Metabolic dysfunction-associated steatotic liver disease (MASLD) is the most common chronic liver disease worldwide, closely linked with the rising prevalence of obesity and type 2 diabetes [
1,
2]. MASLD spans a broad clinical spectrum, from simple hepatic steatosis to its more progressive form metabolic dysfunction-associated steatohepatitis (MASH), which is characterized by liver inflammation and hepatocellular injury. Without intervention, MASH can progress to fibrosis, cirrhosis, portal hypertension and hepatocellular carcinoma [
1]. Clinical guidelines now emphasize the importance of identifying patients with MASH who are at the highest risk for disease progression but have not yet reached irreversible cirrhosis [
3].
Recently, the U.S. Food and Drug Administration (FDA) approved resmetirom and semaglutide as the pharmacologic therapy for MASH with stage 2 or 3 fibrosis [
4,
5]. The American Association for the Study of Liver Diseases (AASLD) subsequently recommended screening for treatment eligibility using vibration-controlled transient elastography (VCTE) or magnetic resonance elastography (MRE) [
3]. In particular, MASLD patients with MRE-derived liver stiffness (LS) 3.1–4.4 kPa are proposed for treatment, while those with LS>5kPa are excluded. This guidance is consistent with that of the American Gastroenterological Association (AGA: treat for 3.3-4.2 kPa, consider treatment for 4.3-4.9 kPa, do not treat for ≥5 kPa) [
6]. MRI-derived proton density fat fraction (PDFF), although not formally included in current treatment eligibility guidelines, remains a highly sensitive marker of steatosis in early MASLD and decreases in advanced “burnt-out” MASH, making it valuable at both ends of the disease spectrum.
This study made retrospective use of a large imaging and histologic dataset to assess the diagnostic performance of the above AASLD guidelines as well as to compute and compare optimized cutoffs for resmetirom and semaglutide treatment eligibility using LS with or without PDFF.
Three prospective MASLD-related research studies conducted at two institutions were assembled into our final dataset [
7-
9]. Histological assessments were performed by experienced hepatopathologists using NASH Clinical Research Network (CRN) criteria for steatosis, inflammation, fibrosis and MASH diagnoses. Patients were included if they had technically adequate MRI-PDFF (PDFF) and 2D MRE-measured LS measurements. N=581 patients met the inclusion criteria, and were broken down into four groups by histology:
Non-MASH: n=218 (38%)
MASH with fibrosis stage 0-1 (MASH+F0/1): n=200 (34%)
MASH with fibrosis stage 2-3 (MASH+F2/3): n=116 (20%)
MASH with fibrosis stage 4 (MASH+F4): n=47 (8%)
The primary aim was to identify MASH+F2/3 patients, the target treatment population. We used the Kruskal-Wallis test to assess group differences with respect to PDFF, LS, and histologic features. Patients were deemed candidates for treatment if their PDFF was greater than a given cutoff value and LS fell within given bounds LSL to LSU. A 10-fold cross-validated grid search was performed to identify PDFF and LS cutoffs that maximize the F-score (i.e., the harmonic mean of sensitivity and positive predictive value [PPV]) for classifying MASH+F2/3 patients. We obtained optimal cutoffs for using LS alone and LS combined with PDFF. Diagnostic statistics for the AASLD guidance cutoffs and the optimized cutoffs were computed and compared to each other; sensitivity was compared by McNemar test while PPV was compared by Kosinski’s test. A final grid search was performed without cross validation to estimate using the entire dataset a set of optimized cutoffs proposed for use. The threshold of statistical significance was chosen to be 0.05.
As shown in
Supplementary Table 1, histologic steatosis grades and fibrosis stages, as well as PDFF and LS measurements, were significantly different across the four groups (
P<0.001). Specifically, LS increased with fibrosis severity, while PDFF decreased with MASH progression. Patients with MASH+F2/3 had median LS 3.6 kPa, with partial interquartile range (IQR) overlap with MASH+F4 but clearer separation from MASH+F0/1. Median PDFF in MASH+F2/3 was 14.0%, approximately 50% higher than that observed in MASH+F4, with minimal IQR overlap.
