MRE is superior to ARFI for the diagnosis of fibrosis in patients with biopsy-proven NAFLD: A prospective study

Abstract

Magnetic resonance elastography (MRE), an advanced MR-based imaging technique, and acoustic radiation force impulse (ARFI), an ultrasound-based imaging technique, have been shown to be accurate for diagnosing nonalcoholic fatty liver disease (NAFLD) fibrosis. However, no head-to-head comparisons between MRE and ARFI for diagnosing NAFLD fibrosis have been performed. We aimed to compare MRE versus ARFI head-to-head for diagnosing fibrosis in well-characterized patients with biopsy-proven NAFLD.

Methods

This cross-sectional analysis of a prospective cohort involved 125 patients (54.4% female) who underwent MRE, ARFI, and contemporaneous liver biopsies scored using the Nonalcoholic Steatohepatitis Clinical Research Network histological scoring system. MRE versus ARFI’s performances for diagnosing fibrosis were evaluated using area under receiver operating characteristic curves (AUROCs).

Results

The mean (±SD) age and BMI were 48.9 (±15.4) years and 31.8 (±7.0) kg/m², respectively. For diagnosing any fibrosis (≥ stage 1), MRE’s AUROC was 0.799 (95% CI, 0.723–0.875), significantly (p=0.012) higher than ARFI’s AUROC of 0.664 (95% CI, 0.568–0.760). In stratified analysis by presence/absence of obesity, MRE was superior to ARFI for diagnosing any fibrosis in obese patients (p<0.001) but not in non-obese patients (p=0.722). MRE’s AUROCs for diagnosing ≥ stages 2, 3, and 4 fibrosis were 0.885 (95% CI, 0.816–0.953), 0.934 (95% CI, 0.863–1.000), and 0.882 (95% CI, 0.729–1.000), and ARFI’s AUROCs were 0.848 (95% CI, 0.776–0.921), 0.896 (95% CI, 0.824–0.968), and 0.862 (95% CI, 0.721–1.000). MRE had higher AUROCs than ARFI for discriminating dichotomized fibrosis stages at all dichotomization cut-points, but the AUROC differences decreased as the cut-points (fibrosis stages) increased.

Conclusions

MRE is more accurate than ARFI for diagnosing any fibrosis in all NAFLD patients and obese NAFLD patients, although not in non-obese NAFLD patients.

INTRODUCTION

Nonalcoholic fatty liver disease (NAFLD) is a spectrum of liver diseases characterized by hepatic steatosis in patients with no significant alcohol use or secondary causes of hepatic steatosis.(1,2) NAFLD is associated with features of the metabolic syndrome, including obesity, type 2 diabetes mellitus, and dyslipidemia, (1,3–6) and is now a major cause of liver disease worldwide affecting up to one hundred million people in the United States alone.(7) Nonalcoholic steatohepatitis (NASH) is a more severe form of NAFLD characterized by hepatocellular ballooning and inflammation.(8) NASH patients with hepatic fibrosis are at greater risk for disease progression to cirrhosis and its associated complications, including hepatocellular carcinoma, and these patients require greater monitoring and therapy.(9,10) Biopsy is currently the gold standard for diagnosing fibrosis in NAFLD patients, but is invasive, has high inter-observer variability, and is associated with various complications, including death.(11) There is a need for noninvasive tests that can diagnose fibrosis in NAFLD patients.

There are no noninvasive tests currently approved to diagnose fibrosis in NAFLD patients. Magnetic resonance elastography (MRE) is a magnetic resonance-based imaging technique shown to accurately diagnose fibrosis in NAFLD patients, although it is associated with high costs and is currently not widely available.(12–15) Ultrasound-based imaging techniques, including transient elastography (FibroScan) (16) and acoustic radiation force impulse (ARFI) elastography, (17) are cheaper and more widely available than MRE. However, ultrasound-based imaging techniques are reported to be operator-dependent and their accuracy may be less than that of MRE, especially in obese patients.(18,19) While ARFI has been shown to be accurate for diagnosing fibrosis in a variety of liver diseases, including NAFLD, (20–22) no head-to-head comparisons between MRE and ARFI have been performed to evaluate their accuracy for diagnosing fibrosis in a prospective cohort of patients with biopsy-proven NAFLD.

