Combination of blood tests for signi fi cant fi brosis and cirrhosis improves the assessment of liver-prognosis in chronic hepatitis C
J. Boursier*,†, C. Brochard*, S. Bertrais†, S. Michalak‡, Y. Gallois†,§, I. Fouchard-Hubert*,†, F. Oberti*,†, M.-C. Rousselet†,‡& P. Cales*,†
*Liver-Gastroenterology Department, University Hospital, Angers, France.
†HIFIH Laboratory, UPRES 3859, SFR 4208, LUNAM University, Angers, France.
‡Pathology Department, University Hospital, Angers, France.
§Biochemistry Department, University Hospital, Angers, France.
Correspondence to:
Dr J. Boursier, Service d’Hepato- Gastroenterologie, CHU, 49933 Angers Cedex 09, France.
E-mail: JeBoursier@chu-angers.fr
Publication data
Submitted 26 December 2013 First decision 15 January 2014 Resubmitted 4 May 2014 Accepted 7 May 2014
This article was accepted for publication after full peer-review.
SUMMARY
Background
Recent longitudinal studies have emphasised the prognostic value of noninvasive tests of liver fibrosis and cross-sectional studies have shown their combination significantly improves diagnostic accuracy.
Aim
To compare the prognostic accuracy of six blood fibrosis tests and liver biopsy, and evaluate if test combination improves the liver-prognosis assessment in chronic hepatitis C (CHC).
Methods
A total of 373 patients with compensated CHC, liver biopsy (Metavir F) and blood tests targeting fibrosis (APRI, FIB4, Fibrotest, Hepascore, FibroMeter) or cirrhosis (CirrhoMeter) were included. Significant liver-related events (SLRE) and liver- related deaths were recorded during follow-up (started the day of biopsy).
Results
During the median follow-up of 9.5 years (3508 person-years), 47 patients had a SLRE and 23 patients died from liver-related causes. For the prediction of first SLRE, most blood tests allowed higher prognostication than Metavir F [Harrell C- index: 0.811 (95% CI: 0.751–0.868)] with a significant increase for FIB4: 0.879 [0.832–0.919] (P = 0.002), FibroMeter: 0.870 [0.812–0.922] (P = 0.005) and APRI:
0.861 [0.813–0.902] (P= 0.039). Multivariate analysis identified FibroMeter, Cir- rhoMeter and sustained viral response as independent predictors offirst SLRE. Cir- rhoMeter was the only independent predictor of liver-related death. The combination of FibroMeter and CirrhoMeter classifications into a new FM/CM classification improved the liver-prognosis assessment compared to Metavir F stag- ing or single tests by identifyingfive subgroups of patients with significantly differ- ent prognoses.
Conclusions
Some blood fibrosis tests are more accurate than liver biopsy for determining liver prognosis in CHC. A new combination of two complementary blood tests, one tar- geted forfibrosis and the other for cirrhosis, optimises assessment of liver-prognosis.
Aliment Pharmacol Ther
INTRODUCTION
The dramatic progress in chronic hepatitis C (CHC) therapy let appear the exciting possibility to cure almost patients because of the high rate of sustained viral response (SVR) and little side effects. As a consequence, determination of fibrosis stage in CHC could appear soon obsolete as it will have much less influence in the treatment decision. Nevertheless, such therapies are not yet available in clinical practice. In addition, for eco- nomic reasons, several countries will probably continue to select patients for treatment according to their liver-related prognosis that is mainly related to liver fibrosis degree. Finally, cirrhosis will remain an impor- tant diagnosis because it is associated with lower rates of SVR with the new therapies and, even if the virus is eradicated, the screening for hepatocellular carcinoma is still required as cirrhosis might not reverse after SVR.
For all these reasons, liver fibrosis evaluation is still relevant in CHC patients.
Compared to staging on liver biopsy, blood fibrosis tests have several advantages for the evaluation of liver fibrosis: they are noninvasive, they can be performed anywhere, there is no measurement failure and their reproducibility is excellent.1 As a rule, bloodfibrosis tests have been constructed for the binary diagnosis of signifi- cant fibrosis, i.e. Metavir F0/1 vs. F ≥2. Because their results are well correlated with the ordinal scale of path- ological fibrosis stages, some fibrosis classifications have been then developed to provide an estimation of the fibrosis stage(s) from the blood test result. Thesefibrosis classifications are particularly helpful for physicians in clinical practice: they give an interpretation of blood test result, they provide a more precise diagnosis of liver fibrosis than the initially proposed binary diagnosis of significantfibrosis and, asfibrosis staging on liver biopsy, they discriminate the patients into several subgroups supposed to have different levels of liverfibrosis.
