Between June 2017 and May 2020, a total of 133 subjects who were either healthy volunteers or patients with NAFLD were recruited from advertisements, as well as from the diabetes and endocrinology, hepatology, metabolic surgery, and health examination centers in the hospital. Study protocols were in accordance with the Declaration of Helsinki and were approved by the Institutional Review Board at the Gachon University Gil Medical Center (GAIRB2017-001, -200, and -310). All participants provided written informed consent and the studies were registered at https://cris.nih.go.kr (registration numbers KCT0005161, KCT0003144, and KCT0003527) in accordance with the World Health Organization International Clinical Trials Registry Platform.

The inclusion criteria were as follows: healthy volunteers or subjects who were diagnosed to have or suspected of having NAFLD (aged from 19 to 70 years). Additionally, we tried to enroll subjects with evidence of NAFLD on a liver biopsy performed within 3 months prior to enrollment at a hepatology clinic of the hospital. All biopsied patients reported no significant changes in body weight or interventions for the treatment of NAFLD or comorbidities between the date of the liver biopsy and enrollment. Patients were excluded if they had excessive alcohol consumption (alcohol intake >20 g/day for women and >30 g/day for men), evidence of another coexistent liver or biliary disease (positive hepatitis B surface antigen, antihepatitis C virus antibody, autoimmune hepatitis, histological evidence of other concomitant chronic liver diseases, primary biliary cirrhosis, or primary sclerosing cholangitis), use of medications known to cause secondary hepatic steatosis (corticosteroids, tamoxifen, amiodarone, or methotrexate) within 1 year, drug abuse, human immunodeficiency viral infection, contraindications for MRI, including the presence of a cardiac pacemaker or other electronic implant devices, a pregnant status, or any conditions that might affect the patient competence or participation as determined by the opinion of the principal investigator.

A wide array of demographic, lifestyle, anthropometric, and clinical characteristics were collected, including age, sex, body weight, waist circumference (WC) and comorbidities such as hypertension and diabetes. After an overnight fast, blood sample analyses for various markers and routine biochemical tests, which included aspartate aminotransferase (AST), alanine aminotransferase (ALT), glucose, insulin, a complete blood count with a platelet count, alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT), total bilirubin, albumin, glycosylated hemoglobin (HbA1c), and lipid panel, were performed within days of the imaging studies. All samples were originally processed to serum and plasma and stored frozen at -80°C. Commercial kits were used for measurement of plasma angiotensinogen (Human Total Angiotensinogen Assay Kit - IBL; Immuno-Biological Laboratories Co. Ltd., Gunma, Japan) and the serum levels of the apoptosis-associated neo-epitope in the C-terminal domain of CK-18 (the M30-Apoptosense ELISA kit, PEVIVA; Alexis, Grünwald, Germany), complement factors C3 and C4 (turbidimetric immunoassay Tinaquant C3c and C4; Roche Diagnostics Ltd., Rotkreuz, Switzerland), glucagon (Mercodia AB, Uppsala, Sweden) and the ELF test (The ADVIA Centaur Enhanced Liver Fibrosis Test; Siemens Healthcare, Erlangen, Germany). The ELF score was produced by the ADVIA Centaur systems, by measuring the serum hyaluronic acid (HA), amino-terminal propeptide of type III procollagen (PIIINP), and tissue inhibitor of matrix metalloproteinase 1 (TIMP 1) levels as follows: ELF score= 2.278+0.851 ln(HA)+0.751 ln(P3NP)+0.394 ln(TIMP 1), where the concentration values in parentheses are in ng/mL. The body fat and lean body mass were measured using the dual-energy X-ray absorptiometry (DXA) technique (GE Healthcare, Wauwatosa, WI, USA), on the same day as the imaging studies. All patients fasted for at least 5 hours before the imaging studies, as described below. Several clinical indices and scores were calculated based on clinical and laboratory data, including the homoeostatic model assessment of insulin resistance (HOMA-IR), NAFLD fibrosis score (NFS), AST-to-platelet ratio index (APRI), and fibrosis-4 (FIB-4) index [8].

The liver biopsy was performed at the discretion of hepatologists in the hospital if indicated or systematically during bariatric surgery. After standard processing, all specimens were subjected to hematoxylin-and-eosin staining and Masson trichrome staining to assess the liver fibrosis stage. A pathologist (D.H.C.) who was blinded to the patients’ clinical and radiologic results assessed the stained specimens. Histological scoring was performed using the Nonalcoholic Steatohepatitis Clinical Research Network histologic (NASH CRN) scoring system [19]. Fibrosis was staged from F0 to F4 (Supplementary Figs 1 and 2) [19].

