Elastography for the diagnosis of severity of fibrosis in chronic liver disease: A meta-analysis of...

10
Elastography for the diagnosis of severity of fibrosis in chronic liver disease: A meta-analysis of diagnostic accuracy E.A. Tsochatzis 1 , K.S. Gurusamy 2 , S. Ntaoula 1 , E. Cholongitas 1 , B.R. Davidson 2 , A.K. Burroughs 1,1 The Royal Free Sheila Sherlock Liver Centre and Division of Surgery, Royal Free Hospital, London NW3 2QG, UK; 2 University Department of Surgery, Royal Free Hospital, London NW3 2QG, UK Background & Aims: Transient elastography is a non-invasive method, for the assessment of hepatic fibrosis, developed as an alternative to liver biopsy. We studied the performance of elas- tography for diagnosis of fibrosis using meta-analysis. Methods: MEDLINE, EMBASE, SCI, Cochrane Library, conference abstracts books, and article references were searched. We included studies using biopsy as a reference standard, with the data necessary to calculate the true and false positive, true and false negative diagnostic results of elastography for a fibrosis stage, and with a 3-month maximum interval between tests. The quality of the studies was rated with the QUADAS tool. Results: We identified 40 eligible studies. Summary sensitivity and specificity was 0.79 (95% CI 0.74–0.82) and 0.78 (95% CI 0.72–0.83) for F2 stage and 0.83 (95% CI 0.79–0.86) and 0.89 (95% CI 0.87–0.91) for cirrhosis. After an elastography result at/ over the threshold value for F2 or cirrhosis (‘‘positive’’ result), the corresponding post-test probability for their presence (if pre-test probability was 50%) was 78%, and 88% respectively, while, if values were below these thresholds (‘‘negative’’ result), the post-test probability was 21% and 16%, respectively. No opti- mal stiffness cut-offs for individual fibrosis stages were validated in independent cohorts and cut-offs had a wide range and overlap within and between stages. Conclusions: Elastography theoretically has good sensitivity and specificity for cirrhosis (and less for lesser degrees of fibrosis); how- ever, it should be cautiously applied to everyday clinical practice because there is no validation of the stiffness cut-offs for the various stages. Such validation is required before elastography is consid- ered sufficiently accurate for non-invasive staging of fibrosis. Ó 2010 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved. Introduction Liver fibrosis represents the final common outcome of chronic liver injury and is often progressive, eventually evolving into cirrhosis [1]. Cirrhosis is the severest form of fibrosis with the worst clinical outcomes. Currently, histological examination of a liver biopsy specimen is the reference standard for the diagnosis, staging, and monitor- ing of liver fibrosis [2]. However, it is invasive and can lead to fatal bleeding [2]. Transient elastography is a non-invasive method of quantify- ing fibrosis developed as an alternative to liver biopsy. Ultra- sound elastography analyses ultrasound frequency waves which are related to the elasticity (deforming capacity) of the liver. It is simple, highly reproducible, and can be completed in 10 min in an outpatient setting with no significant expertise [3]. Magnetic resonance elastography involves measuring the elasticity of liver tissues using complex algorithms [4]. An increasing number of studies have evaluated the accuracy of elastography in the staging of fibrosis and compared it to liver biopsy. In the present study, we used meta-analysis to assess the per- formance of elastography in the diagnosis of severity of liver fibrosis using liver biopsy as the reference standard. Methods Criteria for the selection of studies We included full papers and abstracts, without language restrictions that (1) eval- uated elastography in the diagnosis of severity of liver fibrosis or monitoring thereof, using liver biopsy as the reference standard, and (2) reported on data necessary to calculate the true positive false positive, true negative and false neg- ative diagnostic results of elastography for the diagnosis of a fibrosis stage based on a defined cut-off point for liver stiffness. If such data were unavailable, the cor- responding author was contacted via e-mail to provide them; if he/she failed to reply, the study was excluded. We also excluded studies reporting on <10 patients, or in which the maximum time interval between performing elastogra- phy and liver biopsy was >3 months. Literature search MEDLINE (Pubmed), Embase, Science Citation Index expanded, and The Cochrane Hepato-Biliary Group Controlled Trials Register, the Cochrane Central Register of Controlled Trials in The Cochrane Library were searched until May 2009. Confer- ence abstracts from the American Association for the Study of Liver Diseases (AASLD) and the European Association for the Study of the Liver (EASL) were manually searched from 2007 to 2009. Reference lists of identified studies and reviews were also hand-searched. Search terms for each database are shown in the Appendix. Journal of Hepatology 2011 vol. 54 j 650–659 Received 11 March 2010; received in revised form 16 July 2010; accepted 20 July 2010; available online 24 September 2010 Corresponding author. Address: The Royal Free Sheila Sherlock Liver Centre and Division of Surgery, Royal Free Hospital, Pond Street, Hampstead, London NW3 2QG, UK. Tel.: +44 2074726229; fax: +44 2074726226. E-mail address: [email protected] (A.K. Burroughs). Research Article

Transcript of Elastography for the diagnosis of severity of fibrosis in chronic liver disease: A meta-analysis of...

Research Article

Elastography for the diagnosis of severity of fibrosis in chronicliver disease: A meta-analysis of diagnostic accuracy

E.A. Tsochatzis1, K.S. Gurusamy2, S. Ntaoula1, E. Cholongitas1, B.R. Davidson2, A.K. Burroughs1,⇑

1The Royal Free Sheila Sherlock Liver Centre and Division of Surgery, Royal Free Hospital, London NW3 2QG, UK; 2University Department ofSurgery, Royal Free Hospital, London NW3 2QG, UK

Background & Aims: Transient elastography is a non-invasive cirrhosis [1]. Cirrhosis is the severest form of fibrosis with the

method, for the assessment of hepatic fibrosis, developed as analternative to liver biopsy. We studied the performance of elas-tography for diagnosis of fibrosis using meta-analysis.Methods: MEDLINE, EMBASE, SCI, Cochrane Library, conferenceabstracts books, and article references were searched. Weincluded studies using biopsy as a reference standard, with thedata necessary to calculate the true and false positive, true andfalse negative diagnostic results of elastography for a fibrosisstage, and with a 3-month maximum interval between tests.The quality of the studies was rated with the QUADAS tool.Results: We identified 40 eligible studies. Summary sensitivityand specificity was 0.79 (95% CI 0.74–0.82) and 0.78 (95% CI0.72–0.83) for F2 stage and 0.83 (95% CI 0.79–0.86) and 0.89(95% CI 0.87–0.91) for cirrhosis. After an elastography result at/over the threshold value for F2 or cirrhosis (‘‘positive’’ result),the corresponding post-test probability for their presence (ifpre-test probability was 50%) was 78%, and 88% respectively,while, if values were below these thresholds (‘‘negative’’ result),the post-test probability was 21% and 16%, respectively. No opti-mal stiffness cut-offs for individual fibrosis stages were validatedin independent cohorts and cut-offs had a wide range and overlapwithin and between stages.Conclusions: Elastography theoretically has good sensitivity andspecificity for cirrhosis (and less for lesser degrees of fibrosis); how-ever, it should be cautiously applied to everyday clinical practicebecause there is no validation of the stiffness cut-offs for the variousstages. Such validation is required before elastography is consid-ered sufficiently accurate for non-invasive staging of fibrosis.� 2010 European Association for the Study of the Liver. Publishedby Elsevier B.V. All rights reserved.

