Prospective Evaluation of Acoustic Radiation Force Impulse (ARFI) Elastography and High-Frequency B-Mode Ultrasound in Compensated Patients for the Diagnosis of Liver Fibrosis/Cirrhosis in Comparison to Mini-Laparoscopic Biopsy

Abstract

Purpose: Ultrasound is a well-established noninvasive test for assessing patients with liver disease. This study aims to prospectively compare ultrasound to the new technique elastography (ARFI) for the assessment of liver fibrosis/cirrhosis.

Materials and Methods: High-frequency B-mode ultrasound (liver surface/vein irregularity, liver homogeneity, spleen size), ARFI quantification, mini-laparoscopic liver evaluation including biopsy were prospectively obtained in compensated patients scheduled for liver biopsy. For the diagnosis of cirrhosis, a combined gold standard (cirrhosis at histology and/or at macroscopic liver evaluation) was used.

Results: Out of 157 patients, 35 patients were diagnosed cirrhotic. Ultrasound (combination of liver vein and/or surface irregularity) showed no significant difference compared to ARFI quantification for the diagnosis of significant liver fibrosis (Ishak> = 3) and cirrhosis. Diagnosis of cirrhosis had a sensitivity/specificity/PPV/NPV of 83 %(± 12) / 82 %(± 7) / 57 %(± 14) / 94 %(± 4), respectively, with ultrasound and 86 %(± 12) / 81 %(± 7) / 57 %(± 13) / 95 %(± 4), respectively, with ARFI quantification. The sensitivity/specificity/PPV/NPV for the detection of significant fibrosis were 68 %(± 13) / 86 %(± 7) / 71 %(± 13) / 84 %(± 7), respectively, for ultrasound and 70 %(± 12) / 84 %(± 7) / 69 %(± 12) / 84 %(± 7), respectively, for ARFI quantification.

Conclusion: ARFI elastography and high-frequency B-mode ultrasound show similar and good results for the diagnosis of compensated liver cirrhosis and high-grade fibrosis. A key benefit of both methods is the high NPV suggesting them as noninvasive exclusion tests.

Zusammenfassung

Ziel: Die Ultraschalluntersuchung ist ein etabliertes, nichtinvasives Verfahren bei der Abklärung von Lebererkankungen. Diese Studie soll prospektiv den Nutzen der Ultraschalluntersuchung im Vergleich zur ARFI-Elastografie als neue Methode zur Beurteilung der Leberfibrose/-zirrhose untersuchen.

Material und Methoden: Hochfrequenter B-Bild-Ultraschall (Leberoberflächen-/Venenirregularität, Leberhomogenität, Milzgröße), ARFI-Quantifzierung, Minilaparoskopie mit makroskopischer Beurteilung und Leberbiopsie wurden prospektiv bei kompensierten Patienten mit geplanter Leberbiopsie durchgeführt. Zur Diagnose einer Zirrhose wurde ein kombinierter Goldstandard (Histologie und/oder makroskopische Beurteilung) verwendet.

Ergebnisse: Von 157 Patienten hatten 35 eine Leberzirrhose. Die Ultraschalluntersuchung (Kombination von Oberflächen- und/oder Venenirregularität) zeigte keinen signifikanten Unterschied zur ARFI-Quantifizierung bei der Diagnose einer signifikanten Fibrose (Ishak> = 3) und einer Leberzirrhose. Die Sensitivität/Spezifität/PPV/NPV bei Frage nach Leberzirrhose betrug 83 %(± 12) / 82 %(± 7) / 57 %(± 14) / 94 %(± 4) mit dem Ultraschall und 86 %(± 12) / 81 %(± 7) / 57 %(± 13) / 95 %(± 4) mit der ARFI – Quantifizierung. Die Sensitivität/Spezifität/PPV/NPV für die Detektion einer signifikanten Fibrose bertrug 68 %(± 13) / 86 %(± 7) / 71 %(± 13) / 84 %(± 7) mit dem Ultraschall und 70 %(± 12) / 84 %(± 7) / 69 %(± 12) / 84 %(± 7) mit der ARFI-Quantifizierung.

Zusammenfassung: Die ARFI-Elastografie und der hochfrequente B-Bild-Ultraschall zeigen ähnliche gute Ergebnisse für die Diagnose einer kompensierten Leberzirrhose und höhergradigen Leberfibrose. Ein wesentlicher Vorteil beider Verfahren ist das hohe NPV, welches beide Methoden insbesondere als nichtinvasive Ausschlussverfahren interessant macht.

Key words

liver fibrosis - liver cirrhosis - B-mode ultrasound - ARFI elastography - mini-laparascopy

Introduction

Fibrosis of the liver with end stage cirrhosis is a hallmark of chronic liver disease. It is important to assess the stage of this process for prognosis, surveillance strategies and some treatment decisions. The gold standard, liver biopsy, is not only an invasive procedure, but its quality has also been questioned by several authors due to sampling error and other reasons [1] [2] [3] [4]. Therefore, there is constant effort to develop alternative tools to assess liver fibrosis.