Grid search found that the optimal cutoffs averaged over validation folds were LS
L=3.0kPa and LS
U=7.1kPa if using LS alone, and PDFF≥5.5%, LS
L=3.0kPa and LS
U=7.6kPa if using LS and PDFF jointly. The validation performance of the AASLD guidance cutoffs (sensitivity=47.4%, PPV=45.8%, F=46.6%) was surpassed by that of optimized cutoffs using LS alone (sensitivity=71.6%, PPV=43.7%, F=54.2%) and LS and PDFF jointly (sensitivity=72.4%, PPV=51.5%, F=60.2%). While PPV was not significantly changed (p=0.43, p=0.06 for LS alone and joint LS and PDFF, respectively), sensitivity showed a clear and dramatic increase due to optimization (p<1e-6 in both cases). In comparing the optimized cutoffs with and without PDFF, sensitivity was not statistically differentiated, but PPV was significantly increased when PDFF was included (p<1e-5). The specificity and negative predictive value (NPV) of all three models were approximately the same (AASLD: specificity=83.7%, NPV=89.2%; LS Optimized: specificity=77.0%, NPV=91.6%; LS+PDFF Optimized: specificity=83.0%, NPV=92.3%). In a subcohort analysis excluding patients with high liver enzyme, the optimized cutoffs remained similar and continued to outperform the current guideline window (
Supplementary Table 2).
As shown in
Figure 1, final grid search optimization yielded cutoffs LS
L=3.0kPa and LS
U=6.6kPa if using LS alone, and PDFF≥5.5%, LS
L=3.0kPa and LS
U=7.6kPa if using LS and PDFF jointly.
Our findings demonstrate that optimization greatly improves identification of MASH+F2/3 patients by MRE-measured liver stiffness both with and without the concurrent use of MRI-derived fat fraction. Following optimization, sensitivity was increased 24–25% without sacrificing PPV performance.
Of the 116 MASH+F2/3 patients in our dataset, 55 fell between the guidance cutoffs 3.1–4.4 kPa, 31 had LS≤3.1 kPa, and 30 had LS>4.4 kPa. This suggests a widening of the range for LS may be beneficial for finding patients in need of treatment. Meanwhile, of the 465 patients without MASH+F2/3 in our data, 65 had liver stiffness within the guidance bounds; of these 13 patients (i.e., 20%) had PDFF≤5%, suggesting that a lower bound on PDFF may be beneficial in removing false positives for treatment.
Our optimization process led to cutoffs that align with these preliminary observations. Using optimized cutoffs, which increased the upper limit on LS, 83 and 84 MASH+F2/3 patients were correctly identified using LS alone and LS jointly with PDFF, respectively. Using LS alone led to 107 false positives, but this number was reduced to 79 with the inclusion of PDFF. Our results suggest that using a combination of LS and PDFF, which contribute orthogonal and biologically complementary information, is the most reliable way to identify patients suitable for resmetirom and semaglutide treatment. Our proposed cutoffs represent the optimal choice that balances the complex representation of LS and PDFF measurements across the whole spectrum of MASH progression. The relatively large size of our dataset gives confidence that they are representative of the wider MASLD population.
Compared with VCTE, MRI/MRE offers superior technical performance and reliability. Recent findings from the NIMBLE consortium highlight the limitations of VCTE in clinical practice: the repeatability coefficient (RC) of VCTE-assessed liver stiffness was 38% for same-day, same-operator measurements and worsened to 70% under different-day, different-operator conditions [
10]. In contrast, MRE demonstrated superior technical reliability, with a multi-vendor, different-magnet, different-day reproducibility of 0.75 kPa, and different-day RC of 30% for 2D MRE and 25% for 3D MRE under same-scanner conditions [
11]. These findings are further supported by a head-to-head comparison in 33 patients with cirrhosis, where MRE showed significantly better repeatability than VCTE (21% vs. 38%) [
12]. PDFF also demonstrated outstanding reproducibility (1.2–1.6%) in the same NIMBLE study [
11]. Collectively, these results support the preferential use of MRI/MRE for treatment eligibility assessment, where precision is essential.
