Screening for Metabolic Dysfunction–Associated Steatotic Liver Disease–Related Advanced Fibrosis in Diabetology: A Prospective Multicenter Study Free

Patients with type 2 diabetes and obesity have a high risk of metabolic dysfunction–associated steatotic liver disease (MASLD) compared with the general population (1). They exhibit a higher risk of progression to severe stages of MASLD, such as metabolic dysfunction–associated steatohepatitis (MASH) and advanced fibrosis (AF) (2). The presence of AF is a major determinant of a patient’s overall prognosis because it increases the risk of liver-related outcomes and all-cause mortality (3). Because MASLD often remains asymptomatic even in severe stages, systematic screening for AF is currently widely recommended for high-risk populations such as patients with type 2 diabetes and/or obesity with metabolic risk factors (4–8). Despites these recommendations, MASLD remains largely underdiagnosed and poorly considered in the routine clinical care of patients with type 2 diabetes and obesity outside hepatology clinics (9).

In recent years, strong consensus has emerged, providing clear guidance on noninvasive screening for AF in patients with MASLD (4–7). This two-step strategy, including the fibrosis-4 index (FIB-4) score as a first-line test, followed by a second-line test with higher accuracy in the event of an FIB-4 score ≥1.30, preferably using liver stiffness measurement (LSM) with vibration controlled transient elastography (VCTE) or, alternatively, the Enhanced Liver Fibrosis (ELF) test (4–8,10). There are discrepancies among guidelines regarding whether all patients with an FIB-4 score >1.30 or only those with an FIB-4 score between 1.30 and 2.66 should undergo a second-line assessment before referral to a hepatologist (4–8,10). Furthermore, recently, the use of an age-adapted threshold for the FIB-4 score of <2.0 in patients aged >65 years to rule out AF was suggested (7).

It is noteworthy that the development of these recommended algorithms and thresholds is based mainly on data from large international studies involving patients with biopsy-proven MASLD primarily recruited from hepatology departments (11,12). These studies have significantly advanced our understanding of the utility and diagnostic performance of noninvasive tests (NITs) for detecting AF in patients with MASLD using FIB-4 score and VCTE (11–14); ELF test (11,15,16); and, to a lesser extent, two-dimensional shear-wave elastography (2D-SWE) with SuperSonic Imagine (2D-SWE-SSI) (17,18), FibroMeter (Echosens, Paris, France) (19), or FibroTest (BioPredictive, Paris, France) (20,21). However, because these studies have focused on selected populations of patients with biopsy-proven MASLD, there is a crucial need to assess the diagnostic performance of these NITs in their recommended clinical settings, such as diabetology and nutrition clinics, which include patients who may not require hepatology referral or liver biopsy. Indeed, most patients with type 2 diabetes and/or obesity with MASLD have simple hepatic steatosis without MASH or AF (1,22). Importantly, few studies have conducted head-to-head comparisons between two-step strategies using either FIB-4 score followed by VCTE versus FIB-4 score followed by the ELF test (23) or using the age-adjusted threshold for the FIB-4 score in a population enrolled in systematic screening for AF.

The main aim of this study was to assess the feasibility and diagnostic performance of available NITs in screening for AF in high-risk populations performed in diabetology and nutrition clinics and compare the application and diagnostic performance of different recommended algorithms for referring patients at intermediate or high risk of AF to specialized hepatology care.

This was a prespecified interim analysis from data of the first 664 of 1,000 planned participants, prospectively recruited in four departments of diabetology and nutrition in France at Lyon-South, Lyon-East, Nantes, and Dijon University Hospitals between October 2020 and November 2023 in the Screening for NAFLD-related Advanced Fibrosis in high risk popuLation: optimization of the Diabetology/endoCrinology pAthway Referral using combinations of non-invasive biological and Elastography parameters (NAFLD-Care) study (ClinicalTrials.gov identifier NCT04435054). All participants underwent a standardized research visit including medical history and comprehensive metabolic and liver assessment including blood-based and imaging biomarkers on the same day in the fasting state. In addition, a subgroup of participants underwent a MRI–proton density fat fraction (PDFF) and magnetic resonance elastography (MRE) assessment within 15 days of the inclusion visit. All participants provided written informed consent before enrollment in the study. The study was approved by the Ile de France XI Institutional Review Board (no. 2020-A00477-32). The final study sample included patients with both valid VCTE assessment and ELF test results available.

