Prospective comparison of liver stiffness measurement methods in surveillance biopsies after liver transplantation

Background Liver stiffness measurements (LSMs) have proven useful for non-invasive detection of fibrosis. Previous studies of LSMs after transplantation were performed in cohorts dominated by hepatitis C reinfections and indication biopsies for the evaluation of graft dysfunction. However, the diagnostic fidelity of LSMs for fibrosis is biased by inflammation e.g., during replicative hepatitis C or rejection. Materials and methods The current study aimed for a head-to-head comparison of two different LSMs, acoustic radiation force impulse (ARFI) and transient elastography (TE), and a determination of cut-off values for the detection of advanced fibrosis (any LAF score component ≥2) in grafts undergoing surveillance biopsies (svLbx) without recurrent hepatitis C. Results 103 svLbx were paired with valid LSMs at time of biopsy. AUROC analyses showed significant positive correlation with fibrosis for both methods (TE: AUROC = 0.819 (p < 0.001; 95%CI: 0.717–0.921); ARFI: AUROC = 0.771 (p = 0.001; 95%CI: 0.652–0.890). Patients were randomly assigned to training and validation cohorts for both LSM methods. Cut-off values were determined at 1.29 m/s (ARFI) and at 7.5 kPa (TE) in training cohorts. Sensitivity and specificity in training and validation cohorts were: TE: SEN 0.818 and 0.5; SPE 0.742 and 0.885; ARFI: SEN 0.818 and 1.0; SPE 0.75 and 0.586. LSMs were not associated with BANFF criteria for relevant graft injury. Conclusion LSM is a good non-invasive tool to screen for advanced graft fibrosis but not for relevant graft injury in patients with (near) normal liver enzymes. Fibrosis cut-off values identified and validated in svLbx were lower than in previous cohorts using indication biopsies.


Introduction
Although most complications after orthotopic liver transplantation (OLT), such as infections, kidney failure, and malignancy, are associated with immunosuppressive therapy, graft fibrosis remains a major cause of re-transplantation or death after liver transplantation (1).Graft biopsy is the gold standard for the detection of significant fibrosis and cirrhosis.However, liver biopsies (Lbx) are expensive and carry the risk of complications, such as bleeding or pain, even though these risks have proven to be very low (2,3).For many years, liver stiffness measurements (LSMs) have been used as practical non-invasive surrogate parameters for liver fibrosis.
Transient elastography (TE) is a well-known method for detecting liver fibrosis in patients with different underlying diseases and even after OLT.In addition, it allows the examination of a large volume of liver tissue.However, TE measurements can fail due to a high body mass index, ascites, or narrow intercostal spaces (2).Previous studies have also described correlations to bile tract anomalies and acute hepatitis (3).
Acoustic radiation force impulse (ARFI) has also been evaluated for the detection of liver fibrosis in several different studies (4).It is ultrasound-based and can therefore be combined with regular ultrasonographic examinations before and after liver transplantation.The ultrasound probe generates high-energy acoustic pulses which trigger localised tissue displacements.Through this, shear waves are propagated and their velocity can be measured in a certain region of interest.Results are described in m/s (5).
Several studies have proven a correlation between liver stiffness measurements and fibrosis before and after OLT.However, these studies were performed in cohorts predominantly including patients with chronic hepatitis C before the introduction of direct-acting antivirals (DAA) and mostly in comparison to indication biopsies (4,6).The achievement of sustained virologic response in hepatitis C is, however, associated with an improvement of LSM (7).
The aim of this study was to prospectively compare TE and ARFI by including these examinations in our surveillance biopsy (svLbx) program after OLT in the post-DAA era.

