease at high risk of incident cirrhosis. Gastroenterology 2023;164:966-977.e17.
  • 37. Unalp-Arida A, Ruhl CE. Liver fibrosis scores predict liver disease mortality in the United States population. Hepatology 2017;66:84-95.
  • 38. Lee J, Vali Y, Boursier J, Spijker R, Anstee QM, Bossuyt PM, et al. Prognostic accuracy of FIB-4, NAFLD fibrosis score and APRI for NAFLD-related events: a systematic review. Liver Int 2021;41:261-270.
  • 39. Yi M, Peng W, Teng F, Kong Q, Chen Z. The role of noninvasive scoring systems for predicting cardiovascular disease risk in patients with nonalcoholic fatty liver disease: a systematic review and meta-analysis. Eur J Gastroenterol Hepatol 2022;34:1277-1284.
  • 40. Zelber-Sagi S, Schonmann Y, Weinstein G, Yeshua H. Liver fibrosis marker FIB-4 is associated with hepatic and extrahepatic malignancy risk in a population-based cohort study. Liver Int 2025;45:e70139.
  • 41. Haimi M, Baker FA, Vinker S, Safadi R, Israel A. Association between high FIB-4 score and the risk of malignancy development and mortality: a retrospective longitudinal case-control study. Dig Dis Sci 2025;70:3525-3537.
  • 42. Sripongpun P, Kim WR, Mannalithara A, Charu V, Vidovszky A, Asch S, et al. The steatosis-associated fibrosis estimator (SAFE) score: a tool to detect low-risk NAFLD in primary care. Hepatology 2023;77:256-267.
  • 43. Li G, Lin H, Sripongpun P, Liang LY, Zhang X, Wong VWS, et al. Diagnostic and prognostic performance of the SAFE score in non-alcoholic fatty liver disease. Liver Int 2024;44:15-26.
  • 44. Su TH, Yang SS, Lee MH, Kao WY, Huang SC, Chen FF, et al. High Steatosis-Associated Fibrosis Estimator scores predict hepatocellular carcinoma in viral and non-viral hepatitis and metabolic dysfunction-associated steatotic liver disease. Clin Mol Hepatol 2025;31:796-809.
  • 45. Serra-Burriel M, Juanola A, Serra-Burriel F, Thiele M, Graupera I, Pose E, et al. Development, validation, and prognostic evaluation of a risk score for long-term liver-related outcomes in the general population: a multicohort study. Lancet 2023;402:988-996.
  • 46. Calzadilla-Bertot L, Jeffrey GP, Wang Z, Huang Y, Garas G, Wallace M, et al. Predicting liver-related events in NAFLD: a predictive model. Hepatology 2023;78:1240-1251.
  • 47. Kjaergaard M, Lindvig KP, Thorhauge KH, Andersen P, Hansen JK, Kastrup N, et al. Using the ELF test, FIB-4 and NAFLD fibrosis score to screen the population for liver disease. J Hepatol 2023;79:277-286.
  • 48. Allen AM, Lazarus JV, Alkhouri N, Noureddin M, Wong VW, Tsochatzis EA, et al. Global patterns of utilization of noninvasive tests for the clinical management of metabolic dysfunction-associated steatotic liver disease. Hepatol Commun 2025;9:e0678.
  • 49. Ishiba H, Fujii H, Kamada Y, Sumida Y, Takahashi H, Seko Y, et al. Accuracy of type IV collagen 7S versus Enhanced Liver Fibrosis score for diagnosing fibrosis in patients with metabolic dysfunction-associated steatotic liver disease. Hepatol Commun 2025;9:e0563.
  • 50. Sterling RK, Duarte-Rojo A, Patel K, Asrani SK, Alsawas M, Dranoff JA, et al. AASLD Practice Guideline on imaging-based noninvasive liver disease assessment of hepatic fibrosis and steatosis. Hepatology 2025;81:672-724.
  • 51. Kanwal F, Bril F, Wong VW, Adams LA, Pfotenhauer K, Skolnik N, et al. Clinical care pathway for the risk stratification and management of patients with metabolic dysfunction-associated steatotic liver disease. Gastroenterology 2026;171:164-183.