Limitations include the retrospective design and predominance of mild to moderate disease severity, which may limit generalizability to different populations. Our analysis pooled multiple prospective studies from two institutions, but the optimized cutoffs have not yet been validated across broader patient populations, different vendors, and varied clinical settings before widespread adoption in routine practice. Histologic assessment, while the reference standard, is subject to interobserver variability and sampling bias. Longitudinal imaging and outcome-based validation will be essential in future studies. Lastly, the surprisingly high upper bound for LS in our proposed cutoffs, while enabling the identification of high-LS resmetirom and semaglutide candidates, nonetheless carries with it the risk of falsely suggesting cirrhotic patients for treatment. This tradeoff reflects the complicated progression of high-stiffness MASH. We advise clinicians to exercise particular care with patients whose liver stiffness values fall into traditionally cirrhotic territory, taking into account the patient’s entire body of available clinical data before making a treatment recommendation. Our exploration in
Supplementary Figures 1, 2 supports a pragmatic two-step workflow: (1) MRI-based prescreening using LS+PDFF, followed by (2) targeted review of clinical context, laboratory results, and noninvasive composite cirrhosis scores [
13-
15] in patients with high LS. When these additional data suggest high-risk cirrhosis or decompensation, liver biopsy or a more conservative treatment decision may be warranted even if imaging-based thresholds are met.
By expanding the MRE-derived liver stiffness range to 3.0–7.6 kPa and incorporating a minimum PDFF threshold of 5.5%, we significantly improved the identification of MASH patients with stage 2-3 fibrosis who are eligible for resmetirom or semaglutide therapy. Compared with current AASLD/AGA guideline cutoffs (3.1–4.4 kPa), sensitivity increased by 25% without compromising positive predictive value. These findings support the use of integrated MRE+PDFF screening to more accurately guide patient selection for emerging pharmacologic therapies and to standardize MRI-based eligibility assessments across research and clinical practice.
FOOTNOTES
-
Authors’ contribution
Study concept and design: MY, RLE, AMA, and RL. Data acquisition: JL, MY, AMA, RL, RLE. Data analysis and interpretation: KK, NO and MY. Drafting of the manuscript: NO, KK, JL and MY. Critical revision and approval of the final manuscript: NO, KK, JL, AMA, RL, RLE, and MY.
-
Acknowledgements
We thank Dr. Vijay Shah and Dr. Harmeet Malhi for their outstanding mentorship and support through the NIDDK funded programs (T32DK007198, R13DK14254, R13DK136181). We also thank Dr. Hao Wu and Dr. Caixin Qiu for their assistance in retrieving clinical and laboratory data for the subcohort analyses included in the supplementary materials, which helped address reviewer comments and strengthen our findings.
-
Conflicts of Interest
NO: Dr. Nana Owusu receives support from the T32 training program of multidisciplinary training grant in digestive diseases (T32DK007198) funded by NIDDK, and travel awards to present this work at the midwest Digestive Disease Research Core Center (DDRCC) Alliance conferences (R13DK142549, R13DK136181).
KK: Mr. Kyle Kalutkiewicz reports employment at Resoundant, Inc.
AMA: Dr. Alina M. Allen reports research funding (NIDDK, R01 DK134448) from the National Institutes of Health, Novo Nordisk, Madrigal, Boehringer-Ingelheim, Siemens, Escopics, Oncoustics, and Target Pharma and consulting fees from Novo Nordisk, Madrigal, Boehringer Ingelheim and GSK.
