Evaluation and comparison of six noninvasive tests for prediction of significant or advanced fibrosis in nonalcoholic fatty liver disease

Abstract

Background

In nonalcoholic fatty liver disease (NAFLD), advanced fibrosis has been identified as an important prognostic factor with increased liver-related mortality and treatment need. Due to the high prevalence of NAFLD, noninvasive risk stratification is needed to select patients for liver biopsy and treatment.

Objective

To compare the diagnostic accuracy of several widely available noninvasive tests for assessment of fibrosis among patients with NAFLD with or without nonalcoholic steatohepatitis (NASH).

Methods

We enrolled consecutive patients with NAFLD admitted to two Austrian referral centers who underwent liver biopsy. Liver stiffness measurement (LSM) was obtained by vibration-controlled transient elastography (VCTE, FibroScan) and blood samples were collected for determination of enhanced liver fibrosis (ELF) test, FibroMeter^V2G, FibroMeter^V3G, NAFLD fibrosis score (NFS), and fibrosis-4 index (FIB-4).

Results

Our study cohort contained 186 patients with histologically confirmed NAFLD. On liver histology, NASH was present in 92 patients (50%), significant fibrosis (F ≥ 2) in 71 patients (38%), advanced fibrosis (F ≥ 3) in 49 patients (26%), and F ≥ 3 plus NASH in 35 patients (19%). For diagnosis of F ≥ 2, F ≥ 3, and F ≥ 3 plus NASH, respectively, receiver operating characteristic (ROC) analysis revealed superior diagnostic accuracy of ELF score (area under ROC curve (AUROC) 0.85, 0.90, 0.90), FibroMeter^V2G (AUROC 0.86, 0.88, 0.89), FibroMeter^V3G (AUROC 0.84, 0.88, 0.88), and LSM per protocol (AUROC 0.87, 0.95, 0.91) versus FIB-4 (AUROC 0.80, 0.82, 0.81) or NFS (AUROC 0.78, 0.80, 0.79).

Conclusion

Proprietary fibrosis panels and VCTE show superior diagnostic accuracy for noninvasive diagnosis of fibrosis stage in NAFLD as compared to FIB-4 and NFS.

Keywords

Enhanced liver fibrosis score FibroMeter vibration-controlled transient elastography liver stiffness measurement fibrosis-4 index NAFLD fibrosis score

Introduction

Nonalcoholic fatty liver disease (NAFLD) is highly prevalent in Western countries, closely linked to obesity and the metabolic syndrome, and represents an increasing cause for chronic liver failure, hepatocellular carcinoma, and need for liver transplantation. Within NAFLD, two stages are discerned: (a) nonalcoholic fatty liver (NAFL), which shows low liver-related morbidity, and (b) nonalcoholic steatohepatitis (NASH), which has a higher risk for progressive hepatic fibrosis and shows substantial liver-related mortality.¹

Previous studies have identified advanced fibrosis (F ≥ 3) as the major prognostic factor in NAFLD.^2–4 Besides lifestyle modification, there is currently no accepted treatment for NAFLD, but clinical trials of novel pharmacologic agents are focusing on patients with NASH and advanced fibrosis. Presence of NASH and advanced fibrosis is currently best determined by liver biopsy. However, due to the high prevalence of NAFLD, universal liver biopsy is not feasible and noninvasive triage tests are needed either as an alternative or to select patients for liver biopsy.^5–7

Enhanced liver fibrosis (ELF™) test, a proprietary fibrosis panel based on extracellular matrix proteins containing hyaluronic acid (HA), procollagen-3 N-terminal peptide (P3NP), and tissue inhibitor of metalloproteinase-1 (TIMP-1), has been found useful for staging of NAFLD in adult as well as pediatric populations,^8–11 with cut-offs varying between 9.8 and 10.5 for the diagnosis of advanced fibrosis.^12–14 Additional proprietary fibrosis panels that have been validated in various chronic liver diseases include FibroMeter^V2G (based on platelet count, prothrombin index, aspartate transaminase (AST), alpha-2-macroglobulin, HA, urea, age, and sex),^15,16 FibroMeter^V3G (using gamma-glutamyl transferase (GGT) instead of HA),¹⁷ and FibroTest (based on alpha-2-macroglobulin, haptoglobin, GGT, age, bilirubin, apolipoprotein A1, and sex).¹⁸ On the other hand, simple fibrosis tests based on routine clinical and laboratory parameters might provide similar diagnostic information at much lower cost. Among these, NAFLD fibrosis score (NFS) has been developed specifically for NAFLD considering the parameters age, hyperglycemia, body mass index (BMI), platelet count, albumin, and AST/alanine transaminase (ALT) ratio,¹⁹ while FIB-4 has been derived from a HCV-HIV co-infected cohort and includes age, AST, ALT, and platelet count.²⁰

Vibration-controlled transient elastography (VCTE) now enables accurate staging in a variety of liver diseases, including NAFLD.^21–24 However, this method is limited to referral centers due to high equipment cost and has a substantial failure rate, especially in obese patients. Besides, VCTE results may be influenced by the type of probe used, by BMI, and by hepatic fat content.^25–27

The aim of the present study was to prospectively evaluate and compare the diagnostic accuracy of several blood fibrosis tests including ELF test and the simple tests, NFS and FIB-4, as well as VCTE for detection of fibrosis stage among patients with NAFLD with or without NASH. In addition, we performed a post-hoc comparison with FibroMeter^V2G and FibroMeter^V3G.

