Point Shear Wave Elastography by Acoustic Radiation Force Impulse Quantification in Comparison to Transient Elastography for the Noninvasive Assessment of Liver Fibrosis in Chronic Hepatitis C: A Prospective International Multicenter Study

Abstract

Purpose: The aim of the present prospective European multicenter study was to demonstrate the non-inferiority of point shear wave elastography (pSWE) compared to transient elastography (TE) for the assessment of liver fibrosis in patients with chronic hepatitis C.

Materials and Methods: 241 patients with chronic hepatitis C were prospectively enrolled at 7 European study sites and received pSWE, TE and blood tests. Liver biopsy was performed with histological staging by a central pathologist. In addition, for inclusion of cirrhotic patients, a maximum of 10 % of patients with overt liver cirrhosis confirmed by imaging methods were allowed by protocol (n = 24).

Results: Owing to slower than expected recruitment due to a reduction of liver biopsies, the study was closed after 4 years before the target enrollment of 433 patients with 235 patients in the ‘intention to diagnose’ analysis and 182 patients in the ‘per protocol’ analysis. Therefore, the non-inferiority margin was enhanced to 0.075 but non-inferiority of pSWE could not be proven. However, Paired comparison of the diagnostic accuracy of pSWE and TE revealed no significant difference between the two methods in the ‘intention to diagnose’ and ‘per protocol’ analysis (0.81 vs. 0.85 for F ≥ 2, p = 0.15; 0.88 vs. 0.92 for F ≥ 3, p = 0.11; 0.89 vs. 0.94 for F = 4, p = 0.19). Measurement failure was significantly higher for TE than for pSWE (p = 0.030).

Conclusion: Non-inferiority of pSWE compared to TE could not be shown. However, the diagnostic accuracy of pSWE and TE was comparable for the noninvasive staging of liver fibrosis in patients with chronic hepatitis C.

Zusammenfassung

Ziel: Ziel der vorliegenden prospektiven europäischen Multicenterstudie war es die Nicht-Unterlegenheit der Punkt-Scherwellenelastographie(pSWE) im Vergleich zur transienten Elastographie(TE) für die Beurteilung des Leberfibrosestadiums bei Patienten mit chronischer Hepatitis C nachzuweisen.

Material und Methoden: 241 Patienten mit chronischer Hepatitis C wurden prospektiv an 7 europäischen Zentren eingeschlossen und erhielten eine pSWE, TE und laborchemische Diagnostik. Die Leberhistologie wurde durch einen zentralen Pathologen beurteilt. Zusätzlich wurden maximal 10 % der Patienten ohne Leberhistologie mit gesicherter Leberzirrhose eingeschlossen (n = 24).

Ergebnisse: Bedingt durch einen europaweiten Rückgang der Zahl der Leberbiopsien wurde die Studie nach 4 Jahren Rekrutierung beendet bevor die Ziel-Patientenzahl von 433 erzielt werden konnte. Es konnten 235 Patienten in die ‘intention to diagnose’-Analyse und 182 Patienten in die ‘per protocol’-Analyse eingeschlossen werden. Die statistische Nicht-Unterlegenheitsgrenze wurde angepasst an die geringere Patientienzahl (0,075), die Nicht-Unterlegenheit von pSWE konnte nicht belegt werden. Jedoch ergab der gepaarte Vergleich der diagnostischen Genauigkeit von pSWE und TE keinen signifikanten Unterschied zwischen beiden Methoden sowohl in der ‘intention to diagnose’, als auch in der ‘per protocol’-Analyse (0,81 vs. 0,85 für F ≥ 2, p = 0,15; 0,88 vs. 0,92 für F ≥ 3, p = 0,11; 0,89 vs. 0,94 für F = 4, p = 0,19). Die Rate an Fehlmessungen war mit TE signifikant höher als mit pSWE (p = 0,030).

Schlussfolgerung: Die Nicht-Unterlegenheit von pSWE im Vergleich zu TE konnte nicht nachgewiesen werden. Jedoch war die diagnostische Genauigkeit von pSWE und TE vergleichbar gut für die nicht-invasive Beurteilung des Leberfibrosestadiums.

Clinical Trials Registration: NCT 01113814

Key words

ultrasound - staging - abdomen - hepatic veins

Introduction

Chronic hepatitis C is an important cause of liver cirrhosis with its sequelae including hepatocellular carcinoma [1]. Estimation of the degree of liver fibrosis is important for prognostic and therapeutic decisions in patients infected with chronic hepatitis C [2]. At present, liver biopsy is still most commonly used as the reference standard for the assessment of liver fibrosis. However, it is an invasive method associated with patient discomfort and in rare cases with serious complications [3]. In addition, the accuracy of liver biopsy is limited due to intra- and interobserver variability and sampling error [4] [5]. In recent years, research has focused on the evaluation of noninvasive methods for the assessment of liver fibrosis. In addition, noninvasive methods have been included in the European guidelines for the staging of chronic hepatitis C since 2011 [6] [7]. In the EASL guidelines 2014 it is stated that noninvasive methods can be used for the assessment of fibrosis stage initially, with liver biopsy reserved for cases of uncertainty of potential additional etiologies of liver disease [7]. Besides research on serologic fibrosis markers, the most intensively evaluated and well established elastography method is transient elastography (TE) (FibroScan^®, Echosens, France) [8] [9] [10] [11] [12]. Point shear wave elastography (pSWE) by Acoustic Radiation Force Impulse (ARFI) quantification (Acuson S2000/3000, Virtual Touch^TM tissue quantification, Siemens) is an ultrasound-based elastography method which is integrated in a conventional ultrasound machine and can be performed with conventional ultrasound probes during regular abdominal sonography. In addition, it makes it possible to determine the exact location of pSWE-measurement site during B-mode ultrasound. Meta-analyses have shown comparably good results for pSWE as for transient elastography (TE) for the noninvasive assessment of liver fibrosis [13] [14]. The aim of the present study was to directly compare the diagnostic accuracy of pSWE to TE for the noninvasive assessment of liver fibrosis in patients infected with hepatitis C in a prospective European multicenter study using the non-inferiority approach.

Materials and Methods

Patients

Between February 2009 and March 2012, 241 consecutive patients infected with chronic hepatitis C, who attended the participating hospitals for histological assessment of liver fibrosis or evaluation of liver cirrhosis were included in the study. The patients were recruited from seven European centers in Germany, Rumania, and The Netherlands. All patients received pSWE, TE and determination of serum markers. Liver histology was staged by a central pathologist (V. D.) and used as the reference method in 217 patients. The indication for liver biopsy was the determination of histological fibrosis. In addition, 24 patients with liver cirrhosis confirmed by imaging methods were included, since liver biopsy in patients with evident liver cirrhosis is not justified and F4 patients would have been underrepresented otherwise. In all patients chronic hepatitis C was proven by the presence of HCV-RNA in serum. None of the patients received antiviral therapy at the time of study inclusion. Male patients with more than 40 g alcohol consumption per day and women with more than 20 g/d were excluded from the study.

Written informed consent was obtained from all patients. The multicenter study was approved by the lead ethical committee of the University of Frankfurt with local approval from the ethical committees of the other study centers. The study was registered with clinical-trials.com with registration number NCT 01113814.

Measurements

All patients received pSWE by ARFI quantification (S2000, Siemens, Mountain View, CA) and TE (Echosens, Paris, France) on the same day by experienced physicians blinded to the results of liver biopsy. Patients had been fasting for at least six hours at the time of pSWE and TE examination.