Included patients were aged between 40 and 80 years, had a diagnosis of type 2 diabetes according to the American Diabetes Association standard of care (24) or obesity defined according to the World Health Organization as BMI ≥30 kg/m², had hepatic steatosis determined by conventional abdominal ultrasound, agreed to be included in the study and signed the informed consent form, and were affiliated with a health care insurance plan.

Exclusion criteria are presented in detail in the Supplementary Material and were mainly as follows: 1) evidence of other causes of chronic liver disease, 2) presence of regular and/or excessive use of alcohol (defined as >30 g/day for men and >20 g/day for women) for a period >2 years at any time in the last 10 years, 3) known evidence of cirrhosis, 4) life expectancy <5 years according to the investigator, 5) history of known HIV infection, 6) history of type 1 diabetes, and 7) BMI ≥40 kg/m².

Detailed assessment by NITs is reported in the Supplementary Material. Briefly, fasting blood samples were collected as standard of care in a fasting state at local laboratories. The FIB-4 score, nonalcoholic fatty liver disease fibrosis score (NFS), and Metabolic Dysfunction–Associated Fibrosis 5 (MAF-5) score were calculated using published formulas (25–27). The recommended published low/high cutoffs of −1.455/0.676 for NFS, 0/1 for MAF-5, and 1.30/2.67 for FIB-4 were used to rule out patients with a low probability of AF and rule in patients with a high probability of AF, respectively (4–6,26,27). An age-adapted low threshold for the FIB-4 score of 2.0 was also considered in the analysis (7). FibroTest (21) and FibroMeter results (19) were calculated according to patented formulas. The ELF test was assessed at the Department of Biochemistry in Dijon University Hospital on available serum samples for 661 participants enrolled until November 2023. According to the manufacturer’s indication, low/high cutoffs of 7.7/9.8 (8) or alternative published cutoffs of 9.8/11.3 were used in the analysis (16,23).

LSM using VCTE coupled with controlled attenuation parameter (CAP) was performed using FibroScan (M Probe; XL Probe; Echosens) by well-trained health workers according to previously described methods (28–30). Low and high cutoffs of 8 and 12 kPa were used according to the most recent guidelines (4–8).

2D-SWE-SSI assessment was performed using the Aixplorer ultrasound system (Aix-en-Provence, France) (17). According to previous published data, a low and high cutoff of 8.0 and 10.5 kPa were used (17).

MRI assessment was performed in a subgroup of participants, including liver fat quantification using PDFF and liver stiffness assessment using MRE according to previously described methods (28) using commercially available software and hardware (Resoundant, Inc., Rochester, MN). The presence of cirrhosis (stage F4) was defined by overt imaging diagnosis of cirrhosis (31,32). In addition, in patients without overt diagnosis of cirrhosis, a high cutoff of >3.62 kPa was used to rule in the presence of AF, and a lower threshold of <2.6 kPa was used to rule out with high sensitivity the presence of AF (33).

All liver biopsies were performed if clinically indicated after the noninvasive screening for AF and inclusion in the study. Because clinical practice may differ according to different centers, the study center agreed to follow a standardized clinical algorithm across all study centers specified in the study protocol (Supplementary Fig. 1). All liver biopsies performed in participants enrolled in the NAFLD-Care study were centrally reviewed by a senior liver pathologist expert in MASH using the Nonalcoholic Steatohepatitis Clinical Research Network scoring system at Lyon University Hospital. The median time between NIT assessment and liver biopsy was 22 weeks.

The presence of a high, intermediate, or low risk of AF was defined by a composite hierarchical risk stratification criterion according to clinical practice guidelines as follows:

1. If liver biopsy was performed, high risk of AF was defined by the presence of histological stage of fibrosis F3 or F4, and low risk of fibrosis was defined by histological stage of fibrosis ≤F2.

