Meta-analysis: ARFI elastography versus transient elastography for the evaluation of liver fibrosis

Authors


Correspondence

Markus Peck-Radosavljevic, MD

Department of Gastroenterology/Hepatology

WähringerGürtel, 18-20, A-1090, Vienna, Austria

Tel: +43 1 40400 4744

Fax: +43 1 40400 3403

e-mail: markus.peck@meduniwien.ac.at

Abstract

Aims

This meta-analysis aims to compare the diagnostic performance of acoustic radiation force impulse (ARFI) elastography and transient elastography (TE) in the assessment of liver fibrosis using liver biopsy (LB) as ‘gold-standard’.

Methods

PubMed, Medline, Lilacs, Scopus, Ovid, EMBASE, Cochrane and Medscape databases were searched for all studies published until 31 May 2012 that evaluated the liver stiffness by means of ARFI, TE and LB. Information abstracted from each study according to a fixed protocol included study design and methodological characteristics, patient characteristics, interventions, outcomes and missing outcome data.

Results

Thirteen studies (11 full-length articles and 2 abstracts) including 1163 patients with chronic hepatopathies were included in the analysis. Inability to obtain reliable measurements was more than thrice as high for TE as that of ARFI (6.6% vs. 2.1%, P < 0.001). For detection of significant fibrosis, (F ≥ 2) the summary sensitivity (Se) was 0.74 (95% CI: 0.66–0.80) and specificity (Sp) was 0.83 (95% CI: 0.75–0.89) for ARFI, while for TE the Se was 0.78 (95% CI: 0.72–0.83) and Sp was 0.84 (95% CI: 0.75–0.90). For the diagnosis of cirrhosis, the summary Se was 0.87 (95% CI: 0.79–0.92) and Sp was 0.87 (95% CI: 0.81–0.91) for ARFI elastography, and, respectively, 0.89 (95% CI: 0.80–0.94) and 0.87 (95% CI: 0.82–0.91) for TE. The diagnostic odds ratio of ARFI and TE did not differ significantly in the detection of significant fibrosis [mean difference in rDOR = 0.27 (95% CI: 0.69–0.14)] and cirrhosis [mean difference in rDOR = 0.12 (95% CI: 0.29–0.52)].

Conclusion

Acoustic radiation force impulse elastography seems to be a good method for assessing liver fibrosis, and shows higher rate of reliable measurements and similar predictive value to TE for significant fibrosis and cirrhosis.

Evaluation of liver fibrosis is essential in all chronic hepatopathies because the prognosis of the disease and the treatment decisions often depend on its severity. Liver biopsy (LB) is still considered the ‘gold standard’ for liver fibrosis assessment, despite being an invasive method and not totally risk free [1].

Liver biopsy not only enjoys limited popularity in our patients because of its invasive nature and to the pain it causes, but also there is a risk of complications associated with it, even if it is low [1]. It has also been criticized in the past, because it evaluates only 1/50,000 of the total volume of the liver because of the small volume of the tissue sample [2], which can give birth to a significant sampling error. It has been proven that liver fragments obtained in the same session, by laparoscopic biopsy, from the left and right lobe of the liver, may have different stages of fibrosis in almost half of the cases [3]. The size of the specimen obtained through LB is also very important [4, 5]. The type of needle and the stage of fibrosis are particularly important when performing LB, because the sample is most often fragmented in the case of severe fibrosis or cirrhosis [6].

In the last years, noninvasive methods, aiming to replace LB, were developed for the evaluation of liver fibrosis. These methods are either ultrasound-based methods using Transient elastography (TE) (FibroScan®) [7-9], real-time tissue elastography-Hi-RTE [10-12], Acoustic Radiation Force Impulse elastography-ARFI [13-15], or SuperSonic Shear Imaging [16, 17]; or they were based on magnetic resonance imaging elastography [18] or serological methods [19-21].