Point shear wave elastography by ARFI quantification was performed with a curved array at 4 MHz for B-mode imaging using a Siemens ACUSON S2000 machine. pSWE involves targeting of an anatomic region to be interrogated for elastic properties with a region-of-interest (ROI) cursor while performing real-time B-mode imaging. Tissue at the ROI is mechanically excited using short-duration acoustic pulses to generate localized displacements in tissue. The displacements result in shear wave propagation away from the region of excitation and are tracked using ultrasonic, correlation-based methods [15]. The maximum displacement is estimated for many ultrasound tracking beams laterally adjacent to the single push-beam. By measurement of the time to peak displacement at each lateral location, the shear wave speed of the tissue can be reconstructed [15]. The shear wave propagation velocity is proportional to the square root of tissue elasticity [16] [17]. Results are expressed in m/s.

The patient was in a supine position and was asked to suspend breath during measurement. The examination was performed on the right lobe of the liver through an intercostal approach. An area was chosen where the liver tissue was at least 6 cm thick and free of large blood vessels. A measurement depth of 2 cm below the liver capsule was chosen to standardize the examination. Ten acquisitions within the right liver lobe were performed on each patient. pSWE failure was defined as no successful pSWE measurement after 10 attempts. The pSWE examination took approx. 5 minutes. No limitations were found during the measurement.

TE was performed using FibroScan^® (Echosens, France). FibroScan is a specially developed machine enabling the performance of transient elastography using an M-mode image for the determination of the organ and measurement site location. TE is performed with an ultrasound transducer probe mounted on the axis of a vibrator. A vibration transmitted from the vibrator towards the tissue induces an elastic shear wave that propagates through the tissue. These propagations are followed by pulse-echo ultrasound acquisitions and their velocity is measured which is directly related to tissue stiffness. Results are expressed in kilopascal (kPa) [18]. The details have been published in several studies [9] [10] [11] [19].

The examination was performed on the right lobe of the liver through the intercostal space. After the area of measurement was located, the examiner pressed the button of the probe to start the acquisition. The measurement depth was between 25 and 65 mm. As suggested by the manufacturer, ten successful acquisitions were performed on each patient. FibroScan failure is defined when less than ten valid measurements are obtained. Only TE results obtained with a success rate of at least 60 % and an interquartile range ≤ 30 % were considered reliable and were included in the per protocol analysis. Nevertheless, an intention to diagnose analysis was performed as well including all TE results independent of the success rate and interquartile range.

Liver Histology

Liver biopsy specimens were fixed in 4 %-buffered formalin and embedded in paraffin. Two-micrometer-thick sections were stained with hematoxylin-eosin, Perls-iron-stain, dPAS (periodic-acid-Schiff after digestion with diastase), Masson-Trichrome and Siriusred stain. Biopsy specimens from all centers were analyzed by a central pathologist (VD). The pathologist was blinded to the clinical results of the patients. Liver fibrosis stages were evaluated semi-quantitatively according to the Metavir scoring system [20]. Liver fibrosis was staged on a F0-F4 scale: F0-no fibrosis, F1-portal fibrosis without septa, F2-portal fibrosis with few septa, F3-numerous septa without cirrhosis, F4-cirrhosis. In addition, the Chevallier scoring system [21] was used to semi-quantitatively assess fibrosis using morphometric measurements of fibrosis taking separately into account fibrosis deposits around the centrilobular veins (CLV), perisinusoidal space (PS), portal tract (PT) and septa along with number (NS) and width of septa (WS). Steatosis was assessed according to the percentage of hepatocytes with fatty changes: 0 = no, 1 = 1 – 10 % of hepatocytes, 2 = 11 – 33 % hepatocytes, 2 = 34 – 66 % hepatocytes, 4 = 67 – 100 % hepatocytes. Necroinflammatory activity was graded according to the modified HAI grading system divided into: A = periportal or periseptal interface hepatitis, B = confluent necrosis, C = focal lytic necrosis, apoptosis and focal inflammation, and D = portal inflammation with a maximum score of 18 [22]. Liver biopsies with less than 6 portal tracts were excluded from the study. In addition, a sub-analysis was performed in patients with at least 10 portal tracts.

Blood Markers

The following blood parameters were determined in all patients: aspartate aminotransaminase (AST), alanine aminotransaminase (ALT), γ-glutamyl transpeptidase (GGT), total bilirubin, platelet count, International Normalized Ratio (INR), albumin, cholesterol, triglycerides, fasting glucose, apolipoprotein A1, haptoglobin, and alfa-2-macroglobulin. Enzymatic activity was measured at 37 °C according to International Federation of Clinical Chemistry standards.

The laboratory followed the pre-analytical and analytical recommendations required to obtain the fibrosis marker score FibroTest^® and the serum necroinflammatory marker score ActiTest^® (Biopredictive, France), and the steatosis markers Steatotest^® and Nashtest^® (Biopredictive, France) [23]. These tests were computed on the Biopredictive website (www.biopredictive.com) using the parameters: haptoglobin, alfa-2 macroglobulin, apoprotein A1, bilirubin, GGT, ALT, age, sex with a patented formula. The security algorithms on the industrial website permitting the exclusion of patients with high-risk profile of false positive/negative were respected [23] [24].

In addition 10 mL serum was stored at -20 °C. This stored serum was collected from the centers at the end of the study for the calculation of the Enhanced Liver Fibrosis (ELF) score. The ELF score includes the tissue inhibitor metalloproteinase 1 (TIMP 1), hyaluronoic acid (HA) and N-terminal peptide of procollagen III (P3NP) and was calculated with a patented formula with ADVIA Centaur system by Siemens Healthcare Diagnostics.

Sample Size Calculation

The primary study aim was the statistical proof of the non-inferiority of pSWE in comparison to TE simultaneously for the detection of fibrosis stages equal to or above 2 and equal to or above 3 or of cirrhosis. From a pilot study [25], we know that the variance of AUC_pSWE-AUC_TE using the Delong procedure is around 0.11/n, 0.069/n and 0.098/n, when both are estimated from the same sample of patients with sample size n, respectively. Originally, we aimed to include 433 patients accounting for a dropout rate of 15 % to reach a power of 80 % for the non-inferiority aim if the non-inferiority margin of the estimated AUC is 0.05. Unfortunately, recruiting was slower than expected. Therefore, the study was finished earlier and only data from n = 182 patients are now available for assessing the primary aim. In this situation, the power of the study would only be around 37 % with the previously planned non-inferiority margin of Δ = 0.05. Nevertheless, with an increased non-inferiority margin of Δ = 0.075, a power around 80 % can still be reached.

Statistical Analysis

For TE and pSWE the median of all 10 measurements per subject was calculated and used for further analysis. Values of TE and pSWE were not normally distributed and therefore expressed as median values. The correlations between the different approaches and the histological fibrosis stage were assessed by Spearman’s correlation coefficient. The diagnostic performance of TE, pSWE, and serum fibrosis markers was assessed by receiver operating characteristic (ROC) curves. The ROC curve represents the sensitivity versus 1-specificity for all possible cut-off values for the prediction of the different fibrosis stages. The areas under the ROC curves (AUROC) as well as 95 %-CI of AUROC were calculated (BiAS for Windows, version 10.11, epsilon 2014, Frankfurt, Germany). AUROC values for different diagnostic criteria for the same dataset were compared with the non-parametric DeLong test. Cut-off values defining prediction regions for each fibrosis stage were defined using Youden’s index [26] for predicting advanced stages. An optimized score for AUROC for the diagnosis of significant fibrosis combining elastography and serum marker was calculated. Thereby, linear combinations as well as combinations of logarithmized values and quadratic values were tested. After optimization of cut-off levels, sensitivity, specificity, positive-and negative-predictive-values were calculated without further adjustments (e. g. by cross-validation) from the same data. Logistic regression analysis was used to evaluate reasons for false-negative and false-positive pSWE results. McNemar test was used to compare sensitivities and specificities between the noninvasive tests. For the exclusion of significant fibrosis and liver cirrhosis, specificities were compared only, since cut-offs with a sensitivity > 90 % were chosen for all tests for respective exclusion. For the diagnosis of significant fibrosis and liver cirrhosis, sensitivities were compared only, since cut-offs with a specificity > 90 % were chosen for all tests. Cohen´s kappa test was used to evaluate discordant results of TE and pSWE as compared to liver histology overall and a Q-test for heterogeneity of proportions of discordant results.