  • 52. European Association for the Study of the Liver (EASL); European Association for the Study of Diabetes (EASD); European Association for the Study of Obesity (EASO). EASLEASD-EASO Clinical Practice Guidelines on the management of metabolic dysfunction-associated steatotic liver disease (MASLD). J Hepatol 2024;81:492-542.
  • 53. Rinella ME, Neuschwander-Tetri BA, Siddiqui MS, Abdelmalek MF, Caldwell S, Barb D, et al. AASLD Practice Guidance on the clinical assessment and management of nonalcoholic fatty liver disease. Hepatology 2023;77:1797-1835.
  • 54. Newsome PN, Sasso M, Deeks JJ, Paredes A, Boursier J, Chan WK, et al. FibroScan-AST (FAST) score for the noninvasive identification of patients with non-alcoholic steatohepatitis with significant activity and fibrosis: a prospective derivation and global validation study. Lancet Gastroenterol Hepatol 2020;5:362-373.
  • 55. Noureddin M, Truong E, Gornbein JA, Saouaf R, Guindi M, Todo T, et al. MRI-based (MAST) score accurately identifies patients with NASH and significant fibrosis. J Hepatol 2022;76:781-787.
  • 56. Truong E, Gornbein JA, Yang JD, Noureddin N, Harrison SA, Alkhouri N, et al. MRI-AST (MAST) score accurately predicts major adverse liver outcome, hepatocellular carcinoma, liver transplant, and liver-related death. Clin Gastroenterol Hepatol 2023;21:2570-2577.e1.
  • 57. Pennisi G, Enea M, Pandolfo A, Celsa C, Antonucci M, Ciccioli C, et al. AGILE 3+ score for the diagnosis of advanced fibrosis and for predicting liver-related events in NAFLD. Clin Gastroenterol Hepatol 2023;21:1293-1302.e5.
  • 58. Sanyal AJ, Foucquier J, Younossi ZM, Harrison SA, Newsome PN, Chan WK, et al. Enhanced diagnosis of advanced fibrosis and cirrhosis in individuals with NAFLD using FibroScan-based Agile scores. J Hepatol 2023;78:247-259.
  • 59. Jung J, Loomba RR, Imajo K, Madamba E, Gandhi S, Bettencourt R, et al. MRE combined with FIB-4 (MEFIB) index in detection of candidates for pharmacological treatment of NASH-related fibrosis. Gut 2021;70:1946-1953.
  • 60. Kim BK, Tamaki N, Imajo K, Yoneda M, Sutter N, Jung J, et al. Head-to-head comparison between MEFIB, MAST, and FAST for detecting stage 2 fibrosis or higher among patients with NAFLD. J Hepatol 2022;77:1482-1490.
  • 61. Feng G, Boursier J, de Saint Loup M, Yilmaz Y, Takahashi H, Isoda H, et al. Fibrotic MASH resolutionVCTE index to diagnose fibrotic MASH resolution and predict LREs: a global multicenter study. Med 2026;7:100929.
  • 62. Wagner M, Corcuera-Solano I, Lo G, Esses S, Liao J, Besa C, et al. Technical failure of MR elastography examinations of the liver: experience from a large single-center study. Radiology 2017;284:401-412.
  • 63. Zhang YN, Fowler KJ, Ozturk A, Potu CK, Louie AL, Montes V, et al. Liver fibrosis imaging: a clinical review of ultrasound and magnetic resonance elastography. J Magn Reson Imaging 2020;51:25-42.
  • 64. Kanwal F, Shubrook JH, Adams LA, Pfotenhauer K, Wai-Sun Wong V, Wright E, et al. Clinical care pathway for the risk stratification and management of patients with nonalcoholic fatty liver disease. Gastroenterology 2021;161:1657-1669.
  • 65. Chen VL, Raut C, Oliveri A. 2520: Steatosis-associated fibrosis estimator (SAFE) outperforms other blood-based biomarkers for predicting major adverse liver outcomes in the community. Hepatology 2025;82:S1171-S1172.
  • 66. Graupera I, Thiele M, Serra-Burriel M, Caballeria L, Roulot D, Wong GL, et al. Low accuracy of FIB-4 and NAFLD fibrosis scores for screening for liver fibrosis in the population. Clin Gastroenterol Hepatol 2022;20:2567-2576.e6.