RL: Dr. Rohit Loomba receives funding support from NCATS (5UL1TR001442), NIDDK (U01DK061734, U01DK- 130190, R01DK106419, R01DK121378, R01DK124318, P30DK120515), NHLBI (P01HL147835). RL serves as a consultant to Aardvark Therapeutics, Altimmune, Anylam/Regeneron, Amgen, Arrowhead Pharmaceuticals, AstraZeneca, Bristol-Myer Squibb, CohBar, Eli Lilly, Galmed, Gilead, Glympse bio, Hightide, Inipharma, Intercept, Inventiva, Ionis, Janssen Inc., Madrigal, Metacrine, Inc., NGM Biopharmaceuticals, Novartis, Novo Nordisk, Merck, Pfizer, Sagimet, Theratechnologies, 89 bio, Terns Pharmaceuticals and Viking Therapeutics. In addition his institutions received research grants from Arrowhead Pharmaceuticals, Astrazeneca, Boehringer-Ingelheim, Bristol-Myers Squibb, Eli Lilly, Galectin Therapeutics, Galmed Pharmaceuticals, Gilead, Intercept, Hanmi, Intercept, Inventiva, Ionis, Janssen, Madrigal Pharmaceuticals, Merck, NGM Biopharmaceuticals, Novo Nordisk, Merck, Pfizer, Sonic Incytes and Terns Pharmaceuticals. Co-founder of LipoNexus Inc.
RLE: Dr. Richard L. Ehman reports research funding (NIBIB R37 EB001981) from the National Institutes of Health. RLE and Mayo Clinic have intellectual property rights and a financial interest related to this research.
MY: Dr. Meng Yin reports research funding (NIBIB R01 EB017197) from the National Institutes of Health. MY and Mayo Clinic have intellectual property rights and a financial interest related to this research. This research has been reviewed by the Mayo Clinic Conflict of Interest Review Board and has been conducted in compliance with Mayo Clinic Conflict of Interest policies.
Other co-authors have no relevant conflicts of interest exist.
SUPPLEMENTARY MATERIAL
Supplementary material is available at Clinical and Molecular Hepatology website (
http://www.e-cmh.org).
Supplementary Fig 1.
Comparison of platelet count and Agile 4 scores for ruling out cirrhosis using the current (4.4 kPa) versus optimized (7.6 kPa) MRE thresholds. Subgroup analysis of platelet count (PLT: ×10
3/μL) and Agile 4 score within the optimized LS+PDFF eligibility window (lightgreen rectangular boxes with dashed line border). Histological fibrosis stages were color-coded in green (F2), blue (F3) and red (F4) dots/numbers. Left scatterplot: PLT distribution in patients with histologic F2–F3 versus F4 fibrosis; the bubble size along with number indicates PLT, a commonly used >150 threshold combined with LS for low risk of clinically significant portal hypertension or decompensation[
4]. Right scatterplot: Agile 4 score distribution in patients with histologic F2–F3 versus F4 fibrosis; the bubble size along with number indicates the score value. In these subgroups, most histological F4 cases had PLT >150×10
3/μL and Agile 4 values below the high-risk cutoff of 0.656.
Supplementary Fig 2.
Influence of liver enzyme levels on the ability to rule out cirrhosis using the current (4.4 kPa) versus optimized (7.6 kPa) MRE thresholds. Subgroup analysis of ALT and AST within the optimized LS+PDFF eligibility window (lightgreen rectangular boxes with dashed line border). Histological fibrosis stages were color-coded in green (F2), blue (F3) and red (F4) dots. The bubble size with adjacent numbers indicates the laboratory test values.
Figure 1.The left scatter plot shows all patients with MASH+F2/3 (red points) and without (white points); eligibility cutoffs for PDFF and LS are indicated: (i) Current criterion by black lines and dark green shading—LS from 3.1 to 4.4 kPa; (ii) final optimized cutoffs for LS alone by blue lines and lake blue shading—LS from 3.0 to 6.6 kPa; (iii) final optimized cutoffs for LS+PDFF by green lines and green shading – PDFF≥5.5%, LS from 3.0 to 7.6 kPa. The right diagnostic performance plot shows sensitivity along x-axis, positive predictive value (PPV) along y-axis, and F-score calculated for identifying MASH+F2/3 patients for the current criterion (black point), optimized cutoffs for LS alone (blue point), and optimized cutoffs for LS+PDFF (green point); F-score is defined by F=2×sensitivity×PPV/ (sensitivity+PPV).