Nevertheless, the use of blood fibrosis tests in clinical practice remains under debate.2, 3 Cross-sectional studies that demonstrated the good accuracy of blood fibrosis tests are criticised because they used liver biopsy as refer- ence and were thus impaired by the lack of a perfect gold standard.4 Moreover, the fibrosis classifications may be considered imprecise when their diagnosis is‘Metavir F1/2’ or ‘F3/4’. Keeping in mind that the main interest of liver fibrosis assessment in CHC is to evaluate prog- nosis, the best way to circumvent these drawbacks and assess the practical relevance of blood fibrosis tests is to evaluate their prognostic significance in longitudinal
studies. Previous studies that evaluated the prognostic value of blood fibrosis tests shared several drawbacks:
small samples of patients,5, 6 heterogeneous cohorts with various causes of chronic liver diseases,7 relatively short follow-up,8–10 primary outcome focusing on mortality rather than liver decompensation,6, 9, 10 and no direct comparison between blood fibrosis tests.5, 7, 8, 10 Finally, these studies evaluated the prognostic value of single bloodfibrosis tests.
Consequently, the primary aim of the present study was to compare the prognostic accuracy of several blood fibrosis tests and liver biopsy for the prediction of liver-related events in a large cohort of CHC patients with long-term follow-up. As recent studies demon- strated that combination of fibrosis tests significantly improves the diagnostic accuracy,11–13our secondary aim was to evaluate if a combination of blood fibrosis tests can also improve the liver-prognosis assessment.
PATIENTS AND METHODS
Patients
We used a previously published database of CHC patients with liver biopsy and blood fibrosis tests from the University Hospital of Angers, France (tertiary refer- ral center).14 Patients were included from 1994 to 2007 if they had CHC, defined as both positive anti-hepatitis C virus antibodies and hepatitis C virus RNA in serum.
Exclusion criteria were other causes of chronic hepatitis (hepatitis B or HIV co-infection, alcohol consumption
>30 g/day in men or >20 g/day in women in the 5 years before inclusion, hemochromatosis, autoimmune hepati- tis), or cirrhosis complications (ascites, variceal bleeding, systemic infection, hepatocellular carcinoma). Study pro- tocol conformed to the ethical guidelines of the current Declaration of Helsinki and all patients gave informed consent following an IRB approval.
Data collection during follow-up
Follow-up started the day of liver biopsy and ended January 1, 2011. Patients were followed and received anti-viral therapy according to current EASL and AASLD guidelines. Data recorded during follow-up included: achievement of SVR, date of first significant liver-related event, and date and cause of death. SVR was defined as negative hepatitis C virus RNA 6 months after the end of anti-viral treatment. The significant liver-related events (SLRE) under consideration were those requiring specific therapy or care: ascites, encepha-
lopathy, jaundice (serum bilirubin >50lmol/L), occur- rence of large oesophageal varices (diameter ≥5 mm), variceal bleeding, hepatorenal syndrome and hepatocellu- lar carcinoma. Large oesophageal varices were considered as significant liver-related event because their occurrence significantly impacts treatment and prognosis.15, 16 Fol- low-up data were recorded in patient files; for cases lost to hospital follow-up, the patients or their general practi- tioners were called for follow-up data. Date and cause of death were obtained for all patients by consulting the national death registry. Transplanted patients were cen- sored as dead at the date of liver transplantation. Finally, two clinical outcomes were evaluated in the present study: SLRE as primary endpoint (patients were censored at the date of the first event) and liver-related death as secondary endpoint.
Histological assessment
Liverfibrosis was evaluated according to Metavir fibrosis (F) staging. Initial pathological examinations were inde- pendently performed by two senior experts specialised in hepatology, with a consensus reading for discordant cases. The pathologists were blinded for patient charac- teristics and blood marker results.
Blood tests
Blood collection was performed within the 7 days around the date of liver biopsy in 87.4% of the cases, and no more than 3 months thereafter in the other cases.14 All blood assays were performed in the laboratory of Angers hospital.
Blood fibrosis tests. The following blood fibrosis tests were calculated according to published or patented formu- las: APRI,17FIB4,18 Hepascore,19Fibrotest (Biopredictive, Paris, France),20 FibroMeterV2G (second generation for virus; Echosens, Paris, France).21 We also calculated Cir- rhoMeterV2G(Echosens), which shares the same markers as FibroMeterV2Gbut with specific coefficients targeted for the diagnosis of cirrhosis.22In clinical practice, the results of some bloodfibrosis tests are interpreted by using their fibrosis stage classificationsthat estimate Metavir F stage(s) from the test results (see Figure S1).23Fibrotest classifica- tion includes eight classes (called here FT1 to FT8),24Fib- roMeterV2G includes seven classes (FM1 to FM7) and CirrhoMeterV2Gsix classes (CM1to CM6).