Using a well-characterized, prospective cohort of patients with biopsy-proven NAFLD, we performed comparisons of the accuracy of MRE and ARFI for diagnosing fibrosis in NAFLD patients. Additionally, we compared the accuracy of MRE and ARFI for diagnosing NAFLD fibrosis in sub-groups of obese and non-obese patients.

METHODS

Study Design

This was a cross-sectional analysis of a prospectively-recruited cohort of 125 patients with biopsy-proven NAFLD who underwent MRE and ARFI examinations. All patients had a clinical indication for a liver biopsy and were enrolled in the study after a written informed consent was obtained. Patients were carefully screened and excluded for other causes of liver diseases and secondary causes of hepatic steatosis. Enrolled patients were invited for a clinical research visit and they underwent standardized history, anthropometric exam, physical exam, and biochemical testing at the University of California at San Diego (UCSD) NAFLD Translational Research Unit (23–29) and MRE and ARFI exams at the UCSD MR3T Research Laboratory. Written informed consent was obtained from all patients prior to the study, and the study was approved by the UCSD Institutional Review Board and the UCSD Clinical and Translational Research Institute.

Inclusion/Exclusion Criteria

Patients were included if they were adults ≥18 years old with biopsy-proven NAFLD and provided written informed consent. Patients were excluded if they had regular or excessive alcohol use within two years of recruitment (≥14 drinks/week for men or ≥7 drinks/week for women); use of hepatotoxic drugs or drugs known to cause hepatic steatosis; clinical or laboratory evidence of secondary causes of NAFLD, including nutritional and iatrogenic gastrointestinal disorders and human immunodeficiency virus infection; clinical or laboratory evidence of liver diseases other than NAFLD, including hepatitis B (if positive serum hepatitis B surface antigen), hepatitis C (if positive hepatitis C viral RNA), autoimmune hepatitis, alpha-1 antitrypsin deficiency, Wilson’s disease, hemochromatosis, glycogen storage diseases, and cholestatic or vascular liver diseases; clinical or biochemical evidence of decompensated liver disease (Child-Pugh ≥7 points); active substance abuse; significant systemic illnesses; contraindication(s) to magnetic resonance imaging, including claustrophobia and the presence of metal implants; pregnant status or trying to become pregnant; or any conditions which, in the opinion of the principal investigator, may affect the patient’s competence, compliance, or study participation.

Histological Assessment

Definition of NAFLD: All patients were confirmed to have biopsy-proven NAFLD as defined by the NAFLD practice guidelines. Other causes of secondary NAFLD were excluded as explained in the inclusion and exclusion criteria. All patients underwent a liver biopsy, which was scored using the Nonalcoholic Steatohepatitis Clinical Research Network (NASH CRN) histological scoring system (30) by an experienced hepatopathologist blinded to the patients’ clinical and radiologic data. The histological scoring system is further described in the Supporting Information.

Outcome Measures

The primary outcome was fibrosis (stage 1–4 fibrosis vs. stage 0 fibrosis). Secondary outcomes included dichotomized stages of fibrosis: stage 2–4 vs. stage 0–1 fibrosis, stage 3–4 (advanced fibrosis) vs. stage 0–2 fibrosis, and stage 4 (cirrhosis) vs. stage 0–3 fibrosis.