2. If liver biopsy was not performed and MRE was available, high risk of AF was defined by the presence of overt imaging diagnosis of cirrhosis or MRE >3,6 kPa, intermediate risk of AF was defined by MRE ≥2.6 but <3,6 kPa, and low risk of AF was defined by MRE <2.6 kPa.

3. If liver biopsy and MRE were not performed, high risk of AF was defined by VCTE using FibroScan ≥12 kPa, intermediate risk was defined by VCTE using FibroScan >8 but ≤12 kPa, and low risk of AF was defined by VCTE <8 kPa.

The primary end point was presence of a high risk of AF according to the hierarchical risk stratification for the assessment of the diagnostic performance of NITs. The secondary end point was the specialized hepatology care requirement defined by the presence of intermediate or high risk of AF according to the hierarchical risk stratification. The rationale for using composite criteria in the risk stratification of AF is detailed in the Supplementary Material.

Continuous variables were described by median and interquartile range (IQR); categorical variables were described by the frequency and percentage for each category. Comparisons of the variables according to risk level of AF (high vs. intermediate and high vs. low risk) were performed using the χ² or Fischer exact tests for categorical variables and the nonparametric Wilcoxon test for continuous variables. The diagnostic performance of each NIT for the detection of participants with a high risk of AF was assessed using the area under the empirical receiver operating characteristic curve (AUROC) and 95% CI based on the Delong method. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of each NIT for predefined thresholds were estimated. Statistical analyses were performed using R software (version 2.12.1), SPSS Statistics (version 23; SPSS IBM, New York, NY), and GraphPad Prism (version 9.4.1), and the significance level was set at 5%. This study is reported according to STARD guidelines (Supplementary Table 1) (34).

Of 694 patients screened for the study, 654 adults with MASLD were included (Supplementary Fig. 2). Of those included, 87.2% had type 2 diabetes, and 73.7% had class I–II obesity; 56% were men, and the median age was 60 years (IQR 52–67). The median BMI was 32.7 kg/m² (IQR 29.8–35.6), and 50.8% had abnormal liver enzyme levels (Table 1). The median (IQR) values for various liver fibrosis markers are listed in Table 1.

Using a hierarchical composite criterion, 9.3% (n = 61) of participants were identified as having a high risk of AF, including 24 participants with histological stage of fibrosis ≥F3, 15 with overt diagnosis of cirrhosis or MRE >3.6 kPa, and 22 with VCTE ≥12 kPa. In addition, 8.3% (n = 54) were at intermediate risk, and consequently, 17.6% (n = 115) of participants required specialized hepatology care (Fig. 1).

Participants at high risk of AF were significantly older and had a higher prevalence of type 2 diabetes at 98.4% compared with those at low risk at 85.3%. BMI was not significantly associated with high risk of AF, nor was the proportion of participants receiving glucagon-like peptide 1 receptor agonist (GLP-1 RA) therapy. In the subgroup of 570 participants with type 2 diabetes, 10.5% (n = 60) were at high risk of AF, and 8.8% (n = 50) were at intermediate risk, resulting in 19.3% (n = 110) requiring specialized care in hepatology clinics (Supplementary Fig. 3).

Among the NITs usually performed as first-line tests, FIB-4 had a higher AUROC at 0.78 compared with NFS at 0.71 and MAF-5 at 0.77 (Table 2). In addition, NFS ruled out AF for only 21.9% of participants, and MAF-5 ruled out only 2.7% (Supplementary Fig. 4A and B). At the low cutoff of <1.30, FIB-4 ruled out the risk of AF in 54.1% of participants, with a high NPV of 97.2% and low false-negative rate of 2.5% (Table 2 and Supplementary Fig. 4C). At the high cutoff of ≥2.67, FIB-4 had a PPV of 35.7% (Table 2). The diagnostic performance of the NITs in the subgroup of participants with type 2 diabetes is reported in Supplementary Table 2.