  • 67. Tan DJH, Tamaki N, Kim BK, Wijarnpreecha K, Aboona MB, Faulkner C, et al. Prevalence of low FIB-4 in MASLD-related hepatocellular carcinoma: a multicentre study. Aliment Pharmacol Ther 2025;61:278-285.
  • 68. Golabi P, Paik JM, Arshad T, Younossi Y, Mishra A, Younossi ZM. Mortality of NAFLD according to the body composition and presence of metabolic abnormalities. Hepatol Commun 2020;4:1136-1148.
  • 69. Chen R, Petrazzini BO, Nadkarni G, Rocheleau G, Bansal MB, Do R. Comparison of blood-based liver fibrosis scores in the Mount Sinai Health System, MASLD Registry, and NHANES 2017-2020 study. Hepatol Commun 2024;8:e0515.
  • 70. Hagström H, Adams LA, Allen AM, Byrne CD, Chang Y, Duseja A, et al. The future of International Classification of Diseases coding in steatotic liver disease: an expert panel Delphi consensus statement. Hepatol Commun 2024;8:e0386.
  • 71. Kontrick AV, Alinger JB, Goins EC, Mollo AF, Cruz DS, McCarthy DM. Documentation of incidentally noted hepatic steatosis to emergency department patients: a retrospective study. Acad Emerg Med 2025;32:284-291.
  • 72. Kontrick AV, VanWagner LB, Yeh C, Courtney DM. Hepatic steatosis: an incidental finding that deserves attention. Acad Emerg Med 2021;28:578-581.
  • 73. Kutaiba N, Richmond D, Morey M, Brennan D, Rotella JA, Ardalan Z, et al. Incidental hepatic steatosis on unenhanced computed tomography performed for suspected renal colic: gaps in reporting and documentation. J Med Imaging Radiat Oncol 2019;63:431-438.
  • 74. Stuart A, Dotson A, Swarts K, Granich M, Yeung J, Thirumalai A, et al. Development of an AI algorithm for early identification of MASLD in the electronic health record. Hepatol Commun 2025;9:e0834.
  • 75. Suresh D, Li A, Miller MJ, Wijarnpreecha K, Chen VL. Associations between metabolic hyperferritinaemia, fibrosis-promoting alleles and clinical outcomes in steatotic liver disease. Liver Int 2024;44:389-398.
  • 76. Emir B, Can FE, Kaymaz E, Ozel Z, Efgan MG, Tekindal MA, et al. Sample size and power analysis for ROC AUC differences in diagnostic tests: a methodological evaluation of the Obuchowski-McClish and Hanley-McNeil methods. BMC Med Res Methodol 2026;26:43.
  • 77. Alkhouri N, Cheuk-Fung Yip T, Castera L, Takawy M, Adams LA, Verma N, et al. ALADDIN: a machine learning approach to enhance the prediction of significant fibrosis or higher in metabolic dysfunction-associated steatotic liver disease. Am J Gastroenterol 2026;121:362-374.
  • 78. Nixon J, Dusenberry M, Jerfel G, Nguyen T, Liu J, Zhang L, et al. Measuring calibration in deep learning. arXiv:1904.01685 [Preprint]. 2019 [cited 14 March 2026]. Available from: https://doi.org/10.48550/arXiv.1904.01685.
  • 79. Wang C. Calibration in deep learning: a survey of the state-of-the-art. arXiv:2308.01222 [Preprint]. 2023 [cited 14 March 2026]. Available from: https://doi.org/10.48550/arXiv.2308.01222.
  • 80. Chang D, Truong E, Mena EA, Pacheco F, Wong M, Guindi M, et al. Machine learning models are superior to noninvasive tests in identifying clinically significant stages of NAFLD and NAFLD-related cirrhosis. Hepatology 2023;77:546-557.
  • 81. Goff DC, Lloyd-Jones DM, Bennett G, Coady S, D’Agostino RB, Gibbons R, et al. 2013 ACC/AHA guideline on the assessment of cardiovascular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. J Am Coll Cardiol 2014;63:2935-2959.