Ultrasound elastographic methods have the advantage over serologic methods regarding the fact that the readings can be obtained immediately during a patient visit without having to wait for laboratory results and also that the underlying principle is more obvious and appealing to many physicians and patients, as compared to the mathematical models employed by the serologic tests. At the moment, TE is the most widely used method for the assessment of liver fibrosis, especially in Europe, where it is available in many clinical centers. TE has been validated in chronic hepatitis B and C, nonalcoholic steatohepatitis (NASH), nonalcoholic fatty liver disease (NAFLD), and inpost transplant patients [9, 22-24]. It should be mentioned that in around 20% of the cases, reliable measurements cannot be obtained by TE using the standard M-probe [25]. In obese patients, the rate of reliable measurements is even lower, and also TE (unlike real-time elastographic methods) cannot be performed in patients with ascites.

Transient elastography is performed with the Fibroscan® device (Echosens, Paris, France) which incorporates a 5-MHz ultrasound transducer probe mounted on the axis of a vibrator. The vibrator generates a completely painless vibration (with a frequency of 50 Hz and amplitude of 2 mm), which leads to an elastic shear wave propagating through the skin and the subcutaneous tissue to the liver. The shear wave velocity (expressed in kiloPascals-kPa) is directly related to the stiffness of the tissue [26].

In the last three years, several studies have shown the usefulness of ARFI elastography for the assessment of liver stiffness (LS) [13-15].

Acoustic radiation force impulse elastography is performed with a Siemens AcusonS2000TM (Siemens AG, Erlangen, Germany) ultrasound system. The principle of this method is that of the shearing of the examined tissue which induces a smaller strain in hard tissues than in soft ones. The ultrasound probe automatically produces an acoustic ‘push’ pulse that generates shear-waves which propagate into the tissue. Their speed, measured in meters/second (m/s), is displayed on the screen. The propagation speed increases with fibrosis severity. Using image-based localization and a proprietary implementation of ARFI technology, shear wave speed may be quantified, in a precise anatomical region, focused on a region of interest, with a predefined size, provided by the system. Measurement value and depth are also reported, and the results of the elasticity are expressed in m/s [27, 28].

In 2012, a meta-analysis summarized the value of ARFI elastography for liver fibrosis evaluation [29]. In this meta-analysis 518 patients (from 8 studies) were included, but not in all the patients LB was performed. Only 312 patients (60%; from 4 studies) were evaluated by means of LB, ARFI and TE.

As more information regarding the comparative value of ARFI and TE for liver fibrosis evaluation (considering LB as ‘gold-standard’)is available now, it seems to be the time to evaluate the performance characteristics of the two elastographic methods again.

Methods

Eligibility criteria

All studies (abstracts or full-length articles) published in English until 31 May 2012 that evaluated the LS by ARFI elastography and TE in non-transplant patients with chronic hepatopathies, using LB as ‘gold-standard’ were included. If the data presented in the article were insufficient to perform the statistical analysis, the corresponding author was contacted via e-mail to provide them. The article was excluded from the analysis if the author did not provide the relevant information.

Outcomes

The primary outcome was to assess and compare the summary sensitivity (Se), specificity (Sp), summary receiver operating characteristic (SROC) and diagnostic odds ratio of LS evaluated by ARFI and TE for the diagnosis of significant fibrosis (F ≥ 2, meaning fibrosis from F2 to F4) and cirrhosis (F = 4). The following secondary outcomes were targeted: the summary Se and Spof ARFI and TE for the prediction of significant fibrosis and cirrhosis only in patients with chronic hepatitis C, comparison of cases with failure or unreliable measurements of LS by means of ARFI and TE, and estimation of mean ARFI cut-off values for significant fibrosis and cirrhosis.

Data sources and searches

Relevant studies published until 31 May 2012 were searched in Medline, Lilacs, Scopus, Ovid, EMBASE, Cochrane and Medscape databases using the following keywords: ARFI, TE, Fibroscan®, liver stiffness, liver fibrosis and Virtual Touch Tissue Quantification.

Study selection and data extraction

Two authors independently screened titles and abstracts for potential eligibility and the full texts for final eligibility. The following data were extracted: country of origin, year of publication, number of patients, aetiology of liver disease, technical failures of ARFI and TE, histological score used for the evaluation of fibrosis in LB, the quality of specimen obtained in LB, the Spearman's r correlation coefficient between ARFI and fibrosis, TE and fibrosis, LS cut-offs (of ARFI and TE) for different stages of fibrosis, Se, Sp and receiver operating characteristics (ROC) curves of ARFI elastography and TE for different stages of fibrosis. We also evaluated the number of true positive, true negative, false positive and false negative results of ARFI and of TE for significant fibrosis (F ≥ 2) and cirrhosis (F = 4) in all the aetiologies of liver disease, as well as in patients with chronic hepatitis C.