Subjects
We included all adult liver recipients without a replicative viral hepatitis (HCV-RNA or HBs-Ag negativity) who underwent at least one surveillance liver biopsy and agreed to participate in our prospective liver allograft biorepository data-base and in this prospective observational study from November 2018 to September 2020 (age at time of liver biopsy ≥18 years).A participation in the surveillance biopsy (svLbx) program was voluntary and offered to all liver transplanted patients without contraindications, e.g., severely dilated bile ducts, thrombocytopenia etc.For scientific analysis indication and surveillance biopsies were distinguished based on liver enzyme levels in paired blood samples.While AST and ALT were normal or near normal [<2-fold upper limit of normal (ULN)] and AP and GGT were at stable levels (even though elevated in some patients), indication biopsies had liver enzymes above this threshold.Similar thresholds were used in studies on intentional immunosuppression (IS) withdrawal and in previous analyses of our program (3,8,9).Our svLbx program was established in 2018 and aims to perform svLbx in patients at specific time points after OLT, i.e., one, three, five and ten years post-OLT and then every 5 years.All biopsies are then discussed in monthly interdisciplinary conferences and IS is adapted accordingly.Patients without graft injury are advised to reduce IS in order to minimize adverse effects of the medication.Patients with significant graft fibrosis are mostly advised to switch their IS regimen to a combination of a calcineurin inhibitor (CNI) and an m-TOR inhibitor (i.e., everolimus) due to the antifibrotic effects of m-TOR inhibitors (9)(10)(11)(12).Patients with subclinical graft injury are mainly advised to stay on their current IS regimens, unless there are important reasons to change or reduce IS, i.e., malignant diseases or major infections (9).
This study was approved by the Ethics Committee of our center.Written informed consent was obtained from all subjects.All experiments were performed in accordance with relevant guidelines and regulations.No organs or tissues were procured from prisoners.

Liver stiffness measurements
After an overnight fast, the patients were laid in a supine position with their arms maximally abducted under their head.The success rate was calculated as the number of successful measurements divided by the total number of measurements.Only procedures with at least 10 valid acquisitions, a success rate of greater than 60% and an interquartile range (IQR/M) less than 30% were defined as eligible for the study.
According to study protocol, each patient was to receive both their ARFI and TE measurements from different examiners within 7 days prior to liver biopsy.
ARFI elastography was performed with Virtual Touch tissue quantification (Siemens Acuson S2000, Siemens Healthineers, Erlangen, Germany) with a standard broad-band 4-1 MHz curved array, as described recently (13).This allowed placement of the region of interest under vision in ultrasonographic Bmode.Measurements were taken in liver segment 8 through an intercostal approach without a specific breathing maneuver.Care was taken to minimize the pressure exerted during measurement and to avoid region of interest placement on large vessels, biliary radicles, and focal lesions (if present).The results were given in meters per second (m/s).
Transient elastography (TE) was measured using Fibroscan ® 502 Touch (Echosens, France) with a low-frequency vibrator (50 Hz) for the excitation of shear waves and an ultrasonic single-element transducer operating at 5 MHz on the axis of the vibrator as described recently (13).The results were given in kilopascals (kPa).

Liver biopsy specimens
Liver biopsies were performed percutaneously and ultrasoundguided under local anesthesia with a 16-gauge needle, fixed in 4% neutral buffered formalin and embedded in paraffin wax.

Histological grading and staging
Histological grading and staging was performed as described recently (14).Sections of 2 µm thickness from liver allograft biopsies were stained with hematoxylin and eosin, elastic van Gieson stain, periodic acid-Schiff stain, silver stain, Berlin blue stain and rhodamine stain.Histological examination was performed by experienced liver pathologists in a blinded fashion.Only Lbx regarded as representative by the examining pathologist, including at least 5 portal fields, were included.The liver tissue was examined according to the Ishak scoring system, as well as liver allograft fibrosis score (LAFSc) and Banff schema for grading liver allograft rejection with the rejection activity index (RAI) (15).The RAI score was constituted by examining portal, bile duct and venous endothelial inflammation with a maximum of 3 points, respectively (16).Patients with at least one point in each of the three categories, therefore showing morphological signs of graft rejection, and non-elevated liver enzymes (AST and ALT≤2x ULN) were diagnosed with subclinical T cell-mediated rejection, as described in previous studies (17).LAFSc was scored by separately assessing portal, sinusoidal and centrilobular areas, each with a maximum of 3 points, allowing the maximum score to lie at 9 points for both RAI and LAFSc (15).For this study, significant fibrosis was defined as any LAFSc component ≥2.The much used Ishak scoring system characterizes fibrosis on a scale from 0 to 6 (18).For this study, at least moderate fibrosis was defined as periportal fibrosis (Ishak F ) ≥ 2. Significant fatty degeneration of graft tissue was diagnosed at an augmentation of lipid vacuoles in hepatocytes of >5% (19).Any biopsy not fulfilling the criteria justifying the minimization of immunosuppression ("BanffMini": portal tract inflammation ≤ 1, interface hepatitis ≤ 1, central perivenulitis ≤ 1, lobular inflammation = 0, biliary inflammation = 0, endothelialitis = 0, portal microvasculitis = 0 and periportal fibrosis ≤ 3) was diagnosed with significant subclinical graft injury (16).

