Serological Tests Do Not Predict Residual Fibrosis in Hepatitis C Cirrhotics with a Sustained Virological Response to Interferon

Background and Aim Liver biopsy (LB) has lost popularity to stage liver fibrosis in the era of highly effective anti-hepatitis C virus (HCV) therapy, yet diagnosis of persistent cirrhosis may have important implications following HCV eradication. As performance of serological non-invasive tests (NITs) to predict residual fibrosis in non-viremic HCV patients is unknown, we investigated accuracy of NITs to predict residual fibrosis in cirrhotics after a sustained virological response (SVR) to interferon (IFN). Methods Thirty-eight patients with a pre-treatment histological diagnosis of cirrhosis and a 48–104 months post-SVR LB were tested with APRI, CDS, FIB-4, FibroQ, Forns Score, GUCI Index, King Score, Lok Index, PLF, ELF. In 23 (61%) patients, cirrhosis had histologically regressed. Results All NITs values declined after SVR without any significant difference between regressors and non-regressors (AUROC 0.52–0.75). Using viremic cut-offs, PPV ranged from 34% to 100%, with lower NPV (63% - 68%). NITs performance did not improve using derived cut-offs (PPV: 40% - 80%; NPV: 66% - 100%). PLF, which combines several NITs with transient elastography, had the best diagnostic performance (AUROC 0.75, Sn 61%, Sp 90%, PPV 80%, NPV 78%). After treatment, none of the NITs resulted significantly associated with any of the histological features (activity grade, fibrosis stage, area of fibrosis). Conclusions The diagnostic estimates obtained using both viremic and derived cut-off values of NITs were suboptimal, indicating that none of these tests helps predicting residual fibrosis and that LB remains the gold standard for this purpose.


Introduction
The long-standing dogma that cirrhosis is an irreversible condition has been repeatedly challenged in last decades. Indeed many studies in hepatitis C (HCV) patients with a sustained virological response (SVR) to Interferon (IFN)-based regimens showed cirrhosis regression in 30% to 60% patients 3 to 5 years after SVR [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16]; this histological outcome was associated with reduced risk of developing hepatocellular carcinoma (HCC), hepatic decompensation and variceal bleeding, while, on the contrary, a clinical benefit was not observed in SVR patients with residual cirrhosis [9]. These findings rose the argument about need for assessing residual cirrhosis in HCV patients with an SVR, in order to deliver accurate prognosis and optimize clinical management. Indeed, all SVR cirrhotic patients are currently recommended to keep lifelong surveillance for HCC with abdominal ultrasound (US) every 6 months [17], as these patients still carry a 0.6%-1.2% yearly risk of HCC [13,16,18,19]. The demonstration of cirrhosis regression would allow for tailored surveillance intervals, as the reduced risk of HCC in cirrhosis regressors would make US surveillance every 6 months non cost/effective. Liver biopsy (LB) currently represents the gold standard for accurately assessing liver fibrosis stage and consequently identifying cirrhosis regression, however it has limited application because it is an invasive procedure, poorly accepted by patients and ultimately carries a risk of complications [20,21]. Moreover, LB accuracy is biased by heterogeneous deposition of fibrosis in the liver, sampling error and inter-observer variability [22,23]. These drawbacks highlight the need for alternative non-invasive methods to assess fibrosis stage following an SVR. In this setting, we previously demonstrated that transient elastography (TE) has low diagnostic accuracy in identifying residual cirrhotic patients after an SVR, mainly as a consequence of fibrosis remodeling [24]. The aim of this study was to investigate the diagnostic performance of most popular serological markers of hepatic fibrosis in a unique albeit small group of patients with histologically documented cirrhosis who underwent a second liver biopsy 5 years after achieving an SVR to IFN-based therapies.