Patients, materials, and methods

We conducted a prospective, biopsy-controlled, two-center study to compare the diagnostic accuracy of available noninvasive tools for the detection of F ≥ 2, F ≥ 3, and the combined endpoint of F ≥ 3 plus the presence of NASH. The cohort was part of a larger Austrian study aiming at biomarker development in metabolic syndrome, BioPersMed (Biomarkers for Personalized Medicine), approved by the Ethics Committee of the Medical University of Graz (EC 24-224 ex 11/12) and the Medical University of Vienna (EC #747/2011). The study was performed in accordance with the Declaration of Helsinki including current revisions after written informed consent was obtained. This report follows the Liver-FibroSTARD checklist.²⁸

Patients

Between 2011 and 2016, we enrolled consecutive patients with NAFLD at two Austrian tertiary referral centers (Medical University of Graz and Medical University of Vienna) who had been referred for evaluation of suspected NAFLD based on elevated liver function tests, fatty liver on ultrasound, and/or presence of metabolic risk factors with or without diabetes. Selection criteria were presence of steatosis >5% on liver histology, history of alcohol consumption <30 g/day for men and <20 g/day for women, absence of concomitant viral hepatitis, and absence of other liver disease on histology. All patients underwent liver biopsy, mainly for staging purposes and less often for differential diagnosis (see flowchart in Figure 1). Blood fibrosis tests were obtained on the same day prior to liver biopsy while VCTE was performed with a median interval of two months prior to liver biopsy.

Figure 1.

Study flowchart.

Histology

Histological NASH was defined by the minimal criteria for steatohepatitis in adults, i.e. presence of >5% macrovesicular steatosis, lobular inflammation, and hepatocellular ballooning, typically with a predominantly centrilobular distribution.²⁹ Fibrosis was staged according to the Clinical Research Network (CRN) score.³⁰ All liver biopsies were assessed by either of two experienced specialized hepatopathologists (CL, JS) who were unaware of the results of noninvasive fibrosis tests. Only biopsies with ≥15 mm length were entered into the database.

Noninvasive fibrosis tests

VCTE was performed in patients fasted for at least two hours using a FibroScan® 502 Touch (Echosens, Paris, France) at both centers. Initially all patients were examined using the M-probe; the XL-probe was additionally available at Graz from 2012 and at Vienna from 2014 onwards and was used if probe-to-capsule distance was >25 mm as recommended by the manufacturer. Liver stiffness measurement (LSM) values were considered unreliable if IQR/med was >30% for values above 7.0 kPa.²⁸ For diagnosis of significant or advanced fibrosis, we applied the recently published cut-off values of 8.2 and 9.7 kPa, respectively, based on a large UK study in 383 NAFLD patients.²⁴

Serum samples obtained at the time of liver biopsy were used to perform ELF™ test on an Advia Centaur XP (Siemens Healthcare Diagnostics, Vienna, Austria). According to the manufacturer, coefficient of variation ranges from 5.5% to 7.7% for HA, from 2.8% to 6.6% for P3NP, and from 3.1% to 6.0% for TIMP-1. For diagnosis of significant or advanced fibrosis, the published ELF score cut-offs of 7.7 and 9.8, respectively, were applied.¹²

NFS and FIB-4 were calculated from routine clinical and laboratory parameters obtained at the time of liver biopsy and published cut-offs were used for diagnosis of advanced fibrosis.^19,20,31,32 In addition, following previous reports on the influence of age on FIB-4 and of BMI on LSM, we analyzed the effect of age, BMI, and presence of diabetes/IFG on the results of the noninvasive fibrosis tests evaluated in our study.^33–35

Post-hoc, we obtained the parameters necessary for calculation of FibroMeter^V2G and FibroMeter^V3G (Echosens, Paris, France). Calculation of the fibrometers was kindly provided by Prof. Paul Cales, Angers, France. For FibroMeter^V2G/V3G, cut-offs were derived by Youden index within our cohort, i.e. 0.309/0.378 for F ≥ 2 and 0.385/0.461 for F ≥ 3.

All persons who performed noninvasive fibrosis tests were unaware of the results of liver biopsy.

Statistical analysis

Summary data are reported as median (Q1, Q3) or frequencies. Chi-square test or Mann–Whitney U test was performed for group comparisons. The diagnostic accuracy of noninvasive fibrosis tests for significant or advanced fibrosis was assessed by receiver operating characteristic (ROC) analysis. Areas under the ROC curves (AUROCs) were compared by the method of DeLong et al.³⁶ For assessing the diagnostic accuracy of VCTE, we performed a per protocol (PP) analysis excluding unreliable results as well as an intention-to-diagnose (ITD) analysis based on all VCTE examinations including unreliable and failed measurements. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated using cut-offs previously published or calculated by Youden index in the present study. We used MedCalc Statistical Software 18.11.3 (MedCalc Software, Ostend, Belgium) for AUROC comparisons and SPSS 25 (IBM Statistics, New York, United States) for all other calculations.

Results

Patients

Between 2011 and 2016 we included 192 patients at both centers, 73 from Graz, and 119 from Vienna. Liver biopsy was performed for staging purposes (presence/absence of NASH, determination of fibrosis stage) in 169 patients and (primarily) for differential diagnosis in 23 patients (Figure 1). Liver biopsy was obtained percutaneously in 142 patients and intraoperatively during bariatric surgery in 50 patients. Histology did not confirm NAFLD in four patients and a representative liver biopsy specimen was lacking in two patients, leaving 186 for analysis. Patient characteristics for the whole NAFLD cohort and for subgroups according to fibrosis stage are given in Table 1.