Results

241 patients were evaluated in this prospective multicenter study. 6 patients were excluded with less than 6 portal tracks at liver biopsy. Therefore, 235 patients were included in the “intention to diagnose” analysis. For the “per protocol” analysis, 53 patients had to be excluded for the following reasons: TE failure in 22 patients, invalid TE measurement (success rate < 60 % or IQR ≥ 30 %) in 17 patients, pSWE failure in 9 patients, and both TE and pSWE failure in 5 patients. Therefore, 182 patients were included in the “per protocol” analysis. Patient characteristics are shown in [Table 1].

Table 1 Patient characteristics.

characteristics	intention to diagnose analysis (n = 235)	per protocol analysis (n = 182)
patient age (years)
mean ± SD (range)	50 ± 11 (24 – 79)	48 ± 11 (24 – 77)
median	50	50
male gender, n (%)	131 (56 %)	103 (57 %)
center (no. of patients)
Frankfurt	104	86
Cluj	44	40
Rotterdam	31	25
Hamburg	19	14
Hannover	12	6
Munich	8	6
Düsseldorf	17	5
AST (U/l), mean ± SD (range)	76 ± 69 (13 – 795)	71 ± 51 (13 – 360)
ALT (U/l), mean ± SD (range)	99 ± 84 (19 – 713)	98 ± 79 (19 – 536)
GGT (U/l), mean ± SD (range)	96 ± 98 (9 – 726)	95 ± 102 (9 – 726)
total bilirubin (mg/dl), mean ± SD (range)	0.87 ± 0.81 (0.1 – 9)	0.82 ± 0.76 (0.1 – 9)
platelet count (x10³/mm³), mean ± SD (range)	201 ± 71 (41 – 424)	203 ± 67 (41 – 424)
international ratio (INR), mean ± SD (range)	1.06 ± 0.22 (0.80 – 3.60)	1.05 ± 0.12 (0.85 – 1.72)
HCV-RNA (IU/ml) mean ± SD (range)	3.26*106 ± 6.34*10⁶ (45 – 70.3*10⁶)	3.11*106 ± 6.70*10⁶ (172 – 70.3*10⁶)
biopsy
6 – 10 portal fields (no. of patients)	35	25
> 10 portal fields (no. of patients)	166	133
patients with cirrhosis/hcc without biopsy (no. of patients)	34	24
histological fibrosis stage
F0 (no. of patients)	35	28
F1 (no. of patients)	75	63
F2 (no. of patients)	42	30
F3 (no. of patients)	25	19
F4 (no. of patients)	58	42

SD = standard deviation, pat. = patients; LB = liver biopsy; AST = aspartate aminotransaminase. ALT = alanine aminotransaminase. GGT = gamma-glutamyl transpeptidase; INR = International Normalized Ratio; HCV = hepatitis C virus.

Serum markers to calculate ELF were available in 194 patients and to calculate Fibrotest in 174 patients in the intention to diagnose study population.

Comparison of Noninvasive Methods using Liver Histology as the Reference Method

Measurement failure was significantly higher for TE than for pSWE (27 vs. 14, p = 0.030).

The non-inferiority margin of Δ = 0.075 could not be reached for any fibrosis stage (AUROC of TE – AUROC of pSWE was 0.04 95 % CI: -0.02 – 0.10] for F ≥ 2; 0.03 95 % CI: -0.01 – 0.10] for F ≥ 3, and 0.05 95 % CI: -0.01 – 0.12] for F = 4).

The diagnostic accuracies of pSWE and TE are shown in [Table 2]. No significant difference between pSWE and TE was found in the ‘per protocol’ analysis and the ‘intention to diagnose’ analysis for the diagnosis of mild, significant, severe fibrosis and cirrhosis ([Table 2]). A sub-analysis of patients with more than 10 portal tracts on histology is shown in Supplementary [Table 1]. Also in this subgroup no significant difference was found between pSWE and TE.

Table 2 Area under ROC curve (95 % confidence interval) for pSWE imaging and transient elastography according to Metavir fibrosis stage.

method	F ≥ 1	F ≥ 2	F ≥ 3	F = 4
per protocol analysis (n = 182)
pSWE	0.75 (0.67; 0.83)	0.81 (0.74; 0.88)	0.88 (0.82; 0.94)	0.89 (0.83; 0.96)
TE	0.80 (0.72; 0.88)	0.85 (0.80; 0.91)	0.92 (0.89; 0.97)	0.94 (0.90; 0.98)
p-value	0.28	0.15	0.11	0.19
AUCTE -AUCPSWE	0.14	0.10	0.097	0.12
intention to diagnose analysis (n = 235)
pSWE	0.77 (0.70; 0.84)	0.81 (0.76: 0.87)	0.86 (0.81; 0.92)	0.87 (0.80; 0.93)
TE	0.77 (0.71; 0.84)	0.85 (0.80; 0.90)	0.88 (0.83; 0.93)	0.89 (0.84; 0.95)
p-value	0.90	0.25	0.66	0.48

Comparison of TE and pSWE with serum markers was possible in the ‘per protocol’ analysis and for 155 patients with ELF and 139 patients with Fibrotest; in the ‘intention to diagnose’ analysis these were 194 and 174 patients, respectively. The failure rate of Fibrotest was 0 % following the security algorithms on the industrial website. Since no quality criteria are available for ELF, the failure rate was 0 % here too. Comparison of ELF and Fibrotest was possible in 131 and 159 patients, respectively. Results are shown in [Table 3]. No significant difference was found between pSWE, TE, ELF, and Fibrotest for all fibrosis stages in the ‘intention to diagnose’ analysis. However, TE was significantly better than ELF for the diagnosis of severe fibrosis (F ≥ 3) and liver cirrhosis (F4) and significantly better than Fibrotest for the diagnosis of liver cirrhosis in the ‘per protocol’ analysis ([Table 3, ] [Fig. 1], [2]). When comparing all 4 tests in the ‘per protocol’ analysis using pSWE as the reference, no significant difference was found for F ≥ 1 (p = 0.70), F ≥ 2 (p = 0.34), F ≥ 3 (p = 0.079). Only for F = 4 significant difference was found (p = 0.011), which was due to the difference in AUROC between TE and ELF (94 % vs. 84 %).

Table 3 Area under ROC curve (95 % confidence interval) for pSWE, TE, ELF score and Fibrotest according to Metavir fibrosis stage in per protocol analysis.