Methodological quality

The quality of the studies was assessed using the Quality Assessment of Studies of Diagnostic Accuracy included in Systematic Review (QUADAS-2 tool) [30]. We also assessed if TE was performed according to the manufacturer's instructions: at least 10 successful measurements with a success rate (SR = the ratio of the number of successful acquisitions over the total number of acquisitions) of at least 60% and with interquartile range interval (IQR = the difference between the 75th and the 25th percentile, essentially the range of the middle 50% of the data) <30%.

Data synthesis and analysis

Data are reported as mean ± standard deviation or absolute and relative frequency, if not otherwise stated. We used sensitivity (the proportion of test positives among reference standard positives) and specificity (proportion of test negatives among reference standard negatives) as standard measures and diagnostic odds ratio as a global measure of diagnostic test accuracy. To calculate summary estimates of accuracy, we used bivariate mixed-effects regression models. We aimed at checking for outliers by plotting standardized predicted random effects of sensitivities and specificities. For ARFI, we used meta-regression to test for the influence of cut-offs on diagnostic odds ratios. For both methods, we used meta-regression to test whether publication year or sample size influenced test accuracy. To compare accuracy of TE and ARFI, we tested study-specific differences in diagnostic odds ratios vs. zero using meta-regression. To test the robustness of our results, we compared our estimates between studies with low risk of bias to the remaining studies. We judged the risk of reporting bias from funnel plots where we plotted study log DOR vs. their inverse standard errors. For calculations, we used MS Excel 2011 (Microsoft Corporation, Redmond, WA, USA) and Stata 11. Generally a two-sided P value <0.05 was considered statistically significant.

Results

Of 1092 titles identified during the initial search, 223 were excluded for the duplication of results, while 853 were excluded either because they did not report on both types of elastography, LB was not performed in all the patients, or because the abstracts were subsequently published in full. One abstract was excluded because most of the patients were included in other articles analyzed here (this was a multicentric study that compared ARFI with TE in patients with chronic hepatitis C, using LB as reference method).Two studies could not be included because the corresponding authors did not provide all the necessary data (Fig. 1). Finally, 13 studies (11 full-length articles and 2 abstracts) including 1163 patients in whom LS was evaluated by ARFI elastography and TE were retrieved for the analysis [13, 14, 31-41]. The main characteristics of the studies included in the analysis are presented in Table 1.

Table 1. Characteristics of the studies included in the meta-analysis
StudyNumber of patientsEtiologyHistological score used for liver fibrosis assessmentQuality of liver biopsy specimenCut-off ARFI (m/s) vs. cut-off TE (kPa)AUROC ARFI vs. AUROC TE
  1. CHC, chronic hepatitis C; CHB, chronic hepatitis B; NAFLD, nonalcoholic fatty liver disease; ARFI, acoustic radiation force impulse elastography; TE, transient elastography.

Ebinuma 2011

[13]

113Chronic viral and nonviral hepatopathiesMetavirAt least 1.5 cm in length

F ≥ 1: 1.02 vs. 6.2

F ≥ 2: 1.3 vs. 9.1

F ≥ 3: 1.65 vs. 11.6

F = 4: 1.88 vs. 14.3

F ≥ 1: 0.690 vs. 0.724 (P = 0.748)

F ≥ 2: 0.871 vs. 0.891 (P = 0.512)

F ≥ 3: 0.890 vs. 0.908 (P = 0.344)

F = 4: 0.817 vs. 0.888 (P = 0.90)

Lupsor 2009 [14]112CHCMetavirAt least 6 intact portal tracts

F ≥ 1: 1.19 vs. 5.2

F ≥ 2: 1.34 vs. 8.1

F ≥ 3: 1.61 vs. 9.6

F = 4: 2 vs. 13.1

F ≥ 1: 0.709 vs. 0.902 (P = 0.006)