Table 1.

Patient characteristics.

	All NAFLD	F < 2	F ≥ 2	p-value	F < 3	F ≥ 3	p-value
	N = 186	N = 115	N = 71	F ≥ 2 vs. F < 2	N = 137	N = 49	F ≥ 3 vs. F < 3
Age (years)	52 (39, 60)	46 (35, 55)	57 (52, 64)	<0.001	47 (35, 56)	58 (52, 67)	<0.001
Male sex (%)	57	62	49	0.096	60	49	0.187
BMI	30.5 (27.0, 37.8)	30.4 (27.0, 38.4)	30.7 (27.0, 34.1)	0.355	30.4 (27.0, 38.5)	30.9 (27.0, 33.9)	0.226
AST (U/L)	41 (30, 59)	35 (27, 51)	51 (34, 77)	<0.001	37 (29, 54)	53 (36, 97)	<0.001
ALT (U/L)	55 (36, 84)	51 (36, 80)	60 (37, 87)	0.320	52 (36, 87)	57 (37, 83)	0.742
Albumin (g/dL)	4.5 (4.2, 4.7)	4.5 (4.3, 4.8)	4.3 (4.0, 4.6)	<0.001	4.5 (4.2, 4.8)	4.2 (3.9, 4.5)	<0.001
PLT (10⁹/L)	223 (182, 271)	239 (202, 280)	206 (154, 251)	<0.001	233 (203, 275)	180 (136, 253)	<0.001
Diabetes/IFG (%)	30	17	52	<0.001	21	55	<0.001
HOMA-IR	3.65 (1.91, 6.53)	3.34 (1.55, 5.89)	4.43 (2.64, 8.12)	0.016	3.55 (1.59, 5.98)	4.17 (2.62, 8.39)	0.031
NAS 0–2 / 3–4/5–8 (%)	31/41/28	40/37/23	17/46/37	0.003	36.5/36.5/27	16/53/31	0.026

Data are shown as median (Q1, Q3). BMI, body mass index; AST, aspartate aminotransferase; ALT, alanine aminotransferase; PLT, platelet count; IFG; impaired fasting glucose; HOMA-IR, homeostatic model assessment – insulin resistance; NAS, NAFLD activity score.

Liver histology

On liver histology, CRN fibrosis stage was distributed as follows: F0, 64 patients (34%); F1, 51 patients (27%); F2, 22 patients (12%); F3, 29 patients (16%); F4, 20 patients (11%). NASH was present in 92 patients (50%), significant fibrosis (F ≥ 2) in 71 patients (38%), advanced fibrosis (F ≥ 3) in 49 patients (26%), and F ≥ 3 plus NASH in 38 patients (20%). Median length of biopsy cylinders was 22 mm.

Blood fibrosis tests

ELF score, FibroMeter^V2G, FibroMeter^V3G, FIB-4, and NFS increased stepwise with fibrosis stage (Figure 2, Table 2).

Figure 2.

Boxplots of noninvasive fibrosis tests by histological fibrosis stage.

Table 2.

Noninvasive fibrosis tests in NAFLD.

	N	All NAFLD	F < 2	F ≥ 2	p-value	F < 3	F ≥ 3	p-value
	N	N = 186	N = 115	N = 71	F ≥ 2 vs. F < 2	N = 137	N = 49	F ≥ 3 vs. F < 3
ELF score	181	8.9 (8.2, 9.9)	8.4 (8.0, 9.0)	10.0 (9.2, 11.4)	<0.001	8.5 (8.1, 9.2)	10.5 (9.6, 11.7)	<0.001
FM^V2G	131	0.226 (0.109, 0.562)	0.144 (0.077, 0.274)	0.595 (0.322, 0.935)	<0.001	0.167 (0.093, 0.325)	0.798 (0.464, 0.976)	<0.001
FM^V3G	132	0.267 (0.133, 0.638)	0.178 (0.097, 0.347)	0.661 (0.392, 0.903)	<0.001	0.207 (0.109, 0.421)	0.796 (0.530, 0.956)	<0.001
FIB-4	186	1.17 (0.78, 2.11)	0.92 (0.61, 1.39)	1.74 (1.18, 4.15)	<0.001	1.00 (0.67, 1.45)	2.54 (1.33, 4.66)	<0.001
NFS	167	−1.49 (−2.76, −0.26)	−2.13 (−3.26, −1.17)	−0.31 (−1.58, 0.94)	<0.001	−1.95 (−3.01, −0.93)	−0.22 (−1.20, 1.54)	<0.001
LSM (kPa)	140	8.9 (6.1, 14.8)	6.8 (5.5, 9.2)	16.6 (10.0, 25.9)	<0.001	7.3 (5.8, 10.3)	21.3 (14.8, 33.3)	<0.001

Data are shown as median (Q1, Q3). ELF score, enhanced liver fibrosis score; FM, FibroMeter™; FIB-4, fibrosis-4 index; NFS, NAFLD fibrosis score; LSM, liver stiffness measurement (FibroScan®).

ELF score and FibroMeter^V2G/V3G had the highest accuracies for the diagnosis of fibrosis stage and of the combined diagnostic endpoint of advanced fibrosis plus NASH (Tables 3–7).

Table 3.

AUROCs of noninvasive fibrosis tests for diagnosis of fibrosis stage in NAFLD.