TE was performed using FibroScan 502 (Echosens, Paris, France) by a trained technician blinded to the clinical and histological data. All patients were scanned first using the M probe (3.5 MHz) or, when indicated by equipment, using the XL probe (2.5 MHz) at the right lobe of the liver. At least 10 measurements were made to obtain the median valid LSM in kilopascals (kPa) and the interquartile range (IQR). Liver steatosis values were obtained by the CAP measurement in dB/m. Technical failure was defined as no stiffness measurement obtained or unreliable measurements (defined as a success rate <60% or IQR/median >30%) [20].

All MR examinations, including MRE, were performed with a 3-T scanner (MAGNETOM Skyra; Siemens Healthineers, Erlangen, Germany) using an 18-channel body matrix coil and table-mounted 32-channel spine matrix coil. To quantify the liver fat content, we used a multi-echo three-dimensional gradient-echo sequence to obtain MRI-PDFF from a single breath-hold acquisition. Automated calculation and output of a stack with quantitative MRI-PDFF coding were performed and archived to the picture archiving and communication system. If the automated MRI-PDFF calculation failed, then three circular regions of interest (ROIs) were located in segment VII or VIII on the liver for each MRI-PDFF. The mean value for the three ROIs was considered the representative MRI-PDFF

High-speed T2-corrected multiecho single-voxel ¹H-MRS with stimulated echo acquisition mode readout was performed. Experienced MR technicians placed a spectroscopy voxel ROI (30×30×30 mm) in the superior portion of the right hepatic lobe (segment VII or VIII) with a normal appearance while avoiding large hepatic vessels, bile ducts, focal hepatic lesions, and liver margins.

The two-dimensional SE-EPI-based MRE sequence (WorkIn-Progress package; Siemens Healthineers) was performed. For MRE, a cylindrical passive longitudinal pneumatic driver was attached to the right anterior chest wall using a rubber belt with the center of the driver at the level of the xiphoid process. To produce propagating shear waves in the liver, continuous longitudinal 60 Hz mechanical vibrations were used. An inversion algorithm was used to generate elastograms that depicted the shear stiffness measured in kilopascals (kPa). To obtain LSM values for the liver parenchyma, the reviewer (S.J.C., who was blinded to the patients’ clinical or histologic results) drew a geographical ROI that included the largest part of the liver parenchyma (Supplementary Figs. 1 and 2) [21]. The visceral and subcutaneous fat adipose areas were calculated at the L3-L4 discs from the multi-echo three-dimensional gradient-echo sequence.

Categorical variables were compared as counts and percentages and associations were tested using the chi-square or Fisher’s exact test. Continuous variables are reported as the mean± standard deviation, and differences between groups were analyzed using Student’s t-test (two-tailed) or the Mann-Whitney U test as appropriate. Correlations were evaluated using the Spearman correlation coefficient. A two-tailed P value less than 0.05 was considered statistically significant for all analyses. Univariate and multivariate logistic regression analyses to assess for the potential predicting factors of significant liver fibrosis was performed. Characteristics determined to be statistically significant (P<0.05) by univariate analysis were used as input variables for multivariate logistic regression analysis. The diagnostic accuracy was assessed by the standard area under the receiver operating characteristic curve (AUROC) and the weighted AUROC. For each AUROC, 95% confidence intervals were measured using its standard error. The AUROC curves were compared according to DeLong et al. [22]. The AUROC and the Youden index were used to determine the optimal thresholds. Statistical analyses were performed using R software/environment (R-2.9.1; R Foundation for Statistical Computing, Vienna, Austria).

A total of 133 subjects were included in this cross-sectional study, after excluding three patients due to chronic hepatitis B (n=2) and C (n=1). Table 1 presents the characteristics of the study subjects. The mean age of the participants was 39.4 years, with a prevalence of males (46%). Of the 130 TE assessments, 82 (63%) were performed with the M probe and the XL probe was applied for 48 (37%) subjects whose BMI was at least 30 kg/m². Three patients failed TE (failure rate, 2.2% [3/133]).