Magnetic Resonance Elastography

MRE was performed with commercially available software and hardware (Resoundant Inc., Rochester, MN) using previously described techniques.(12,13,31–33) Briefly, shear waves at 60 Hz were generated in the liver by an acoustic passive driver attached to the body wall anterior to the liver and coupled to an acoustic active driver outside of the MRI room. The shear waves were imaged using a 2D gradient-recalled-echo MRE pulse sequence with oscillating motion-sensitizing gradients synchronized to the shear wave frequency. Four noncontiguous axial slices (10 mm thick with 10 mm interslice gap) were acquired, each in a separate 16-second breath hold, through the widest transverse dimension of the liver with the following acquisition parameters: repetition time (TR), 50 ms; echo time (TE), 20.2 ms; flip angle (FA), 30 degrees; matrix, 256 × 64; field of view (FOV), 48 × 48 cm; one-signal average; receiver bandwidth (BW) ± kHz (confirm); and parallel imaging accelerating factor, 2. The total acquisition time was approximately two minutes for the four breath-holds with short recovery times in between.

The wave images generated at each slice location were automatically processed on the scanner computer using specialized software and algorithm (called an inversion algorithm) to generate quantitative cross-sectional maps (called elastograms) depicting tissue stiffness. The elastograms were transferred offline for analysis (34,35) by a trained imaging analyst (at least six months of experience with MRE) in the MR3T research laboratory. The imaging analyst manually drew regions of interest on the elastograms using a custom software package in portions of the liver where the corresponding wave images showed clearly observable wave propagation, and liver edges, large blood vessels, and artifacts were avoided. The mean liver stiffness, calculated by averaging the per-pixel stiffness values across regions of interests at the four slice locations, were automatically outputted to an electronic spreadsheet.

Acoustic Radiation Force Impulse

ARFI (36,37) was performed using the Siemens Acuson S3000 scanner (Siemens AG, Mountain View, CA) by a research physician (E.H.) blinded to clinical and histology data, using the 4C1 curved vector array transducer (1- 4.5 MHz). Subjects were asked to fast 4 hours prior to their ultrasound examinations. Scanning was done in dorsal decubitus position with the right arm fully abducted, through a right intercostal approach with the transducer placed at ~90° to the skin surface and the patient instructed to suspend breathing. The region of interest (ROI) was placed in a homogenous area, away from major vasculature in the right lobe of the liver 2–5 cm deep from the proximal liver capsule. 11 valid shear wave speed (SWS) acquisitions were acquired for each subject, with one acquisition per breath hold. For each acquisition, if the proscribed measurement quality threshold is not achieved by the ultrasound scanner, it shows an error message, and no SWS value is displayed and the acquisition is considered invalid. ARFI was considered successful if at least seven measurements were obtained from a patient, and subjects with fewer than seven successful measurements were considered to have failed ARFI exams. A monitor screen shot of the ROI location, depth, and SWS value was saved for each acquisition (Supplementary Figure 1). SWS values and ROIs depth were manually recorded to our database for all subjects, and the mean SWS was then calculated for each subject.

Duration between MRE and Liver Biopsy

The median time interval between MRE and biopsy was 46.5 (interquartile range: 62) days.

Duration between ARFI and Liver Biopsy

The median time interval between ARFI and biopsy was 47.5 (interquartile range: 87) days.

Duration between MRE and ARFI

The median time interval between MRE and ARFI was 0 (interquartile range: 1.5) days with the majority of MRE and ARFI done on the same day.

Clinical Research Assessment

All patients underwent a clinical research assessment at the UCSD NAFLD Translational Research Unit, which included standardized history, anthropometric exam including body mass index (BMI), physical exam, and biochemical testing (see Supporting Information for details).

Statistical Analyses

All statistical analyses was performed using SPSS version 22.0 and “R” statistical computing software version 3.1.1 (Vienna, Austria).(38) Patients’ demographic, laboratory, histological, and imaging data were summarized with mean and standard deviation for continuous variables and numbers and percentages for categorical variables. A two-tailed p-value ≤ 0.05 was considered statistically significant. Based on previously published results for MRE (13,14) and ARFI, (21) we anticipated a difference of about 0.19 between the AUROCs of MRE and ARFI for diagnosing fibrosis. A sample of at least 118 patients with MRE and ARFI was needed to achieve a power of 90% with an alpha value of 0.05.