	N	AUROC F ≥ 2	AUROC F ≥ 3	AUROC F ≥ 4	AUROC NASH & F ≥ 3
	N	n = 71	n = 49	n = 20	n = 35
ELF	181	0.85 (0.80–0.91)	0.90 (0.85–0.95)	0.92 (0.88–0.97)	0.90 (0.85–0.95)
FM^V2G	131	0.86 (0.79–0.93)	0.88 (0.80–0.96)	0.95 (0.90–0.99)	0.89 (0.81–0.96)
FM^V3G	132	0.84 (0.77–0.92)	0.88 (0.80–0.95)	0.94 (0.89–0.99)	0.88 (0.81–0.95)
FIB-4	186	0.80 (0.74–0.87)	0.82 (0.75–0.90)	0.86 (0.78–0.95)	0.81 (0.73–0.90)
NFS	167	0.78 (0.71–0.85)	0.80 (0.72–0.88)	0.79 (0.67–0.92)	0.79 (0.70–0.89)
LSM	140^a	0.85 (0.78–0.91)	0.91 (0.86–0.97)	0.95 (0.91–0.99)	0.87 (0.80–0.94)
LSM_PP	122^b	0.87 (0.80–0.94)	0.95 (0.90–0.99)	0.98 (0.95–1.00)	0.91 (0.85–0.98)

ELF, enhanced liver fibrosis score; FM, FibroMeter™; FIB-4, fibrosis-4 index; NFS, NAFLD fibrosis score; LSM, liver stiffness measurement.

All cases with LSM results (including 18 cases with unreliable results, excluding 10 cases with failed measurements). ^bLSM per protocol (reliable results only)

LSM per protocol (reliable results only).

Table 4.

AUROC comparisons of noninvasive fibrosis tests in NAFLD.

	N	AUROC	p-value^a	AUROC	p-value^a	AUROC	p-value^a
	N	F ≥ 2	vs. NFS	F ≥ 3	vs. NFS	F ≥ 3 & NASH	vs. NFS
ELF	181	0.85	0.036	0.90	0.003	0.90	0.010
FM^V2G	131	0.86	0.025	0.88	0.004	0.89	0.020
FM^V3G	132	0.84	0.058	0.88	0.004	0.88	0.031
FIB-4	186	0.80	0.308	0.82	0.179	0.82	0.360
NFS	167	0.78	—	0.79	—	0.79	—
LSM	140	0.85	0.051	0.91	0.017	0.87	0.142
LSM_PP	122	0.87	0.008	0.95	0.008	0.91	0.027

AUROCs of ELF, FM^V2G, FM^V3G, LSM, and LSM_PP were compared with that of NFS according to DeLong et al.³⁶

ELF, enhanced liver fibrosis score; FM, FibroMeter™; FIB-4, fibrosis-4 index; NFS, NAFLD fibrosis score; LSM, liver stiffness measurement (all cases with LSM results); LSM_PP, LSM per protocol (reliable results only).

For diagnosis of significant fibrosis, the published ELF score cut-off of 7.7 was not found useful. Instead an optimized cut-off of 9.1 derived by Youden index within our cohort was calculated (Table 5).³⁷

Table 5.

Accuracy of noninvasive fibrosis tests for diagnosis of F ≥ 2 in NAFLD. Calculations were based on published cut-offs for ELF, FIB-4, and LSM as well as optimized cut-offs determined in our patient cohort.

	Cut-off	Sensitivity	Specificity	PPV	NPV	DA
ELF score	7.7	100%	11%	40%	100%	44%
	9.1^a	82%	76%	67%	88%	78%
FM^V2G	0.309^a	80%	80%	72%	86%	80%
FM^V3G	0.378^a	78%	80%	71%	86%	80%
FIB-4	1.30	69%	72%	60%	79%	71%
	2.67	38%	97%	90%	72%	75%
LSM_PP (kPa)	8.2	83%	68%	56%	89%	73%
LSM_ITD (kPa)	8.2	84%	58%	51%	88%	67%

Cut-off for F ≥ 2 calculated by Youden index in the present study.

DA, diagnostic accuracy; LSM_PP, per protocol (reliable results only); LSM_ITD, intention-to-diagnose (based on 122 reliable results, 18 unreliable results, and 10 failed measurements coded as false-negative or false-positive).

Vibration-controlled transient elastography

VCTE was attempted in 150 patients and successfully performed in 140 patients (M probe, 88; XL probe, 52). The XL probe was available in 97 of the 150 VCTE examinations (65%). VCTE failed due to technical reasons in 10 patients (7%) and produced unreliable values in 18 patients (12%). Interestingly, the proportion of unreliable/failed VCTE measurements was similar independently of whether the XL probe was available (20%) or not (17%).

In parallel to blood fibrosis tests, LSM increased stepwise with fibrosis stage (Figure 2, Table 2). As expected, diagnostic accuracy for F ≥ 3 was higher in LSM PP (n = 122, AUROC 0.95 (0.90–0.99)) than in unreliable LSM (n = 18, AUROC 0.64 (0.37–0.91)) (p = 0.030 by DeLong test).

For diagnosis of advanced fibrosis, the recently published LSM cut-off of 9.7 kPa was found less useful than an optimized cut-off of 11.0 kPa derived by Youden index within our cohort (Tables 6 and 7).³⁷

Table 6.

Accuracy of noninvasive fibrosis tests for diagnosis of F ≥ 3 in NAFLD. Calculations were based on published cut-offs for ELF, FIB-4, and LSM as well as optimized cut-offs determined in our patient cohort.