Introduction

Liver fibrosis represents the final common outcome of chronicliver injury and is often progressive, eventually evolving into

Journal of Hepatology 20

Received 11 March 2010; received in revised form 16 July 2010; accepted 20 July2010; available online 24 September 2010⇑ Corresponding author. Address: The Royal Free Sheila Sherlock Liver Centre andDivision of Surgery, Royal Free Hospital, Pond Street, Hampstead, London NW32QG, UK. Tel.: +44 2074726229; fax: +44 2074726226.E-mail address: [email protected] (A.K. Burroughs).

worst clinical outcomes.Currently, histological examination of a liver biopsy specimen

is the reference standard for the diagnosis, staging, and monitor-ing of liver fibrosis [2]. However, it is invasive and can lead tofatal bleeding [2].

Transient elastography is a non-invasive method of quantify-ing fibrosis developed as an alternative to liver biopsy. Ultra-sound elastography analyses ultrasound frequency waveswhich are related to the elasticity (deforming capacity) of theliver. It is simple, highly reproducible, and can be completedin 10 min in an outpatient setting with no significant expertise[3]. Magnetic resonance elastography involves measuring theelasticity of liver tissues using complex algorithms [4]. Anincreasing number of studies have evaluated the accuracy ofelastography in the staging of fibrosis and compared it to liverbiopsy.

In the present study, we used meta-analysis to assess the per-formance of elastography in the diagnosis of severity of liverfibrosis using liver biopsy as the reference standard.

Methods

Criteria for the selection of studies

We included full papers and abstracts, without language restrictions that (1) eval-uated elastography in the diagnosis of severity of liver fibrosis or monitoringthereof, using liver biopsy as the reference standard, and (2) reported on datanecessary to calculate the true positive false positive, true negative and false neg-ative diagnostic results of elastography for the diagnosis of a fibrosis stage basedon a defined cut-off point for liver stiffness. If such data were unavailable, the cor-responding author was contacted via e-mail to provide them; if he/she failed toreply, the study was excluded. We also excluded studies reporting on <10patients, or in which the maximum time interval between performing elastogra-phy and liver biopsy was >3 months.

Literature search

MEDLINE (Pubmed), Embase, Science Citation Index expanded, and The CochraneHepato-Biliary Group Controlled Trials Register, the Cochrane Central Register ofControlled Trials in The Cochrane Library were searched until May 2009. Confer-ence abstracts from the American Association for the Study of Liver Diseases(AASLD) and the European Association for the Study of the Liver (EASL) weremanually searched from 2007 to 2009. Reference lists of identified studies andreviews were also hand-searched. Search terms for each database are shown inthe Appendix.

11 vol. 54 j 650–659

1289 studies identified

398 duplicates618 excluded by title and/or abstract

306 references retrieved formore detailed evaluation

94 no comparison with liver biopsy41 published in full paper42 long interval between tests80 not enough data on sensitivity-specificity5 no original data2 less than 10 patients2 no elastography performed3 did not use untrasound elastography

40 studies for analysis

Fig. 1. Flow diagram of search results and study selection.

JOURNAL OF HEPATOLOGY

Study selection and data extraction

The identified studies were screened independently by E.T. and S.N., and thenverified reciprocally. The data were extracted independently by E.T. and S.N. usinga predefined form. Any differences in study selection or data extraction wereresolved by K.S.G. and A.K.B. The following data were extracted: country ofstudy’s origin, year of publication, patient number, patients’ epidemiologicaland laboratory characteristics, aetiology of liver disease, technical failures inundertaking liver biopsy or elastography, liver stiffness cut-offs for stages offibrosis, histological score used, true positive, false positive, true negative, andfalse negative elastography results and methodological quality.

Methodological quality

The quality of the studies was assessed independently by E.T. and S.N. using theQuality Assessment of Studies of Diagnostic Accuracy included in SystematicReview (QUADAS) assessment tool which contains 14 questions [5]. We alsoassessed if the index test (elastography) was performed according to the manu-facturer’s instructions i.e. at least ten successful measurements with a successrate of at least 60% and was thus likely to correctly classify the target condition.Liver biopsy was rated as an acceptable reference standard if the specimen wasP15 mm long and included P6 portal tracts [2].

Statistical analysis and data synthesis

Data were combined using the hierarchical summary receiver operator character-istics (HSROC) method and the bivariate normal random-effects analysis of sen-sitivity and specificity within the METANDI module [6,7] in the STATA 10statistical software (Statacorp LP, Texas, USA). METANDI performs meta-analysisof diagnostic test accuracy studies in which both the index test under study andthe reference test (gold standard) are dichotomous. It fits a two-level mixed logis-tic regression model, with independent binomial distributions for the true posi-tives and true negatives within each study, and a bivariate normal model forthe logit transforms of sensitivity and specificity between studies [6]. Currently,these methods are considered more reliable than the Littenberg and Mosesmeta-analytical method of diagnostic accuracies [8]. The Metandiplot, which usesa HSROC plot, was used for graphical representation.

We evaluated pre-test probabilities of 25%, 50%, and 75% versus correspond-ing post-test probabilities following a ‘‘positive’’ or ‘‘negative’’ elastography resultbased on the summary sensitivity and specificity. ‘‘Positive’’ elastography resultswere defined as all results above the optimal liver stiffness threshold for a givenfibrosis stage, given in each individual study, while ‘‘negative’’ test results wereall results below the same threshold.

Exploration of heterogeneity, subgroup and sensitivity analysis

We planned to perform the following subgroup analyses: high versus low meth-odological quality, different stages of fibrosis (scoring systems were converted tocomparable stages in METAVIR [9]), aetiological diagnoses, different stiffness cut-offs for a specific stage of fibrosis, initial diagnosis versus monitoring of fibrosis,different specific treatments such as interferon therapy which may alter the scor-ing of liver biopsy, different ranges of body mass index (BMI) (<18.5, 18.5–24.9,25–29.9, and P30), different ranges of transaminases (normal range, betweennormal, and P3 or >3 times the upper limit of normal), and country of the study’sorigin.

