Performance of transient elastography and serum fibrosis biomarkers for non-invasive evaluation of recurrent fibrosis after liver transplantation: A meta-analysis

Recurrent fibrosis after liver transplantation (LT) impacts on long-term graft and patient survival. We performed a meta-analysis to compare the accuracy of non-invasive methods to diagnose significant recurrent fibrosis (stage F2-F4) following LT. Studies comparing serum fibrosis biomarkers, namely AST-to-platelet ratio index (APRI), fibrosis score 4 (FIB-4), or transient elastography (TE) with liver biopsy in LT recipients were systematically identified through electronic databases. In the meta-analysis, we calculated the weighted pooled odds ratio and used a fixed effect model, as there was no significant heterogeneity between studies. Eight studies were included for APRI, four for FIB-4, and twelve for TE. The mean prevalence of significant liver fibrosis was 37.4%. The summary odds ratio was significantly higher for TE (21.17, 95% CI confidence interval 14.10–31.77, p = 1X10-30) as compared to APRI (9.02, 95% CI 5.79–14.07; p = 1X10-30) and FIB-4 (7.08, 95% CI 4.00–12.55; p = 1.93X10-11). In conclusion, TE performs best to diagnose recurrent fibrosis in LT recipients. APRI and FIB-4 can be used as an estimate of significant fibrosis at centres where TE is not available. Longitudinal assessment of fibrosis by means of these non-invasive tests may reduce the need for liver biopsy.


Introduction
The development of hepatic fibrosis adversely affects the prognosis of any chronic liver disease, regardless of etiology. In the specific setting of liver transplantation (LT), recurrent liver fibrosis portends worse graft survival and need for re-transplantation, thus impacting on overall survival [1][2][3]. LT recipients may have multiple cofactors leading to liver fibrosis recurrence. Patients who undergo LT for hepatitis C virus (HCV) infection have universal recurrence, with development of cirrhosis in up to 30% by 5 years [4][5][6]. The landscape of antiviral therapy for hepatitis C has been dynamic, and has led to highly effective active, interferon-free regimens [7]. However, concerns about costs and availability in the post-transplant setting may a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 • Manual review of reference lists of relevant articles.
A combination of relevant text words and MeSH terms were applied, namely liver transplantation AND one of the following: fibrosis, non-invasive fibrosis markers, serum fibrosis markers, APRI, FIB-4, TE, and liver stiffness measurement. For management of the searched literature, Endnote version X7 (Thomson Reuters, New York, NY) was used. We chose these three non-invasive tests as they are widely used and available, as well as the literature search revealed more studies than other scores. For example, there has been limited evaluation of the Fibrotest in the detection of significant fibrosis after LT, with only 2 prospective studies comprising 82 patients, both of which reported a low diagnostic accuracy [18,19]. Another patented non-invasive test, the Enhanced Liver Fibrosis (ELF) Score has been evaluated in two prospective studies comprising 113 patients with all etiologies of liver disease as indication for LT, with variable accuracy reported [19,20].

Inclusion criteria
Studies were included if they met the following criteria: they evaluated APRI, FIB-4 or TE; they used liver biopsy as the reference standard; they employed comparable histologic staging systems: METAVIR, Ishak, Brunt, Ludwig, Knodell; they evaluated the diagnostic accuracy of the test expressed as area under the curve (AUC) for significant liver fibrosis (stage F2-4); and/ or they evaluated sensitivity, specificity, positive predictive value (PPV) or negative predictive value (NPV) for the diagnosis of significant liver fibrosis based on cut-off values for APRI, FIB-4 or TE. Both adult and pediatric studies were included.