Patient population
This study is a post-hoc analysis of a cohort study of 38 HCV cirrhotic patients, who underwent a LB following an SVR to assess histological modifications associated with viral clearance [12]. The first LB was performed within 12 months before anti-HCV treatment, while the second at least 4 years after the SVR [12]. Exclusion criteria were HBV or HIV co-infection, >40 g/day alcohol intake, malignancies, contraindication to LB. At the time of post-SVR LB, all patients had liver stage concomitantly assessed through TE and the following serological non-invasive tests (NITs): APRI, CDS, FIB-4, FibroQ, Forns Score, GUCI Index, King Score, Lok Index, PLF and ELF (see Material and Methods Section). All patients gave their written informed consent to use of their medical records and perform the post-treatment LB for the study, which was approved by the Institutional Review Board of the Department of Internal Medicine.

Histological assessment
All pre-treatment LBs were performed with a cutting needle (Tru-cut 16 G, TSK Laboratories, Tokyo, Japan), whilst a Menghini-like semiautomatic biopsy device (Biomol 16 G, HS Hospital Service, Latina, Italy) was used for post-treatment samples. Liver biopsies were considered adequate for histological analysis if longer than 10 mm and/or carrying at least 12 portal tracts. All biopsies were read independently by two expert Pathologists blinded to any clinical information. Fibrosis stage was assessed according to the METAVIR scoring system [25], while hepatic inflammation was evaluated according to both METAVIR and Ishak grading systems [25,26]. Cirrhosis regression was defined as decrease from F4 to F3 or less when comparing pre-and post-treatment LBs [12]. Area of fibrosis was calculated by using the HISTOLAB1 software (ALPHELYS, Plaisir, France), and was defined as ratio of fibrous tissue area stained with picrosirius red (%) upon the total tissue surface, as previously described [12].

Non-invasive measurements
Transient elastography. Transient elastography was performed as already described [27]; liver stiffness measurement (LSM) was expressed in kiloPascal (kPa) as the median value of the successful measurements. As previously published [28] TE threshold of ! 12 kPa was used for the diagnosis of cirrhosis.
Serological non-invasive tests. The following serological indirect and direct markers of liver fibrosis [29] were tested according to their formula: 1. Indirect serological markers APRI (AST to platelet ratio): AST levels (× ULN)/platelets count (10 3 /l) × 100 CDS (Cirrhosis Discriminate Score): calculated by summing the scores awarded for the following laboratory results (Table 1)

Statistical analysis
Comparisons between groups were made by using the Mann-Whitney U or the Student's-t tests for continuous variables, and the χ2 or Fisher test for categorical data. A probability value of p<0.05 was considered statistically significant. The diagnostic accuracy of non-invasive markers in SVR patients was expressed as sensitivity, specificity, positive and negative likelihood ratios (LR+ and LR-), positive and negative predictive values (PPV and NPV), using receiver operating characteristic (ROC) curves analysis. The accuracy of NIT and TE to identify cirrhosis has been assessed. A receiver operating characteristic (ROC) curve analysis was performed and the area under the ROC curve (AUC) has been reported. Categorical data were reported as counts (percentages) and continuous variables as mean (standard deviation) or median (interquartile range, IQR), as appropriate.
Uni-and multivariate linear regression analyses were performed to assess the ability of the NITs (APRI, CDS, FIB-4, FIBRO-Q, FORNS score, GUCI index, King score, Lok index, ELF, PLF) in predicting the histological features (i.e. activity grade, fibrosis stage, area of fibrosis). Also TE was included in the analysis. Only those parameters that showed a statistically significant result at the univariate analysis were entered in the multivariate analysis. The performance of the linear models was evaluated by the determination coefficient (R 2 ). Two analyses were performed considering, in separate, pre-and post-treatment (SVR) data. P values lower than 0.05, two sided, were considered statistically significant. All statistical analyses were performed with SAS statistical software (release 9.4; SAS Institute Inc, Cary, NC).