  • 82. Chen VL, Morgan TR, Rotman Y, Patton HM, Cusi K, Kanwal F, et al. Resmetirom therapy for metabolic dysfunction-associated steatotic liver disease: October 2024 updates to AASLD Practice Guidance. Hepatology 2025;81:312-320.
  • 83. Bansal MB, Patton H, Morgan TR, Carr RM, Dranoff JA, Allen AM. Semaglutide therapy for metabolic dysfunction-associated steatohepatitis: November 2025 updates to AASLD Practice Guidance. Hepatology 2026;83:1326-1340.
  • 84. Lin H, Lee HW, Yip TC, Tsochatzis E, Petta S, Bugianesi E, et al. Vibration-controlled transient elastography scores to predict liver-related events in steatotic liver disease. JAMA 2024;331:1287-1297.
  • 85. Møller S, Sjöstedt SMS, Hobolth L, Mortensen C, Kimer N. The pathophysiological role of portal hypertension in metabolic dysfunction-associated steatotic liver disease. Hepatol Commun 2025;9:e0817.
  • 86. Wieland AC, Mettler P, McDermott MT, Crane LA, Cicutto LC, Bambha KM. Low awareness of nonalcoholic fatty liver disease among patients at high metabolic risk. J Clin Gastroenterol 2015;49:e6-e10.
  • 87. Singh A, Dhaliwal AS, Singh S, Kumar A, Lopez R, Gupta M, et al. Awareness of nonalcoholic fatty liver disease is increasing but remains very low in a representative US cohort. Dig Dis Sci 2020;65:978-986.
  • 88. Polanco-Briceno S, Glass D, Stuntz M, Caze A. Awareness of nonalcoholic steatohepatitis and associated practice patterns of primary care physicians and specialists. BMC Res Notes 2016;9:157.
  • 89. Le MH, Yeo YH, Cheung R, Wong VW, Nguyen MH. Ethnic influence on nonalcoholic fatty liver disease prevalence and lack of disease awareness in the United States, 2011-2016. J Intern Med 2020;287:711-722.
  • 90. Kim D, Cholankeril G, Loomba R, Ahmed A. Prevalence of nonalcoholic fatty liver disease and hepatic fibrosis among US adults with prediabetes and diabetes, NHANES 2017- 2018. J Gen Intern Med 2022;37:261-263.
  • 91. Man S, Deng Y, Ma Y, Fu J, Bao H, Yu C, et al. Prevalence of liver steatosis and fibrosis in the general population and various high-risk populations: a nationwide study with 5.7 million adults in China. Gastroenterology 2023;165:1025-1040.
  • 92. Díaz LA, Lazarus JV, Fuentes-López E, Idalsoaga F, Ayares G, Desaleng H, et al. Disparities in steatosis prevalence in the United States by Race or Ethnicity according to the 2023 criteria. Commun Med (Lond) 2024;4:219.
  • 93. Wong VW, Chu WC, Wong GL, Chan RS, Chim AM, Ong A, et al. Prevalence of non-alcoholic fatty liver disease and advanced fibrosis in Hong Kong Chinese: a population study using proton-magnetic resonance spectroscopy and transient elastography. Gut 2012;61:409-415.
  • 94. Kim MN, Han JW, An J, Kim BK, Jin YJ, Kim SS, et al. KASL clinical practice guidelines for noninvasive tests to assess liver fibrosis in chronic liver disease. Clin Mol Hepatol 2024;30:S5-S105.
  • 95. Eslam M, Fan JG, Yu ML, Wong VW, Cua IH, Liu CJ, et al. The Asian Pacific association for the study of the liver clinical practice guidelines for the diagnosis and management of metabolic dysfunction-associated fatty liver disease. Hepatol Int 2025;19:261-301.
  • 96. Cusi K, Abdelmalek MF, Apovian CM, Balapattabi K, Bannuru RR, Barb D, et al. Metabolic dysfunction-associated steatotic liver disease (MASLD) in People With Diabetes: The Need For Screening And Early Intervention. A Consensus Report of the American Diabetes Association. Diabetes Care 2025;48:1057-1082.
  • 97. Yip TC, Lee HW, Lin H, Tsochatzis E, Petta S, Bugianesi E, et al. Prognostic performance of the two-step clinical care pathway in metabolic dysfunction-associated steatotic liver disease. J Hepatol 2025;83:304-314.