Among available patented blood-based NITs, the ELF test had the highest AUROC for the detection of high risk of AF at 0.82 compared with the AUROCs of FibroTest and FibroMeter at 0.78 and 0.74, respectively (Table 2). FibroTest classified 14% of participants with F3–F4, and FibroMeter classified 11.6% of participants as F3–F4 (Supplementary Fig. 4G and H). The ELF test at the recommended low threshold of 7.7 ruled out AF for only 1.8% of participants (Supplementary Fig. 4D). However, using alternative thresholds of 9.8 and 11.3, the ELF test ruled out the risk of AF in 74.9% of participants with an NPV of 95.6% and ruled in 1.7% of participants with a PPV of 61.8% (Table 2 and Supplementary Fig. 4E). In this study population, the optimal threshold for the detection of a high risk of AF was 9.6, and thresholds at fixed 90% sensitivity and specificity were 9.2 and 10.2, respectively (Table 2).

Finally, 2D-SWE-SSI had an AUROC of 0.84 for the detection of a high risk of AF. At the low cutoff of <8 kPa, the NPV was 97.1%, ruling out a high risk of AF for 75% of participants (Supplementary Fig. 4F), whereas at the high cutoff of ≥10.5, the PPV was 39.1% (Table 2).

We assessed the different recommended strategies for the referral of patients with intermediate/high risk of AF (Supplementary Table 3). The strategies using FIB-4/VCTE led to specialized care referral of 12.8–15.4% of participants, including 2% (n = 11) who did not have AF confirmed by liver biopsy (n = 10) or MRE <2.6 kPa (n = 1) (Fig. 2A). In the subgroup of participants classified using liver biopsy or MRE, the use of VCTE for those with FIB-4 score >1.30 or FIB-4 score between 1.30 and 2.66 had the same NPV of 81% and a PPV of 75% (Supplementary Table 4). However, the use of an age-adjusted low threshold for the FIB-4 score resulted in a lower NPV at 75% and lower PPV at 70% (Supplementary Table 4).

The alternative use of 2D-SWE-SSI using a low threshold of <8 kPa resulted in a similar proportion of specialized care referral of 14.1–11.6% compared with the use of VCTE (Fig. 2B), with the NPV ranging from 89% to 91% and the PPV from 58% to 62%.

Finally, the alternative use of ELF as a second-line test, using the low threshold of <9.8, resulted in a higher proportion of patients with false-negative results with high/intermediate risk not referred to specialized care (8–10%), with an NPV of 88–89% and PPV of 39–48% (Supplementary Table 3 and Fig. 2C). Use of an adapted low threshold of <9.2 resulted in an improved NPV for specialized care referral of 89–92%, at the cost of a higher proportion of participants receiving referrals (32–33%) and a lower PPV of 35–39%. These results were similar in the subgroup of participants classified using liver biopsy or MRE (Supplementary Table 4).

In this prospective multicenter study conducted in diabetology and nutrition clinics, the systematic screening for AF in patients with type 2 diabetes and/or class I–II obesity identified 9.3% as having a high risk of AF and 17.6% as needing specialized care in hepatology. The recommended strategy using a FIB-4 score followed by VCTE had good diagnostic performance. Alternative strategies may be used depending on the clinical setting at the cost of a higher proportion of patients referred to hepatology with a lower PPV for the presence of intermediate and high risk of AF.

Our findings indicate that use of an age-adjusted low threshold for the FIB-4 score of <2.0 in participants aged >65 years seems unnecessary, because it decreased the NPV and PPV, with a similar referral rate to hepatology. Furthermore, the use of ELF as a second-line test is a reasonable alternative strategy, but our data demonstrate that an adapted threshold of 9.8 rather than 7.7 to rule out the risk of AF is required. This approach may be relevant in settings with low accessibility to VCTE but would result in a higher proportion of patients being referred to hepatology clinics, including a higher rate of false-positive cases. The alternative use of 2D-SWE as a second-line test, using the same threshold as VCTE (<8 kPa), also had a high NPV and good PPV and may serve as an alternative to VCTE. Overall, our data demonstrate that the recommended two-step screening strategy for MASLD-related AF using FIB-4 score/VCTE is applicable and appropriate for risk stratification of MASLD-related AF in diabetology. The use of FIB-4 score/2D-SWE or FIB-4 score/ELF test with adjusted thresholds <9.8 may also be an alternative, depending on the clinical setting and availability of these second-line NITs, but would lead to higher referral and lower PPV, which would need to be anticipated by hepatologists when designing the clinical referral pathway.