When assessing patients with known or suspected chronic liver disease, ultrasound is usually one of the first tests performed. As the diagnostic accuracy is very high in advanced decompensated cirrhosis, it is considered less accurate in compensated cirrhosis or in earlier stages of fibrosis. In the last years noninvasive elastography methods with promising results evolved to assess the degree of fibrosis, nicely summarized in the recent EFSUMB Guidelines and Recommendations on the clinical use of ultrasound elastography [5] [6]. A promising technique is point shear wave elastography (pSWE) embedded in ultrasound machines, which has been described in the literature as acoustic radiation force impulse (ARFI) quantification. Studies have shown a good correlation between liver fibrosis and ARFI quantification measurements [7]. However, the diagnostic benefit of the new elastography methods directly compared to the widely accepted and established B-mode ultrasound examination is not sufficiently defined with regard to the diagnosis of liver fibrosis and compensated cirrhosis.

Therefore, we conducted this study to evaluate and prospectively compare conventional B-mode liver ultrasound, including high-frequency ultrasound, with ARFI elastography in clinical use for the assessment of liver fibrosis and cirrhosis in patients with compensated liver disease. For the diagnosis of cirrhosis, we used a combined gold standard of liver histology and/or macroscopic appearance at mini-laparoscopy (nodular surface) as this has been proposed to improve the imperfect gold standard, liver biopsy, by reducing the false negatives regarding the diagnosis of cirrhosis [8] [9] [10] [11] [12].

Materials and Methods

Patients

Consecutive patients scheduled for liver biopsy for the staging of liver disease including the assessment of the degree of fibrosis were recruited between February 2011 and August 2014 at our department. Laboratory tests included blood count, AST, ALT, INR, γGT, albumin, and bilirubin. In a stepwise approach the examiner decided for or against the diagnosis of cirrhosis after liver ultrasound, ARFI quantification and macroscopic appearance in mini-laparoscopy. Patients with liver decompensation as well as signs of portal hypertension (esophageal varices, recanalized umbilical vein, ascites) were excluded. Also patients with acute liver damage (elevated transaminases > 7 ULN), obstructive cholestasis at ultrasound and severe heart insufficiency (NYHA III-IV) were not included, since liver stiffness might be increased in those cases [13] [14] [15]. The diagnostic accuracy of liver ultrasound and ARFI quantification for the diagnosis of liver cirrhosis was compared to a combined gold standard (cirrhosis at histology defined by an Ishak 5 – 6, and/or macroscopic liver surface nodularity assessed during mini-laparoscopy). Informed consent was obtained from all patients. Elastography studies were approved by the local ethics committee and the protecol conforms to the ethical guidelines of the 1975 Declaration of Helsinki. The study was registered at ClinicalTrials.gov (Identifier: NCT01 807 013).

Ultrasound

Ultrasound was performed with a mean time difference of 9 days before mini-laparoscopy. B-mode ultrasound was performed by doctors who regularly performed abdominal ultrasound for at least 4 years, using a 4 MHz convex probe for conventional liver ultrasound and a linear 9 MHz probe (ultrasound system S2000, Siemens Medical Solutions, Erlangen, Germany) for high frequency evaluation of the liver surface. B-mode liver ultrasound included the assessment of liver surface irregularity, vein irregularity, angle of inferior border greater than 45 degrees, splenomegaly (defined as a longitudinal diameter over 13 cm), homogeneity of parenchyma and steatosis. Ultrasound diagnosis of liver cirrhosis was based on the presence of an irregular liver surface ([Fig. 1c, d]) and/or irregular liver veins ([Fig. 1e, f]).

Fig. 1 Mini-laparoscopic appearance of a healthy liver with a normal plain surface a and a cirrhotic liver with a nodular surface b. Ultrasonographic appearance of the liver surface (c, d, marked with white arrows) and the middle liver vein (e, f, marked with white arrows) for a healthy c, e and a cirrhotic d, f liver, respectively.

Abb. 1 Mini-laparoskopische Darstellung einer gesunden Leber mit einer normalen glatten Oberfläche a und einer zirrhotisch veränderten Leber mit einer nodulären Oberfläche b. Ultrasonografische Darstellung der Leberoberfläche (c, d, markiert durch weiße Pfeile) und der mittleren Lebervene (e, f, markiert durch weiße Pfeile) am Beispiel einer gesunden c, e und einer zirrhotisch d, f veränderten Leber.

Shear wave liver elastography

Liver stiffness was assessed by ARFI quantification (ultrasound system S2000, Siemens Medical Solutions, Erlangen, Germany) at a frequency of 4 MHz. In brief, ARFI quantification uses a short (approximately 100 microseconds), high-energy, focused acoustic pulse to generate a radiation force which displaces the tissue [16]. This generates lateral-travelling shear waves, which can be detected by ultrasound. In linear, isotropic, elastic solids, the speed of shear wave propagation c is related to shear modulus G and the tissue’s density ρ by: c = √^G/_p [6] [17] [18]. The shear modulus G is the tissue property that determines the amount of shear strain (relative change in shape) produced by a given directionally applied stress. In simple terms shear waves travel faster the “stiffer” the tissue is.