method “per protocol”	F ≥ 1 (F1,2,3,4)	F ≥ 2 (F2,3,4)	F ≥ 3 (F3,4)	F = 4
pSWE (n = 155)	0.77 (0.68; 0.86)	0.82 (0.75; 0.89)	0.89 (0.82; 0.95)	0.89 (0.81; 0.96)
ELF (n = 155)	0.76 (0.67; 0.86)	0.83 (0.76; 0.89)	0.85 (0.79; 0.92)	0.85 (0.77; 0.93)
p-value	0.91	0.96	0.49	0.45
TE (n = 155)	0.79 (0.70; 0.88)	0.87 (0.81; 0.92)	0.92 (0.88; 0.97)	0.94 (0.89; 0.98)
ELF (n = 155)	0.76 (0.66; 0.86)	0.83 (0.76; 0.89)	0.85 (0.79; 0.92)	0.85 (0.77; 0.93)
p-value	0.58	0.27	0.044[1]	0.0221
pSWE (n = 139)	0.78 (0.70; 0.87)	0.85 (0.77; 0.92)	0.89 (0.82; 0.96)	0.88 (0.80; 0.96)
Fibrotest (n = 139)	0.80 (0.71; 0.89)	0.85 (0.79; 0.92)	0.87 (0.81; 0.94)	0.86 (0.79; 0.93)
p-value	0.83	0.83	0.75	0.69
TE (n = 139)	0.77 (0.68; 0.87)	0.88 (0.83; 0.94)	0.94 (0.90; 0.98)	0.94 (0.90; 0.99)
Fibrotest (n = 139)	0.80 (0.71; 0.89)	0.85 (0.79; 0.92)	0.87 (0.81; 0.94)	0.86 (0.80; 0.93)
p-value	0.66	0.42	0.062	0.0151
ELF (n = 131)	0.75 (0.64; 0.85)	0.83 (0.76; 0.90)	0.85 (0.77; 0.92)	0.84 (0.75; 0.93)
Fibrotest (n = 131)	0.80 (0.71; 0.90)	0.86 (0.79; 0.92)	0.87 (0.80; 0.94)	0.85 (0.78; 0.93)
p-value	0.24	0.47	0.63	0.79
“intention to diagnose”	F ≥ 1 (F1,2,3,4)	F ≥ 2 (F2,3,4)	F ≥ 3 (F3,4)	F = 4
pSWE (n = 194)	0.78 (0.70; 0.85)	0.82 (0.76; 0.88)	0.87 (0.81; 0.93)	0.87 (0.80; 0.94)
ELF (n = 194)	0.80 (0.72; 0.88)	0.83 (0.78; 0.89)	0.86 (0.80; 0.92)	0.86 (0.80; 0.93)
p-value	0.71	0.77	0.83	0.85
TE (n = 194)	0.76 (0.68; 0.84)	0.86 (0.81; 0.91)	0.88 (0.82; 0.94)	0.90 (0.85; 0.96)
ELF (n = 194)	0.80 (0.72; 0.88)	0.83 (0.78; 0.89)	0.86 (0.80; 0.92)	0.86 (0.80; 0.93)
p-value	0.44	0.44	0.61	0.24
pSWE (n = 174)	0.79 (0.72; 0.87)	0.83 (0.77; 0.90)	0.88 (0.82; 0.94)	0.87 (0.80; 0.94)
Fibrotest (n = 174)	0.82 (0.75; 0.90)	0.86 (0.80; 0.91)	0.88 (0.82; 0.93)	0.86 (0.80; 0.92)
p-value	0.58	0.54	0.99	0.85
TE (n = 174)	0.74 (0.66; 0.83)	0.87 (0.82; 0.93)	0.89 (0.84; 0.95)	0.89 (0.83; 0.95)
Fibrotest (n = 174)	0.82 (0.75; 0.90)	0.86 (0.80; 0.91)	0.88 (0.82; 0.93)	0.86 (0.80; 0.92)
p-value	0.12	0.60	0.66	0.49
ELF (n = 159)	0.78 (0.69; 0.87)	0.83 (0.76; 0.89)	0.85 (0.78; 0.92)	0.85 (0.77; 0.93)
Fibrotest (n = 159)	0.82 (0.74; 0.91)	0.85 (0.79; 0.91)	0.87 (0.81; 0.93)	0.86 (0.79; 0.92)
p-value	0.32	0.49	0.55	0.85

¹ significant (p < 0.05)

Fig. 1 Receiver-operating characteristic (ROC) curves for ARFI imaging, TE (Fibroscan), Fibrotest, ELF score and the combined score of TE+Fibrotest and ARFI+ELF for the diagnosis of significant fibrosis (F ≥ 2) in the ‘per protocol’ analysis; no significant difference was observed when comparing all 4 methods (p = 0.34).

Abb. 1 Receiver-operating characteristic (ROC)-Kurven für pSWE (ARFI), TE (Fibroscan), Fibrotest, ELF-Score und den kombinierten Score TE+Fibrotest und pSWE+ ELF für die Diagnose einer signifikanten Fibrose (F ≥ 2) in der ‘per protocol’-Analyse; es zeigte sich kein signifikanter Unterschied zwischen den 4 Methoden (p = 0,34).

Fig. 2 Receiver-operating characteristic (ROC) curves for ARFI imaging, TE (Fibroscan), Fibrotest, ELF score and the combined score of TE+Fibrotest and ARFI+ELF for the diagnosis of liver cirrhosis (F = 4) in the 'per protocol' analysis; a significant difference was observed when comparing all 4 methods (p = 0.011), which was due to the difference in AUROC between TE and ELF (94 % vs. 84 %).

Abb. 2 Receiver-operating characteristic (ROC)-Kurven für pSWE (ARFI), TE (Fibroscan), Fibrotest, ELF-Score und den kombinierten Score TE+Fibrotest und pSWE+ ELF für die Diagnose einer Leberzirrhose (F = 4) in der ‘per protocol’-Analyse; es zeigte sich ein signifikanter Unterschied zwischen den 4 Methoden (p = 0,011), dieser war bedingt durch den Unterschied der AUROC von TE und ELF (94 % vs. 84 %).

Furthermore, optimization for combining pSWE and ELF as well as TE and Fibrotest with respect to diagnostic accuracy for significant fibrosis (F ≥ 2) lead to the following scores:

Score AE = ELF² + 25 * pSWE² and Score TF = TE² + 85 * Fibrotest.

The diagnostic accuracy was comparable between both scores and slightly improved when compared to that of the single diagnostic procedures. Details are shown in [Fig. 1], [2].

pSWE and TE were in agreement in 74/182 patients (40.7 %) concerning the diagnosis or exclusion of significant fibrosis and both methods had values within the gray zone in agreement in 39/182 patients (21.4 %). Therefore, they were discordant concerning the diagnosis of significant fibrosis in 69/182 (37.9 %). Cohen’s kappa coefficient was 0.42 (Z = 8.04, p < 0.000001). A Q-test for heterogeneity was not significant (p > 0.20). For the diagnosis and exclusion of liver cirrhosis, pSWE and TE coincided in 139/182 patients (76.4 %) and both methods had values within the gray zone in 1/182 patients (0.5 %). Therefore, they were discordant concerning the diagnosis of cirrhosis in 42/182 (23.1 %). A Q-test for heterogeneity was significant (p = 0.011) and a random effect estimator for discordance rate yielded a rate of about 33 %. Cohen’s kappa coefficient was 0.55 (Z = 9.48, p < 0.000 001).

Cut-Offs and Measurement Failure

Cut-offs of pSWE, TE, Fibrotest and ELF measurements with respective sensitivity, specificity, PPV, NPV, positive likelihood ratio and negative likelihood ratio for the diagnosis and exclusion of significant fibrosis and liver cirrhosis are shown in [Table 4, ]Supplementary [Table 2], [3], [4]. Comparison of sensitivity and specificity between TE and pSWE revealed significantly higher specificity of TE for exclusion of significant fibrosis (p = 0.017) and liver cirrhosis (p = 0.0026), while no significant difference was found between the respective sensitivities for the diagnosis of significant fibrosis (p = 0.45) and liver cirrhosis (p = 0.23). Comparison between pSWE and ELF revealed no significant difference in sensitivity and specificity for the diagnosis and exclusion of significant fibrosis (p = 0.46 and p = 0.12) and for the diagnosis of liver cirrhosis (p = 1.00), while the specificity for the exclusion of liver cirrhosis was significantly higher for pSWE as compared to ELF (p = 0.00001). Comparison between pSWE and Fibrotest revealed no significant difference between sensitivities for the diagnosis of significant fibrosis and liver cirrhosis (p = 0.15 and p = 0.27). However, for the exclusion of significant fibrosis, the specificity of Fibrotest was significantly better (p = 0.00029), while for the exclusion of liver cirrhosis the specificity of pSWE was significantly higher (p = 0.00070). Comparison between TE and ELF revealed no significant difference in the sensitivity and specificity for the diagnosis and exclusion of significant fibrosis (p = 0.12 and p = 0.18) and for the diagnosis of liver cirrhosis (p = 0.13), while the specificity for the exclusion of liver cirrhosis was significantly higher for TE as compared to ELF (p < 0.000001). Comparison between TE and Fibrotest revealed no significant difference of specificity and sensitivity for the exclusion and diagnosis of significant fibrosis (p = 0.076 and p = 0.12) and of sensitivity for the diagnosis of liver cirrhosis (p = 0.063), while the specificity for exclusion of liver cirrhosis was significantly higher for TE (p < 0.000001). Comparison between ELF and Fibrotest revealed no significant difference of specificity for exclusion and sensitivity for the diagnosis of significant fibrosis (p = 0.52 and p = 1.00) and liver cirrhosis (p = 0.47 and p = 0.34).