Main Analysis

Receiver operating characteristic (ROC) curve analysis was performed for both MRE and ARFI as classifiers of fibrosis (stage 1–4 fibrosis) using biopsy as the gold standard. The overall performance of MRE and ARFI for diagnosing fibrosis was evaluated using the area under ROC curve (AUROC). AUROCs of MRE and ARFI for diagnosing fibrosis were also separately calculated for obese (BMI ≥ 30.0) and non-obese patients. 95% confidence intervals were calculated for each AUROC using its standard error. The optimal thresholds of MRE and ARFI for diagnosing fibrosis was selected using the Youden Index, (39) and the following performance parameters were calculated: sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). The DeLong test (40) was used to perform head-to-head comparisons of accuracy of MRE vs. ARFI for diagnosing biopsy-proven fibrosis.

Secondary Analyses

Additional ROC analyses were performed for the overall cohort and separately in obese and non-obese patients and their corresponding AUROCs, diagnostic thresholds, and performance parameters were calculated for the accuracy of MRE and ARFI for diagnosing other dichotomized stages of fibrosis (stage 2–4 vs. stage 0–1, stage 3–4 vs. stage 0–2, and stage 4 vs. stage 0–3). The DeLong test was used to perform head-to-head comparisons between MRE and ARFI for diagnosing each clinical outcome.

RESULTS

Baseline Characteristics

125 consecutive patients (54.4% female) with biopsy-proven NAFLD, MRE, and ARFI were prospectively enrolled between December 2013 and March 2015. The mean (±SD) age and BMI were 48.9 (±15.4) and 31.8 (±7.0), respectively. 71 patients were obese (BMI ≥30.0) and 54 patients were not obese (BMI <30.0). Patients’ baseline demographic, biochemical, histological, and imaging data are summarized in Table 1. A total of 184 patients with biopsy-proven NAFLD were initially evaluated at the UCSD NAFLD Translational Research Unit, although 16 patients did not undergo ARFI due to patient schedule, 11 patients did not undergo MRE due to patient schedule, 28 patients did not undergo both ARFI and MRE due to patient schedule, three patients failed ARFI, and one patient failed MRE (Supplementary Figure 2).

Table 1.

Demographic, biochemical, and histological characteristics of subjects with biopsy-proven NAFLD

	Patients with biopsy, 2D-MRE, and ARFI (n = 125)
Demographic
Female patients (%)	68 (54.4%)
Age (years) at biopsy (SD)	48.9 (15.4)
Height (cm) mean (SD)	167.7 (9.5)
Weight (kg) mean (SD)	89.8 (24.0)
BMI (kg/m2) mean (SD)	31.8 (7.0)
Ethnic origin:
White (%)	61 (48.8%)
Black (%)	0 (0%)
Asian (%)	22 (17.6%)
Hispanic (%)	37 (29.6%)
Multi-racial (%)	1 (0.8%)
Other (%)	1 (0.8%)
Refused to disclose	3 (2.4%)
Diabetes (%)	32 (26.0%)
Metabolic syndrome (%)	31 (24.8%)
Biochemical Profile
AST U/L mean (SD)	41.9 (35.0)
ALT U/L mean (SD)	56.4 (51.8)
AST/ALT ratio mean (SD)	0.87 (0.63)
Alk Phos U/L mean (SD)	75.1 (32.0)
GGT U/L mean (SD)	62.1 (70.3)
Total Bilirubin mg/dL mean (SD)	0.56 (0.40)
Direct Bilirubin mg/dL mean (SD)	0.14 (0.08)
Albumin g/dL mean (SD)	4.8 (3.5)
Glucose mg/dL mean (SD)	106.4 (31.6)
Hgb A1c mean (SD)	6.1 (0.9)
Triglycerides mg/dL mean (SD)	157.2 (80.8)
Total Cholesterol mg/dL mean (SD)	185.0 (38.9)
HDL mg/dL mean (SD)	50.1 (19.4)
LDL mg/dL mean (SD)	106.0 (32.6)
Platelet count 109/L mean (SD)	250.1 (70.4)
Protime mean (SD)	10.7 (0.9)
INR mean (SD)	1.0 (0.1)
Histology
Steatosis
0	4 (3.3%)
1	41 (33.3%)
2	39 (31.7%)
3	39 (31.7%)
Lobular Inflammation
0	4 (3.3%)
1	57 (46.7%)
2	60 (49.2%)
3	1 (0.8%)
Ballooning
0	32 (26.7%)
1	70 (58.3%)
2	18 (15.0%)
Fibrosis
0	53 (42.4%)
1	39 (31.2%)
2	12 (9.6%)
3	12 (9.6%)
4	9 (7.2%)
NAS mean (SD)	4.29 (1.54)
Imaging
2D-MRE and biopsy date interval median (IQR)	46.5 (62)
ARFI and biopsy date interval median (IQR)	47.5 (87)
2D-MRE mean (SD)	3.13 (1.27)
ARFI mean (SD)	1.48 (0.68)