The study protocol included at least 7 ARFI quantification measurements with an intercostal approach to the right liver lobe during relaxed breathing arrest in a supine position. On average, 8.1 intercostal measurements were recorded in each patient. The region of interest (10*5 mm²) was placed in the liver parenchyma with a distance of more than 2 cm from the liver capsule making sure not to include visible vessels. In most cases ultrasound and ARFI exams were done by the same doctor on the same day, performing the ultrasound exam before the ARFI measurements.

Mini-laparoscopy

Mini-laparoscopy was performed under analgosedation with 50 mg pethidin i. v. and 2 – 7 mg midazolam i. v. as required during the investigation and local skin anesthesia with lidocaine (5 ml s. c.). The word “mini” refers to the fact that a small-diameter laparoscope is used for a minimally invasive procedure [12]. A 1.9-mm diameter end-viewing 0° optical instrument, which was inserted through the same 2.75-mm trocar as the Veress needle (Richard Wolf GmbH, Knittlingen, Germany) was used for inflating the pneumoperitoneum with about 2 l NO₂. The liver surface was macroscopically evaluated in three categories: 1 – normal: smooth surface of the liver without recoveries, unbroken light reflex ([Fig. 1a]); 2 – intermediate: many recoveries of surface, broken light reflex, rounded lower margin, but no nodular changes; 3 – cirrhosis: nodular changes of liver surface ([Fig. 1b]). Liver biopsy was obtained under video-optic control using an 18 G Bard Monopty needle (Bard, Covington, Georgia, USA), yielding a 17 mm tissue cylinder. In total, 154 biopsy samples of the right liver lobe and 81 samples of the left liver lobe were obtained. After biopsy, the puncture site was inspected carefully. In the case of bleeding lasting longer than two minutes, coagulation of the biopsy site was performed. There were no major complications during or after mini-laparoscopy. The doctor performing the mini-laparoscopy was not blinded to the ultrasound exam as pronounced hepatosplenomegaly or the presence of collateral vessels makes laparoscopy more dangerous.

Histology/gold standard

Liver specimens were cut according to standard protocols and stained with hematoxylin, eosin and Gieson’s stain. Liver specimens with a length of less than 10 mm and less than 6 portal fields were excluded (n = 5 patients). The mean length of liver specimens was 22 mm with 9 portal fields on average. Hepatic fibrosis was classified using the Ishak scoring system [19]. Ishak 1: Fibrous expansion of some portal areas, with or without short fibrous septa; Ishak 2: Fibrous expansion of most portal areas, with or without short fibrous septa; Ishak 3: Fibrous expansion of most portal areas with occasional portal-portal bridging; Ishak 4: Fibrous expansion of most portal areas with marked bridging; Ishak 5: Marked bridging with occasional nodules; Ishak 6: Cirrhosis, probable or definite.

A combined gold standard for the diagnosis of cirrhosis defined by an Ishak fibrosis score ≥ 5 and/or by nodular changes of the liver surface at mini-laparoscopy was used. Thus, if referring to the group of individuals with an Ishak of 0 – 4, those with nodular changes of the liver surface are excluded, as they are regarded as cirrhotic. Significant fibrosis was defined as an Ishak fibrosis score > = 3 analogously to former studies [20].

Statistical analysis

Statistical analysis was performed using R (version 2.15.3, www.r-project.org) and Graph Pad Prism (version 5, GraphPad Software Inc., La Jolla, USA). The Wilcoxon-Mann-Whitney U-test was used for the comparison of means. For the gender distribution the chi-square test was used. All tests were considered significant if p < 0.05. Mean values are shown with 95 %-CI. Median values are shown with the 25th and 75th percentile in brackets in the text if not otherwise stated. The McNemar test was used to compare the accuracy of the tests [21]. P-values were corrected for multiple tests according to Benjamini and Hochberg [22]. A diagnostic test is regarded as perfect if the AUROC is 100 %, excellent if the AUROC is greater than 90 % and good if it is greater than 80 % [24] [25].

Results

Patients

A total number of 213 patients were scheduled for liver biopsy in the recruitment period. The study’s flow diagram including exclusion reasons and the underlying etiologies of liver disease are shown in [Fig. 2]. Patient characteristics of the 157 individuals included in the final analysis are displayed in [Table 1] subdivided into non-cirrhotic and cirrhotic individuals. The etiologies were HCV (n = 34), HBV (n = 22), PBC (n = 20), autoimmune hepatitis (n = 13), NASH (n = 16), PSC (n = 5), hemochromatosis (n = 3), toxic (n = 5), Wilson’s disease (n = 2), alcohol (n = 3) and unclear (n = 34).

Fig. 2 Flow diagram of the study with reasons for exclusion as well as etiologies of underlying liver diseases.

Abb. 2 Flussdiagramm mit Ausschlusskriterien und zugrundeliegenden Ätiologien der Lebererkrankungen.