F ≥ 2: 0.851 vs. 0.941 (P = 0.02)

F ≥ 3: 0.869 vs. 0.926 (P = 0.15)

F = 4: 0.911 vs. 0.945 (P = 0.3)

Sporea 2011 [31]197CHCMetavirAt least 6 intact portal tracts

F ≥ 2: 1.2 vs. 6.7

F = 4: 1.8 vs. 12.2

F ≥ 2: 0.84 vs. 0.87

F = 4: 0.91 vs. 0.97

Sporea 2010 [32]71CHC, CHBMetavirAt least 2 cm in length

F ≥ 2: 1.33 vs. 7.6

F = 4: 1.8 vs. 13.2

F ≥ 2: 0.649 vs. 0.731 (P = 0.47)

F = 4: 0.868 vs. 0.936 (P = 0.29)

Friedrich-Rust 2009 [33]86Chronic viral and nonviralhepatopathiesMetavir

Mean length: 22.9 ± 9.4 mm

Median length: 20 mm (10–48)

F ≥ 2: 1.37 vs. 6.3

F ≥ 3: 1.45 vs. 9.5

F = 4: 1.75 vs. 9.8

F ≥ 2: 0.84 vs. 0.86

F ≥ 3: 0.91 vs. 0.90

F = 4: 0.91 vs. 0.91

Salzl 2010 [34]48Chronic viral and nonviralhepatopathiesMetavirAt least 1.5 cm in lengthF = 4: 1.71 vs. 16F = 4: 0.849 vs. 0.955
Rizzo 2011 [35]139CHCMetavirAt least 1.5 cm in length and 10 portal tracts

F ≥ 2: 1.3 vs. 6.5

F ≥ 3: 1.7 vs. 8.8

F = 4: 2 vs. 11

F ≥ 2: 0.86 vs. 0.78 (P = 0.02)

F ≥ 3: 0.94 vs. 0.83 (P = 0.002)

F = 4: 0.89 vs. 0.80 (P = 0.09)

Jang 2011 [36]74CHC, CHBMetavirNot specified

F ≥ 2: 1.19 vs. 7.5

F = 4: 1.45 vs. 9.6

F ≥ 2: 0.701 vs. 0.712

F ≥ 3: 0.822 vs. 0.821

F = 4: 0.811 vs. 0.747

Yoneda 2010 [37]54NAFLDBruntAt least 2 cm in length

F ≥ 3: 1.77 vs. 9,9

F = 4: 1.9 vs. 16

F ≥ 3: 0.930 vs. 0.937

F = 4: 0.970 vs. 0.991

Colombo 2012

[38]

45Chronic viral and nonviral hepatopathiesMetavir

At least 2 cm in length and 11 portal tracts

Median length: 38 mm (20–48)

F = 4: 1.7 vs. 9.1

F = 4: 0.903 vs. 0.850

(P = 0.41)

Crespo 2012

[39]

59

Chronic viral and nonviral

hepatopathies

Scheuer classification

At least 1.5 cm in length and 6 portal tracts

Median length: 19 mm (15–40)

F ≥ 2: 1.43 vs. 8.2

F = 4: 2.05 vs. 16.5

Not specified
Kircheis 2012 [40]68Chronic viral and nonviralhepatopathiesMetavirNot specified

F ≥ 2: 1.32 vs. 7.6

F = 4: 1.62 vs. 13

F ≥ 2: 0.929 vs. 0.920

F = 4: 0.934 vs. 0.958

Yoon 2012 [41]97Chronic viral and nonviral hepatopathiesBatts and Ludwing scoring systemAt least 1 cm in length

F ≥ 2: 1.23 vs. 6.1

F = 4: 1.77 vs. 12.2

F ≥ 2: 0.62 vs. 0.87

F = 4: 0.81 vs. 0.86

Figure 1.

Flowchart of the selection of studies.

In Table 2, we presented the risk of bias and the concerns regarding the applicability of the studies included in this meta-analysis.