Metabolic syndrome was calculated using the International Diabetes Federation criteria, which included central obesity and at least two of the following: elevated triglycerides, reduced HDL-cholesterol, elevated blood pressure, and elevated fasting plasma glucose.

Abbreviations: BMI: body mass index, AST: aspartate aminotransferase, ALT alanine aminotransferase, Alk phos: alkaline phosphatase, GGT: gamma-glutamyl transpeptidase, Hgb A1c: hemoglobin A1c, HDL: high-density lipoprotein, LDL: low-density lipoprotein, INR: international normalized ratio, NAS: nonalcoholic fatty liver disease activity score, ARFI: acoustic radiation force impulse, 2D-MRE: 2-dimensional magnetic resonance elastography, ARFI: acoustic radiation force impulse

Distribution of Fibrosis Stages

53, 39, 12, 12, and 9 patients had stage 0, 1, 2, 3, and 4 fibrosis, respectively.

Accuracy of MRE vs. ARFI for the Diagnosis of Fibrosis

MRE had an AUROC of 0.799 (95% CI: 0.723–0.875) for diagnosing fibrosis (Figure 1). At a threshold of 2.99 kPa (Figure 2A), MRE had sensitivity 0.583, specificity 0.906, PPV 0.894, and NPV 0.615 for diagnosing fibrosis. ARFI had an AUROC of 0.664 (95% CI: 0.568–0.760) for diagnosing fibrosis. At a threshold of 1.29 meters/second (Figure 2B), ARFI had sensitivity 0.542, specificity 0.774, PPV 0.765, and NPV 0.554 for diagnosing fibrosis. In a head-to-head comparison using the DeLong test, MRE had significantly higher accuracy than ARFI (p=0.012) for diagnosing fibrosis (Table 2).

Figure 1.

MRE had an AUROC of 0.799 for diagnosing fibrosis and ARFI had an AUROC of 0.664 for diagnosing fibrosis.

Figure 2.

Figure 2A–2B: Distribution of MRE (Figure 2A) and ARFI (Figure 2B) readings by fibrosis stage. The dichotomization cut points for any fibrosis was 2.99 for MRE and 1.29 for ARFI.

Table 2.

Diagnostic test characteristics of 2D-MRE and ARFI for the diagnosis of fibrosis

		N Positive	N Negative	AUROC (95% Confidence Interval)	Cutoff	Sensitivity	Specificity	PPV	NPV	MRE vs. ARFI (p-value)
Primary Analysis
Stage 1–4 vs. Stage 0	MRE	72	53	0.799 (0.723, 0.875)	2.99	0.583	0.906	0.894	0.615	0.012
	ARFI			0.664 (0.568, 0.760)	1.29	0.542	0.774	0.765	0.554
Secondary Analyses
Stage 2–4 vs. Stage 0–1	MRE	33	92	0.885 (0.816, 0.953)	3.62	0.667	0.957	0.846	0.889	0.343
	ARFI			0.848 (0.776, 0.921)	1.34	0.818	0.783	0.574	0.923
Stage 3–4 vs. Stage 0–2	MRE	21	104	0.934 (0.863, 1.000)	3.62	0.905	0.933	0.731	0.980	0.113
	ARFI			0.896 (0.824, 0.968)	1.34	0.952	0.740	0.426	0.987
Stage 4 vs. Stage 0–3	MRE	9	116	0.882 (0.729, 1.000)	4.15	0.889	0.914	0.444	0.991	0.418
	ARFI			0.862 (0.721, 1.000)	2.48	0.778	0.931	0.467	0.982