Table 1 Characteristics of cirrhotic and non-cirrhotic patients by the combined gold standard (Ishak 5/6 or laparoscopic nodular changes of liver surface) are shown.

characteristics	total	cirrhosis	non-cirrhotic	p-value
total	157	35	122
gender (m/f)	75 (48 %)/82 (52 %)	17 (49 %)/18 (51 %)	58 (48 %)/64 (52 %)
age (years)	50 (+/– 13)	58 (+/– 11)	47 (+/– 12)	< 0.001
BMI (kg/m2)	27 (+/– 5)	29 (+/– 5)	26 (+/– 5)	0.002
bilirubin (mg/dl)	1.0 (+/– 1.6)	1.2 (+/– 1.2)	1.0 (+/– 1.7)	0.003
ALT (U/l)	57 (+/– 37)	76 (+/– 45)	51 (+/– 32)	< 0.001
AST (U/l)	74 (+/– 53)	76 (+/– 48)	73 (+/– 54)	0.545
γGT (U/l)	184 (+/– 318)	364 (+/– 572)	132 (+/– 161)	< 0.001
platlet count (103/μL)	193 (+/– 65)	144 (+/– 60)	207 (+/– 59)	< 0.001
INR	1.0 (+/– 0.1)	1.1 (+/– 0.1)	1.0 (+/– 0.1)	< 0.001
albumin (g/l)	43 (+/– 5)	40 (+/– 6)	44 (+/– 5)	0.003
portal fields	8.9 (+/– 3.8)	9.9 (+/– 4.0)	8.7 (+/– 3.7)
length of biopsy (mm)	22 (+/– 9)	22 (+/– 9)	21 (+/– 9)
F0	39 (25 %)	1 (3 %)	38 (31 %)
F1	49 (31 %)	0 (0 %)	49 (40 %)
F2	18 (11 %)	1 (3 %)	17 (14 %)
F3	14 (9 %)	2 (6 %)	12 (10 %)
F4	12 (8 %)	6 (17 %)	6 (5 %)
F5	14 (9 %)	14 (40 %)	0 (0 %)
F6	11 (7 %)	11 (31 %)	0 (0 %)

B-mode ultrasound and ARFI quantification for the diagnosis of fibrosis/compensated liver cirrhosis

For B-mode ultrasound parameters, liver surface irregularity, vein irregularity, inhomogeneity of parenchyma, splenomegaly and a combination of surface and/or vein irregularity, the sensitivity, specificity, PPV and NPV are shown for the different stages of fibrosis in [Table 2] (column 2 – 6). The highest accuracies (defined by a maximum of sensitivity + specificity) of B-mode parameters were observed for irregularity of the liver surface and/or liver veins, classifying 82 % of patients correctly when examining for liver cirrhosis. A combination of both with the “believe-the-positive” rule, which means that an individual is regarded as having a disease if at least one test is positive, was used for the ultrasonographic diagnosis of fibrosis/cirrhosis. For lower stages of fibrosis, the sensitivity of B-mode ultrasound distinctly declined, whereas the specificity showed only minor changes.

Table 2 Sensitivity, specificity, PPV and NPV are shown for ARFI elastography and several B-mode ultrasound signs for different stages of fibrosis with corresponding 95 %-CI.

	ARFI quantification			B-mode ultrasound			combination
sens./spec.	ARFI[1]	irregular surface	irregular liver veins	inhomogenity	splenomegaly	irregular surface or veins1	surface or veins or ARFI	p-values1
cirrhosis	86 (± 12)/81 (± 7)	69 (± 15)/92 (± 5)	71 (± 15)/84 (± 6)	83 (± 12)/60 (± 9)	23 (± 14)/89 (± 5)	83 (± 12)/82 (± 7)	94 (± 08)/69 (± 8)	1
ishak > = 4	78 (± 13)/81 (± 7)	61 (± 15)/92 (± 5)	68 (± 14)/86 (± 6)	78 (± 13)/60 (± 9)	24 (± 13)/91 (± 5)	78 (± 13)/84 (± 7)	90 (± 09)/71 (± 8)	0.87
ishak > = 3	70 (± 12)/84 (± 7)	53 (± 13)/94 (± 4)	58 (± 13)/88 (± 6)	75 (± 12)/63 (± 9)	21 (± 11)/90 (± 6)	68 (± 13)/86 (± 7)	83 (± 10)/74 (± 8)	1
ishak > = 2	79 (± 10)/71 (± 10)	43 (± 12)/95 (± 4)	47 (± 12)/87 (± 7)	66 (± 11)/63 (± 10)	19 (± 09)/91 (± 6)	54 (± 12)/85 (± 7)	87 (± 08)/62 (± 10)	0.59
ishak > = 1	71 (± 08)/68 (± 15)	28 (± 08)/97 (± 5)	34 (± 09)/92 (± 9)	55 (± 09)/68 (± 15)	13 (± 06)/87 (± 11)	39 (± 09)/89 (± 10)	79 (± 07)/61 (± 16)	0.004
PPV/ NPV	ARFI	irregular surface	irregular liver veins	inhomogenity	splenomegaly	irregular surface or veins	surface or veins or ARFI	ARFI cut-off (m/s)
cirrhosis	57 (± 13)/95 (± 4)	71 (± 15)/91 (± 5)	57 (± 15)/91 (± 5)	37 (± 11)/92 (± 6)	38 (± 21)/80 (± 7)	57 (± 14)/94 (± 4)	46 (± 12)/98 (± 3)	1.85
ishak > = 4	59 (± 13)/91 (± 5)	74 (± 15)/87 (± 6)	64 (± 14)/88 (± 6)	41 (± 11)/89 (± 7)	48 (± 21)/77 (± 7)	63 (± 13)/92 (± 5)	52 (± 12)/95 (± 4)	1.82
ishak > = 3	69 (± 12)/84 (± 7)	82 (± 13)/80 (± 7)	70 (± 13)/81 (± 7)	51 (± 11)/84 (± 8)	52 (± 21)/69 (± 8)	71 (± 13)/84 (± 7)	62 (± 11)/90 (± 6)	1.82
ishak > = 2	68 (± 10)/80 (± 9)	88 (± 11)/67 (± 8)	75 (± 13)/67 (± 9)	59 (± 11)/70 (± 10)	62 (± 21)/58 (± 8)	75 (± 12)/70 (± 9)	65 (± 10)/86 (± 9)	1.30
ishak > = 1	87 (± 7)/43 (± 13)	97 (± 6)/30 (± 8)	93 (± 7)/31 (± 9)	85 (± 8)/33 (± 10)	76 (± 18)/24 (± 7)	92 (± 7)/32 (± 9)	86 (± 6)/48 (± 14)	1.19