We computed the diagnostic odds ratio (DOR) as a single indicator of elastog-raphy performance in each study and disease stage [10]. We used the test ofinteraction to assess if the diagnostic odds ratio (DOR) was statistically differentbetween subgroups by computing the DOR ratio and the 95% confidence intervalsof the DOR ratio between the subgroups [11].

We explored heterogeneity in each fibrosis stage by computing Higgin’s I2

and v2 tests for heterogeneity using the generic inverse variance method ofmeta-analysis of DOR [12]. We considered an I2 value of more than 30% and av2 p value of 0.10 to indicate statistically significant heterogeneity. We furtherexplored heterogeneity by using meta-regression of logarithmically transformedDOR (ln DOR) for continuous variables that could potentially cause heterogeneityand used test of interaction for binary variables in each fibrosis stage. The numberof studies in each fibrosis stage was not sufficient to perform a multiple meta-regression analysis.

We performed sensitivity analyses by excluding studies of low methodolog-ical quality and studies solely published as abstracts from the main analysis.

Journal of Hepatology 201

Publication bias

We performed a funnel plot; visual assessment was used to assess the bias, aswell as the methods by Deeks et al. [13].

Results

Description of studies

We identified 1289 references. The reference flow is shown inFig. 1. The inclusion criteria were fulfilled in 43 studies (35 fullpapers, eight abstracts) [3,4,14–54]. From these, three studieswere excluded, as they evaluated magnetic resonance elastogra-phy [4], real time elastography [27] or used a probe especiallybuilt for children [23]. Important characteristics of the includedstudies are shown in Table 1. The risk of bias of these studies isdetailed in Supplementary Fig. 1.

Transient elastography for the detection of fibrosis METAVIR stages1–4

The diagnostic accuracy of transient elastography for cirrhosis(METAVIR F = 4) was evaluated in 30 studies: the summary sen-sitivity was 0.83 (95% CI 0.79–0.86) and specificity was 0.89(95% CI 0.87–0.91). The mean optimal cut-off point of liver stiff-ness was 15 ± 4.1 kPa (median 14.5), however, values rangedfrom 9.0–26.5 kPa.

The diagnostic accuracy for F = 3 was evaluated in 24 studies,with summary sensitivity 0.82 (95% CI 0.78–0.86) and summaryspecificity 0.86 (0.82–0.89). The mean optimal cut-off was10.2 ± 1.9 kPa (median 9.6), with a range of 7.3–15.4 kPa.

The detection of F = 2 was evaluated in 31 studies, with sum-mary sensitivity 0.79 (95% CI 0.74–0.82) and specificity 0.78 (95%CI 0.72–0.83). The mean optimal cut-off was 7.3 ± 1.4 kPa (med-ian 7.2) with a range of 4.0–10.1.

1 vol. 54 j 650–659 651

Table 1. Main characteristics of the studies included in the meta-analysis.

Abbreviations: BMI: body mass index, CHC: chronic hepatitis C, CHB: chronic hepatitis B, ALD: alcoholic liver disease, NAFLD: non-alcoholic fatty liver disease.

Research Article

Finally, the detection of F = 1 was evaluated in 10 studies withsummary sensitivity 0.78 (95% CI 0.731–0.830) and summaryspecificity 0.83 (95% CI 0.72–0.0.90). The mean optimal cut-offwas 6.5 ± 1.1 kPa (median 6.0), with a range of 4.9–8.8 kPa.

The above is summarised in Table 2, while the correspondingHSROC plots are shown in Fig. 2.

652 Journal of Hepatology 201

Transient elastography and aetiology of liver disease

The aetiology of liver disease was documented in 17 studies evalu-ating patients with chronic hepatitis C (CHC), 10 studies withchronic hepatitis B (CHB), 3 studies with alcoholic liver disease,two studies with non-alcoholic fatty liver disease, and one study

1 vol. 54 j 650–659

Table 2. Summary sensitivity and specificity of: (A) all studies evaluating the diagnostic accuracy of elastography and (B) those studies with optimal biopsy qualityand elastography measurements. ‘‘Positive’’ elastography results were considered all results over the optimal liver stiffness threshold for a given fibrosis stage in eachindividual study, while ‘‘negative’’ test results all results below that threshold. Post-test (+) denotes the probability following a positive test, while post-test (-) denotes theprobability following a negative test.

≥1

≥2

≥3

4

10

31

24

30

0.78 (0.73-0.83)

0.79 (0.74-0.82)

0.82 (0.78-0.86)

0.83 (0.79-0.86)

0.83 (0.72-0.90)

0.78 (0.72-0.83)

0.86 (0.82-0.89)

0.89 (0.87-0.91)

0.250.50.750.250.50.750.250.50.750.250.50.75

0.600.820.930.550.780.920.660.850.950.720.880.96

0.080.210.440.080.210.450.070.170.390.060.160.36

Fibrosisstage

StudiesN

Sensitivity(95%CI)

Specificity(95%CI)

All studies Studies with optimal biopsy and elastographyPre-test

probabilityPost-test

(+)Post-test

(-)<4

6

7

6

-

0.75(0.65-0.83)

0.82(0.77-0.87)

0.90(0.78-0.96)

-

0.79(0.71-0.85)

0.86(0.80-0.91)

0.91(0.84-0.95)

0.250.50.750.250.50.750.250.50.750.250.50.75

0.550.780.920.670.860.950.760.910.97

0.090.240.480.060.170.380.030.100.24

StudiesN

Sensitivity(95%CI)

Specificity(95%CI)

Pre-testprobability

Post-test(+)

Post-test(-)

Abbreviations: CI: confidence interval.

JOURNAL OF HEPATOLOGY

with CHC/HIV co-infection. Further analysis was possible for CHCand CHB, as P4 studies are needed for a metanalysis using Metandi.

The diagnostic accuracy of elastography did not significantlydiffer between CHC and CHB in any METAVIR stage. The liverstiffness cut-offs had wider ranges and were higher for a givenfibrosis stage in CHC: they were 7.6 (range 5.1–10.1), 10.9 (8.0–15.4), and 15.3 (11.9–26.5) kPa for F = 2, 3, and 4, respectively,in CHC, and 7.0 (range 6.9–7.2), 8.2 (7.3–9.0), and 11.3 (9.0–13.4) kPa for F = 2, 3, and 4 respectively, in CHB. All the aboveinformation is summarised in Tables 3 and 4.