Table 2. The risk of bias and concerns regarding the applicability of the studies included in the analysis (QUADAS 2 criteria)
StudyRisk of biasApplicability Concerns
Patients selectionIndex testReference standardFlow and timingPatients selectionIndex testReference standard
Lupsor 2009 [14]LowLowLowLowLowLowLow
Friedrich-Rust 2009 [33]LowLowLowUnclearLowLowLow
Ebinuma 2011 [13]UnclearLowUnclearUnclearLowLowLow
Sporea 2011 [31]LowLowLowLowLowLowLow
Sporea 2010 [32]LowLowLowLowLowLowLow
Salzl 2010 [34]LowLowLowHighLowLowLow
Rizzo 2011 [35]LowLowLowLowLowLowLow
Jang 2011 [36]UnclearUnclearLowUnclearUnclearLowUnclear
Yoneda 2010 [37]LowLowUnclearHighLowLowUnclear
Colombo 2012 [38]LowLowUnclearUnclearLowLowLow
Crespo 2012 [39]LowLowLowLowLowLowLow
Kircheis 2012 [40]UnclearUnclearLowUnclearLowLowLow
Yoon 2012 [41]LowLowLowLowLowLowLow

From eight studies [13, 14, 32-35, 38, 39], we extracted the information about the percentage of reliable measurements for the two elastographic methods. Inability to obtain reliable measurements was more than thrice as high for TE as for ARFI elastography (6.6% vs. 2.1%, P < 0.0001). ‘Failure on TE measurement' means the impossibility to obtain any valid measurement and ‘unreliable TE measurements’ means the impossibility to obtain 10 valid measurements or improper technical parameters (IQR ≥ 30% and/or SR < 60%). Because the manufacturer did not recommend the use of technical parameters as IQR and SR for ARFI elastography, ‘failure’ here means the impossibility to obtain any valid measurement and ‘unreliable ARFI measurement’ means the impossibility to obtain 10 valid measurements. Five studies [31, 36, 37, 40, 41] included only patients with valid ARFI and TE.

The diagnostic accuracy of ARFI and TE for the prediction of significant fibrosis (F ≥ 2) was evaluated in 10 studies [13, 14, 31-33, 35, 36, 39-41]. For ARFI elastography, the summary Se was 0.74 (95% CI: 0.66–0.80), the summary Sp was 0.83 (95% CI: 0.75–0.89), the summary likelihood positive ratio (LR+) was 4.29 (95% CI: 2.89–6.37), the summary likelihood negative ratio (LR-) was 0.31 (95% CI: 0.24–0.41), the summary DOR was 13.5 (95% CI: 7.6–24) and the SROC curve was 0.85 (95% CI: 0.82–0.88). (Fig. 2A, Fig. S1A). For TE, the summary Se was 0.78 (95% CI: 0.72–0.83), the summary Sp was 0.84 (95% CI: 0.75–0.90), the summary LR+ was 4.79 (95% CI: 2.92–7.88), the summary LR- was 0.26 (95% CI: 0.19–0.35), the summary DOR was 18.3 (95% CI: 8.8–38.1) and the SROC curve was 0.87 (95% CI: 0.83–0.89) (Fig. 2B, Fig. S1B). The diagnostic odds ratio of ARFI and TE did not differ significantly for the detection of significant fibrosis [mean difference in rDOR = 0.27 (95% CI −0.69 to 0.14)].

Figure 2.

Summary sensitivity and specificity of elastographic methods for predicting significant fibrosis (A) Acoustic radiation force impulse elastography (ARFI) and (B) Transient elastography (TE).

The diagnostic accuracy of ARFI and TE for the prediction of cirrhosis (F = 4) was evaluated in 13 studies [13, 14, 31-41]. For ARFI elastography, the summary Se was 0.87 (95% CI: 0.79–0.92), the summary Sp was 0.87 (0.81–0.91), the summary LR+ was 6.48 (95% CI: 4.43–9.49), the summary LR− was 0.15 (95% CI: 0.09–0.24), the summary DOR was 42.9 (95% CI: 21.1–87.1) and the SROC curve was 0.93 (95% CI: 0.91–0.95) (Fig. 3A, Fig. S2A). For TE, the summary Se was 0.89 (95% CI: 0.80–0.94), the summary Sp was 0.87 (95% CI: 0.82–0.91), the summary LR+ was 6.79 (95% CI: 4.68–9.84), the summary LR− was 0.13 (95% CI: 0.07–0.24), the summary DOR was 51.9 (95% CI: 21.4–125.7) and the SROC curve was 0.93 (95% CI: 0.91–0.95) (Fig. 3B, Fig. S2B). The diagnostic odds ratio of ARFI and TE did not differ significantly for detection of cirrhosis [mean difference in rDOR = 0.12 (95% CI −0.29 to 0.52)].