	Cut-off	Sensitivity	Specificity	PPV	NPV	DA
ELF score	9.8	72%	90%	70%	90%	85%
FM^V2G	0.385^a	81%	81%	58%	93%	81%
FM^V3G	0.461^a	84%	78%	55%	94%	80%
FIB-4	1.30	76%	68%	46%	89%	70%
	2.67	49%	96%	80%	84%	83%
LSM_PP (kPa)	9.7	92%	77%	52%	97%	80%
	11.0^a	92%	85%	63%	98%	87%
LSM_ITD (kPa)	9.7	91%	65%	41%	96%	71%
	11.0^a	91%	73%	48%	97%	77%

Cut-off for F ≥ 3 calculated by Youden index in the present study.

Table 7.

Accuracy of noninvasive fibrosis tests for diagnosis of F ≥ 3 and NASH. Calculations were based on published cut-offs for ELF, FIB-4, and LSM as well as optimized cut-offs determined in our patient cohort.

	Cut-off	Sensitivity	Specificity	PPV	NPV	DA
ELF score	9.8	78%	85%	53%	95%	84%
FM^V2G	0.385^a	88%	78%	49%	97%	80%
FM^V3G	0.461^a	88%	75%	45%	96%	77%
FIB-4	1.30	74%	64%	32%	91%	66%
	2.67	54%	93%	63%	90%	85%
LSM_PP (kPa)	9.7	90%	73%	39%	97%	75%
	11.0^a	90%	80%	47%	98%	82%
LSM_ITD (kPa)	9.7	87%	61%	29%	96%	65%
	11.0^a	87%	68%	33%	97%	71%

Cut-off for F ≥ 3 calculated by Youden index in the present study.

Comparison of noninvasive fibrosis tests

On ROC analysis, ELF score, FibroMeter^V2G/V3G, and LSM showed superior diagnostic accuracy than FIB-4 and NFS for diagnosis of F ≥ 2, F ≥ 3, and the combined endpoint F ≥ 3 plus NASH (Table 3, Figure 3(a) to (c)). On DeLong test, ELF score, Fibrometer^V2G/V3G, and LSM (PP) outperformed NFS for diagnosis of both F ≥ 3 and F ≥ 3 & NASH (Table 4). Likewise, ELF score, Fibrometer^V2G, and LSM (PP) outperformed NFS for diagnosis of F ≥ 2 (Table 4).

Figure 3.

Receiver operating characteristic (ROC) curves of six noninvasive tests for diagnosis of CRN fibrosis stage: (a) significant fibrosis (F ≥ 2); (b) advanced fibrosis (F ≥ 3); (c) combined endpoint F ≥ 3 plus NASH. Areas under the ROC curves (AUROCs) are shown in Table 3.

For diagnosis of significant fibrosis, best results with ELF score were obtained at an optimized cut-off of 9.1 whereas the cut-off of 7.7 recommended by the manufacturer was not found useful (Table 5). Similar diagnostic accuracies could be demonstrated for the published LSM cut-off of 8.2 kPa and the calculated cut-offs for FibroMeter^V2G/V3G of 0.309/0.378 (Table 5). For diagnosis of advanced fibrosis, using the published cut-offs of 9.8 for ELF score and of 9.7 kPa for LSM,^12,24 and the calculated cut-offs for FibroMeter^V2G/V3G of 0.385/0.461, all showed moderate to good PPV and excellent NPV (≥90%) (Table 6). For LSM, diagnosis of advanced fibrosis was slightly improved using an optimized cut-off of 11.0 kPa obtained by Youden index (Table 6). When the combined endpoint F ≥ 3 plus NASH was analyzed using the same cut-offs as for F ≥ 3, PPVs were somewhat lower while NPVs remained excellent (Table 7).

Sequential analysis of noninvasive fibrosis tests

Aiming to optimize cost-effectiveness of noninvasive fibrosis testing in NAFLD, we designed a post hoc sequential strategy, first testing the inexpensive (non-commercial) FIB-4 index followed by the costlier (proprietary) ELF test in cases with FIB-4 values ≥1.30 only. However, diagnostic accuracy of this sequential approach for F ≥ 3 (sensitivity 57%, specificity 94%, PPV 77%, NPV 86%, DA 84%) was not better than that of FIB-4 alone at the ≥2.67 cut-off value (Table 6).

Effect of age, BMI, and diabetes/IFG on diagnostic accuracy of noninvasive fibrosis tests

As expected, age per se was associated with advanced fibrosis (AUROC 0.77 (0.69–0.84)). However, when our cohort was split at an age of <60 years (n = 143) vs. ≥60 years (n = 43), no major differences were evident for AUROCs of noninvasive fibrosis tests for F ≥ 3: ELF, 0.87 (0.79–0.95) vs. 0.92 (0.81–1.00), p = 0.517; FIB-4, 0.76 (0.65–0.87) vs. 0.85 (0.73–0.97), p = 0.315; LSM, 0.88 (0.80–0.96) vs. 0.96 (0.89–1.00), p = 0.151.

When the cohort was split by BMI < 30 (n = 86) vs. ≥30 (n = 100), blood fibrosis tests showed similar AUROCs for diagnosis of F ≥ 3: ELF, 0.92 (0.87–0.98) vs. 0.89 (0.81–0.97), p = 0.469; FIB-4, 0.82 (0.73–0.92) vs. 0.82 (0.72–0.92), p = 0.959. In contrast, accuracy of LSM for F ≥ 3 decreased with increasing body mass: BMI < 30, AUROC 0.98 (0.96–1.00) vs. BMI ≥ 30, AUROC 0.84 (0.74–0.94), p = 0.005.