The MR studies could not be finished in one patient due to unexpected claustrophobia. A MRE-LSM result from another patient was not included due to a disorganized wave pattern on wave images and the patient refused to receive MRE reexamination. A total of 131 patients underwent MRE. Liver biopsy data were available for 54 patients. A total of 16, 23, five, six, and four patients had fibrosis stages 0, 1, 2, 3, and 4, respectively. In the present study, the minimal diagnostic criterion for the diagnosis of NAFLD was defined as the presence of ≥5% hepatic steatosis on MRI-PDFF or NAFLD evidence based on liver biopsy, after the exclusion of secondary causes [23]. Patients with NAFLD had higher BMI, WC, and whole body and abdominal fat measured by DXA and MRI, respectively. They were more likely to have hypertension or type 2 diabetes mellitus than the non-NAFLD group. Additionally, compared with the values of the subjects without NAFLD, the NAFLD group had higher AST, ALT, HbA1c, fasting glucose, fasting insulin, glucagon, HOMA-IR, C3, C4, CK-18, low density lipoprotein cholesterol, triglycerides, and white blood cell and platelet counts in the blood tests (Table 1).

For the evaluation of hepatic steatosis, the MRI-PDFF study was performed in the same session as the ¹H-MRS examination and on the same day or within 1 or 2 days of the TE examination. The MRI-PDFF correlated perfectly with the ¹H-MRS, whereas the TE-CAP showed only a modest correlation with the ¹H-MRS or MRI-PDFF (Fig. 1). Compared with those of the TE-CAP, the MRI-PDFF showed better correlations with fasting circulating AST, ALT, glucose, glucagon, triglycerides, C3, and CK-18 levels as well as the HOMA-IR value (Table 2). However, the TE-CAP showed higher correlations with BMI, WC, DXA total body fat (%), and MRI-measured abdominal fat areas than the MRI-PDFF (Table 2).

The optimal cutoff values of TE-CAP and ALT for the prediction of hepatic steatosis (≥5% on MRI-PDFF) were 264 dB/m and 31 U/L, respectively (Fig. 2). When the upper limits of normal for ALT were defined as 35 and 25 U/L for the male and female subjects, respectively [24], the optimal discriminating cutoff values of MRI-PDFF and TE-CAP for abnormal ALT were 9.9% and 270 dB/m, respectively, with a higher AUROC with MRI-PDFF (Fig. 2).

The MRE-LSM showed significant correlations with the blood AST, ALT, GGT, ALP, and CK-18 levels; the platelet count (negative); and the clinical and ELF scores, whereas the TE-LSM showed poor or weak correlations with those parameters (Table 3, Fig. 1).

Both MRE- and TE-measured LSM values showed a higher correlation with the APRI and FIB-4 index than with the NFS (Table 3). The NFS, FIB-4 index, and APRI scores were categorized using their respective lower and higher cutoff values, which have been suggested for the exclusion or prediction of advanced fibrosis (for the NFS <-1.455 and >0.676; for the FIB-4 index <1.3 and >2.67; and for the APRI <0.5 and >1.5, respectively) [8]. There was less overlap in both the MRE-LSM and TE-LSM values between the low-, intermediate-, and high-risk fibrosis groups based on the FIB-4 index than in those groups based on the NFS or APRI (Supplementary Fig. 3).

When we analyzed liver-biopsied patients (n=54), the MRE, TE and the ELF score showed statistical significance in differentiating significant fibrosis (≥F2) from no or minimal hepatic fibrosis (F0/1). However, the ELF score showed more overlap between the fibrosis grades and wider variation within each fibrosis grade than the MRE- or TE-LSM (Supplementary Fig. 4). In addition, MRE performed better in differentiating advanced fibrosis (≥F3) from lower grade hepatic fibrosis (<F3) than TE and the ELF test (Fig. 3). The cutoff values for advanced fibrosis were 3.9 kPa for MRE-LSM, 8.1 kPa for TE-LSM and 9.2 for the ELF test.

When we analyzed liver-biopsied patients (n=54), the optimal discriminating cutoff value of significant fibrosis (≥F2) based on MRE from those with lower grade fibrosis (<F2) was 3.8 kPa. We dichotomized the subjects between significant fibrosis and non-significant fibrosis groups by utilizing this cutoff value (3.8 kPa) (Supplementary Table 1). Based on this cutoff value, patients with high MRE-LSM (≥3.8 kPa) had higher age and other various unfavorable body and metabolic parameters, and lower platelet count compared to those who represented low MRE-LSM (<3.8 kPa). Factors associated with high MRE-LSM on univariate analysis were age, type 2 diabetes mellitus, AST, ALT, GGT, ALP, platelet count, MRI-PDFF, and MRI-visceral and subcutaneous fat areas. Multivariate analyses showed that age (odds ratio [OR], 1.058; P<0.05), ALP (OR, 1.023; P<0.05), and platelet count (OR, 0.990; P<0.05) were found the potential predictors for significant fibrosis based on the MRE-LSM cutoff value (Supplementary Table 2).