In our study, the proportion of patients with a high risk of AF was 9.3%, and that of patients with an intermediate/high risk of AF was 17.6%. Interestingly, in the subgroup of patients with type 2 diabetes, the prevalence of AF was similar at 10.5%, which is lower than the reported prevalence in MASLD biopsy–proven cohorts with type 2 diabetes, ranging from 40% to 46% (13,35). However, our results are in line with the 12.5% reported prevalence in the Ajmera et al. (23) study using MRE as reference, which limits the selection bias associated with liver biopsy indications. This clearly indicates that biopsy-proven cohorts from hepatology settings may not be well suited for assessment of screening strategies, because the patients with MASLD receiving care in hepatology clinics are enriched with AF cases and do not reflect its overall prevalence in the type 2 diabetes population. This is crucial, because the performance of NITs is highly dependent on the prevalence of the condition being tested. Furthermore, the MAF-5 score, developed for AF screening in the general population (27), seems not to be clinically relevant in our study population enriched with patients with type 2 diabetes or obesity. These observations reinforce the importance of clinical context in the use of NITs, which can affect diagnostic performance. Interestingly, the estimated global prevalence of AF in patients with type 2 diabetes reported in the meta-analysis by Younossi et al. (1) was 15.9%, higher than in our study. This difference may be due to disparities across countries, including lifestyle, diet, and glucose-lowering therapies, which can influence the progression toward MASLD-related AF. Indeed, our study population included a high proportion of patients treated with GLP-1 RA, known to exert a beneficial effect on MASLD and MASH (2), potentially accounting for a lower prevalence of AF, even though GLP-1 RA treatment was not significantly different between patients with high and low risk of AF.

Our findings align with those of Ajmera et al. (23), indicating that the application of a screening algorithm using FIB-4 score followed by VCTE in patients with type 2 diabetes had a low false-negative rate and was applicable. Our study also reports the diagnostic performance of available patented blood-based NITs, including the ELF test, FibroMeter, and FibroTest. Among these, the ELF test showed the highest diagnostic accuracy at 0.82, confirming good diagnostic performance for the detection of AF previously reported in MASLD (11,15,16). These data also confirm the possible use of the ELF test as an alternative to VCTE in second-line testing, as suggested by the society guidelines (4–8). Furthermore, our data confirm that the recommended low threshold for the ELF test of <7.7 is unsuitable and would lead to an excessive proportion of unnecessary referrals in hepatology (23,36). Interestingly, we found the optimal threshold for the ELF test in the detection of AF to be 9.6, which corresponds to the optimal threshold in MASLD reported in a previous meta-analysis (37). Finally, data on using SWE as a potential alternative to VCTE are limited. Here, we provide a head-to head comparison of the use of 2D-SWE compared with VCTE, demonstrating that 2D-SWE using the same low threshold of 8 kPa could be an alternative to rule out AF in clinical settings with better accessibility to 2D-SWE than VCTE.

Current recommended screening strategies aim to identify patients with AF. However, resmetirom was recently approved for the treatment of patients with MASH and fibrosis stage F2 and F3. In this context, Nourredine et al. (38) proposed an expert panel recommendation for resmetirom initiation and monitoring in patients with MASH and moderate to noncirrhotic AF based on NITs. Interestingly, the thresholds for VCTE were 10–15 kPa; for MRE, 3.3–4.9 kPa; and ELF, 9.8–10.4; therefore, a majority of patients with LSM >8 kPa or alternative ELF score >9.8 will be identified by the current recommended screening strategies as shown in this study.