Indirect markers of fibrosis
NITs values at post-SVR LB and their reference cut-offs for the diagnosis of cirrhosis are shown in Table 3. Forns Score and PLF were calculated in 33 (87%) patients with available cholesterol and TE values.
Apart from PLF, NITs values did not significantly differ in regressed and non-regressed patients (Table 3), and also when stratifying data according to post-SVR fibrosis stage [(F1 vs.  Table 3, Fig 2).
When analyzing diagnostic performances according to ROC curves generated on reference cut-offs, all NITs demonstrated low diagnostic accuracy in identifying patients with residual cirrhosis after the SVR. Indeed, due to low sensitivity and specificity, AUROC ranged from  0.51 (CDS) to 0.75 (PLF) ( Table 4). NITs accuracy remained low also when using the derived thresholds, obtained by combining the best sensitivity/specificity, with PPV values ranging between 40% (CDS) and 80% (FIB-4 and PLF), and NPV between 66% (APRI and Lok Index) and 100% (CDS) ( Table 4). Therefore, 5% to 40% of SVR patients were misclassified as regressors (<F4) by NITs, despite cirrhosis persistence at post-SVR LB. NITs values at baseline are reported in Table 5. Most NITs showed a significant post-treatment decrease (Table 5; Figs 3 and 4), except for CDS, FibroQ and Lok Index.
Before treatment, none of the NITs showed any correlation with the most important histological features (i.e. fibrosis, activity, area of fibrosis). After treatment, some of the NITs (CDS, FIB-4, FibroQ, PLF) showed a correlation with the morphometric area of fibrosis, only; however, this correlation was lost at multivariate analysis (Table 6).
At linear regression analysis, ELF values did not correlate with any of the histological parameters as assessed at LB (activity, fibrosis, area of fibrosis), both at baseline and after the achievement of an SVR (Table 6).