  • 98. Zhang X, Yip TC, Wong GL, Leow WX, Liang LY, Lim LL, et al. Clinical care pathway to detect advanced liver disease in patients with type 2 diabetes through automated fibrosis score calculation and electronic reminder messages: a randomised controlled trial. Gut 2023;72:2364-2371.
  • 99. Singh S, Allen AM, Wang Z, Prokop LJ, Murad MH, Loomba R. Fibrosis progression in nonalcoholic fatty liver vs nonalcoholic steatohepatitis: a systematic review and meta-analysis of paired-biopsy studies. Clin Gastroenterol Hepatol 2015;13:643-654.e1-e9 quiz e39-e40.
  • 100. Huang DQ, Wilson LA, Behling C, Kleiner DE, Kowdley KV, Dasarathy S, et al. Fibrosis progression rate in biopsy-proven nonalcoholic fatty liver disease among people with diabetes versus people without diabetes: a multicenter study. Gastroenterology 2023;165:463-472.e5.
  • 101. Goyal T, Song MW, Suresh D, Jasty VSJ, Urias E, Wijarnpreecha K, et al. Serial liver stiffness measurement and fibrosis-4 scores in metabolic dysfunction-associated steatotic liver disease. Dig Dis Sci 2024;69:4250-4258.
  • 102. Lee HW, Wong GL, Kwok R, Choi KC, Chan CK, Shu SS, et al. Serial transient elastography examinations to monitor patients with type 2 diabetes: a prospective cohort study. Hepatology 2020;72:1230-1241.
  • 103. Lai M, Chen IY, Lai JC, Ge J. Generative AI in hepatology: transforming multimodal patient-generated data into actionable insights. Hepatol Commun 2025;9:e0683.
  • 104. Sherman MS, Challa PK, Przybyszewski EM, Wilechansky RM, Uche-Anya EN, Ott AT, et al. A natural language processing algorithm accurately classifies steatotic liver disease pathology to estimate the risk of cirrhosis. Hepatol Commun 2024;8:e0403.
  • 105. Fox RK, Chu JN, Goldman ML, Islam KB, Brandman D. Prospective study of a case-finding algorithm to detect NAFLD with advanced fibrosis in primary care patients. Hepatol Commun 2023;7:e0024.
  • 106. Vong T, Rizer N, Jain V, Thompson VL, Dredze M, Klein EY, et al. Automated identification of incidental hepatic steatosis on Emergency Department imaging using large language models. Hepatol Commun 2025;9:e0638.
  • 107. Persaud CK, Beckman D, Cheatham C. 1110-P: effect of fibrosis-4 index clinic template inclusion on fibrosis screening in diabetes mellitus. Diabetes 2025;74:1110-P.
  • 108. Berinstein E, Antonios K, Priyata D, Ammar A. S1338 Quality improvement: improving the screening of fatty liver disease in patients with type 2 diabetes mellitus. Am J Gastroenterol 2023;118:S1023-S1024.
  • 109. Mohanty A, Austad K, Bosch NA, Long MT, Nolen-Doerr E, Walkey AJ, et al. Assessing clinician engagement with a passive clinical decision support system for liver fibrosis risk stratification in a weight management clinic. Endocr Pract 2025;31:899-905.
  • 110. Douthit B, Richesson RL, Marsolo K, Melnick ER, Grudzen CR, Curtis L. Real world evidence: clinical decision support. In: NIH Pragmatic Trials Collaboratory, ed. Rethinking clinical trials: a living textbook of pragmatic clinical trials. Bethesda, MD: NIH Pragmatic Trials Collaboratory; 2025.
  • 111. Spann A, Bishop KM, Weitkamp AO, Stenner SP, Nelson SD, Izzy M. Clinical decision support automates care gap detection among primary care patients with nonalcoholic fatty liver disease. Hepatol Commun 2023;7:e0035.