Figure 3.

Summary sensitivity and specificity of elastographic methods for predicting cirrhosis (A) Acoustic radiation force impulse elastography (ARFI) and (B) Transient elastography (TE).

In four studies [14, 30, 32, 34] only data for patients with chronic hepatitis C were presented. For the detection of significant fibrosis, the summary Se was 0.75 (95% CI: 0.69–0.81) and summary Sp was 0.87 (95% CI: 0.73–0.94) for ARFI elastography. For TE a stable summary estimate could not be yielded.

For the detection of cirrhosis in patients with chronic hepatitis C, for ARFI the summary Se was 0.88 (95% CI: 0.79–0.91) and the summary Sp was 0.91 (95% CI: 0.86–0.94), while for TE the summary Se was 0.92 (95% CI: 0.78–0.97) and the summary Sp was 0.86 (95% CI: 0.82–0.90).

The mean optimal cut-off value of LS assessed by ARFI for the detection of significant fibrosis was 1.30 ± 0.07 m/s (median 1.31 m/s) and values ranged between 1.19–1.43 m/s.

The mean optimal cut-off value of LS assessed by ARFI for the detection of cirrhosis was 1.80 ± 0.16 m/s (median 1.8 m/s) and values ranged between 1.45–2.05 m/s.

As by meta-regression, publication year, sample size or cut-off values did not explain heterogeneity (data not shown). Estimates of diagnostic test accuracy were robust against individual study quality. Funnel plots did not show major asymmetry, which indicates the absence of relevant reporting bias (funnel plots not presented).

Discussion

Over the last 10–15 years, a major effort has been made by various groups all over the world to establish reliable and reproducible noninvasive markers of liver fibrosis to substitute LB.

Although noninvasive, these methods have also disadvantages compared to LB, as they can neither determine the aetiology (Mallory bodies, etc.) nor evaluate the presence and severity of hepatocytic fatty infiltration so far.

Our pooled analysis shows that failure to obtain a reliable reading was more than thrice as common with TE compared to ARFI elastography. This is of relevance because with the increased recognition of liver damage in overweight and obese individuals, a noninvasive, quick, and reliable office-based test to evaluate liver fibrosis would be highly desirable and very helpful in early diagnosis of potentially progressive liver disease. TE is clearly underperforming in this population, but this problem may have been partially solved by the recent development of a more sensitive ultrasound sensor for TE (XL probe) that allows liver stiffness measurement in overweight and obese patients [42, 43]. On the other hand, the studies included in this analysis did not use the technical parameters IQR and SR in case of ARFI measurements, because the manufacturer hasn't recommended their use. Recently published studies [44] show that the correlation of LS assessed by ARFI with fibrosis is significantly higher in cases of measurements with good technical parameters.

It would be interesting to compare the failure rate of ARFI and TE in obese patients and in those with severe steatosis, but unfortunately these data were not available in the included studies.

When following the example of several meta-analyses of TE in the assessment of liver fibrosis [8, 9] and grouping patients into those with fibrosis ≥2 (significant fibrosis) and with F4-cirrhosis, 7 of 9 studies showed numerical superiority of AUROC curve for significant fibrosis, and 8 of 12 studies showed a numerical superiority in case of cirrhosis for TE as compared with ARFI elastography in patients with chronic liver disease of various aetiologies. The statistical analysis performed on the all studies included in this meta-analysis showed similar summary Se, Sp, SROC and diagnostic odds ratio for ARFI elastography and TE for diagnosis of significant fibrosis and cirrhosis. Our results are different from those published recently in a meta-analysis [29], but only 312 patients from four studies were included in that meta-analysis. Regarding the best ARFI cut-offs for predicting significant fibrosis and cirrhosis, our results match well with the data published [29].