Similar performance of all noninvasive fibrosis tests was observed when the cohort was split by absence vs. presence of diabetes/IFG, respectively (data not shown).

Discussion

In this study we demonstrate superior diagnostic accuracy of ELF score, Fibrometer^V2G/V3G, and LSM (using PP analysis) compared to the simple fibrosis tests, FIB-4 and NFS, in a large cohort of adult NAFLD patients. As expected, LSM analyzed PP showed higher diagnostic accuracies than on ITD analysis. Of note, accuracy of FIB-4 was numerically higher than that of NFS in all analyses and thus proved very useful given its simple calculation.

Comparisons of noninvasive fibrosis tests in NAFLD patients have shown variable results. Guha et al. reported only marginally superior accuracy of ELF score compared to NFS for the diagnosis of F ≥ 3 (AUROC 0.93 vs. 0.89).⁸ On the other hand, consistent with the results of our study, diagnostic accuracy of VCTE was found clearly superior to NFS ²² or to FIB-4.¹⁶ We recently reported an AUROC of 0.876 by ELF test (n = 162) and of 0.935 by VCTE (n = 75) for the diagnosis of F ≥ 3 in various cohorts of NAFLD patients.³⁸ However, this study did not provide direct comparison of ELF and VCTE in the same patients, did not consider intent-to-diagnose analysis for VCTE and did not account for the presence or absence of NASH. Karlas et al. compared VCTE and ELF test in a small cohort of morbidly obese patients (n = 41, median BMI 47) prior to bariatric surgery with intraoperative liver biopsy and obtained valid LSM data in only 50% of the patients while ELF test was feasible in all patients.³⁵

Determination of advanced fibrosis in NAFLD is crucial for prognosis estimation and treatment need in general, and is now especially useful to determine eligibility of patients for clinical trials with novel anti-inflammatory/antifibrotic drugs focusing on patients with fibrosis stage F ≥ 3 and NASH. Furthermore, a diagnosis of advanced fibrosis/cirrhosis establishes the need for HCC surveillance. Our patient cohort represents the spectrum of NAFLD typically admitted to referral centers, with a relatively low prevalence of advanced fibrosis (F ≥ 3) (26%). Hence any of the fibrosis tests studied show good NPV at the selected cut-offs, allowing to safely spare liver biopsy, whereas patients with test results above the respective cut-off should undergo liver biopsy to determine whether NASH with advanced fibrosis is truly present.

Among the tests investigated in the present study, FIB-4 tended to show higher diagnostic accuracy than NFS and seems therefore preferable, given its simple calculation which may be easily implemented into automated lab reports. We confirmed the previously published low cut-off of 1.30 for ruling out advanced fibrosis, while the published high cut-off for NAFLD of 2.67 proved good accuracy to rule in advanced fibrosis (PPV 80%).³¹ However, the diagnostic utility of FIB-4 is limited considering the proportion of patients who lie within the “grey zone” of 1.30–2.67 (27% of the patients in the present study). Likewise, the chosen single cut-offs for ELF score (9.8), FibroMeter^V2G/V3G (0.385/0.461), and LSM (9.7 or 11.0 kPa) allow for safe exclusion of advanced fibrosis while they provide moderate to good PPVs. Of note, the published LSM cut-offs for F ≥ 3 in NAFLD are lower than the 15.0 kPa cut-off recently reported for diagnosis of advanced fibrosis in alcoholic liver disease,³⁹ which may be explained by different fibrosis patterns and/or spectrum bias.⁴⁰

Considering VCTE, it should be kept in mind that this method is not always feasible, with unreliable LSM values or failed measurements in approximately 20% of our study patients, irrespective of the availability of the XL probe. Other studies investigating VCTE in NAFLD using the M probe have reported similar failure rates of VCTE.^41,42 Studies comparing values obtained by both probes demonstrate lower LSM values and suggest lower cut-offs for the XL probe.⁴¹ In addition, recent data indicate that LSM values should be corrected for simultaneously measured controlled attenuation parameter (CAP).²⁶ Of note, published LSM cut-off values for F ≥ 3 in NAFLD vary between 7.9 and 12.5 kPa for the M probe and between 5.7 and 9.3 kPa for the XL probe.^21,22,41,42

For FIB-4, significant effects of age on its diagnostic accuracy have been reported. Boursier et al. noted a loss of specificity for FIB-4 in patients above age 60,³³ while McPherson et al. proposed a different FIB-4 cut-off of 2.00 for patients aged >65 years.³⁴ We could not clearly reproduce these observations in our cohort probably due to the limited number of patients in the respective subgroups defined by age and fibrosis stage.

Strengths of our study include (a) evaluation of several commonly used noninvasive fibrosis tests in a large cohort of unselected NAFLD patients in a clinical real life setting; (b) direct comparison of ELF score, FibroMeter^V2G/V3G, FIB-4, and LSM in the same patients; (c) definition of liver histology by two experienced specialized hepatopathologists; (d) evaluation of various endpoints of high clinical relevance (F ≥ 2, F ≥ 3, F ≥ 3 and NASH); (e) providing optimized cut-offs for ELF, FibroMeter^V2G/V3G, and LSM. Limitations of our study include the lack of an internal validation of our findings, the shortage of data on some fibrosis tests evaluated (VCTE, FibroMeter^V2G/V3G), the time interval between VCTE and liver biopsy, and the lack of availability of XL probes in the early phase of the study. Nevertheless, diagnostic accuracy of LSM was excellent and superior to that of the simple fibrosis tests on per-protocol analysis.