Exclusion criteria
Studies were excluded if they met the following criteria: they were not conducted in the post-LT setting; they did not use liver biopsy as the standard of reference; they did not report data on diagnostic accuracy (AUC), sensitivity, specificity, PPV or NPV for the diagnosis of significant liver fibrosis based on cut-off values for APRI, FIB-4 or TE; they were review articles, letters to the editor or editorials not reporting own original data; they were abstracts presenting data from the same study at different meetings.
Data extraction. One reviewer (M.B.) searched the various databases listed above, and identified pertinent studies (both articles and abstracts) for further assessment. Both reviewers (M.B. and G.S.) reviewed the abstracts and full text versions of the articles independently, evaluating study eligibility, grading study quality, and extracting data. We took note of study design, number of participants, non-invasive tool adopted among APRI, FIB-4 and TE. From the selected studies, we extracted the following data: study characteristics such as authors, year of publication, hospital or medical school and study design; demographic and clinical characteristics of participants; characteristics of liver biopsy; non-invasive method adopted among APRI, FIB-4 and TE; correlation with liver biopsy staging of fibrosis as data categories and relative cut-off value of the non-invasive test. In case of disagreement regarding the selected studies or data appropriate for extraction, studies were refereed by a third reviewer (M.T.) and resolved by consensus. Methodological quality was assessed using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) Score [21]. The primary outcome was the identification of significant fibrosis, defined as stage F2-4 according to the METAVIR staging system. This threshold was chosen because it indicates a progressive liver disease eventually leading to cirrhosis and requiring interventions, such as antiviral therapy (hepatitis C) and life style modifications (NAFLD) [7,22]. In all the included studies, APRI and FIB-4 were computed as per original formulas: APRI as [100 x (AST/upper limit of normality)/platelet count (10 9 /L) [23]; FIB-4 as age (years) x AST /platelet count (10 9 /L)] x ALT 1/2 [24].

Data analysis
We estimated the diagnostic odds ratios and AUC with their corresponding 95% confidence intervals (95% CIs) for each study based on the diagnostic tests. In the meta-analysis we calculated the weighted pooled odds ratio. The heterogeneity statistics calculated the summary odds ratio under random effects model [25]. The summary receiving operating characteristics (ROC) curves (SROC) shows the overall diagnostic test including each study's test and the sample size of each study.
Pooled measures for diagnostic performance, such as sensitivity, specificity, diagnostic odds ratios (DORs) with their corresponding 95% confidence intervals (95% CIs), and AUC were calculated. The DORs combine sensitivity and specificity into one measure for diagnostic performance. A DOR of 1 means that the test has no ability to discriminate between two outcomes. In the context of this study, the higher the DOR, the better the diagnostic accuracy of the non-invasive test for assessing significant liver fibrosis.
The forest plot was graphed for odds ratios, as a summary and for individual studies. A Chi-square-based test of homogeneity was performed, and the Cochran's Q inconsistency index (I2) statistics were determined. The heterogeneity between studies was calculated by I 2 values. Publication bias was assessed by constructing funnel plots for DOR by Egger's test [26]. The absence of publication bias was indicated by the data points forming a symmetric funnelshaped distribution and significance level p>0.05. We used the SAS statistical software 9.4 (SAS Institute, Cary, NC) to perform the statistical analysis at p<0.05 level of significant test and R program for SROC.

Sensitivity analyses
Sensitivity analyses were employed to examine the impact of the following factors on test performance for identifying significant liver fibrosis: study quality; prevalence of significant fibrosis; and quality of the reference standard for assessing fibrosis. The reference standard was deemed adequate if a study excluded liver biopsies <10 mm. If sufficient data were not reported in the manuscript, the reference standard was considered inadequate.

Results
The literature search by the reviewers resulted in 2399 studies identified with our search terms (S1-S12 Files). Of these, 55 included original data. 48 were full-text manuscripts, and 7 studies were available in abstract form only. The systematic literature search and study selection are depicted in the flowchart of Fig 1. Ultimately, studies were excluded as they were not pertinent to non-invasive evaluation of fibrosis after LT, due to insufficient data and review articles/ commentary. A total of 171 studies described TE, APRI or FIB-4, of which 33 included original data. 33 studies compared TE, APRI or FIB-4 versus liver biopsy in LT recipients. Finally, 24 studies were included in the meta-analysis. The studies excluded from the meta-analysis and reasons for exclusion are provided in Table 1. Our final data set for the meta-analysis included 8 studies assessing APRI, 4 assessing FIB-4 and 12 assessing TE ( Table 2).