Statistical analysis
Statistical analysis was performed using SPSS 15.0.The Mann-Whitney U test was used to compare quantitative data between two groups.The Chi 2 test was used to prepare contingency tables with two groups.Correlation analyses were calculated with Spearman's rank correlation.Area under the receiver operating characteristic (AUROC) analyses and the Youden's index were used to guide identification of cut-off values.P-values below 0.05 (two-tailed) were considered significant in all analyses.

Results
From 11/2018 to 09/2020, 293 Lbx were performed on 287 patients at our center.All biopsies regarded in this study were surveillance Lbx (svLbx), performed to adjust the immunosuppression regimen without relevantly elevated liver enzymes or previously elevated LSMs (Table 1).Twenty-eight (10%) Lbx were excluded from this study due to nonrepresentative graft biopsy.In total, 226 LSMs were carried out.Baseline patient characteristics are shown in Table 1.Valid LSMs were performed within 4 days prior to Lbx (median: 0; range: 0-4).Overall, one hundred three surveillance biopsies could be paired with at least one valid LSM (ARFI and/or TE).In 42 (41%) cases, both ARFI and TE were matched (Figure 1).
To determine and validate a cut-off value for each LSM, the cohorts for ARFI and TE were divided into a training and a validation group respectively for each LSM methodology.Assignment to either training or validation cohort was performed randomly, aiming for 50%-66% of patients to be in the respective training cohorts.Clinical characteristics of training and validation cohorts are outlined in Supplementary Table S1.
For ARFI, the ideal cut-off value for relevant graft fibrosis (any LAFSc component ≥2), was determined at 1.29 m/s using the Youden's index in the ARFI training cohort.Test characteristics describing the predictive fidelity of ARFI with this cut-off value in the training and validation cohort are outlined in Table 3.With the same approach, the ideal cut-off value for any LAFSc component ≥2 was determined at 7.5 kPa in the TE training cohort.Test characteristics of when this cut-off value was applied are also outlined in Table 3.