One problem of the elastographic techniques can be the influence of severe steatosis on the accuracy of LS measurements. Some TE studies [45] obtained a significant correlation of LS measurements to steatosis grade or hepatocyte ballooning, while in other studies the steatosis grade did not affect the LS measurements [46, 47]. A recently published study [48] analyzed the influence of steatosis on TE measurements using novel elastography fat-mimicking materials in phantoms for simulating different steatosis grades; a consistent overestimation of LS in steatosis was observed as compared with nonsteatosis phantoms. Studies published in the field of ARFI elastography did not show a significant correlation of LS measurements with steatosis grade [49, 50]. Another ARFI study [51] that analyzed various factors that may influence ARFI accuracy for liver fibrosis assessment in chronic hepatitis C patients showed that moderate/severe steatosis (≥30% of hepatocytes affected by steatosis on biopsy) is not a significant error factor for ARFI elastography. Thus, published data suggest that ARFI is influenced less by steatosis, but unfortunately, similar to other published meta-analyses [8, 9, 29, 52, 53], we were not able to analyze the diagnostic performance of each elastographic technique according to the steatosis severity as these data were not available in the included studies.

Even if different scoring systems were used for histological liver fibrosis evaluation in various studies, all scoring systems defined as stage 2 fibrosis, the presence of portal or periportalfibrosis with rare portal–portal septa, as stage 3 the presence of bridging fibrosis with numerous septa, and as stage 4 the presence of cirrhosis. So, it is not an important limitation of our meta-analysis. Also a previously published meta-analysis in the field of ARFI elastography [29] included patients in which the LB specimens were interpreted by various scoring systems.

The quality of LB specimen varies between the studies included in our meta-analysis. They were, however, within the performance characteristics mandated by international guidelines [1].

From the current data, one can only estimate with greater certainty the diagnostic performance of ARFI elastography only for hepatitis C patients, while data on other aetiologies such as alcoholic liver disease (ALD) or NAFLD are still quite limited [37, 49, 50]. It will also be interesting to see whether variables like an acute hepatitis flare [54], differences in central venous pressure [55], extrahepatic cholestasis [56], or the use of beta-blockers [57] will have an influence on ARFI-elastography, as demonstrated for TE. Regarding the acute hepatic flare, Karlas et al. [58] evaluated the LS by means of ARFI in three patients with acute liver failure and the values obtained in these patients were similar to the values obtained in cirrhotic patients.

The obvious advantage of ARFI elastography is that it can be performed with an ultrasound machine with which the sonographic evaluation of the liver can also be performed, thereby providing an ideal ‘one-stop shop’ of noninvasive liver evaluation even in patients with a significant amount of ascites. This might be accompanied by a slightly reduced accuracy of liver stiffness evaluation and, at least for the moment, it is bound to a specific build of ultrasound machine. TE with a dedicated machine might be better for high volume elastrography measurements, such as in a busy outpatient clinic, with a slightly better performance, the new types of probes having the additional ability to noninvasively quantify liver steatosis as well.

Acoustic radiation force impulse elastography seems to be a good method for assessing liver fibrosis (0.85 SROC curve for detecting significant fibrosis and 0.93 for diagnosing cirrhosis), especially in chronic hepatitis C patients and shows similar predictive value for significant fibrosis and cirrhosis compared to TE. A 6.6% rate of failed and unreliable measurements for TE in this meta-analysis, and up to 20% in the published studies [25] using the standard M-probe is a drawback that has to be considered but is expected to be significantly improved by the XL-probe available for the current generation of TE-machines [42, 43]. As a result of the different inclusion criteria into multiple studies, the various aetiologies of chronic liver diseases (many patients with chronic viral hepatopathies, especially chronic hepatitis C, but relatively few with NAFLD and ALD), some very preliminary reporting of the results only in abstract form, a definitive conclusion regarding fibrosis evaluation cannot be drawn from the available data. As the size of published studies is still limited, a large prospective and ideally multicenter comparative evaluation of both methods seems warranted. Such an evaluation should be performed using the new XL-probe for TE, able not only to perform better in obese patients, but also to evaluate steatosis, in addition to liver fibrosis.

Acknowledgements

Financial support: none

Conflict of interests: none

Ancillary