Characteristics of the included studies
We included eight studies investigating the ability of APRI to diagnose significant fibrosis (F2-4). These studies included a total of 448 LT recipients, of whom 30 were children. Five out of  Non-invasive methods for recurrent liver fibrosis the eight studies focused on HCV patients, who comprised 329 out of the 448 patients. The overall prevalence of significant liver fibrosis was 36.2% (range 20-68.1%). Biopsy quality was considered acceptable in three of these studies [18,27,28]. None of the studies included patients with acute rejection. A total of 435 patients were included in the four FIB-4 studies. The overall prevalence of significant liver fibrosis was 39.8% (range 28-68.1%). Two of these studies comprised only HCVinfected patients, while two included patients of mixed etiology of liver disease. Biopsy quality was considered acceptable in one of these studies [27].
In the twelve TE studies, a total of 1,196 LT recipients were included. The overall prevalence of significant liver fibrosis was 37.4% (range 27.1-68.1%). Six of these studies comprised only HCV-infected patients, while six included patients of mixed etiology of liver disease. Biopsy quality was considered acceptable in seven of these studies [18,20,27,29,30].
According to the QUADAS scale, the methodological quality of the included studies was good. Two studies met all 14 requirements of this scale, four studies met all but one of these requirements.

Diagnostic accuracy of APRI
The AUC for significant fibrosis ranged from a low of 0.50 to a high of 0.83 for HCV, as reported in Table 3. The APRI test fared better in the studies of patients with all etiologies, with an AUC ranging from 0.74 to 0.87. As illustrated in Fig 2, the summary odds ratio for APRI was 9.20 (95% CI 5.79-14.07, p = 1X10 -30 ).

Diagnostic accuracy of TE
TE detected significant fibrosis with an AUC ranging from 0.75 to 0.96, as reported in Table 3. Two small prospective pediatric studies comprising 50 patients with various indications for LT revealed different results: the study of 24 patients negative for viral hepatitis had an AUC of 0.71, whereas a study of 26 patients with all etiologies had an excellent AUC of 0.93. As illustrated in Fig 4, the summary odds ratio for TE was the best, at 21.17 (95% CI 14.10-31.77, p = 1X10 -30 ), having excluded one study due to publication bias as elucidated below. A comparison of diagnostic accuracy of the three tests is illustrated in Fig 5. There was no significant difference between APRI and FIB4 (p>0.05), but there was a significant difference between TE and FIB4 (p<0.05) and between TE and APRI (P<0.05).

Publication bias
Funnel plots of diagnostic accuracy versus effective sample size for the prediction of significant liver fibrosis are illustrated in Fig 6. For APRI and FIB-4 ( Fig 6A and 6B), the funnel plot did not detect a significant publication bias (p>0.05). For TE (Fig 6C), there was evidence of publication bias with a single study. The Egger test for the remaining 11 studies showed no evidence of publication bias (p>0.05). The sensitivity analysis including study quality, prevalence of significant fibrosis, and quality of the reference standard, did not explain this asymmetry.