¹ P-values are shown for a comparison of ARFI quantification to a combination of surface and vein irregularity at B-mode ultrasound.

ARFI results were successfully obtained in all included patients with a median shear wave velocity of 1.32 m/s [1.08 – 2.25]. ARFI measurements had a median of 1.09 m/s [0.97 – 1.31], 1.22 m/s [0.94 – 1.53], 1.39 m/s [1.25 – 1.8], 1.41 m/s [1.15 – 2.38], 1.51 m/s [1.23 – 1.97], 2.70 m/s [2.07 – 3.49] for Ishak scores of 0 – 4 and liver cirrhosis (combined gold standard), respectively. The corresponding ROC curves including the AUROC are shown in [Fig. 3a–d] with the cut-offs optimized by the Youden index [23]. The highest AUROC is obtained for cirrhosis with 0.86 (95 %-CI 0.79 – 0.94) and declines for lower stages of fibrosis to 0.80 (95 %-CI 0.73 – 0.87) in the case of an Ishak fibrosis score of > = 2. The sensitivity, specificity, PPV and NPV for the different stages of fibrosis, calculated with the optimized cut-offs, are shown for ARFI quantification in [Table 2] (first column).

Fig. 3 ROC curves for different stages of fibrosis with corresponding AUROCs and cut-offs optimized by the Youden Index are shown.

Fig. 3 Darstellung der ROC-kurven für die verschiedenen Fibrosegrade mit den dazugehörigen AUROCs und den durch den Youden-Index optimierten Trennwerten.

Interestingly, there was no significant difference of diagnostic accuracy for detecting Ishak > = 2 (p = 0.59), Ishak > = 3 (p = 1), Ishak > = 4 (0.87) and liver cirrhosis (p = 1) between ARFI quantification and ultrasound (using the B-mode parameters: liver surface and/or liver vein irregularity). Only for the differentiation between no fibrosis to Ishak > = 1, ARFI quantification was significantly better than ultrasound (p < 0.001). The combination of ultrasound and ARFI quantification using the “believe-the-positive rule” was more sensitive but less specific as shown in [Table 2] and classifies overall fewer patients correctly than ultrasound (p = 0.014) or ARFI elastography (p < 0.001) alone.

Combined gold standard of liver histology and macroscopic evaluation during mini-laparoscopy

Nodular changes to the liver surface leading to the macroscopic diagnosis of liver cirrhosis as shown in [Fig. 1] were seen in 29 out of 157 patients (18 %). In 26 of those 29 patients (90 %), an irregular surface and/or liver veins were seen at ultrasound. However, on the other hand, an irregular surface and/or liver veins were also described in 22 of 122 patients (18 %) with a non-cirrhotic surface at mini-laparoscopy and a non-cirrhotic histology. There were 10 cases with the diagnosis of cirrhosis at mini-laparoscopy not confirmed by histology including an Ishak of 4 (n = 6), Ishak of 3 (n = 2), Ishak of 2 (n = 1) and an Ishak of 0 (n = 1). Characteristics including mini-laparoscopic pictures of those 10 patients are shown in [Table 3]. On the other hand, there were 6 patients (5 with a macroscopic intermediate but not nodular surface and 1 with a normal smooth liver surface) without macroscopic diagnosis of cirrhosis but with histologic diagnosis of cirrhosis. Compared to histology alone, the macroscopic appearance at mini-laparoscopy classified significantly more patients correctly (90 %) compared to the noninvasive tests ultrasound (77 %, p = 0.004) and elastography (80 %, p < 0.001).