Discussion
In this proof of concept study about diagnostic accuracy of serological assays for liver fibrosis, all tests failed to separate SVR patients with residual cirrhosis from minor levels of fibrosis. Cumulatively, the diagnostic accuracy of these tests by AUROC ranged from 0.51 and 0.75, whereas, individually, none had a cut-off value that could distinguish between different stages of fibrosis according to METAVIR.
In HCV patients, serological markers of liver fibrosis are becoming increasingly popular, as they have shown satisfactory accuracy for liver fibrosis staging, with AUROC ranging between 0.78 and 0.89 according to different studies [29,45]. However, the performance of serological tests in HCV patients is lower than TE [28,29,45]. We previously demonstrated that TE failed to identify residual cirrhosis after the SVR using the classical viremic TE cut-off of 11.9 kPa [24], speculating that the reduced TE accuracy in non-viremic patients could be the consequence of fibrosis remodeling occurring after the SVR. The present study focuses on this same subgroup of patients who were evaluated with both direct and indirect serological markers of fibrosis. Nine indirect tests failed to identify patients with residual cirrhosis, showing lower diagnostic accuracy in comparison with their performance reported in viremic patients [29]. This poor diagnostic accuracy was in line with the reduced TE performance in this non-viremic population (AUROC serum assays vs. TE: 0.51-0.75 vs. 0.75) [24]. The loss in diagnostic accuracy was probably driven by the normalization of biochemical parameters that are considered in each test assessed, thus confirming a previous report by Sebastiani et al, who showed poor performance of APRI, AAR, Forns' Index, Fibrotest and Fibroindex in 80 viremic patients with persistently normal transaminase values [46]. On the other hand, a recent study in 115 patients tested with APRI, FIB-4 and Forns Score 6 years after the SVR found increased values in parallel with histological fibrosis stage (p<0.0001), however with low specificity and PPV [47]. These findings discrepant from our study could be mainly explained by differences in patient demography, stage of fibrosis and cumulative analysis of residual F3 and F4 stages.
To overcome the pitfalls of a single-test approach in a relatively small study, we tested whether the combination of different serological tests and TE resulted in improved diagnostic accuracy. This notwithstanding, PLF, which combines four indirect markers of fibrosis and TE, showed poor diagnostic accuracy, although it carried the best performance among all tests. Differences observed in PLF values between regressors and non-regressors (p = 0.01) and across all stages of fibrosis (p = 0.01) were likely consequence of the inclusion of TE values in the test formula, since TE values significantly separated F4 from <F4 patients (p = 0.01) and correlated  [24].
Since indirect serological tests are mostly based upon biochemical parameters of liver inflammation, which do not directly reflect extracellular matrix (ECM) remodeling, we also tested the diagnostic accuracy of ELF test, a direct marker of liver fibrosis that could be an ideal candidate for our study purposes. Indeed, ELF formula includes specific compounds of the ECM (TIMP-1, PIIINP, HA), which could be more sensitive to hepatic fibrosis rearrangement occurring after HCV eradication. Although we were not able to calculate ELF values for all the patients included in this study, we observed that these values were significantly reduced in post-SVR assessment compared to pre-treatment sampling. ELF diagnostic performance for identifying residual cirrhosis was sub-optimal in line with other indirect tests. Even the 8.1 derived cut-off did not further improve ELF diagnostic accuracy (AUROC 0.63). Another study in 81 SVR patients (22% with cirrhosis) found lower ELF values at follow-up than at baseline, while median ELF values in SVR patients were lower than in non-responder patients [48]. However, this study was flawed by the lack of a histological reference standard.
While a reduction of serum tests for fibrosis following an SVR has already been reported in HCV patients [47,49], the present study has the merit to correlate serological tests with posttreatment stages of liver fibrosis. Indeed, identification of residual cirrhotics in SVR patients bears clinical interest, since cirrhotic patients who achieve an SVR following antiviral treatments are currently maintained under surveillance due to their residual risk of HCC [17]. A few years ago, Mallet and colleagues demonstrated that patients who achieved cirrhosis regression after an SVR remained free from liver-related complications in the following 5 years [9] whilst HCC developed in patients with persistent cirrhosis, only. Thus, assessment of post-SVR fibrosis stage might help individualizing the surveillance algorithm in patients with a pre-treatment diagnosis of cirrhosis who achieved viral eradication, although this individual approach is not recommended by international societies [17]. A point which needs clarifications is also whether serum assays of liver fibrosis predict long-term outcome and prognosis of SVR patients as they do with non-viremic patients [41,50]. By the same token, it could be of interest to identify post-SVR cut-offs of fibrosis tests predicting clinical events during surveillance. In our small study, we recorded no liver-related events over a 5-year follow-up, thus preventing the possibility to correlate post-SVR fibrosis stage with clinical outcomes, or to identify those patients at risk of complications according to their post-treatment values of serum tests. Due to the lack of robust studies assessing the accuracy of non-invasive tests in staging post-SVR fibrosis and/or predicting clinical events after HCV eradication, international guidelines do not recommend their use in tailoring the management of post-treatment follow-up [45].
We acknowledge weaknesses of our study that may ultimately affect the strength of our conclusions. No doubts that sample size is one such weakness, but it was to some extent the result of our stringent criteria of patient selection in terms of well-compensated patients fitting IFNbased therapy, who gave their informed consent to perform a second liver biopsy and the long term surveillance of 5 year on average post-SVR. In addition, for some of these patients Forns Score, PLF and ELF tests were not available, thus probably limiting the accuracy of their predictive value among this cohort. Finally, the retrospective design of the study prevented the dynamic study of serum tests of fibrosis. On the other hand, we think that these cons are counterbalanced by several strengths including the single center design in which liver biopsies were collected before treatment and 5 years following an SVR, and read by two expert pathologists blinded to any clinical features.
In conclusion, this study demonstrates that serological biomarkers of liver fibrosis failed to stratify SVR patients by degree of residual fibrosis and, most important, did not allow identification of patients with residual cirrhosis following an SVR.

Author Contributions
Conceived and designed the experiments: RD AA PL MC. Performed the experiments: RD PB WR MC. Analyzed the data: RD ED PL WR GC MGR MF. Contributed reagents/materials/ analysis tools: FF EG. Wrote the paper: RD ED AA PL.