Table 4 Cut-off values of pSWE for the diagnosis of significant liver fibrosis (FS≥ 2) and liver cirrhosis (F4). Two cut-offs are given: one with high sensitivity to exclude fibrosis/cirrhosis and one with high specificity to diagnose fibrosis/cirrhosis.

	significant fibrosis (F≥ 2)		liver cirrhosis (F4)
value	exclusion of F≥ 2	diagnosis of F≥ 2	exclusion of F4	diagnosis of F4
cut-off (m/s)	< 1.035	≥ 1.435	< 1.405	≥ 1.755
sensitivity	90.11 % (82 %; 95 %)	64.84 % (54 %; 75 %)	90.48 % (77 %; 97 %)	73.81 % (58 %; 86 %)
specificity	25.27 % (17 %; 35 %)	90.11 % (82 %; 95 %)	75.71 % (68 %; 83 %)	90.00 % (84 %; 94 %)
positive predictive value	54.67 % (46 %; 63 %)	86.76 % (76 %; 94 %)	52.78 % (41 %; 65 %)	68.89 % (53 %; 82 %)
negative predictive value	71.88 % (53 %; 86 %)	71.93 % (63 %; 80 %)	96.36 % (91 %; 99 %)	91.97 % (86 %; 96 %)
positive likelihood ratio	1.206 (1.051; 1.384)	6.556 (3.462; 12.412)	3.725 (2.737; 5.072)	7.381 (4.351; 12.522)
negative likelihood ratio	0.391 (0.192; 0.799)	0.390 (0.293; 0.520)	0.126 (0.049; 0.321)	0.291 (0.175; 0.485)
efficiency	57.69 % (50 %; 65 %)	77.47 % (71 %; 83 %)	79.12 % (73 %; 85 %)	86.26 % (80 %; 91 %)

Using TE, 65 patients (35.7 %) were in the gray zone for the diagnosis and exclusion of F ≥ 2 and 7 patients (3.8 %) for the diagnosis and exclusion of F = 4. Using pSWE, 82 patients (45.1 %) were in the gray zone for the diagnosis and exclusion of F ≥ 2 and 27 patients (14.8 %) for the diagnosis and exclusion of F = 4. Using the ELF test, 53 patients (34.2 %) were in the gray zone for the diagnosis and exclusion of F ≥ 2 and 59 patients (38.1 %) for the diagnosis and exclusion of F = 4. Using Fibrotest, 42 patients (30.2 %) were in the gray zone for the diagnosis and exclusion of F ≥ 2 and 48 patients (34.5 %) for the diagnosis and exclusion of F = 4.

To evaluate reasons for false-negative and false-positive pSWE results, logistic regression analysis was performed including the following variables: pSWE-IQR/pSWE median, pSWE success rate, AST, ALT, GGT, and BMI. GGT levels (p = 0.00052) remained the only independent risk factors for failure to exclude significant fibrosis while no independent risk factors were found for failure to diagnose significant fibrosis. IQR/pSWE (p = 0.032) and ALT (p = 0.046) remained independent risk factors for failure to exclude liver cirrhosis, while only IQR/pSWE (p = 0.033) remained an independent risk factor to diagnose liver cirrhosis.

The success rate in the ‘per protocol analysis’ was > 60 % for all pSWE measurements. To evaluate whether IQR has an influence on pSWE results, a sub-analysis was performed calculating AUROC of pSWE in patients with IQR/pSWE median < 0.30 only (n = 155). While the AUROC of pSWE for the diagnosis of F ≥ 1 (72 % vs. 75 %) and F ≥ 2 (81 % vs. 81 %) did not improve, the AUROC of pSWE for the diagnosis of F ≥ 3 (90 % vs. 88 %) and F = 4 (92 % vs. 89 %) improved.

Correlations of Noninvasive Methods with Histology and Biomarkers

A significant correlation was found between pSWE imaging, TE, ELF score and Fibrotest with the histological fibrosis stage (0.67, p < 0.0001; 0.65, p < 0.0001; 0.64, p < 0.0001; 0.64, p < 0.0001). Also a significant correlation was found between all noninvasive methods (pSWE and TE: 0.74, p < 0.0001; pSWE and ELF: 0.52, p < 0.0001; pSWE and Fibrotest: 0.49; p < 0.0001; TE and ELF: 0.60, p < 0.0001; TE and Fibrotest: 0.51; p < 0.0001; Fibrotest and ELF: 0.58, p < 0.001).

Histological necroinflammation assessed with the HAI score significantly correlated with pSWE (0.24, p < 0.0026), TE (0.19, p = 0.015), ELF (0.25, p = 0.0038), Fibrotest (0.23, p = 0.015), Actitest (0.49, p < 0.0001), ALT (0.45, p < 0.0001), AST (0.43, p < 0.0001), and histological fibrosis stage (0.38, p < 0.0001). Actitest and AST levels significantly correlated with pSWE (0.39, p < 0.0001; 0.37, p < 0.0001), TE (0.36, p < 0.0001; 0.39, p < 0.0001) ELF (0.36, p < 0.001; 0.37, p < 0.001) and Fibrotest (0.49, p = 0.024; 0.37, p < 0.0001). There was no significant correlation between pSWE, TE, ELF, Fibrotest and histological fibrosis with ALT.

Histological steatosis significantly correlated with pSWE (0.17, p = 0.036), TE (0.24, p = 0.0023), Fibrotest (0.21; p = 0.024), Steatotest (0.20, p = 0.031), triglycerides (0.26, p < 0.0019), and histological fibrosis (0.25, p = 0.0017). No significant correlation was found between histological steatosis and ELF score. Furthermore, pSWE, TE and histological fibrosis correlated significantly with Steatotest with correlations of 0.23 (p = 0.0065), 0.24 (p = 0.0041) and 0.19 (p = 0.023). No significant correlation was found for pSWE, TE and histological fibrosis with NASH test. In addition, pSWE, TE and histological fibrosis correlated significantly with fasting blood sugar with correlations of 0.38 (p < 0.0001), 0.33 (p < 0.0001) and 0.29 (p = 0.0001). There was no significant correlation between pSWE, TE and histological fibrosis with triglyceride and an inverse correlation with cholesterol.

pSWE correlated significantly with the semiquantitative scoring system of Chevalier (0.67, p < 0.0001) and with four components (0.25, p = 0.0005 with perisinusoidal space; 0.59, p < 0.0001 with portal tract; 0.64, p < 0.0001 with number of septa; 0.62, p < 0.0001 with width of septa). TE correlated significantly with the semiquantitative scoring system of Chevalier (0.66, p < 0.0001) and with four components (0.25, p = 0.0015 with perisinusoidal space; 0.54, p < 0.0001 with portal tract; 0.59, p < 0.0001 with number of septa; 0.61, p < 0.0001 with width of septa).