In conclusion, fibrosis stage in NAFLD is best assessed by ELF test, FibroMeter^V2G/V3G, and/or VCTE. When VCTE produces reliable results (as in 81% in the present study), it shows superior diagnostic accuracy. Accuracy of VCTE decreases with increasing BMI while blood fibrosis tests are not affected by BMI. Among the blood fibrosis tests studied, ELF test and FibroMeter^V2G show the best performance. In settings where neither VCTE nor proprietary blood fibrosis tests are available, FIB-4 is a reasonably accurate and cost-effective alternative to assess advanced fibrosis in NAFLD. Further research is needed (a) to noninvasively delineate presence of NASH from fibrosis stage and (b) to evaluate noninvasive markers as treatment endpoints in NAFLD.

Footnotes

Acknowledgments

The authors wish to thank Prof. Paul Calès for calculation of FibroMeter^V2G/V3G, Ms. Andrea Streit for excellent technical assistance, and Dr. Tatjana Stojakovic and the team of the Endo-Lab-Platform for analytical assistance.

Declaration of conflicting interests

R.M. received speaker honoraria from Siemens.

Funding

This work was in part supported by BioPersMed (COMET K-project 825329) funded by the Austrian Research Promotion Agency (FFG).

Ethics approval

This study was approved by the Ethics Committee of the Medical University of Graz (EC 24-224 ex 11/12) and the Medical University of Vienna (EC #747/2011) and was performed in accordance with the 1975 Declaration of Helsinki.

Informed consent

Written informed consent was obtained from all study participants.

References

Singh

Allen

Wang

, et al. Fibrosis progression in nonalcoholic fatty liver vs nonalcoholic steatohepatitis: a systematic review and meta-analysis of paired-biopsy studies. Clin Gastroenterol Hepatol 2015; 13: 643–654.

Angulo

Kleiner

Dam-Larsen

, et al. Liver fibrosis, but no other histologic features, is associated with long-term outcomes of patients with nonalcoholic fatty liver disease. Gastroenterology 2015; 149: 389–397.

Ekstedt

Hagstrom

Nasr

, et al. Fibrosis stage is the strongest predictor for disease-specific mortality in NAFLD after up to 33 years of follow-up. Hepatology 2015; 61: 1547–1554.

Hagstrom

Nasr

Ekstedt

, et al. SAF score and mortality in NAFLD after up to 41 years of follow-up. Scand J Gastroenterol 2017; 52: 87–91.

Kwok

Tse

Wong

, et al. Systematic review with meta-analysis: non-invasive assessment of non-alcoholic fatty liver disease–the role of transient elastography and plasma cytokeratin-18 fragments. Aliment Pharmacol Ther 2014; 39: 254–269.

Younossi

Loomba

Anstee

, et al. Diagnostic modalities for nonalcoholic fatty liver disease, nonalcoholic steatohepatitis, and associated fibrosis. Hepatology 2018; 68: 349–360.

Vilar-Gomez

Chalasani

. Non-invasive assessment of non-alcoholic fatty liver disease: clinical prediction rules and blood-based biomarkers. J Hepatol 2018; 68: 305–315.

Guha

Parkes

Roderick

, et al. Noninvasive markers of fibrosis in nonalcoholic fatty liver disease: validating the European Liver Fibrosis Panel and exploring simple markers. Hepatology 2008; 47: 455–460.

Miele

De Michele

Marrone

, et al. Enhanced liver fibrosis test as a reliable tool for assessing fibrosis in nonalcoholic fatty liver disease in a clinical setting. Int J Biol Markers 2017; 32: e397–e402.

10.

Lopez

Aroca

Bernal

MDF

, et al. Utility of the ELF test for detecting steatohepatitis in morbid obese patients with suspicion of nonalcoholic fatty liver disease. Obes Surg 2017; 27: 2347–2353.

11.

Nobili

Parkes

Bottazzo

, et al. Performance of ELF serum markers in predicting fibrosis stage in pediatric non-alcoholic fatty liver disease. Gastroenterology 2009; 136: 160–167.

12.

Lichtinghagen

Pietsch

Bantel

, et al. The Enhanced Liver Fibrosis (ELF) score: normal values, influence factors and proposed cut-off values. J Hepatol 2013; 59: 236–242.

13.

Fagan

Pretorius

Horsfall

, et al. ELF score >/=9.8 indicates advanced hepatic fibrosis and is influenced by age, steatosis and histological activity. Liver Int 2015; 35: 1673–1681.

14.

Glen

Floros

Day

, et al. Non-alcoholic fatty liver disease (NAFLD): summary of NICE guidance. BMJ 2016; 354: i4428–i4428.

15.

Leroy

Halfon

Bacq

, et al. Diagnostic accuracy, reproducibility and robustness of fibrosis blood tests in chronic hepatitis C: a meta-analysis with individual data. Clin Biochem 2008; 41: 1368–1376.

16.

Boursier

Vergniol

Guillet

, et al. Diagnostic accuracy and prognostic significance of blood fibrosis tests and liver stiffness measurement by FibroScan in non-alcoholic fatty liver disease. J Hepatol 2016; 65: 570–578.

17.