Discussion
In this meta-analysis, we have evaluated the diagnostic accuracy of three non-invasive diagnostic tests to stage liver fibrosis in the post-LT setting. The identification of significant liver fibrosis is pivotal since it is a clear indication of recurrence of the primary disease (e.g. hepatitis C) Table 3 or the occurrence of de novo liver disease (e.g. NAFLD) triggered by multiple risk factors related to immunosuppression or metabolic disorders which occur frequently after LT. Our data indicate that TE has good accuracy in detecting significant liver fibrosis in LT recipients and outperforms simple serum biomarkers, such as APRI and FIB-4. Our work has the strength of including all etiologies of liver disease and comparing TE with simple serum biomarkers. To our knowledge, there is only a single systematic review that included TE, APRI and FIB-4, however it was conducted only in HCV-positive recipients [31]. The development and validation of non-invasive diagnostic methods to diagnose liver fibrosis has gained much attention in recent years. These efforts have proceeded in parallel with increased evidence of the limitations and drawbacks of liver biopsy. Indeed, liver biopsy is an invasive and costly procedure, which could potentially lead to serious complications including bleeding and pneumothorax [17]. Moreover, there are intrinsic limitations in histological determination. Indeed, the quality and size of the specimen obtained through a percutaneous liver biopsy greatly affect the diagnostic reliability. Intra-observer and inter-observer variability have also been extensively studied. It has been suggested that liver biopsy is the best available standard of reference for fibrosis evaluation, although an imperfect gold standard [32].