Discussion

The present study is the largest prospective study comparing the elastography methods pSWE and TE as well as the serum marker Fibrotest (combination of indirect fibrosis biomarkers), and ELF (combination of direct fibrosis biomarkers) in a study population infected with chronic hepatitis C. The strength of the present study is the homogeneity of chronic hepatitis C patients as compared to many elastography studies mixing different liver diseases in the same study cohort. This is the first study with the primary aim of demonstrating the non-inferiority of pSWE compared to TE. Owing to slower recruitment than expected, the study was closed before the target enrollment of 433 patients. A reason is the significant reduction of liver biopsies in patients with chronic hepatitis C in Europe since noninvasive methods have been implemented. Therefore, the non-inferiority margin for the comparison of the AUROCs was enhanced to Δ = 0.075. Nevertheless, the non-inferiority of pSWE to TE could not be demonstrated in the present study.

Paired comparison of the diagnostic accuracy of pSWE and TE, however, revealed no significant difference between both methods for the diagnosis of liver fibrosis for all fibrosis stages in the ‘per protocol’ analysis and the ‘intention to diagnose’ analysis. The diagnostic accuracy of both methods was good for the diagnosis of significant fibrosis and excellent for the diagnosis of severe fibrosis and liver cirrhosis. Nevertheless, the primary aim of the study was to prove the non-inferiority of pSWE compared to TE, but this goal could not be reached. Further improvement in pSWE technique and quality criteria seems to be necessary for subsequent reevaluation of non-inferiority.

From the results of the present study, one can conclude that in a scenario of 53 – 50 % prevalence of significant fibrosis (METAVIR fibrosis stage ≥ 2) in the ‘intention to diagnose’ and ‘per-protocol’ analyses, respectively, pSWE is not useful to exclude (LR- 0.39), TE is fairly good to exclude (LR- 0.184) and both methods are moderately useful to confirm (LR+ 6.556 and 7.000) significant fibrosis. In a scenario of 25 – 23 % prevalence of cirrhosis (METAVIR stage 4) in the ‘intention to diagnose’ and ‘per-protocol’ analyses, respectively, pSWE and TE are fairly good in excluding (LR- 0.126 and 0.110) and moderately useful in confirming (LR+ 7.381 and 8.571) liver cirrhosis. The discordant rate of pSWE and TE was 38 % for the diagnosis of significant fibrosis and 23 % to 33 % for the diagnosis of liver cirrhosis.

Direct comparison of pSWE and TE revealed significantly higher specificities of TE for the exclusion of significant fibrosis and liver cirrhosis, while the sensitivities were comparable for the diagnosis of significant fibrosis and liver cirrhosis. In addition, as a result of choosing a cut-off with > 90 % sensitivity to exclude and > 90 % specificity to diagnose significant fibrosis and/or liver cirrhosis, more patients fell within the gray zone using pSWE as compared to TE. This has to be taken into account and weighed against the significantly higher measurement failure rate of TE as compared to pSWE. For the serological markers Fibrotest and ELF, a comparable number of patients as for TE fell within the gray zone for the diagnosis and exclusion of significant fibrosis, while for the diagnosis and exclusion of liver cirrhosis TE and pSWE were superior to the serum markers.

The advantage of pSWE as compared to TE is that it is integrated in a routine ultrasound machine and can be performed with conventional ultrasound probes. The measurement site can be visualized with B-mode ultrasound and this allows more exact measurement of liver tissue by excluding small non-parenchymatous areas such as biliary structures and blood vessels within the measurement site.

In addition, significantly more patients can be examined using pSWE due to the lower measurement failure rate. In a recently published meta-analysis the inability to obtain reliable measurements was more than thrice as high for TE than for pSWE (6.6 % vs. 2.1 %, p < 0.001) [27]. However, as discussed above, more incongruent values are obtained with pSWE.

An advantage of TE is the larger measurement area of 4 cm in length as compared to only 1 cm for pSWE. However, the possibility to measure in different areas of the liver might overcome these shortcomings of pSWE. In addition, no clear quality criteria are available for pSWE at present, while for TE quality criteria have been developed in large prospective studies [28]. A previous study with predominantly patients with liver cirrhosis reported better results for pSWE using both quality criteria of TE (success rate > 60 % and IQR< 0.30) [29]. In the present study, IQR/pSWE median was a significant independent risk factor for pSWE failure to diagnose and exclude liver cirrhosis. In addition, GGT was an independent risk factor for pSWE failure to exclude significant fibrosis and ALT levels for pSWE failure to exclude liver cirrhosis, supporting studies that suggest adapting the cut-off to ALT levels [30] [31] [32].

No significant difference in diagnostic accuracy was found between pSWE, TE and the serum markers Fibrotest and ELF in the ‘intention to diagnose’ analysis. In addition, no significant difference was found between Fibrotest, a marker combining indirect fibrosis biomarkers and ELF, a marker combining direct (hyaluronic acid) and indirect fibrosis biomarkers.

Only in the ‘per protocol’ analysis was TE significantly better than ELF for the diagnosis of severe fibrosis (F ≥ 3) and liver cirrhosis (F4) and significantly better than Fibrotest for the diagnosis of liver cirrhosis. This is in accordance with a previously published study including

1307 patients with chronic viral hepatitis which reported comparable results for TE and different serum markers for the diagnosis of significant fibrosis, but better results for TE in advanced fibrosis/liver cirrhosis [34]. However, while the inability to obtain reliable measurements was 6.6 % using TE, it was 0 % using the serum markers Fibrotest and ELF. Therefore, significantly more patients can be successfully evaluated with serum markers than TE. This is supported by a study with 1289 patients with chronic hepatitis C, reporting lower performance rates of TE compared to Fibrotest when the spectrum effects and applicability rates were taken into account [33]. However, the diagnostic performance using different cut-offs with a gray zone in between, as performed in the present study, revealed significantly fewer patients within the gray zone for TE as compared to the serological markers.

In a previous study, the diagnostic accuracy of TE for the diagnosis of significant fibrosis could be improved by approx. 10 % when TE was combined with serum markers [35]. Several algorithms have proposed the combination of different serum markers or the combination of TE with serum markers to optimize the diagnosis of significant fibrosis in hepatitis C patients [36] [37]. In a recent study the combination of TE with serum markers improved the diagnostic yield reducing the number of liver biopsies significantly more than the combination of different serum markers [36]. Also in the present study the combination of pSWE with ELF and TE with Fibrotest further improved diagnostic accuracy for the diagnosis of significant fibrosis and severe liver fibrosis as compared to each test alone.

All noninvasive methods were significantly influenced by histological necroinflammatory activity which correlated with AST and Actitest. Interestingly also the histological fibrosis stage assessed with Metavir and Chevallier Score significantly correlated with histological necroinflammatory activity in the present study implying that increased inflammatory activity induced more severe fibrosis. The best correlation with histological necroinflammation was found for Actitest^®, AST and ALT. As reported in previous studies, cut-offs of elastography (pSWE and TE) should be adapted to the levels of transaminases to prevent overestimation of fibrosis stage [30] [31] [32].

A limitation of the present study is that 10 % of patients with overt liver cirrhosis without liver biopsy were included in the study to increase the number of patients with F4 fibrosis (cirrhosis). This can bias the results of noninvasive tests and increase the diagnostic accuracy. Nevertheless, overestimation affects all noninvasive tests equally and therefore does not influence the comparison of these methods. Another limitation is the quality of liver biopsy. Biopsies including more than 6 portal tracts were included as a reference in the present study. However, sub-analysis including only patients with more than 10 portal tracts on histology did not show any difference in results.

Another limitation is that the cut-off values of pSWE shown in [Table 4] were derived from the results of the present study and not prior to the study. However, at the time the study protocol was written and the study started, no cut-off values were defined for pSWE. Therefore, to prevent bias between the noninvasive tests, cut-off values were also derived from the results for TE, Fibrotest and ELF.