Discussion
This is the first study to compare two common LSMs in a svLbx cohort and to generate and validate cut-off values in separate patient cohorts.In the current study, TE showed a more stringent correlation with graft fibrosis, but looked to be more strongly influenced by cholestasis.The validated cut-off values for the prediction of relevant liver graft fibrosis (Ishak F ≥ 2) were 1.29 m/s in ARFI and 6.9 kPa in TE and lower than in a previous study that was still dominated by HCV reinfections (cut-off values: 1.39 m/s (ARFI) and 8.4 kPa (TE)) (20).Cut-off values for advanced fibrosis were 1.29 m/s in ARFI and 7.5 kPa in TE when liver graft fibrosis was quantified according to LAFSc, which is more accurate for fibrosis quantification than the Ishak fibrosis score (15).
Other studies also determined cut-off values for each histological fibrosis stage even after transplantation, but these studies mostly did not validate these detailed cut-off values.We, however, preferred to use the sample size to validate the cut-offs predicting relevant graft fibrosis, rather than to generate several cut-offs.Furthermore, other studies rather included indication biopsies, mostly looking for signs of acute graft rejection, while our svLbx and LSMs were performed in patients with normal/ near normal liver values.(4,(23)(24)(25).With regard to TE, some studies have analyzed the correlation with fibrosis after OLT with larger patient numbers, all of them, however, focused on patients with HCV reinfection (75%-100%) (26)(27)(28)).If we had not excluded the indication biopsies in the present study, the optimal cut-off values for relevant liver graft fibrosis would have been higher for ARFI (1.37 m/s) and similar (6.9 kPa) for TE.[Data previously published elsewhere] Compared to other studies, the cut-off value established for TE in this study was relatively low, since other results have ranged from 4.7 to 12.3 kPa (20,(29)(30)(31).In this study, the cut-off values at 1.29 m/s and 7.5 kPa for advanced fibrosis according to LAFSc (and 1.29 m/s and 6.9 kPa for advanced fibrosis according to Ishak score) showed excellent negative predictive values, so that ruling out significant graft fibrosis using LSMs can be deemed considerably safe (Tables 3, 4).While sensitivity was a little lower, the detection of graft injury is more important for OLT recipients than confirmation of health in graft tissue.
Regular testing of liver enzymes paired with TE and/or ARFI can be helpful in between biopsies and in patients declining offers of svLbx.According to our results, since TE was more stringently associated with graft fibrosis, we recommend primarily performing TE.However, if TE does not deliver valid results or if TE is not available at the transplant center, ARFI ROC curves showing correlation of liver stiffness measurements with any kind of histological graft injury (beyond banffMini).(A) ARFI: the ROC curve was calculated for valid ARFI measurements (n = 69) and any kind of histological graft injury (anything beyond BanffMini).Histopathological evaluation yielded BanffMini criteria fulfilled for 25 (36%) and not fulfilled for 44 (64%) patients.(CI = 95% confidence interval) (B) TE: the ROC curve was calculated for valid TE measurements (n = 76) and any kind of histological graft injury (anything beyond BanffMini).Histopathological evaluation yielded BanffMini criteria fulfilled for 26 (34%) and not fulfilled for 50 (66%) patients.(CI = 95% confidence interval).This study showed significant correlation between LSMs and graft fibrosis.This non-invasive prediction can help to screen liver graft recipients in order to explore the causes of liver graft fibrosis such as insufficiently controlled alloreactivity or disease recurrence (17,(33)(34)(35).When LSM is above the defined cut-off value (1.29 m/s and 7.5 kPa respectively; the respective cut-offs for advanced fibrosis according to Ishak score were 1.29 m/s and 6.9 kPa), an indication graft biopsy should follow to securely diagnose graft fibrosis or other kinds of graft injury.LSM can be considered a pre-biopsy screening program to prioritize patients for Lbx.Personalized immunosuppression programs based on LSM screening and subsequent svLbx can help to improve the balance of necessary immunosuppression and immunosuppression associated side effects without relevant biopsy risk in OLT recipients (9).For example, knowledge of Ishak F2 fibrosis in a liver graft with normal liver values could lead to changes in immunosuppression, i.e., switching from high-dosage tacrolimus and MMF to low-dose tacrolimus and an mTOR inhibitor.
Although LSM is associated with inflammation and fibrosis before and after transplantation, we were not able to see an association of relevant liver graft injury in terms of 2016 BANFF criteria for the reduction of immunosuppression (BanffMini) and LSM.These findings match the recently published study by Vionnet et al. (8).However, the predictive fidelity for liver graft injury, defined by an elevated expression of rejection associated transcripts, could be increased by the combination of LSM with aminotransferase levels (8).Unfortunately, this combined prediction of LSM and aminotransferases could not be assessed in this cohort, because gene expression was not assessed in our clinical study.
Since all the performed Lbx were svLbx, an obvious limitation to this study was lack of sufficient availability of graft cirrhosis, so that safe cut-off values between various grades of graft fibrosis and cirrhosis could not be determined.Also, the patient number is at the lower limit of what is necessary to determine valid cut-off values for the identification of advanced fibrosis.Follow-up LSMs remain to be performed to evaluate changes in their results over time.Another important limitation was the small number of valid ARFI measurements for the left liver lobe, making it impossible to propose a specific cut-off value.ARFI measurements of the left lobe, however, are needed, since especially children regularly receive left-lobe split transplantations.
In conclusion, this study showed comparatively low cut-off values for the non-invasive detection of graft fibrosis in OLT recipients regardless of the underlying diseases in the absence of HCV reinfection.Transient elastography is primarily recommended due to lower rates of non-valid measurements and better correlation with graft fibrosis.In patients with LSM above 1.29 m/s and 7.5 kPa respectively, Lbx should be offered to determine the histopathological stage of fibrosis and evaluate changes in the immunosuppression regimen (9).
Compared to other studies, the sample size for ARFI (n = 69) was quite large (Abdelhaleem et al. 2018: n = 70; Liao et al. 2014: n = 57, Schmillevitch et al. 2016: n = 33) (4, 5, 21, 22).The cutoff at 1.29 m/s was more within the lower range of most other studies on ARFI in OLT recipients dominated by HCV reinfection, in which cut-off values ranged between 1.29 m/s and 1.75 m/s
Percentages are shown in brackets.

TABLE 1
Baseline characteristics of patients included in the study (n = 103).

TABLE 3
Predictive accuracy of cut-off values for ishak F ≥ 2.