p-value: AUROC of MRE vs. ARFI → DeLong test

Abbreviations: ARFI: acoustic radiation force impulse, MRE: magnetic resonance elastography, AUROC: area under receiver operating characteristic curve

For MRE, the AUROC for diagnosing fibrosis in obese (n=71) patients was 0.850 (95% CI: 0.761–0.939) and in non-obese patients (n=54) was 0.729 (95% CI: 0.586–0.872). For ARFI, the AUROC for diagnosing fibrosis in obese patients was 0.603 (95% CI: 0.457–0.749) and in non-obese patients was 0.694 (95% CI: 0.537–0.850). MRE had significantly higher accuracy than ARFI for diagnosing fibrosis in obese patients (p<0.001) but did not have significantly higher accuracy than ARFI for diagnosing fibrosis in non-obese patients (p=0.722) (Table 3).

Table 3.

AUROCs and AUROC comparisons in obese and non-obese patients

		Obese (n=71)				Non-Obese (n=54)
		N Positive	N Negative	AUROC (95% Confidence Interval)	MRE vs. ARFI (p-value)	N Positive	N Negative	AUROC (95% Confidence Interval)	MRE vs. ARFI (p-value)
Primary Analysis
Stage 1–4 vs. Stage 0	MRE	50	21	0.850 (0.761, 0.939)	<0.001	22	32	0.729 (0.586, 0.872)	0.722
	ARFI			0.603 (0.457, 0.749)				0.694 (0.537, 0.850)
Secondary Analyses
Stage 2–4 vs. Stage 0–1	MRE	24	47	0.885 (0.796, 0.974)	0.274	9	45	0.867 (0.753, 0.980)	1.000
	ARFI			0.829 (0.730, 0.928)				0.867 (0.762, 0.971)
Stage 3–4 vs. Stage 0–2	MRE	17	54	0.917 (0.823, 1.000)	0.235	4	50	0.955 (0.871, 1.000)	0.607
	ARFI			0.869 (0.772, 0.966)				0.935 (0.861, 1.000)
Stage 4 vs. Stage 0–3	MRE	6	65	0.821 (0.582, 1.000)	1.000	3	51	1.000 (1.000, 1.000)	0.185
	ARFI			0.821 (0.612, 1.000)				0.961 (0.905, 1.000)

p-value: AUROC of MRE vs. ARFI → DeLong test

Abbreviations: ARFI: acoustic radiation force impulse, MRE: magnetic resonance elastography, AUROC: area under receiver operating characteristic curve

27 patients in our cohort had class II obesity (BMI 35.0–39.9 kg/m²), with 20 patients in the class II obesity group having fibrosis and 7 patients having no fibrosis. For diagnosing any fibrosis in patients with class II obesity, the AUROC of MRE remained robust at 0.914 (95% CI: 0.805–1.000), whereas the AUROC of ARFI declined further to 0.536 (95% CI: 0.262–0.810). MRE had significantly higher accuracy than ARFI for diagnosing fibrosis in patients with class II obesity (p=0.007).

Accuracy of MRE vs. ARFI for the Diagnosis of other Stages of Fibrosis

The AUROCs of MRE and ARFI for diagnosing stage 2–4 vs. stage 0–1 fibrosis, stage 3–4 vs. stage 0–2 fibrosis, and stage 4 vs. stage 0–3 are shown in Table 2. MRE had higher AUROCs than ARFI at each dichotomization cutpoint, although the differences in AUROCs decreased as the dichotomization cutpoint (i.e., the stage of fibrosis) increased. Additionally, no significant difference existed between MRE and ARFI between obese and non-obese patients for diagnosing other stages of fibrosis (Table 3).