A common limitation of studies evaluating noninvasive methods is that histology does not present the best reference method [4]. A study reported that even in the best scenario a perfect noninvasive method can only reach a diagnostic accuracy of 90 % using liver biopsy as the reference method [38]. The ultimate validation of liver fibrosis as a marker of liver injury is its prognostic value in terms of morbidity and mortality. In recent studies transient elastography and the serum fibrosis markers FibroTest and ELF have shown a 4 – 10-year prognostic value similar or even better to that of liver biopsy [39] [40] [41] [42]. However, pSWE is still a very novel method, so that long-term follow-up studies are not available yet. Prospective studies assessing survival without complications related to liver disease are awaited to directly compare the value of transient elastography and pSWE.

In conclusion, the non-inferiority of pSWE compared to TE for the assessment of liver fibrosis could not be shown in the present study.

Supplementary tables online as PDF file ([Table 5 – 8]).

Supplementary Table 5 Area under ROC curve (95 % confidence interval) for pSWE and transient elastography according to Metavir fibrosis stage in patients with > 10 portal fields in liver biopsy.

method	F≥ 1	F≥ 2	F≥ 3	F = 4
per protocol analysis (n = 157)
pSWE	0.77 (0.68; 0.87)	0.81 (0.74; 0.88)	0.88 (0.81; 0.94)	0.88 (0.81; 0.95)
TE	0.78 (0.69; 0.87)	0.85 (0.79; 0.91)	0.92 (0.87; 0.96)	0.93 (0.89; 0.97)
p-value	0.96	0.21	0.19	0.22
AUCFibroscan -AUCARFI	0.12	0.10	0.10	0.12
intention to diagnose analysis (n = 200)
pSWE	0.80 (0.72; 0.88)	0.83 (0.77; 0.89)	0.88 (0.82; 0.93)	0.89 (0.82; 0.95)
TE	0.79 (0.70; 0.87)	0.87 (0.83; 0.92)	0.92 (0.88; 0.96)	0.93 (0.89; 0.97)
p-value	0.83	0.10	0.071	0.16

Supplementary Table 6 Cut-off values of TE for the diagnosis of significant liver fibrosis (F≥ 2) and liver cirrhosis (F4). Two cut-offs are given: one with high sensitivity to exclude fibrosis/cirrhosis and one with high specificity to diagnose fibrosis/cirrhosis.

	significant fibrosis (f≥ 2)		liver cirrhosis (f4)
value	exclusion of f≥ 2	diagnosis of f≥ 2	exclusion of f4	diagnosis of f4
cut-off (m/s)	< 5.35	≥ 8.75	< 10.85	≥ 12.40
sensitivity	92.31 % (85 %, 97 %)	69.23 % (59 %, 78 %)	90.48 % (77 %, 97 %)	85.71 % (71 %, 95 %)
specificity	41.76 % (32 %, 53 %)	90.11 % (82 %, 95 %)	86.43 % (80 %, 92 %)	90.00 % (84 %, 94 %)
positive predictive value	61.31 % (53 %, 70 %)	87.50 % (78, 94 %)	66.67 % (53 %, 79 %)	72.00 % (58 %, 84 %)
negative predictive value	84.44 % (71 %, 94 %)	74.55 % (65 %, 82 %)	96.80 % (92 %, 99 %)	95.45 % (90 %, 98 %)
positive likelihood ratio	1.585 (1.319, 1.905)	7.000 (3.709, 13.211)	6.667 (4.339, 10.242)	8.571 (5.137, 14.303)
negative likelihood ratio	0.184 (0.087, 0.391)	0.341 (0.249, 0.468)	0.110 (0.043, 0.281)	0.159 (0.076, 0.334)
efficiency	67.03 % (60 %, 74 %)	79.67 % (73 %, 85 %)	87.36 % (82 %, 92 %)	89 % (84 %, 93 %)

Supplementary Table 7 Cut-off values of Fibrotest for the diagnosis of significant liver fibrosis (F≥ 2) and liver cirrhosis (F4). Two cut-offs are given: one with high sensitivity to exclude fibrosis/cirrhosis and one with high specificity to diagnose fibrosis/cirrhosis.

	significant fibrosis (F≥ 2)		liver cirrhosis (F4)
value	exclusion of F≥ 2	diagnosis of F≥ 2	exclusion of F4	diagnosis of F4
cut-off (m/s)	< 0.415	≥ 0.695	< 0.475	≥ 0.805
sensitivity	90.00 % (80 %, 96 %)	58.57 % (46 %, 70 %)	91.43 % (77 %, 98 %)	60.00 % (42 %, 76 %)
specificity	62.32 % (50 %, 74 %)	91.30 % (82 %, 97 %)	54.81 % (45 %, 65 %)	90.38 % (83 %, 95 %)
positive predictive value	70.79 % (60 %, 80 %)	87.23 % (74 %, 95 %)	40.51 % (30 %, 52 %)	67.74 % (49 %, 83 %)
negative predictive value	86.00 % (73 %, 94 %)	68.48 % (58 %, 78 %)	95.00 % (86 %, 99 %)	87.04 % (79 %, 93 %)
positive likelihood ratio	2.388 (1.746, 3.267)	6.736 (3.058, 14.835)	2.023 (1.600, 2.558)	6.240 (3.263, 11.933)
negative likelihood ratio	0.160 (0.078, 0.332)	0.454 (0.340, 0.605)	0.156 (0.052, 0.468)	0.443 (0.294, 0.667)
efficiency	76.26 % (68 %, 83 %)	74.82 % (67 %, 82 %)	64.03 % (55 %, 72 %)	82.73 % (75 %, 89 %)

Supplementary Table 8 Cut-off values of ELF test for the diagnosis of significant liver fibrosis (F≥ 2) and liver cirrhosis (F4). Two cut-offs are given: one with high sensitivity to exclude fibrosis/cirrhosis and one with high specificity to diagnose fibrosis/cirrhosis.

	significant fibrosis (F≥ 2)		liver cirrhosis (F4)
value	exclusion of F≥ 2	diagnosis of F≥ 2	exclusion of F4	diagnosis of F4
cut-off (m/s)	< 8.270	≥ 9.455	< 8.550	≥ 10.225
sensitivity	90.00 % (81 %, 96 %)	60.00 % (48 %, 71 %)	92.31 % (79 %, 98 %)	71.79 % (55 %, 85 %)
specificity	52.00 % (40 %, 64 %)	90.67 % (82 %, 96 %)	46.55 % (37 %, 56 %)	90.52 % (84 %, 95 %)
positive predictive value	66.67 % (57 %, 75 %)	87.27 % (76 %, 95 %)	36.73 % (27 %, 47 %)	71.79 % (55 %, 85 %)
negative predictive value	82.98 % (69 %, 92 %)	68.00 % (58 %, 77 %)	94.74 % (85 %, 99 %)	90.52 % (84 %, 95 %)
positive likelihood ratio	1.875 (1.465, 2.399)	6.429 (3.105, 13.309)	1.727 (1.425, 2.094)	7.571 (4.173, 13.736)
negative likelihood ratio	0.192 (0.096, 0.384)	0.441 (0.334, 0.583)	0.165 (0.055, 0.499)	0.312 (0.188, 0.516)
efficiency	71.61 % (64 %, 79 %)	74.84 % (67 %, 81 %)	58.06 % (50 %, 66 %)	85.81 % (79 %, 91 %)