Table 3 Characteristics of patients with macroscopic, but not histologic diagnosis of cirrhosis are shown including mini-laparoscopic picture and ARFI, biopsy and ultrasound results. Additional information which supports the diagnosis of cirrhosis is stated.

id	age, gender, etiology	mini-laparoscopic picture	ARFI values in m/s	ishak fibrosis score (length of biopsy in mm/number or portal fields)	irregular liver surface and/or liver veins at ultrasound	additional reason(s) supporting the diagnosis of cirrhosis
1	64, female, NASH		2.94 +/– 0.57	4 (19, 8)	yes	(+) suspected portal hypertensive gastropathy in gastroscopy (+) platelet count reduced (127*103/μL) (+) INR increased (1.37) (+) Albumin reduced (26 g/l)
2	48,male, HCV		2.7 +/– 0.76	4 (30, 17)	no	(+) suspected portal hypertensive gastropathy in gastroscopy
3	61, male, alcohol		2.56 +/– 0.54	4 (16, 10)	no	(+) platelet count reduced (107*103/μL)
4	38, male, HBV		1.3 +/– 0.54	4 (20, 12)	yes	(+) platelet count reduced (134*103/μL)
5	52, female, unclear		1.25 +/– 0.29	4 (36, 15)	yes	none
6	59, male, HCV	not available	1.15 +/– 0.61	4 (14, 10)	no	(+) at mini-laparoscopy a cirrhotic surface was only diagnosed at the left but not the right liver lobe. ARFI of the left liver lobe was 2.8 m/s (+) platelet count reduced (127*103/μL)
7	71, female PBC/AHI		2.8 +/– 0.65	3 (15, 7)	yes	none
8	74, male unclear		2.57 +/– 0.76	3 (17, 5)	yes	(+) platelet count reduced (80*103/μL)
9	55, female, unclear		3.48 +/– 0.79	2 (17, 6)	yes	none
10	51, male, NASH		3.58 +/– 0.73	0 (16, 7)	yes	none

Discussion

For ARFI elastography we found good diagnostic accuracy for the detection of significant fibrosis and cirrhosis, with a tendency towards higher AUROCs in the case of higher stages of fibrosis. To our knowledge, this is the first study that used a combined gold standard with mini-laparoscopic biopsy to evaluate ARFI quantification, which might reflect an overall better gold standard than conventional percutaneous biopsy alone (see discussion below). We collected data prospectively with a small time difference between ARFI measurements and biopsy compared to other studies assessing the accuracy of ARFI technology [26] [27]. Also we included only patients without signs or history of liver decompensation and portal hypertension (ascites, history of esophageal varices). This might partly explain the slightly lower AUROCs compared to a recent meta-analysis by Nierhoff et al. with an AUROC of 0.91 for the detection of cirrhosis [7]. In this meta-analysis the AUROC declined similar to our data in the case of lower stages of fibrosis (0.89 for Metavir > = F3, 0.84 for Metavir > = F2).

For ultrasound a study by Aube et al. concludes that cirrhosis can be correctly classified in 82 – 88 % of cases using ultrasound alone, which is similar to our results [28]. Another study on 212 compensated patients using liver surface evaluation and portal flow velocity reports correct classification of 82 % of patients when examining for liver cirrhosis [29]. For the diagnosis of liver fibrosis, B-mode ultrasound is less accurate. One recent study reports poor results on ultrasonographic surface evaluation with a sensitivity/specificity of 57 %/62 %, respectively, for significant fibrosis (defined as Metavir > = 2) and 58 %/59 %, respectively, for severe fibrosis (defined as Metavir > = 3). However, in this study also the sensitivity/specificity for the detection of cirrhosis (74 %/59 %, respectively) were lower compared to other studies which might be partly due to the fact that ultrasound pictures were evaluated retrospectively. A study by Colli et al. showed a sensitivity/specificity of 54 %/95 %, respectively, for the detection of severe fibrosis (defined as Metavir > = 3) using liver surface evaluation [30]. This is comparable to our results when examining for Ishak > = 3 looking at surface irregularity. As expected, our data shows that the accuracy of ultrasound declines further in the case of lower stages of fibrosis with a very poor sensitivity for detecting any fibrosis.

Interestingly, this prospective comparison of ARFI quantification and B-mode ultrasound shows no significant difference for the detection of Ishak fibrosis scores of > = 2, > = 3, > = 4 and > = 5. ARFI quantification was significantly better only when examining for Ishak fibrosis score > = 1. For experienced examiners ARFI elastography might not be necessary to decide for or against high-grade fibrosis or cirrhosis. Also it seems possible that in borderline cases of surface or vein evaluation the examiner might subconsciously be influenced by the clinical picture of the patient. ARFI quantification is probably easier to learn and less subjective and might be especially helpful for less experienced sonographers. A combination of ARFI elastography and B-mode ultrasound is more sensitive for detecting high-grade fibrosis/cirrhosis but less specific and classified overall significantly fewer patients correctly in this studied population.