DISCUSSION

Main Findings

Utilizing a prospective cohort of patients with biopsy-proven NAFLD, our study showed that MRE was significantly more accurate than ARFI for diagnosing fibrosis in NAFLD patients. The difference in diagnostic accuracy between MRE and ARFI decreased as the dichotomization cutpoint increased, and MRE and ARFI were both highly accurate for diagnosing advanced fibrosis (stage 3–4 fibrosis) and cirrhosis (stage 4 fibrosis). Apart from any fibrosis (stage 1–4 fibrosis), there were no significant differences between MRE and ARFI for diagnosing other stages of fibrosis at higher dichotomization cutpoints. Additionally, MRE was significantly more accurate than ARFI for diagnosing fibrosis in obese, but not non-obese, patients. In obese individuals, MRE should be considered for diagnosis of fibrosis in patients with NAFLD. However, in non-obese individuals ARFI measurements may provide adequate diagnostic accuracy. These findings may have important clinical implications in non-invasive assessment of NAFLD and developing an optimal strategy for resource utilization. Cost-effectiveness studies are needed to determine the optimal strategy but we believe that an approach using an MRE in those with obesity and ARFI in non-obese may be applicable if these findings are confirmed in studies published by independent investigators or in a multicenter setting. Additionally, MRE may be preferable to ARFI if diagnosing the earliest stages of fibrosis is needed, although ARFI may be sufficient in patients with higher risks of advanced fibrosis, even if obese.

In Context of Published Literature

This is the first study to make head-to-head comparisons between the accuracy of MRE vs. ARFI for diagnosing liver fibrosis in a large, prospective cohort of NAFLD patients with strict inclusion and exclusion criteria and liver biopsies within one year of ARFI.(41,42) The results of our study are consistent with those of previous studies showing MRE to be highly accurate for diagnosing fibrosis in NAFLD patients.(12–14) Our study is also consistent with previous studies showing ARFI to be accurate for diagnosing fibrosis, especially at higher dichotomization cut-points.(17,20–22) The accuracy of ARFI becomes less for diagnosing fibrosis at lower dichotomization cut-points.(21)

In a recent study by Angulo et al of NAFLD patients, any fibrosis (stage 1–4 fibrosis) was shown to be a significant, independent predictor of increased mortality or liver transplantation compared to no fibrosis.(43,44) Additionally, any fibrosis was associated with increased mortality or liver transplantation compared to no fibrosis even when individuals with more severe stage 3–4 fibrosis were removed from the analysis.(43) Therefore, early detection and screening of any fibrosis before its progression into more severe stage 3–4 fibrosis may be beneficial in NAFLD patients. These patients may require increased monitoring and management, and may also be candidates for treatment trials targeting NAFLD patients. Our study shows that MRE is superior to ARFI for identifying these NAFLD patients with any fibrosis who are at increased risk for mortality and liver transplantation.

While ultrasound-based imaging techniques for detecting hepatic fibrosis may be less accurate in obese patients, earlier studies have suggested that ARFI may still be feasible in obese NAFLD patients with good accuracy.(45) Our study shows that MRE was significantly more accurate than ARFI for diagnosing fibrosis in obese patients, but not in non-obese patients. Our study also shows that for patients with class II obesity (BMI 35.0–39.9 kg/m²), the AUROC of MRE remained robust at 0.914 whereas the AUROC of ARFI declined further to 0.536, compared to 0.603 for ARFI in the overall cohort of obese patients (BMI ≥ 30.0). As the population becomes more obese, the difference between the AUROCs of MRE versus ARFI for diagnosing fibrosis widens, with a relative increase in the diagnostic accuracy of MRE and a relative decrease in the diagnostic accuracy of ARFI. Therefore, obesity and body habitus may negatively affect the accuracy of ARFI for diagnosing fibrosis in NAFLD patients. Additionally, three patients in our study, all of whom were obese, failed their ARFI examinations due to difficulty in obtaining accurate reads, and obesity may have played a role in the failure to obtain accurate readings in these patients. Only one patient in our study failed MRE due to poor wave progression from an iron-overloaded liver. Obesity appears to have less of an effect on the accuracy of MRE than on the accuracy of ARFI.