Author, Year Test Cut-off Sensitivity (%) Specificity (%) PPV (%) NPV (%) AUC (95% CI)
Non-invasive methods for liver fibrosis have been extensively validated for the main etiologies of chronic liver diseases in the pre-transplant setting. TE and serum fibrosis biomarkers are the most validated tests and are recommended by recent guidelines. The recently endorsed guidelines of the European Association for the Study of the Liver (EASL) on the non-invasive diagnosis of liver fibrosis suggest that non-invasive methods could substitute liver biopsy when combined in the pre-transplant setting [17]. However, fewer data are available in LT recipients. After careful selection of available studies through a quality control check, we included 24 studies evaluating the performance of non-invasive diagnostic tools for liver fibrosis against liver histology. Although some of them included only patients with HCV, several included mixed etiologies of liver disease, such as alcoholic liver disease, NAFLD, cholestasis.
There have been few studies employing other non-invasive tools, such as NAFLD fibrosis score, Magnetic Resonance Elastography (MRE) or Acoustic Radiation Force Impulse imaging (ARFI). However, we excluded those non-invasive tests either for poor quality of the studies or for the limited number of studies conducted, insufficient to perform a meta-analysis.
The primary outcome of our meta-analysis was the identification of significant liver fibrosis (stage F2-4), which indicates a progressive liver fibrosis that will eventually lead to liver cirrhosis if appropriate interventions are not implemented. In the specific setting of LT, significant liver fibrosis may mean both recurrence of primary liver disease (hepatitis C, NASH) or a de novo liver disease. In any case, it is a clear signal that further investigations are needed. Only a previous meta-analysis reported APRI's performance in LT recipients and this was superior to other simple non-invasive diagnostic tests [31]. In our experience, there was no significant difference between the two simple fibrosis biomarkers APRI and FIB-4 in terms of diagnostic Non-invasive methods for recurrent liver fibrosis performance. Importantly, we did not find a significant publication bias in the studies including these two biomarkers, and this was confirmed by absence of statistically significant asymmetry on the funnel plots. APRI and FIB-4 have been employed serially in LT recipients not only for diagnostic purposes, but also for prognosis. The fact that both biomarkers and their longitudinal changes predict long-term outcomes in the recipient, including death and graft loss, confirms their pathophysiologic link with liver damage [33]. A major issue for both APRI and FIB-4 remains the wide range of cut-off values adopted in different studies to diagnose significant liver fibrosis, which may limit their applicability in clinical practice. Patented panels of fibrosis biomarkers, such as Fibrotest and ELF, have not been included in the present metaanalysis. Indeed, there are only a few studies which investigated the diagnostic accuracy of these tests in LT and they reported an unsatisfactory performance. This could be due to suboptimal adherence of the laboratory to pre-analytic recommendations to obtain reliable results [34] or on specific alterations of the individual components of the patented panels that could occur following LT due to chronic inflammation (elevation of alpha-2-macroglobulin, hyaluronan), hemolytic anemia (haptoglobin), multidrug induction (GGT) or cholestasis (bilirubin).
TE had the highest number of included studies among the three non-invasive methods we evaluated. In the pre-transplant setting, TE has been shown to outperform simple serum biomarkers to diagnose liver cirrhosis [17]. This gap in diagnostic accuracy is smaller for the detection of significant liver fibrosis. In the present study, we found that TE had higher diagnostic accuracy than APRI and FIB-4 to detect significant liver fibrosis in the post-transplant setting. Diagnostic accuracy was consistent in both HCV patients and other etiologies of liver disease. There was a tendency for a significant publication bias for TE, which was driven by only one study. However, we were not able to find any predictor of this marginally significant asymmetry with the sensitivity analysis. Interestingly, there have been several studies investigating the use of TE for serial assessment in LT recipients in order to identify liver fibrosis progression. On this topic, Rigamonti et al conducted two studies, one in patients transplanted for HCV and one in those transplanted for non-viral etiologies [30,35]. In 95 patients transplanted for end-stage liver disease due to HCV and undergoing serial liver biopsies, the authors found an AUC of 0.85 for significant liver fibrosis, concluding that protocol liver biopsy could be avoided in patients with stable TE during follow-up. Along the same lines, in 69 LT recipients the same group found that TE allow accurate discrimination between presence or absence of liver graft damage, thus helping the selection of patients most in need of liver biopsy.
The EASL guidelines on non-invasive evaluation of liver fibrosis suggest combining TE with a serum biomarker in order to reduce the number of liver biopsies needed for correct clinical management [17]. The avoidance of a liver biopsy becomes a more complicated matter when dealing with the post-transplant setting. Indeed, in LT recipients, it is not only liver fibrosis stage that may cause graft damage, but also multiple liver pathologies that arise from the complex interaction between primary liver disease recurrence (hepatitis C, NAFLD), development of metabolic syndrome in the context of immunosuppression (calcineurin and mTOR inhibitors), and possible acute or chronic rejection. However, given the good accuracy we found in this meta-analysis, especially for TE, we feel that non-invasive methods for liver fibrosis could be used to reduce the number of liver biopsies in LT recipients. We acknowledge that non-invasive methods should always be used having carefully ascertained the clinical context and correlated with liver biochemistry and function tests.
Our study has certain limitations, including the higher proportion of HCV patients represented. We did also include pediatric studies, with various reasons for development of fibrosis post-transplant. Finally, the reported range of cut-off values for TE is wide (from 7.3 kPa to 12.3 kPa) in LT recipients, rendering potentially complicated the use in clinical practice for the individual patients. This finding is likely due to highly heterogeneous study populations due to etiology of liver disease, different post-LT time points for the assessment of liver fibrosis, study design and sample size and variable interval range between liver biopsy and the non-invasive assessment of liver fibrosis. We could not account for these different cut-offs for transient elastography for the various etiologies of CLD in our meta-analysis, as these were not provided in the studies. Nonetheless, the validity of our findings across this heterogeneous population suggests the usefulness of these non-invasive tools for fibrosis even in a complex context such as that of LT.

Conclusions
In conclusion, given non-invasiveness and feasibility for serial measurements, non-invasive tests for liver fibrosis could be used in the post-transplant clinical setting as an additive tool for suspected recurrent or de novo liver disease. The high accuracy we found in our meta-analysis, especially for TE, suggests that these tests have similar diagnostic value as in the pre-transplant setting. It should be underlined that liver biopsy remains a cornerstone for the clinical management of LT recipients, as non-invasive tests cannot differentiate between different liver pathologies that can coexist in this setting, such as acute or chronic rejection. Nonetheless, when the etiology of recurrent or de novo disease has been established, these non-invasive tests are helpful in following fibrosis progression longitudinally and implementing preventive therapies in a timely manner. Further studies aimed at defining optimal cut-off values for definition of significant liver fibrosis are warranted.