  • 112. Spann A, Bishop K, Marbach S, Ji X, Slaughter J, Weitkamp A, et al. Electronic decision aids enhance management of primary care patients with steatotic liver disease: proof of concept pilot study. Hepatol Commun 2025;9:e0794.
  • 113. Kwan JL, Lo L, Ferguson J, Goldberg H, Diaz-Martinez JP, Tomlinson G, et al. Computerised clinical decision support systems and absolute improvements in care: meta-analysis of controlled clinical trials. BMJ 2020;370:m3216.
  • 114. Lobach DF. Evaluation of clinical decision support. In: Berner E, ed. Clinical decision support systems. Health informatics. Cham: Springer; 2016. p. 147-161.
  • 115. Ge J, Fontil V, Ackerman S, Pletcher MJ, Lai JC. Clinical decision support and electronic interventions to improve care quality in chronic liver diseases and cirrhosis. Hepatology 2025;81:1353-1364.
  • 116. Stempniewicz N, Vassalotti JA, Cuddeback JK, Ciemins E, Storfer-Isser A, Sang Y, et al. Chronic kidney disease testing among primary care patients with type 2 diabetes across 24 U.S. health care organizations. Diabetes Care 2021;44:2000-2009.
  • 117. Amoroso M, Augustin S, Moosmang S, Gashaw I. Noninvasive biomarkers prognostic of decompensation events in NASH cirrhosis: a systematic literature review. J Mol Med (Berl) 2024;102:841-858.
  • 118. Park J, Kim G, Kim BS, Han KD, Kwon SY, Park SH, et al. The associations of hepatic steatosis and fibrosis using fatty liver index and BARD score with cardiovascular outcomes and mortality in patients with new-onset type 2 diabetes: a nationwide cohort study. Cardiovasc Diabetol 2022;21:53.
  • 119. Sripongpun P, Kaewdech A, Udompap P, Kim WR. Characteristics and long-term mortality of individuals with MASLD, MetALD, and ALD, and the utility of SAFE score. JHEP Rep 2024;6:101127.
  • 120. Leite NC, Villela-Nogueira CA, Santos LV, Cardoso CRL, Salles GF. Prognostic value of changes in vibration-controlled transient elastography parameters for liver, cardiovascular and mortality outcomes in individuals with type 2 diabetes and metabolic dysfunction-associated steatotic liver disease: the Rio de Janeiro type 2 diabetes cohort. Diabetes Obes Metab 2025;27:2024-2034.
  • 121. Vilar-Gomez E, Vuppalanchi R, Gawrieh S, Samala N, Chalasani N. CAP and LSM as determined by VCTE are independent predictors of all-cause mortality in the US adult population. Hepatology 2023;77:1241-1252.
  • 122. Trebicka J, Gu W, de Ledinghen V, Aubé C, Krag A, Praktiknjo M, et al. Two-dimensional shear wave elastography predicts survival in advanced chronic liver disease. Gut 2022;71:402-414.
  • 123. Kim BK, Bergstrom J, Loomba R, Tamaki N, Izumi N, Nakajima A, et al. Magnetic resonance elastography-based prediction model for hepatic decompensation in NAFLD: a multicenter cohort study. Hepatology 2023;78:1858-1866.
  • 124. Higuchi M, Tamaki N, Kurosaki M, Inada K, Kirino S, Yamashita K, et al. Longitudinal association of magnetic resonance elastography-associated liver stiffness with complications and mortality. Aliment Pharmacol Ther 2022;55:292-301.
  • 125. Jasty VSJ, Urias E, Le Ashley Tiong K, Aboona MB, Song M, Faulkner C, et al. Prognostic accuracy of transient elastography-based predictors in diabetes and obesity: a multicenter international cohort study. Dig Dis Sci 2024;69:4240-4249.
  • 126. Noureddin N, Ajmera V, Bergstrom J, Bettencourt R, Huang DQ, Siddiqi H, et al. MEFIB-Index and MAST-Score in the assessment of hepatic decompensation in metabolic dysfunction-associated steatosis liver disease-Individual participant data meta-analyses. Aliment Pharmacol Ther 2023;58:856-865.

Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.

Format:

Include:

Predictive modeling and clinical decision tools for risk stratification in steatotic liver disease
Clin Mol Hepatol. 2026;32(Suppl):e2026.0204  Published online April 24, 2026
DOI: https://doi.org/10.3350/cmh.2026.0204
Download Citation

Download a citation file in RIS format that can be imported by all major citation management software, including EndNote, ProCite, RefWorks, and Reference Manager.
Format:
Include:
Predictive modeling and clinical decision tools for risk stratification in steatotic liver disease
Clin Mol Hepatol. 2026;32(Suppl):e2026.0204  Published online April 24, 2026
DOI: https://doi.org/10.3350/cmh.2026.0204
Close

Figure

  • 0
  • 1

Related articles

Predictive modeling and clinical decision tools for risk stratification in steatotic liver disease
Image Image
Figure 1. Liver risk stratification algorithms recommended by major professional societies. Society recommendation statements are from the Korean Association for the Study of the Liver (KASL) [94], Asian Pacific Association for the Study of the Liver (APASL) [95], American Association for the Study of Liver Diseases (AASLD) [53], European Association for the Study of the Liver (EASL) [52], American Gastroenterological Association (AGA) [64], and American Diabetes Association (ADA) [96]. Note that APASL recommends that patients with type 2 diabetes, overweight/obesity, or at least two metabolic syndrome components should be screened for MASLD and that those with MASLD should undergo FIB4 screening; FIB4 screening in those not known to have MASLD is not explicitly recommended. Other societies do not explicitly require hepatic steatosis to undergo screening. Unless otherwise noted, recommendations are common to all societies; discrepancies between recommendations are shown in parentheses. ELF, enhanced liver fibrosis score; FIB4, Fibrosis-4 score; LSM, liver stiffness measurement; MASLD, metabolic dysfunction-associated steatotic liver disease; MRE, magnetic resonance elastography; NFS, nonalcoholic fatty liver disease fibrosis score; NIT, non-invasive test; SWE, shear wave elastography; VCTE, vibration-controlled transient elastography. “Consensus” is defined as three or more of the non-APASL recommendation statements recommending screening. *AASLD and ADA: every 1–2 years if type 2 diabetes/prediabetes or at least two metabolic syndrome components and every 2–3 years otherwise; AGA: every 2 years; APASL: every 2–3 years; EASL and KASL: every 1–3 years. **Only AASLD provides specific ELF cutoffs, but we show cutoffs here for clarity.
Figure 2. Proposed consensus risk stratification algorithm for liver fibrosis screening. Screening is recommended in individuals with cardiometabolic risk factors, most notably type 2 diabetes or prediabetes, metabolic syndrome, or obesity complicated by at least one cardiometabolic risk factor. Fibrosis assessment is also recommended in those with known hepatic steatosis or elevated liver enzymes. If not already done, universal screening for other key risk factors for liver disease such as hepatitis B and C, alcohol use disorder, diabetes, hypertension, and dyslipidemia should be performed. In at-risk individuals, we recommend Fibrosis-4 (FIB4) for initial risk stratification, though aspartate aminotransferase-platelet ratio index (APRI) and steatosis-associated fibrosis estimator (SAFE) are reasonable alternatives depending on clinical context. In patients with low FIB4 (<1.3 if <65 years old or <2 if 65+ years old), no additional risk stratification is routinely required and the NIT should be repeated. In those with intermediate FIB4 (1.3–2.67 if <65 or 2–2.67 if ≥65 years), we suggest initial risk stratification with the enhanced liver fibrosis (ELF) test. However, it may be reasonable depending on availability, local expertise, and patient characteristics, to proceed directly to imaging-based noninvasive tests preferably vibration-controlled transient elastography (VCTE), though shear wave elastography (SWE) and even magnetic resonance elastography (MRE) are alternatives. In individuals with low ELF, the noninvasive test (NIT) should be repeated, while those with elevated ELF should be referred to subspecialty care. Those with high FIB4 (>2.67) should proceed to imaging-based NITs for risk stratification though direct referral to subspecialty care can be considered. In those with normal liver stiffness measurement (LSM) <8 kPa, the non-invasive test should be repeated; if LSM is elevated, referral to subspecialty care should be pursued. We suggest that non-invasive tests be repeated annually if FIB4 >2.67, every 1–2 years if there is diabetes, metabolic syndrome, or markedly elevated liver enzymes, and every 2–3 years for other patients. NIT, non-invasive test.