The gold standard, liver biopsy, has been questioned by several authors mainly due to sampling error [1] [2] [3] [4], which has been described in the literature for decades [3] [4] [31] [32]. For instance, a study using surgical liver samples of patients infected with hepatitis C found, by taking multiple virtual biopsies, that the fibrosis score of only 75 % of biopsy specimens with a length of 25 mm was classified correctly in comparison to a sample of at least 2 × 3 cm² [33]. In another study 32 patients out of 212 were diagnosed as cirrhotic at ultrasound with a non-cirrhotic histology and 25 % of these developed signs of decompensated liver disease or signs of portal hypertension at a 6-month follow-up [29]. Inaccurate gold standards may overestimate (in the case of statistically dependent variables) or underestimate (in the case of statistically independent variables) test accuracy [34]. Mehta et al. report that in the “best” scenario (a biopsy sensitivity and specificity of 90 %), very good accuracy (meaning AUROC> 0.90) cannot be achieved even by a perfect test and conclude that “an alternative gold standard is needed” [35]. Mini-laparoscopy is an alternative to obtain a biopsy of the liver, being more invasive and expensive compared to percutaneous liver biopsy but showing a similar safety profile [10]. It has the advantage of visualizing the liver surface. Some authors claim that the macroscopic liver appearance combined with histology leads to a reduction of the false negatives, therefore reflecting a better gold standard [8] [9] [10] [11] [12]. In our study the macroscopic appearance at mini-laparoscopy was significantly better than ARFI and ultrasound for predicting cirrhosis, underlining the validity of macroscopic evaluation at mini-laparoscopy. 10 patients that were not diagnosed by histology were diagnosed as cirrhotic by macroscopic appearance, suggesting a sampling error, which tends to be mostly a difference of one fibrosis stage [4] [9]. As shown by the laparoscopic pictures in [Table 3] (except for #6 where no pictures were available), the liver surface is nodular in those patients. Taking this together with clinical information (ARFI results, ultrasound, blood values), as shown in [Table 3], cirrhosis is likely in most patients. Interestingly, in one case, where elastometry of the right liver lobe was not suggestive of cirrhosis, a typical cirrhotic surface was described at mini-laparoscopy only in the left liver lobe, corresponding to the observation that elastometry values were suggestive of cirrhosis in the left lobe of the liver (2.8 m/s), but not in the right lobe of the liver ([Table 3], patient #6). However, it has been shown that measurements of the left liver lobe are generally less accurate than measurements of the right liver lobe compared to histology [27]. One other case is debatable since the Ishak fibrosis score is 0 and might therefore reflect a false-positive patient ([Table 3], patient #10).

Overall it is important to note that this and previous data do not definitely prove that the addition of macroscopic information improves the gold standard, due to a lack of a perfect gold standard for the diagnosis of compensated liver cirrhosis. Our data, however, indicates as suggested by others that the gold standard histology might be improved (with only minor drawbacks regarding specificity) by adding macroscopic information obtained by mini-laparoscopic visualization of the liver. This is in line with a large study on 442 patients with mini-laparoscopic biopsy suggesting an underestimation of 26 % (28 % in our study) of cirrhotic patients by histology alone [10]. As the risk profile between mini-laparoscopic biopsy and percutaneous biopsy has been shown to be similar in a large prospective study, mini-laparoscopically guided biopsy should be considered as the first choice when assessing patients for liver cirrhosis [10]. However, in order not to reduce specificity, the final diagnosis of cirrhosis might be worth reconsidering if the histological stage of fibrosis is very low, since a sampling error tends to be primarily a difference of one stage [4] [9].

One drawback of this study is that it was performed in one single center and therefore the transferability to other facilities has not been proven. This might be especially relevant for the B-mode ultrasound examination as this is probably the most subjective parameter. A bias of single ultrasound parameters in borderline cases, subconsciously influenced by the clinical picture of the patient, can also not be ruled out. And again, we still lack a perfect gold standard for the evaluation of liver fibrosis, which should be kept in mind when interpreting noninvasive as well as invasive tests.

In summary the data of this prospective single center study suggests that shear wave liver elastometry and assessment of irregularity of the liver surface and veins by ultrasound showed comparable results for the diagnosis of compensated liver cirrhosis and high-grade fibrosis. The key benefit seems to be the high NPV suggesting both tests for the noninvasive exclusion of high-grade fibrosis/cirrhosis.

Abbreviations

ALT: alanine transaminase

ARFI: acoustic radiation force impulse

ARFI: acoustic radiation force impulse quantification

AST: aspartate transaminase

CI: confidence interval

γGT: γ-glutamyl transferase

PBC: primary biliary cirrhosis

PSC: primary sclerosing cholangitis

PPV: positive predictive value

NPV: negative predictive value

AUROC: area under the receiver operating curve

TP: true positives

FP: false positives

TN: true negatives

FN: false negatives

Acknowledgments

Lukas Pfeifer received a study grant from the GFGB (Gesellschaft für Gastroenterologie in Bayern e. V.)