Strengths and Limitations

The strength of this study lies in its use of a well-characterized, prospective cohort of patients with biopsy-proven NAFLD in which all patients had a clinical indication for a liver biopsy and underwent a paired MRE and ARFI evaluation. Liver biopsy assessments were done using the NASH CRN Histologic Scoring System. All patients underwent a dedicated research visit at the UCSD NAFLD Translational Research Unit, and other causes of liver diseases were systematically ruled out before enrollment into the study. The study was performed by an experienced group of investigators at a center that is highly specialized for NAFLD research, including imaging with MRE and ARFI. The median time interval between MRE and biopsy was only 46.5 days and that between ARFI and biopsy was only 47.5 days.

However, there are also a few limitations to this study. It was performed at a center highly specialized for both clinical and radiological NAFLD research, and the generalizability of its results to other settings requires validation in a multicenter setting. The study is cross-sectional, and the utility of MRE and ARFI for monitoring longitudinal changes in fibrosis progression remains unknown. The diagnostic accuracy of ARFI is reported to be operator dependent and may be subject to inter- and intra-operator variability, although at our center all ARFI was performed by a single experienced investigator. MRI-based techniques, including MRE, may be expensive to perform, even though at our center the price of an MRE is less than that of a liver biopsy without the morbidity associated with liver biopsies. Although MRE is more accurate than ARFI, it is more expensive to perform and has limited availability compared to ARFI. Further research is needed to evaluate the cost-effectiveness of MRE vs. ARFI for diagnosing NAFLD-associated fibrosis before utilizing these competing strategies for the non-invasive assessment of NAFLD in routine clinical practice.

Implication for Future Research

This study provides a prospective, head-to-head comparison of MRE vs. ARFI for diagnosing fibrosis in a cohort of patients with biopsy-proven NAFLD. We show MRE to have significantly higher accuracy than ARFI for diagnosing fibrosis in NAFLD patients, and this difference exists in obese patients but not in non-obese patients. Therefore, while both MRE and ARFI may be accurate for diagnosing fibrosis in non-obese NAFLD patients, MRE may be more preferable than ARFI for diagnosing fibrosis in obese NAFLD patients. Future studies will be needed to assess the utility of MRE and ARFI for diagnosing fibrosis in multicenter, longitudinal settings, the cost-effectiveness of MRE vs. ARFI for diagnosing fibrosis in both obese and non-obese NAFLD patients, and the cost-effectiveness of using MRE to diagnosis early stages of fibrosis and ARFI to diagnose later stages of fibrosis, even in obese patients. Additional studies are also needed to evaluate the diagnostic utility of more novel imaging techniques, including three-dimensional MRE (3D-MRE), for diagnosing NAFLD-associated fibrosis.

Supplementary Material

Supp infoS1

Supplementary Figure 1: ARFI screen save showing a B-mode image with region of interest (ROI) placed in the right lobe of the liver. The shear wave speed was 1.26 meters/second at a depth of 5.0 cm.

Supplementary Figure 2: Derivation of cohort.

Abbreviations

NAFLD

nonalcoholic fatty liver disease

NASH

nonalcoholic steatohepatitis

MRE

magnetic resonance elastography

ARFI

acoustic radiation force impulse

UCSD

University of California at San Diego

CRN

Clinical Research Network

TR

repetition time

TE

echo time

FA

flip angle

FOV

field of view

BW

bandwidth

ROI

region of interest

SWS

shear wave speed

BMI

body mass index

ROC

receiver operating characteristic

AUROC

area under receiver operating characteristic

PPV

positive predictive value

NPV

negative predictive value