Figure 1.

Figure 2.

Predictive modeling and clinical decision tools for risk stratification in steatotic liver disease
Model Components Outcomes predicted
Major adverse liver outcomes Major adverse cardiovascular events Overall mortality
Blood-based predictors
 AST-platelet ratio index (APRI) AST, platelets ↑ [37] ↑/= [39] ↑/= [38]
 Fibrosis-4 (FIB4) AST, platelets, ALT, age ↑ [34,37] ↑ [39] ↑ [38]
 Nonalcoholic fatty liver disease fibrosis score (NFS) AST, platelets, ALT, age, albumin, body mass index, diabetes/pre-diabetes ↑ [37] ↑ [39] ↑ [38]
 BMI-AST/ALT Ratio-Diabetes (BARD) AST, ALT, body mass index, diabetes ↑ [117] ↑ [118] ↑ [118]
 Steatosis-associated fibrosis estimator (SAFE) AST, platelets, ALT, age, globulin, body mass index, diabetes ↑ [65] ↑ [42,119]
 LiverRisk score (LRS) AST, ALT, platelets, gamma glutamyl transferase, cholesterol, fasting glucose ↑ [45] ↑ [45]
 Nonalcoholic fatty liver disease outcomes score (NOS) Age, diabetes, albumin, bilirubin, platelet count, international normalized ratio ↑ [46]
Imaging-based predictors
 Vibration-controlled transient elastography (VCTE) Liver stiffness measurement (LSM), controlled attenuation parameter (CAP) ↑ [84] ↑ [120] ↑ [121]
 Shear wave elastography (SWE) LSM ↑ [122] ↑ [122]
 Magnetic resonance elastography (MRE) LSM, liver fat content (LFC) ↑ [123] ↑ [124]
Imaging- and blood-based
 FAST VCTE LSM, CAP, AST ↑ [125]
 MAST MRE LSM, LFC, AST ↑ [126]
 Agile 3+ and Agile 4 VCTE LSM, multiple laboratory values, comorbidities ↑ [84,125]
 MEFIB MRE LSM, FIB4 ↑ [126]
Current state Goal
Diagnostic accuracy High false negative and false positive rates Low false negative and false positive rates
Calibration in diverse settings Most models have been studied in tertiary care and a small number of community cohorts (e.g., UK Biobank) Validation in community and primary care settings, and in South America and Africa
Quantitative estimates “Low” vs. “intermediate” vs. “high” risk Estimated risk of an event occurring within 5–10 years
Interpretability, simplicity, and transparency Acceptable for most widely-used models, but at risk with increased adoption of machine learning Maintain in the era of machine learning models
Actionable findings Fibrosis (measured directly or indirectly) as the key determinant of pharmacotherapy Predicted risk as the primary determinant of pharmacotherapy
Capacity to implement and act on findings Limited screening (likely due to low disease awareness) and limited specialist capacity to act on findings Increased outreach to primary care providers and health systems, and management of high-risk patients by advanced practice providers
Keys to successful CDS tool implementation
Integration with charting or order entry system
Automatic provision of decision support as part of clinician workflow
No need for additional clinician data entry
Provision of decision support at the time and location of decision making
Provision of a recommendation, not just an assessment
Justification of decision support via provision of research evidence
Promotion of action rather than inaction
Local user involvement in the development process
Provision of decision support results to patients as well as providers
Table 1. Risk prediction models

ALT, alanine aminotransferase; AST, aspartate aminotransferase; FAST, Fibroscan-AST; MAST, MRE-AST; MEFIB, MRE-FIB4.

↑ implies that higher values of a score are associated with increased risk of the outcome.

Table 2. Current state of the art and eventual goals in risk prediction algorithm development
Table 3. Keys to successful clinical decision support (CDS) tool implementation

Table 1.

Table 2.

Table 3.

Clin Mol Hepatol : Clinical and Molecular Hepatology
Twitter

ABOUT

BROWSE ARTICLES

FOR CONTRIBUTORS

© The Korean Association for the Study of the Liver.