There are several strengths of the current study, including a large study population of well-characterized participants with type 2 diabetes and/or obesity and MASLD, prospectively enrolled in screening for AF in a multicenter setting. All participants underwent blood-based NITs and LSM using VCTE and SWE on the same day in the fasting state. In addition, other etiologies of liver disease were carefully excluded. Another major strength of this study was the use of a composite hierarchical criterion to determine the risk of AF, enabling assessment of the diagnostic performance of NITs in their actual context of use, which is crucially needed to demonstrate the feasibility and applicability of the recommended screening algorithm in nonhepatology settings (39). Indeed, although liver biopsy is still considered the gold standard for the diagnosis of AF, it is rarely performed in clinical practice and is usually performed in a highly selected population. Therefore, the use of a composite criterion allows a comprehensive assessment, including participants who require liver biopsy while also considering patients who do not have any clinical indication for liver biopsy either because of overt imaging diagnosis of cirrhosis by MRI or conversely because of low LSM and normal liver enzymes. More importantly, there remain major barriers to the implementation of active case finding of AF in clinical practice; providing evidence of feasibility and application of the screening strategies will enhance their dissemination and acceptance in diabetology. In addition, accessibility to the second-line NITs remains a major issue in health care systems globally. Therefore, by providing robust data on these screening strategies, the current study will help influence health care providers and stakeholders to support the implementation of clinical guidelines (39).

Finally, we acknowledge that this study was performed in a tertiary university hospital in France by expert investigators in the field; therefore, the generalizability of our study must be confirmed in other clinical settings, such as secondary care and private practice. Particularly, the presence of hepatic steatosis was determined using abdominal ultrasound, which may not be widely available in many centers; CAP may be considered as an alternative, although it is less validated in this setting (7). The lack of historical data before the implementation of the screening strategy at the study sites limits the assessment of its impact in the clinical care of patients. However, retrospective data collected at the Lyon site indicate that the number of patients referred to hepatology before the implementation of screening using VCTE was higher (44%), including a majority of patients with a low risk of AF (69%) (29).

In addition, patients with class III obesity were excluded from the study, because ultrasound transient elastography may lack accuracy in this population, and dedicated studies are warranted in this population (28). Despite the large sample size, this study lacked statistical power to investigate the impact of age and GLP-1 RA therapies on the diagnostic performance of the NITs. Therefore, international studies are required with larger sample sizes, because differences may occur across countries in terms of the prevalence of AF and patients characteristics. Finally, longitudinal studies are needed to assess the impact of the implementation of this screening strategy on clinical outcomes. Therefore, we plan to perform a follow-up assessment at 4 years in a subgroup of participants with type 2 diabetes enrolled in the current study (ClinicalTrials.gov identifier: NCT06567990).

Funding. The assay kits for the ELF test assessment were provided by Siemens Healthineers.

Duality of Interest. The study was funded by an investigator-initiated research grant from Gilead Sciences. C.C. has received consultant fees from Gilead Sciences, Novo Nordisk, AstraZeneca, Lilly, E-Scopics, MSD, Bayer, Corcept Therapeutics, and Echosens and grant support from Gilead Sciences. P.Mou. has received speaker and/or consultant fees from Amarin, Amryt Pharma, Ionis Pharmaceuticals, Ultragenyx Pharmaceutical, and Viatris and fees for clinical trials from Amgen, Amryt Pharma, Ionos Pharmaceuticals, and Novo Nordisk. BC has served on scientific advisory boards and/or received honoraria or consulting fees from Amgen, Abbott, AstraZeneca, Lilly, MSD, Novartis, Novo Nordisk, Sanofi, and Ultragenyx Pharmaceutical. S.C. has received honoraria or consulting fees from Amarin, Amgen, AstraZeneca, BioMarin, Boehringer Ingelheim, Ionis Pharmaceuticals, MSD, Novartis, and Novo Nordisk. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. C.C. was responsible for the study concept and design, analysis and interpretation of data, drafting of the manuscript, and critical revision of the manuscript. B.V., B.S., A.R., P.Mor., S.H., C.P., J.-M.P., S.C., P.Mou., M.L., B.C., and E.D. were responsible for study visits and data collection. D.L. and L.D. performed ELF test analyses and critical revision of the manuscript. F.S. and A.A.-K. were responsible for statistical analysis. V.H. was responsible for pathology analysis and data collection. L.M. was responsible for imaging analysis and data collection. S.B. was responsible for methodology and study design. D.D. performed study site coordination and data collection. All authors provided critical revision of the manuscript and approved the final version of this article. C.C. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Steven E. Kahn and Amalia Gastaldelli.