Cales

Boursier

Bertrais

, et al. Optimization and robustness of blood tests for liver fibrosis and cirrhosis. Clin Biochem 2010; 43: 1315–1322.

18.

Imbert-Bismut

Ratziu

Pieroni

, et al. Biochemical markers of liver fibrosis in patients with hepatitis C virus infection: a prospective study. Lancet 2001; 357: 1069–1075.

19.

Angulo

Hui

Marchesini

, et al. The NAFLD fibrosis score: a noninvasive system that identifies liver fibrosis in patients with NAFLD. Hepatology 2007; 45: 846–854.

20.

Sterling

Lissen

Clumeck

, et al. Development of a simple noninvasive index to predict significant fibrosis in patients with HIV/HCV coinfection. Hepatology 2006; 43: 1317–1325.

21.

Wong

Vergniol

Wong

, et al. Diagnosis of fibrosis and cirrhosis using liver stiffness measurement in nonalcoholic fatty liver disease. Hepatology 2010; 51: 454–462.

22.

Tapper

Challies

Nasser

, et al. The performance of vibration controlled transient elastography in a US cohort of patients with nonalcoholic fatty liver disease. Am J Gastroenterol 2016; 111: 677–684.

23.

Siddiqui

Vuppalanchi

Van Natta

, et al. Vibration-controlled transient elastography to assess fibrosis and steatosis in patients with nonalcoholic fatty liver disease. Clin Gastroenterol Hepatol 2019; 17: 156–163.

24.

Eddowes

Sasso

Allison

, et al. Accuracy of FibroScan controlled attenuation parameter and liver stiffness measurement in assessing steatosis and fibrosis in patients with nonalcoholic fatty liver disease. Gastroenterology 2019; 156: 1717–1730.

25.

Petta

Maida

Macaluso

, et al. The severity of steatosis influences liver stiffness measurement in patients with nonalcoholic fatty liver disease. Hepatology 2015; 62: 1101–1110.

26.

Petta

Wong

Camma

, et al. Improved noninvasive prediction of liver fibrosis by liver stiffness measurement in patients with nonalcoholic fatty liver disease accounting for controlled attenuation parameter values. Hepatology 2017; 65: 1145–1155.

27.

Petta

Di Marco

Camma

, et al. Reliability of liver stiffness measurement in non-alcoholic fatty liver disease: the effects of body mass index. Aliment Pharmacol Ther 2011; 33: 1350–1360.

28.

Boursier

de Ledinghen

Poynard

, et al. An extension of STARD statements for reporting diagnostic accuracy studies on liver fibrosis tests: the Liver-FibroSTARD standards. J Hepatol 2015; 62: 807–815.

29.

Sanyal

Brunt

Kleiner

, et al. Endpoints and clinical trial design for nonalcoholic steatohepatitis. Hepatology 2011; 54: 344–353.

30.

Kleiner

Brunt

Van Natta

, et al. Design and validation of a histological scoring system for nonalcoholic fatty liver disease. Hepatology 2005; 41: 1313–1321.

31.

Shah

Lydecker

Murray

, et al. Comparison of noninvasive markers of fibrosis in patients with nonalcoholic fatty liver disease. Clin Gastroenterol Hepatol 2009; 7: 1104–1112.

32.

McPherson

Stewart

Henderson

, et al. Simple non-invasive fibrosis scoring systems can reliably exclude advanced fibrosis in patients with non-alcoholic fatty liver disease. Gut 2010; 59: 1265–1269.

33.

Boursier

de Ledinghen

Leroy

, et al. A stepwise algorithm using an at-a-glance first-line test for the non-invasive diagnosis of advanced liver fibrosis and cirrhosis. J Hepatol 2017; 66: 1158–1165.

34.

McPherson

Hardy

Dufour

, et al. Age as a Confounding Factor for the Accurate Non-Invasive Diagnosis of Advanced NAFLD Fibrosis. Am J Gastroenterol 2017; 112: 740–751.

35.

Karlas

Dietrich

Peter

, et al. Evaluation of transient elastography, acoustic radiation force impulse imaging (ARFI), and Enhanced Liver Function (ELF) score for detection of fibrosis in morbidly obese patients. PLoS One 2015; 10: e0141649–e0141649.

36.

DeLong

Clarke-Pearson

. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988; 44: 837–845.

37.

Youden

. Index for rating diagnostic tests. Cancer 1950; 3: 32–35.

38.

Staufer

Dengler

Huber

, et al. The non-invasive serum biomarker soluble Axl accurately detects advanced liver fibrosis and cirrhosis. Cell Death Dis 2017; 8: e3135–e3135.

39.

Thiele

Madsen

Hansen

, et al. Accuracy of the enhanced liver fibrosis test vs FibroTest, elastography, and indirect markers in detection of advanced fibrosis in patients with alcoholic liver disease. Gastroenterology 2018; 154: 1369–1379.

40.

European Association for Study of the Liver. EASL-ALEH clinical practice guidelines: non-invasive tests for evaluation of liver disease severity and prognosis. J Hepatol 2015; 63: 237–264.

41.

De Ledinghen

Wong

Vergniol

, et al. Diagnosis of liver fibrosis and cirrhosis using liver stiffness measurement: comparison between M and XL probe of FibroScan(R). J Hepatol 2012; 56: 833–839.

42.

Cassinotto

Boursier

de Ledinghen

, et al. Liver stiffness in nonalcoholic fatty liver disease: A comparison of supersonic shear imaging, FibroScan, and ARFI with liver biopsy. Hepatology 2016; 63: 1817–1827.