Transient elastography and stiffness cut-offs

All but three studies reported on optimal stiffness cut-offs for thediagnosis of specific fibrosis stages [28,29,41]. However, thesestudies derived these cut-offs for each stage statistically, onlyafter data collection, and no study prospectively validated themin an independent cohort. There was a wide range and overlapof cut-offs within and between different fibrosis stages. Roundingup the cut-offs to the nearest 0.5, a metanalysis was only possiblefor cut-offs P7, P9.5, and P12 Kpa in stages 2, 3, and 4, respec-tively, due to a wide distribution that resulted in small studynumbers for other cut-offs. The corresponding summary sensitiv-ity and specificity were: for F = 2, 0.70 (95% CI 0.66–0.75) and0.81 (95% CI 0.77–0.85), for F = 3 0.80 (95% CI 0.72–0.85) and0.85 (95% CI 0.79–0.90), and for F = 4, 0.86 (95% CI 0.80–0.91)and 0.88 (95% CI 0.82–0.91), respectively. No direct comparisonsof different cut-offs within the same fibrosis stage were possible.

We performed a meta-regression analysis of ln DOR and stiff-ness cut-offs in fibrosis stages 1–4. There was a statistically sig-nificant negative correlation (p <0.05) for F1, positivecorrelations for F2 and 3, and non-significance for cirrhosis. Thescatter plot is shown in Supplementary Fig. 2.

Methodological bias risk

No study was identified as being free of any bias risk. Therefore,all our estimates may be biased.

Journal of Hepatology 201

Transient elastography in studies with acceptable reference (liverbiopsy) and index (elastography) test quality

There were 12 studies with biopsies P15 mm long and containingP6 portal tracts and 21 studies where elastography measure-ments were made according to the manufacturer’s recommenda-tion. In the diagnosis of F4, there was a significant difference inDOR between studies with acceptable and unreported elastogra-phy measurements (DOR ratio 2.69, 95% CI 1.24–5.84). The sum-mary sensitivity and specificity according to reference and indextest quality are summarised in Tables 3 and 4.

Nine studies had both acceptable reference and index testquality. The summary sensitivity and specificity are summarisedin Table 2.

Transient elastography and type of publication

There was a statistical difference in the diagnostic accuracy ofF = 2 among studies published in full-paper and abstract forms(DOR ratio 2.57, 95% CI 1.44–4.58), but not in the rest of the fibro-sis stages. These results are summarised in Tables 3 and 4.

Transient elastography and ALT levels

There were no studies with a mean ALT level within normal lim-its. Data were available for comparisons only within fibrosis F2and F4 (Tables 3 and 4), for which DOR was statistically and sig-nificantly different for F = 2 (DOR ratio 1.87, 95% CI 1.01–3.55).Linear regression between logarithm of the diagnostic accuracyand ALT values did not demonstrate significant relation withinany fibrosis stage.

Further subgroup analyses

Only one study involved monitoring of fibrosis [44], and anotherscreening for fibrosis in apparently healthy individuals [55]. Nostudies evaluated the effect of treatment. There were no signifi-cant differences in summary sensitivity and specificity when

1 vol. 54 j 650–659 653

Study estimate

A B

C D

Summary point

HSROC curve 95% confidenceregion

95% prediction

0.0

0.2

0.4

0.6

0.8

1.0S

ensi

tivity

0.00.20.40.60.81.0Specificity

0.0

0.2

0.4

0.6

0.8

1.0

Sen

sitiv

ity

0.00.20.40.60.81.0Specificity

0.0

0.2

0.4

0.6

0.8

1.0S

ensi

tivity

0.00.20.40.60.81.0Specificity

0.0

0.2

0.4

0.6

0.8

1.0

Sen

sitiv

ity

0.00.20.40.60.81.0Specificity

region

Study estimate Summary point

HSROC curve 95% confidenceregion

95% predictionregion

Study estimate Summary point

HSROC curve 95% confidenceregion

95% predictionregion

Study estimate Summary point

HSROC curve 95% confidenceregion

95% predictionregion

Fig. 2. The hierarchical summary receiver operating characteristic (HSROC) plot for the sensitivity and specificity in METAVIR stages 1, 2, 3, and 4. The summarypoint represents the summary sensitivity and specificity, the 95% confidence region represents the 95% confidence intervals of the summary sensitivity and specificity andthe 95% prediction region represents the 95% confidence interval of sensitivity and specificity of each individual study included in the analysis. The plot also includes studyestimates indicating the sensitivity and specificity estimated using the data from each study separately. The size of the marker is scaled according to the total number ineach study. (A) HSROC plot for METAVIR F1, (B) HSROC plot for METAVIR F2, (C) HSROC plot for METAVIR F3 (D) HSROC plot for METAVIR F4.

Research Article

evaluating different histological scoring systems, country of ori-gin or different ranges of BMI. However, high BMI was the mostcommon cause of unsuccessful elastography measurements(Data not shown).

Exploration of heterogeneity

There was statistically significant heterogeneity in DOR in fibro-sis stages 2 (v2 p <0.001, I2 = 67%) and 4 (v2 p = 0.002, I2 = 49) butnot in stages 1 (v2 p = 0.32, I2 = 13) and 3 (v2 p = 0.27, I2 = 13). Asalready mentioned, potential sources of heterogeneity include

654 Journal of Hepatology 201

liver stiffness cut-offs in fibrosis stages 1–3, ALT levels, and typeof publication in F = 2 and index test quality in F = 4. Finally, therewas no significant correlation of ln DOR with cohort size in anyfibrosis stage.

Publication bias

The funnel plots for METAVIR 1–4 are shown in SupplementaryFig. 3. Asymmetry was visually present for MEATVIR 2 and 4.However, the studies causing the asymmetry were of high biasrisk, as 3 outlier studies overestimated the diagnostic odds ratio

1 vol. 54 j 650–659

Table 3. Subgroup analysis in METAVIR stage 2. ‘‘Positive’’ elastography results were considered all results over the optimal liver stiffness threshold for a given fibrosisstage in each individual study, while ‘‘negative’’ test results all results below that threshold. Post-test (+) denotes the probability following a positive test, while post-test (-)denotes the probability following a negative test.

CHC

CHB

Acceptable biopsy quality

Unreported biopsy quality

Low biopsy quality

Acceptable elastography quality

Unreported elastography quality

Low elastography quality

Full-paper publication

Abstract publication

ULN<ALT<3 ULN

ALT>3 ULN

14

4

11

8

11

15

7

9

32

8

13

4

0.78 (0.71-0.84)

0.84 (0.67-0.93)

0.76 (0.68-0.82)

0.81 (0.75-0.86)

0.79 (0.71-0.86)

0.79 (0.73-0.84)

0.76 (0.68-0.83)

0.81 (0.70-0.88)

0.80 (0.75-0.85)

0.74 (0.64-0.82)

0.80 (0.74-0.85)

0.65 (0.60-0.70)

0.80 (0.71-0.86)

0.78 (0.68-0.85)

0.76 (0.69-0.82)

0.78 (0.63-0.88)

0.80 (0.70-0.87)

0.82 (0.79-0.84)

0.71 (0.56-0.83)

0.75 (0.60-0.86)

0.80 (0.75-0.84)

0.69 (0.54-0.80)