• References

• 1 Castera L, Pinzani M. Biopsy and non-invasive methods for the diagnosis of liver fibrosis: does it take two to tango?. Gut 2010; 59: 861-866
• 2 Lohse AW. Rolls Royce for everybody? Diagnosing liver disease by mini-laparoscopy. J Hepatol 2011; 54: 584-585
• 3 Bedossa P, Carrat F. Liver biopsy: the best, not the gold standard. J Hepatol 2009; 50: 1-3
• 4 Regev A, Berho M, Jeffers LJ et al. Sampling error and intraobserver variation in liver biopsy in patients with chronic HCV infection. Am J Gastroenterol 2002; 97: 2614-2618
• 5 Cosgrove D, Piscaglia F, Bamber J et al. EFSUMB Guidelines and Recommendations on the Clinical Use of Ultrasound Elastography. Part 2 : Clinical Applications. Ultraschall in Med 2013; 238-253
• 6 Bamber J, Cosgrove D, Dietrich CF et al. EFSUMB guidelines and recommendations on the clinical use of ultrasound elastography. Part 1: Basic principles and technology. Ultraschall in Med 2013; 34: 169-184
• 7 Nierhoff J, Chávez 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
• 8 Poniachik J, Bernstein DE, Reddy KR et al. The role of laparoscopy in the diagnosis of cirrhosis. Gastrointest Endosc 1996; 43: 568-571
• 9 Nudo CG, Jeffers LJ, Bejarano PA et al. Correlation of Laparoscopic Liver Biopsy to Elasticity Measurements (FibroScan) in Patients With Chronic Liver Disease. Gastroenterol Hepatol (N Y) 2008; 4: 862-870
• 10 Denzer U, Arnoldy A, Kanzler S et al. Prospective randomized comparison of minilaparoscopy and percutaneous liver biopsy: diagnosis of cirrhosis and complications. J Clin Gastroenterol 2007; 41: 103-110
• 11 Helmreich-Becker I, Schirmacher P, Denzer U et al. Minilaparoscopy in the diagnosis of cirrhosis: superiority in patients with Child-Pugh A and macronodular disease. Endoscopy 2003; 35: 55-60
• 12 Hoffman A, Rahman F, Murthy S et al. Mini-laparoscopy in the endoscopy unit. Curr Opin Gastroenterol 2012; 28: 461-466
• 13 Sagir A, Erhardt A, Schmitt M et al. Transient elastography is unreliable for detection of cirrhosis in patients with acute liver damage. Hepatology 2008; 47: 592-595
• 14 Millonig G, Friedrich S, Adolf S et al. Liver stiffness is directly influenced by central venous pressure. J Hepatol 2010; 52: 206-210
• 15 Pfeifer L, Strobel D, Neurath MF et al. Liver Stiffness Assessed by Acoustic Radiation Force Impulse (ARFI) Technology Is Considerably Increased in Patients with Cholestasis. Ultraschall in Med 2014; 35: 364-367
• 16 Nightingale K, Palmeri M, Trahey G. Analysis of contrast in images generated with transient acoustic radiation force. Ultrasound Med Biol 2006; 32: 61-72
• 17 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
• 18 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
• 19 Ishak K, Baptista A, Bianchi L et al. Histological grading and staging of chronic hepatitis. J Hepatol 1995; 22: 696-699
• 20 Mohamadnejad M, Montazeri G, Fazlollahi A et al. Noninvasive markers of liver fibrosis and inflammation in chronic hepatitis B-virus related liver disease. Am J Gastroenterol 2006; 101: 2537-2545
• 21 Everitt BS. The analysis of contingency tables. London: Chapman and Hall; 1977
• 22 Benjamini YHY. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J Roy Stat SocSer B 1995; 289-300
• 23 Youden WJ. Index for rating diagnostic tests. Cancer 1950; 3: 32-35
• 24 Swets JA. Measuring the accuracy of diagnostic systems. Science 1988; 240: 1285-1293
• 25 Ebell MH. Probability of cirrhosis in patients with hepatitis C. Am Fam Physician 2003; 68: 1831-1833
• 26 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
• 27 Pfeifer L, Goertz RS, Sturm J et al. Acoustic Radiation Force Impulse (ARFI) and High-Frequency Ultrasound of the Liver Surface for the Diagnosis of Compensated Liver Cirrhosis. Ultraschall in Med 2014; 35: 44-50
• 28 Aube C, Oberti F, Korali N et al. Ultrasonographic diagnosis of hepatic fibrosis or cirrhosis. J Hepatol 1999; 30: 472-478
• 29 Gaiani S, Gramantieri L, Venturoli N et al. What is the criterion for differentiating chronic hepatitis from compensated cirrhosis? A prospective study comparing ultrasonography and percutaneous liver biopsy. J Hepatol 1997; 27: 979-985
• 30 Colli A, Fraquelli M, Andreoletti M et al. Severe liver fibrosis or cirrhosis: accuracy of US for detection--analysis of 300 cases. Radiology 2003; 227: 89-94
• 31 Abdi W, Millan JC, Mezey E. Sampling variability on percutaneous liver biopsy. Arch Intern Med 1979; 139: 667-669
• 32 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
• 33 Bedossa P, Dargere D, Paradis V. Sampling variability of liver fibrosis in chronic hepatitis C. Hepatology 2003; 38: 1449-1457
• 34 Phelps CE, Hutson A. Estimating diagnostic test accuracy using a “fuzzy gold standard”. Med Decis Making 1995; 15: 44-57
• 35 Mehta SH, Lau B, Afdhal NH et al. Exceeding the limits of liver histology markers. J Hepatol 2009; 50: 36-41