Abbreviations
The American Association for the Study of Liver Diseases
The American Gastroenterological Association
Clinical Research Network
U.S. Food and Drug Administration
metabolic dysfunction-associated steatohepatitis
metabolic dysfunction-associated steatotic liver disease
magnetic resonance elastography
proton density fat fraction
positive predictive value
repeatability coefficient
vibration-controlled transient elastography
REFERENCES
- 1. Younossi ZM, Razavi H, Sherman M, Allen AM, Anstee QM, Cusi K, et al. Addressing the High and Rising Global Burden of Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD) and Metabolic Dysfunction-Associated Steatohepatitis (MASH): From the Growing Prevalence to Payors’ Perspective. Aliment Pharmacol Ther 2025;61:1467-1478.
- 2. Huang DQ, Wong VWS, Rinella ME, Boursier J, Lazarus JV, Yki-Järvinen H, et al. Metabolic dysfunction-associated steatotic liver disease in adults. Nat Rev Dis Primers 2025;11:14.
- 3. 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.
- 4. Harrison SA, Bedossa P, Guy CD, Schattenberg JM, Loomba R, Taub R, et al. A Phase 3, Randomized, Controlled Trial of Resmetirom in NASH with Liver Fibrosis. N Engl J Med 2024;390:497-509.
- 5. Sanyal AJ, Newsome PN, Kliers I, Østergaard LH, Long MT, Kjær MS, et al. Phase 3 Trial of Semaglutide in Metabolic Dysfunction-Associated Steatohepatitis. N Engl J Med 2025;392:2089-2099.
- 6. Noureddin M, Charlton MR, Harrison SA, Bansal MB, Alkhouri N, Loomba R, et al. Expert Panel Recommendations: Practical Clinical Applications for Initiating and Monitoring Resmetirom in Patients With MASH/NASH and Moderate to Noncirrhotic Advanced Fibrosis. Clin Gastroenterol Hepatol 2024;22:2367-2377.
- 7. Allen AM, Shah VH, Therneau TM, Venkatesh SK, Mounajjed T, Larson JJ, et al. The Role of Three-Dimensional Magnetic Resonance Elastography in the Diagnosis of Nonalcoholic Steatohepatitis in Obese Patients Undergoing Bariatric Surgery. Hepatology 2020;71:510-521.
- 8. Chen J, Yin M, Talwalkar JA, Oudry J, Glaser KJ, Smyrk TC, et al. Diagnostic Performance of MR Elastography and Vibration-controlled Transient Elastography in the Detection of Hepatic Fibrosis in Patients with Severe to Morbid Obesity. Radiology 2017;283:418-428.
- 9. Loomba R, Cui J, Wolfson T, Haufe W, Hooker J, Szeverenyi N, et al. Novel 3D Magnetic Resonance Elastography for the Noninvasive Diagnosis of Advanced Fibrosis in NAFLD: A Prospective Study. Am J Gastroenterol 2016;111:986-994.
- 10. Pierce TT, Ozturk A, Sherlock SP, Moura Cunha G, Wang X, Li Q, et al. Reproducibility and Repeatability of US Shear-Wave and Transient Elastography in Nonalcoholic Fatty Liver Disease. Radiology 2024;312:e233094.
- 11. Fowler KJ, Venkatesh SK, Obuchowski N, Middleton MS, Chen J, Pepin K, et al. Repeatability of MRI Biomarkers in Nonalcoholic Fatty Liver Disease: The NIMBLE Consortium. Radiology 2023;309:e231092.
- 12. Siddiqi H, Huang DQ, Mittal N, Nourredin N, Bettencourt R, Madamba E, et al. Repeatability of vibration-controlled transient elastography versus magnetic resonance elastography in patients with cirrhosis: A prospective study. Aliment Pharmacol Ther 2024;60:484-491.
- 13. 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.
- 14. Noureddin M, Mena E, Vuppalanchi R, Samala N, Wong M, Pacheco F, et al. Increased accuracy in identifying NAFLD with advanced fibrosis and cirrhosis: independent validation of the Agile 3+ and 4 scores. Hepatol Commun 2023;7:e0055.
- 15. 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.