0.80 (0.71-0.87)

0.82 (0.76-0.87)

13.9 (8.5-22.8)

17.9 (7.7-41.7)

10.1 (6.4-16.0)

15.4 (8.4-28.3)

15.5 (10.0-24.0)

15.0 (7.6-29.5)

8.0 (5.0-12.6)

12.7 (7.0-23.2)

16.2 (12.3-21.2)

6.3 (3.8-10.5)

16.1 (10.8-24.7)

8.6 (5.3-14.1)

0.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.75

0.560.790.920.550.790.920.520.760.910.550.790.920.570.800.920.590.810.930.470.730.890.520.770.910.520.760.910.440.700.880.570.800.920.550.780.92

0.080.220.450.070.170.390.100.240.490.070.190.420.080.200.440.080.200.430.100.250.500.080.200.440.080.210.440.110.270.530.080.200.430.120.300.56

Subgroup Studies,N

Sensitivity(95%CI)

Specificity(95%CI)

DOR (95%CI)

Pre-testprobability

Post-test(+)

Post-test(-)

Abbreviations: DOR: diagnostic odds ratio, CI: confidence intervals, CHC: chronic hepatitis C, CHB: chronic hepatitis B, ULN: upper limits of normal.

JOURNAL OF HEPATOLOGY

in F = 2, and 2 outlier studies underestimated it in F = 4. By usingthe linear regression approach [56], a possible publication biaswas present in F = 2 (p = 0.077), but not in F = 4. However, whenwe excluded the F2 and F4 outlier studies, the summary sensitiv-ity and specificity did not significantly change.

Discussion

In our meta-analysis, we evaluated the diagnostic accuracy oftransient elastography in the staging of liver fibrosis, as reportedin 40 studies. Although three meta-analyses have already beenpublished on this subject [57–59], our meta-analysis evaluates29 different studies compared to these, with an overlap of 25%or less. Moreover, none of these meta-analyses has used the opti-

Journal of Hepatology 201

mal statistical methods of combining the studies, i.e. HSROC orbivariate model [8]. Furthermore, none has incorporated in theinclusion criteria a maximum interval between performing liverbiopsy and transient elastography. This is important, as compar-ing the two techniques at a similar point minimises differencesdue to progression of fibrosis, which may not be constant overtime [1]. Efficacy of drug therapies in major trials is assessed aschanges in 6-month or 1-year biopsies [60], whereas elastogra-phy studies have included patients with as long as a 3-year inter-val between tests, making comparisons meaningless [61]. Mostimportantly, our conclusions are different and have a potentialimpact on everyday clinical practice, as we will further discuss.

Our results show that elastography could be used to diag-nose cirrhosis (when pre-test probability = 50%), with 90% prob-ability of correctly diagnosing cirrhosis following a ‘‘positive’’

1 vol. 54 j 650–659 655

Table 4. Subgroup analysis in cirrhosis (METAVIR stage 4). ‘‘Positive’’ elastography results were considered all results over the optimal liver stiffness threshold for a givenfibrosis stage in each individual study, while ‘‘negative’’ test results all results below that threshold. Post-test (+) denotes the probability following a positive test, whilepost-test (-) denotes the probability following a negative test.

CHC

CHB

Acceptable biopsy quality

Unreported biopsy quality

Low biopsy quality

Acceptable elastography quality

Unreported elastography quality

Low elastography quality

Full-paper publication

Abstract publication

ULN<ALT<3 ULN

ALT>3ULN

11

6

9

8

13

14

5

11

32

8

12

6

0.83 (0.77-0.88)

0.80 (0.61-0.91)

0.84 (0.67-0.93)

0.83 (0.74-0.89)

0.82 (0.79-0.84)

0.84 (0.78-0.89)

0.73 (0.47-0.89)

0.82 (0.78-0.85)

0.83 (0.79-0.87)

0.80 (0.59-0.92)

0.80 (0.76-0.83)

0.84 (0.77-0.90)

0.90 (0.87-0.93)

0.89 (0.82-0.94)

0.92 (0.86-0.95)

0.86 (0.82-0.89)

0.90 (0.86-0.93)

0.89 (0.86-0.92)

0.90 (0.81-0.95)

0.89 (0.84-0.92)

0.90 (0.87-0.92)

0.87 (0.90-0.92)

0.89 (0.85-0.93)

0.89 (0.82-0.93)

46.5 (26.7-91.0)

34.3 (17.0-69.2)

58.4 (27.7-122.7)

29.6 (19.4-45.0)

40.8 (26.6-62.7)

65.7 (38.0-113.)

24.4 (14.1-42.4)

35.5 (24.2-51.9)

43.1 (30.5-61.1)

28.2 (15.2-52.1)

33.5 (21.3-52.8)

43.0 (27.1-68.2)

0.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.750.250.50.75

0.740.900.960.710.880.960.770.910.970.660.860.950.730.890.960.730.890.960.710.880.960.710.880.960.730.890.960.680.860.950.710.880.960.700.870.95

0.060.160.360.070.180.400.050.150.340.060.170.380.060.170.380.060.150.350.090.230.470.060.170.380.060.160.360.070.180.400.0070.180.400.060.150.35

Subgroup Studies,N

Sensitivity(95%CI)

Specificity(95%CI)

DOR (95%CI)

Pre-testprobability

Post-test(+)

Post-test(-)

Abbreviations: DOR: diagnostic odds ratio, CI: confidence intervals, CHC: chronic hepatitis C, CHB: chronic hepatitis B, ULN: upper limits of normal.

Research Article

measurement. However, results are less promising for lesserfibrosis stages, as the probability of correctly diagnosing stages1–3 ranges from 78% to 85% following a ‘‘positive’’ measure-ment. A ‘‘negative’’ measurement is less informative, as cirrho-sis might still be present in 16%, or F2 in 20%, of patients(Table 2). When there is a high pre-test index of suspicion(pre-test probability = 75%), then the probability of a correctdiagnosis following a ‘‘positive’’ measurement exceeded 90%for all fibrosis stages, but with a ‘‘negative’’ measurement, thediagnosis would still be wrong in 24–44% of patients.