• Reference

• 1 El Serag HB. Hepatocellular carcinoma and hepatitis C in the United States. Hepatology 2002 36: S74-S83
• 2 National Institutes of Health Consensus Development Conference Statement. Management of hepatitis C 2002 (June 10–12, 2002). Hepatology 2002; 36: S3-S20
• 3 Castera L, Negre I, Samii K et al. Pain experienced during percutaneous liver biopsy. Hepatology 1999; 30: 1529-1530
• 4 Bedossa P, Dargere D, Paradise V. Sampling variability of liver fibrosis in chronic hepatitis C. Hepatology 2003; 38: 1449-1457
• 5 Maharaj B, Maharaj RJ, Leary WP et al. Sampling variability and its influence on the diagnostic yield of percutaneous needle biopsy of the liver. Lancet 1986; 1: 523-525
• 6 EASL Clinical Practice Guidelines: management of hepatitis C virus infection. J Hepatol 2011; 55: 245-264
• 7 EASL Clinical Practice Guidelines: Management of hepatitis C virus infection. J Hepatol 2014; 60: 392-420
• 8 Tsochatzis EA, Gurusamy KS, Ntaoula S et al. Elastography for the diagnosis of severity of fibrosis in chronic liver disease: a meta-analysis of diagnostic accuracy. J Hepatol 2011; 54: 650-659
• 9 Friedrich-Rust M, Ong MF, Martens S et al. Performance of transient elastography for the staging of liver fibrosis: a meta-analysis. Gastroenterology 2008; 134: 960-974
• 10 Talwalkar JA, Kurtz DM, Schoenleber SJ et al. Ultrasound-based transient elastography for the detection of hepatic fibrosis: systematic review and meta-analysis. Clin Gastroenterol Hepatol 2007; 5 (10) 1214-1220
• 11 Stebbing J, Farouk L, Panos G et al. A meta-analysis of transient elastography for the detection of hepatic fibrosis. J Clin Gastroenterol 2010; 44: 214-219
• 12 Chon YE, Choi EH, Song KJ et al. Performance of transient elastography for the staging of liver fibrosis in patients with chronic hepatitis B: a meta-analysis. PLoS One 2012; 7: e44930
• 13 Friedrich-Rust M, Nierhoff J, Lupsor M et al. Performance of Acoustic Radiation Force Impulse imaging for the staging of liver fibrosis: a pooled meta-analysis. J Viral Hepat 2012; 19: e212-e219
• 14 Nierhoff J, Chavez Ortiz AA, Herrmann E et al. The efficiency of acoustic radiation force impulse imaging for the staging of liver fibrosis: a meta-analysis. Eur Radiol 2013; 23: 3040-3053
• 15 Palmeri ML, Wang MH, Dahl JJ et al. Quantifying hepatic shear modulus in vivo using acoustic radiation force. Ultrasound Med Biol 2008; 34: 546-558
• 16 Nightingale K, McAleavey S, Trahey G. Shear-wave generation using acoustic radiation force: in vivo and ex vivo results. Ultrasound Med Biol 2003; 29: 1715-1723
• 17 Sarvazyan AP, Rudenko OV, Swanson SD et al. Shear wave elasticity imaging: a new ultrasonic technology of medical diagnostics. Ultrasound Med Biol 1998; 24: 1419-1435
• 18 Sandrin L, Fourquet B, Hasquenoph JM et al. Transient elastography: a new non-invasive method for assessment of hepatic fibrosis. Ultrasound Med Biol 2003; 29: 1705-1713
• 19 Shaheen AA, Wan AF, Myers RP. FibroTest and FibroScan for the Prediction of Hepatitis C-Related Fibrosis: A Systematic Review of Diagnostic Test Accuracy. Am J Gastroenterol 2007; 102: 2589-2600
• 20 Bedossa P, Poynard T. An algorithm for the grading of activity in chronic hepatitis C. The METAVIR Cooperative Study Group. Hepatology 1996; 24: 289-293
• 21 Chevallier M, Guerret S, Chossegros P et al. A histological semiquantitative scoring system for evaluation of hepatic fibrosis in needle liver biopsy specimens: comparisaon with morphometric studies. Hepatology 1994; 20: 349-355
• 22 Ishak K, Baptista A, Bianchi L et al. Histological grading and staging of chronic hepatitis. J Hepatol 1995; 22: 696-699
• 23 Imbert-Bismut F, Messous D, Thibault V et al. Intra-laboratory analytical variability of biochemical markers of fibrosis (Fibrotest) and activity (Actitest) and reference ranges in healthy blood donors. Clin Chem Lab Med 2004; 42: 323-333
• 24 Poynard T, Ratziu V, Naveau S et al. The diagnostic value of biomarkers (SteatoTest) for the prediction of liver steatosis. Comp Hepatol 2005; 23: 4-10
• 25 Friedrich-Rust M, Wunder K, Kriener S et al. Liver fibrosis in viral hepatitis: noninvasive assessment with acoustic radiation force impulse imaging versus transient elastography. Radiology 2009; 252: 595-604
• 26 Youden WJ. Index for rating diagnostic tests. Cancer 1950; 3: 32-35
• 27 Bota S, Herkner H, Sporea I et al. Meta-analysis: ARFI elastography versus transient elastography for the evaluation of liver fibrosis. Liver Int 2013; 33: 1138-1147
• 28 Castera L, Foucher J, Bernard PH et al. Pitfalls of liver stiffness measurement: a 5-year prospective study of 13,369 examinations. Hepatology 2010; 51: 828-835
• 29 Bota S, Sporea I, Sirli R et al. Factors that influence the correlation of acoustic radiation force impulse (ARFI), elastography with liver fibrosis. Med Ultrason 2011; 13: 135-140
• 30 Chen YP, Liang XE, Dai L et al. Improving transient elastography performance for detecting hepatitis B cirrhosis. Dig Liver Dis 2012; 44: 61-66
• 31 Yoon KT, Lim SM, Park JY et al. Liver stiffness measurement using acoustic radiation force impulse (ARFI) elastography and effect of necroinflammation. Dig Dis Sci 2012; 57: 1682-1691
• 32 Bota S, Sporea I, Peck-Radosavljevic M et al. The influence of aminotransferase levels on liver stiffness assessed by Acoustic Radiation Force Impulse Elastography: a retrospective multicentre study. Dig Liver Dis 2013; 45: 762-768
• 33 Poynard T, de L V et al. FibroTest and Fibroscan performances revisited in patients with chronic hepatitis C. Impact of the spectrum effect and the applicability rate. Clin Res Hepatol Gastroenterol 2011; 35 (11) 720-730
• 34 Degos F, Perez P, Roche B et al. Diagnostic accuracy of FibroScan and comparison to liver fibrosis biomarkers in chronic viral hepatitis: a multicenter prospective study (the FIBROSTIC study). J Hepatol 2010; 53: 1013-1021
• 35 Zarski JP, Sturm N, Guechot J et al. Comparison of nine blood tests and transient elastography for liver fibrosis in chronic hepatitis C: the ANRS HCEP-23 study. J Hepatol 2012; 56: 55-62
• 36 Castera L, Sebastiani G, Le BB et al. Prospective comparison of two algorithms combining non-invasive methods for staging liver fibrosis in chronic hepatitis C. J Hepatol 2010; 52: 191-198
• 37 Sebastiani G, Halfon P, Castera L et al. SAFE biopsy: a validated method for large-scale staging of liver fibrosis in chronic hepatitis C. Hepatology 2009; 49: 1821-1827
• 38 Mehta SH, Lau B, Afdhal NH et al. Exceeding the limits of liver histology markers. J Hepatol 2009; 50: 36-41
• 39 Ngo Y, Munteanu M, Messous D et al. A prospective analysis of the prognostic value of biomarkers (FibroTest) in patients with chronic hepatitis C. Clin Chem 2006; 52 (10) 1887-1896
• 40 Parkes J, Roderick P, Harris S et al. European Liver Fibrosis (ELF) panel of serum markers can predict clinical outcome in a cohort of patients from England with mixed aetiology chronic liver disease. Hepatology 2007; 46: S1
• 41 Vergniol J, Foucher J, Terrebonne E et al. Noninvasive tests for fibrosis and liver stiffness predict 5-year outcomes of patients with chronic hepatitis C. Gastroenterology 2011; 140: 1970-1979
• 42 de Ledinghen V, Vergniol J, Barthe C et al. Non-invasive tests for fibrosis and liver stiffness predict 5-year survival of patients chronically infected with hepatitis B virus. Aliment Pharmacol Ther 2013; 37 (10) 979-988