The above results, especially those following a ‘‘positive’’ mea-surement, have been considered encouraging in individual stud-ies and in some reviews. However, the major drawback is that thestiffness cut-off value, for which a particular measurement islabelled as ‘‘positive’’ for a particular stage, is different across

656 Journal of Hepatology 201

studies and has not been validated. Studies have derived thesedifferent liver stiffness cut-offs for each stage statistically onlyafter data collection, and none has been prospectively validatedin an independent cohort. These cut-offs not only have wideranges (in cirrhosis the range is from 9.0 to 26.5 Kpa) but alsooverlap between different fibrosis stages in different studies.Pooling such ‘‘optimal’’ results from these studies may artificiallyincrease the summary sensitivity and specificity. In our meta-analysis, we showed that the DOR was better in higher stiffnesscut-offs in F2 and F3, and in lower cut-offs in F1, but no optimalcut-off could be computed. Moreover, the fact that different aeti-ologies of liver disease might have different stiffness cut-offs for agiven fibrosis stage is likely, but has not been tested. Further-more, our meta-analysis confirms that higher transaminase levelsinfluence the accuracy of elastography measurements [57] and

1 vol. 54 j 650–659

JOURNAL OF HEPATOLOGY

also exposes a possible publication bias. Therefore, the sensitivityand specificity values obtained from the meta-analysis might notbe easily applicable in everyday clinical practice. In clinical prac-tice, in patients with an established cause of chronic liver disease,‘‘positive’’ elastography results for the diagnosis of F4 and espe-cially those above 15.4 kPa, which is the highest published opti-mal stiffness cut-off for the diagnosis of a lesser stage, wouldmean a probability for cirrhosis of 90% and would warrant furthertesting and screening for varices and HCC. ‘‘Negative’’ results donot sufficiently exclude cirrhosis and this is an equally importantmessage.

Although liver biopsy is currently the best-accepted standardagainst which the performance of any non-invasive marker ismade, it is not a gold standard. Apart from sample size, bothintra- and inter-observer variability influence fibrosis staging byhistological assessment [62,63]. The potential error in thehistological staging makes it difficult to correctly evaluate anon-invasive marker: even with a ‘‘high’’ liver biopsy sensitivityand specificity of 90%, the calculated AUROC for a perfect markerwould be 0.90 (versus 0.99 actual accuracy) [64]. Therefore, a dif-ferent gold standard or surrogate marker is urgently neededagainst which non-invasive markers of fibrosis can be evaluated.

Another issue is that the elastographic measurements, whichare quantitative and continuous, are compared to liver biopsystaging scores, which are categorical and descriptive and whichdo not have a quantitative relationship between them [62]. How-ever, histology could still represent a gold standard. The collagenproportionate area [65] quantifies liver collagen and could be abetter candidate for comparison with transient elastography.Hepatic venous pressure gradient, another continuous measure-ment, has been used to further characterise patients withcirrhosis in relation to their stiffness measurement [16]. Thediagnostic accuracy of elastography in various disease stagesshould be evaluated with respect to continuous measurementsof fibrosis and prospectively in terms of clinical outcomes.

A potential limitation of our metanalysis is that results werepooled for all aetiologies of liver disease. We chose to do so asno validated data are available on whether aetiology impactson elastography measurements. However, to address this issue,we performed subgroup analyses in which no significant differ-ences were noted, at least for CHB and CHC, as there were insuf-ficient data for other aetiologies. Moreover, most publishedstudies had a high bias risk and therefore all the estimates wederived may also have significant bias. Overall, only nine studies,comprising 1364 patients, had acceptable standards for both liverbiopsy and elastography, limiting the robustness of the conclu-sions that can be reached. Finally, significant heterogeneity waspresent in the evaluation of F2 and F4 stages, and therefore inter-pretation should be cautious.

In conclusion, our meta-analysis shows that elastographycould be used as a good screening test for cirrhosis, with a 90%disease probability following a ‘‘positive’’ measurement, but notfor lesser fibrosis degrees. A ‘‘negative’’ measurement is less accu-rate and informative, with disease being present in 15% ofpatients. Our meta-analysis exposes the problem of the absenceof validated cut-off values for the various fibrosis stages and fordifferent aetiologies of liver disease. These are a prerequisitebefore elastography results can be generally applicable in every-day clinical practice. This is a very important message to be con-veyed, particularly as the use of transient elastography isexpanding and decisions on patients’ management are made

Journal of Hepatology 201

based on unverified and potentially inaccurate fibrosis classifica-tions. Based on existing evidence, we would not recommenddiagnosing stages other than cirrhosis with the use of elastogra-phy. Better quality studies are needed as current studies have ahigh bias risk. Future studies should compare elastography mea-surements with quantitative histological assessments of fibrosis,such as collagen proportionate area or evaluate these withrespect to clinical outcomes.

Conflict of interest

The authors who have taken part in this study declare that theydo not have anything to disclose regarding funding or conflictof interest with respect to this manuscript.

Acknowledgements

E.A. Tsochatzis has received an educational grant from the Hel-lenic Association for the Study of the Liver.

Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.jhep.2010.07.033.

References

[1] Bataller R, Brenner DA. Liver fibrosis. J Clin Invest 2005;115:209–218.[2] Bravo AA, Sheth SG, Chopra S. Liver biopsy. N Engl J Med 2001;344:495–500.[3] Fraquelli M, Rigamonti C, Casazza G, Conte D, Donato MF, Ronchi G, et al.

Reproducibility of transient elastography in the evaluation of liver fibrosis inpatients with chronic liver disease. Gut 2007;56:968–973.

[4] Huwart L, Sempoux C, Vicaut E, Salameh N, Annet L, Danse E, et al. Magneticresonance elastography for the noninvasive staging of liver fibrosis. Gastro-enterology 2008;135:32–40.

[5] Whiting PF, Weswood ME, Rutjes AW, Reitsma JB, Bossuyt PN, Kleijnen J.Evaluation of QUADAS, a tool for the quality assessment of diagnosticaccuracy studies. BMC Med Res Methodol 2006;6:9.

[6] Harbord R. METANDI: Stata module to perform meta-analysis of diagnosticaccuracy. Boston College Department of Economics IS – S456932 http://ideasrepecorg/c/boc/bocode/s456932 html (last accessed on 16/12/08) 2008.

[7] Harbord RM, Deeks JJ, Egger M, Whiting P, Sterne JA. A unification of modelsfor meta-analysis of diagnostic accuracy studies. Biostatistics 2007;8:239–251.

[8] Leeflang MM, Deeks JJ, Gatsonis C, Bossuyt PM. Systematic reviews ofdiagnostic test accuracy. Ann Intern Med 2008;149:889–897.

[9] Intraobserver and interobserver variations in liver biopsy interpretation inpatients with chronic hepatitis C. The French METAVIR Cooperative StudyGroup. Hepatology 1994;20:15–20.

[10] Glas AS, Lijmer JG, Prins MH, Bonsel GJ, Bossuyt PM. The diagnostic oddsratio: a single indicator of test performance. J Clin Epidemiol 2003;56:1129–1135.

[11] Altman DG, Bland JM. Interaction revisited: the difference between twoestimates. BMJ 2003;326:219.

[12] Higgins J, Green S. Cochrane Handbook for Systematic Reviews of Iterven-tions Version 5.0.2 [updated September 2009]. The Cochrane Collaboration,Available from <www.cochrane-handbook.org>, 2008.

[13] Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication biasand other sample size effects in systematic reviews of diagnostic testaccuracy was assessed. J Clin Epidemiol 2005;58:882–893.

[14] Alric L, Kamar N, Bonnet D, Danjoux M, Abravanel F, Lauwers-Cances V, et al.Comparison of liver stiffness, fibrotest and liver biopsy for assessment of

1 vol. 54 j 650–659 657

Research Article

liver fibrosis in kidney-transplant patients with chronic viral hepatitis.Transplant Int 2009;22:568–573.

[15] Arena U, Vizzutti F, Abraldes JG, Corti G, Stasi C, Moscarella S, et al.Reliability of transient elastography for the diagnosis of advanced fibrosis inchronic hepatitis C. Gut 2008;57:1288–1293.

[16] Carrion JA, Navasa M, Bosch J, Bruguera M, Gilabert R, Forns X. Transientelastography for diagnosis of advanced fibrosis and portal hypertension inpatients with hepatitis C recurrence after liver transplantation. Livertransplant 2006;12:1791–1798.

[17] Castera L, Vergniol J, Foucher J, Le Bail B, Chanteloup E, Haaser M, et al.Prospective comparison of transient elastography, fibrotest, APRI, and liverbiopsy for the assessment of fibrosis in chronic hepatitis C. Gastroenterology2005;128:343–350.

[18] Castera L, Le BB, Roudot-Thoraval F, Bernard PH, Foucher J, Merrouche W,et al. Early detection in routine clinical practice of cirrhosis and oesophagealvarices in chronic hepatitis C: comparison of transient elastography(FibroScan) with standard laboratory tests and non-invasive scores. JHepatol 2009;50:59–68.

[19] Chan HLY, Wong GLH, Choi PCL, Chan AWH, Chim AML, Yiu KKL, et al.Alanine aminotransferase-based algorithms of liver stiffness measurementby transient elastography (Fibroscan) for liver fibrosis in chronic hepatitis B.J Viral Hepatitis 2009;16:36–44.

[20] Chang PE, Lui HF, Chau YP, Lim KH, Yap WM, Tan CK, et al. Prospectiveevaluation of transient elastography for the diagnosis of hepatic fibrosis inAsians: comparison with liver biopsy and aspartate transaminase plateletratio index. Aliment Pharmacol Ther 2008;28:51–61.

[21] Corradi F, Piscaglia F, Flori S, Errico-Grigioni A, Vasuri F, Tame MR, et al.Assessment of liver fibrosis in transplant recipients with recurrent HCVinfection: usefulness of transient elastography. Digest Liver Dis2009;41:217–225.

[22] de Ledinghen V, Douvin C, Kettaneh A, Ziol M, Roulot D, Marcellin P, et al.Diagnosis of hepatic fibrosis and cirrhosis by transient elastography in HIV/hepatitis C virus-coinfected patients. J Acquir Immune Defic Syndr2006;41:175–179.

[23] de Ledinghen V, Le Bail B, Rebouissoux L, Fournier C, Foucher J, Miette V,et al. Liver stiffness measurement in children using FibroScan: feasibilitystudy and comparison with fibrotest, aspartate transaminase to plateletsratio index, and liver biopsy. J Pediatr Gastroenterol Nutr 2007;45:443–450.

[24] Degos F, Perez P, Roche B, Mahmoudi A, Asselineau J, Voitot H, et al. J Hepatol2009;50:S40.

[25] Foucher J, Chanteloup E, Vergniol J, Castera L, Le Bail B, Adhoute X, et al.Diagnosis of cirrhosis by transient elastography (FibroScan): a prospectivestudy. Gut 2006;55:403–408.

[26] Harada N, Soejima Y, Taketomi A, Yoshizumi T, Ikegami T, Yamashita YI, et al.Assessment of graft fibrosis by transient elastography in patients withrecurrent hepatitis C after living donor liver transplantation. Transplantation2008;85:69–74.

[27] Kanamoto M, Shimada M, Ikegami T, Uchiyama H, Imura S, Morine Y, et al.Real time elastography for noninvasive diagnosis of liver fibrosis. J Hepa-tobiliary Pancreat Surg 2009.

[28] Kettaneh A, Marcellin P, Douvin C, Poupon R, Ziol M, Beaugrand M, et al.Features associated with success rate and performance of fibroscanmeasurements for the diagnosis of cirrhosis in HCV patients: A prospectivestudy of 935 patients. J Hepatol 2007;46:628–634.

[29] Kim SU, Ahn SH, Park JY, Kang W, Kim dY, Park YN, et al. Liver stiffnessmeasurement in combination with noninvasive markers for the improveddiagnosis of B-viral liver cirrhosis. J Clin Gastroenterol 2009;43:267–271.

[30] Kim KM, Choi WB, Park SH, Yu E, Lee SG, Lim YS, et al. Diagnosis of hepaticsteatosis and fibrosis by transient elastography in asymptomatic healthyindividuals: a prospective study of living related potential liver donors. JGastroenterol 2007;42:382–388.

[31] Kim SU, Kim JK, Park JY, Ahn SH, Lee JM, Baatarkhuu O, et al. Variability inliver stiffness values from different intercostal spaces. Liver Int2009;29:760–766.

[32] Lewin M, Poujol-Robert A, Boelle PY, Wendum D, Lasnier E, Viallon M, et al.Diffusion-weighted magnetic resonance imaging for the assessment offibrosis in chronic hepatitis C. Hepatology 2007;46:658–665.

[33] Lupsor M, Badea R, Stefanescu H, Grigorescu M, Sparchez Z, Serban A, et al.Analysis of histopathological changes that influence liver stiffness in chronichepatitis C. Results from a cohort of 324 patients. J Gastrointest Liver Dis2008;17:155–163.

[34] Marcellin P, Ziol M, Bedossa P, Douvin C, Poupon R, de Ledinghen V, et al.Non-invasive assessment of liver fibrosis by stiffness measurement inpatients with chronic hepatitis B. Liver Int 2009;29:242–247.

658 Journal of Hepatology 201

[35] Masuzaki R, Tateishi R, Yoshida H, Goto E, Sato T, Ohki T, et al. Comparison ofliver biopsy and transient elastography based on clinical relevance. CanadianJournal of Gastroenterology 2008;22:753–757.

[36] Mueller S, Millonig G, Sarovska L, Friedrich S, Reimann S, Pritsch M, et al. JHepatol 2009;50:S366.

[37] Nahon P, Kettaneh A, Tengher-Barna I, Ziol M, de Ledinghen V, Douvin C,et al. Assessment of liver fibrosis using transient elastography in patientswith alcoholic liver disease. Journal of Hepatology 2008;49:1062–1068.

[38] Nguyen-Khac E, Chatelain D, Tramier B, Decrombecque C, Robert B, JolyJP, et al. Assessment of asymptomatic liver fibrosis in alcoholic patientsusing fibroscan: prospective comparison with seven non-invasive labo-ratory tests. Alimentary Pharmacology & Therapeutics 2008;28:1188–1198.

[39] Nitta Y, Kawabe N, Hashimoto S, Harata M, Komura N, Kobayashi K, et al.Liver stiffness measured by transient elastography correlates with fibrosisarea in liver biopsy in patients with chronic hepatitis C. Hepatol Res 2009.

[40] Nudo C, Jeffers L, Bejarano P, Servin-Abad L, Leibovici Z, De Medina M, et al.Correlation of laparoscopic liver biopsy to elasticity measurements (Fibro-scan) in patients with chronic liver disease. Gastroenterol Hepatol2008;12:862–870.

[41] Ogawa E, Furusyo N, Toyoda K, Takeoka H, Otaguro S, Hamada M, et al.Transient elastography for patients with chronic hepatitis B and C virusinfection: non-invasive, quantitative assessment of liver fibrosis. HepatolRes 2007;37:1002–1010.

[42] Petta S, Camma C, Di Marco V, Calvaruso V, Enea M, Bronte F, et al. J Hepatol2009;50:S155.

[43] Reiberger T, Ferlitsch A, Ulbrich G, Homoncik M, Peck-Radosavljevic M.Identification of preclinical portal hypertension and and staging of fibrosisby transient elastography. Hepatology 2009;48 (Suppl 4):620A–621A.

[44] Rigamonti C, Donato MF, Fraquelli M, Agnelli F, Ronchi G, Casazza G, et al.Transient elastography predicts fibrosis progression in patients with recur-rent hepatitis C after liver transplantation. Gut 2008;57:821–827.

[45] Schlosser B, Felder A, Biermer M, Muller HP, Van Bommel F, Weich V, et al.Hepatology 2008;48:990A.

[46] Sporea I, Sirli R, Deleanu A, Popescu A, Cornianu M. Liver stiffnessmeasurement by transient elastography in clinical practice. J GastrointestLiver Dis 2008;17:395–399.

[47] Sporea I, Sirli R, Deleanu A, Tudora A, Curescu M, Cornianu M, et al.Comparison of the liver stiffness measurement by transient elastographywith the liver biopsy. World J Gastroenterol 2008;14:6513–6517.

[48] Tanwandee T, Charatcharoenwitthaya P, Viboolsirikul V, Chotiyaputta W,Chainuvati S, Maneerattanaporn M, et al. Utility of liver stiffness measuredby transient elastography for determining significant liver fibrosis inpatients with chronic hepatitis B. Hepatology 2008;48 (Suppl 1):709A.

[49] Verveer C, Hansen BE, Zondervan PE, Janssen HL, de Knegt RJ. Diagnosticaccuracy of transient elastography. A comparison between chronic hepatitisB and C correlated with optimal-length liver biopsies. Hepatology 2008;48(Suppl 1):1107A–1108A.

[50] Wang JH, Changchien CS, Hung CH, Eng HL, Tung WC, Kee KM, et al.FibroScan and ultrasonography in the prediction of hepatic fibrosis inpatients with chronic viral hepatitis. J Gastroenterol 2009;44:439–446.

[51] Wong GL, Wong VW, Choi PC, Chan AW, Chim AM, Yiu KK, et al. Metabolicsyndrome increases the risk of liver cirrhosis in chronic hepatitis B. Gut2009;58:111–117.

[52] Wong VW, Wong GL, Chim AM, Yiu K, Chan AW, Choi PC, et al. J Hepatol2008;48:S367.

[53] Yoneda M, Mawatari H, Fujita K, Endo H, Iida H, Nozaki Y, et al. Noninvasiveassessment of liver fibrosis by measurement of stiffness in patients withnonalcoholic fatty liver disease (NAFLD). Dig Liver Dis 2008;40:371–378.

[54] Tatsumi C, Kudo M, Ueshima K, Kitai S, Takahashi S, Inoue T, et al.Noninvasive evaluation of hepatic fibrosis using serum fibrotic markers,transient elastography (FibroScan) and real-time tissue elastography. Inter-virology 2008;51 (Suppl 1):27–33.

[55] Kim KM, Choi WB, Park SH, Yu E, Lee SG, Lim YS, et al. Diagnosis of hepaticsteatosis and fibrosis by transient elastography in asymptomatic healthyindividuals: a prospective study of living related potential liver donors. JGastroenterol 2007;42:382–388.

[56] Egger M, Davey SG, Schneider M, Minder C. Bias in meta-analysis detectedby a simple, graphical test. BMJ 1997;315:629–634.

[57] Friedrich-Rust M, Ong MF, Martens S, Sarrazin C, Bojunga J, Zeuzem S, et al.Performance of transient elastography for the staging of liver fibrosis: ameta-analysis. Gastroenterology 2008;134:960–974.

[58] Stebbing J, Farouk L, Panos G, Anderson M, Jiao LR, Mandalia S, et al. A meta-analysis of transient elastography for the detection of hepatic fibrosis. J ClinGastroenterol 2009;1:1–2.

1 vol. 54 j 650–659

JOURNAL OF HEPATOLOGY

[59] Talwalkar JA, Kurtz DM, Schoenleber SJ, West CP, Montori VM. Ultrasound-

based transient elastography for the detection of hepatic fibrosis: systematicreview and meta-analysis. Clin Gastroenterol Hepatol 2007;5:1214–1220.

[60] Belfort R, Harrison SA, Brown K, Darland C, Finch J, Hardies J, et al. A placebo-controlled trial of pioglitazone in subjects with nonalcoholic steatohepatitis.N Engl J Med 2006;355:2297–2307.

[61] Baldaia C, Serejo F, Marinho R, Costa A, Velosa J, Moura MC. Transientelastography in chronic hepatitis C: comparison between different nonin-vasive methods for liver fibrosis assessment. Hepatology 2006;44 (Suppl1):276A.

Journal of Hepatology 201

[62] Cholongitas E, Senzolo M, Standish R, Marelli L, Quaglia A, Patch D, et al. Asystematic review of the quality of liver biopsy specimens. Am J Clin Pathol2006;125:710–721.

[63] Bedossa P, Dargere D, Paradis V. Sampling variability of liver fibrosis inchronic hepatitis C. Hepatology 2003;38:1449–1457.

[64] Mehta SH, Lau B, Afdhal NH, Thomas DL. Exceeding the limits of liverhistology markers. J Hepatol 2009;50:36–41.

[65] Calvaruso V, Burroughs AK, Standish R, Manousou P, Grillo F, Leandro G,et al. Computer-assisted image analysis of liver collagen: relationship toIshak scoring and hepatic venous pressure gradient. Hepatology2009;49:1236–1244.

1 vol. 54 j 650–659 659