Statistical analysis
Survival curves were determined using the Kaplan–Meier method and compared with the log rank test. Discrimina-
tive ability for the prediction of clinical outcomes was eval- uated by the C-index of Harrell.25The C-index of Harrell is an extension of the AUROC for time-to-event (survival) data and evaluates the concordance between the predicted risk of event and the observed survival time. Its results vary from 0 to 1;1 shows a perfect concordance (discrimi- native power of the risk score), 0.5 shows random predic- tion and a value less than 0.5 indicates discrimination in the opposite direction to that expected. For each test eval- uated, the 95% confidence interval of the Harrell C-index was calculated by a bootstrap method, using 1000 random samples (with replacement) of the same size as the original data set. Paired comparisons of the C-indexes were per- formed using a bootstrap resampling procedure, as previ- ously described.26To compare two tests, we calculated the difference between their C-indexes on each of the 1000 random samples chosen with replacement from the origi- nal data set. Then, the bootstrap 95% confidence interval of the difference was obtained, and we used the estimate of the standard error from this bootstrap distribution to cal- culate the standardised difference between the C-Indexes andfind the correspondingPvalue. Most statistical analy- ses were performed usingSPSSversion 18.0 software (IBM, Armonk, NY, USA) andSAS 9.1 (SAS Institute Inc., Cary, NC, USA) by a professional statistician (SB).
RESULTS
Patients
The main baseline characteristics of the 373 included patients are detailed in Table 1. Median liver biopsy length at inclusion was 20 mm (first quartile: 15 mm, third quar- tile: 25 mm). 47 patients (12.6%) experienced at least one SLRE during follow-up. First SLRE were: large varices (n =20), hepatocellular carcinoma (n =13), ascites (n =4), jaundice (n =4), hepatic encephalopathy (n =1), variceal bleeding (n= 1), ascites and jaundice (n =2), ascites and jaundice and encephalopathy (n= 2). Median follow-up for SLRE was 7.6 years (interquartile range: 4.0– 12.1; 2972 person-years). Cumulative incidence of first SLRE at median follow-up was 11.5% (Figure S2a).
Fifty patients died during follow-up; cause of death was unknown in two cases and liver-related in 17. Also, six patients had liver transplantation. Death was thus considered as liver-related in 23 patients. As death occurred after the first SLRE, the median follow-up for death was longer than for SLRE: 9.5 years (interquartile range: 5.8–13.0; 3508 person-years). Survival without liver-related death and overall survival at median fol- low-up were, respectively: 95.1% (Figure S2b) and 88.8%.
Of the total patients, 37.2% patients achieved SVR during follow-up, and 62.8% had either no or unsuccess- ful anti-viral therapy. There was no liver-related death during the follow-up of patients who achieved SVR, and only five of them had a SLRE. Three of these five patients (two with Metavir F4 and 1 with F3 at inclu- sion) developed large oesophageal varices before achiev- ing SVR. One patient (Metavir F2 at inclusion) had large oesophageal varices 5 years after achieving SVR in a context of metabolic syndrome, and thus attributed to an ongoing active nonalcoholic steatohepatitis. The last patient (Metavir F3 at inclusion) had jaundice and asci- tes before achieving SVR in a context of recovery of excessive alcohol consumption during follow-up.
Prediction offirst SLRE
Liver biopsy vs. blood tests. Figure 1a shows the cumu- lative incidence of first SLRE as a function of Metavir F stages. The difference was highly significant between F4 and F3 stages (P <0.001), but not among F0 to F3 stages (P =0.174).
Harrell C-indexes of liver biopsy (Metavir F) and blood fibrosis tests for the prediction of the first SLRE are depicted in Table 2 (paired comparisons between tests are detailed in Table S1). Briefly, Metavir F and the six blood fibrosis tests had good discriminative ability for the prediction of the first SLRE (with C-indexes
>0.8). Five of six blood fibrosis tests had a higher C-index than liver biopsy; this increase being significant for FIB4 (P =0.002), FibroMeter (P =0.005) and APRI (P =0.039).
To evaluate each blood fibrosis test against liver biopsy, we performed successive multivariate stepwise forward Cox models including each blood test with Metavir F, adjusted on age, sex, HCV genotype (one vs.
others) and SVR for the prediction of first SLRE (Table S2). Each blood fibrosis test was an independent predic- tor of SLRE (at the first step, except for Hepascore).
Metavir F was not selected as an independent predictor of SLRE when introduced with FibroMeter or Cirrho- Meter in the multivariate analysis. This showed that Fib- roMeter or CirrhoMeter entirely caught the independent prediction without any significant additional prediction by liver biopsy.
Taken together, all the previous results show that blood fibrosis tests are predictors at least as accurate, and for some significantly better than Metavir F staging for the prediction of SRLE in CHC patients.
Fibrosis classifications. As our results showed blood fibrosis tests as prognostic markers, they should individ- ualise subgroups of patients with different prognosis. A robust a posteriori determination of these subgroups would require a large cohort (≥1000 patients) with high Table 1 | Patient characteristics at inclusion
All SVR Others* P
Age (years) 42.8 (36.1–52.9) 40.4 (35.5–48.9) 44.4 (36.8–57.3) 0.002
Male sex (%) 235 (63.0) 95 (68.3) 140 (59.8) 0.120
Genotype 1 (%) 247 (66.2) 69 (49.6) 178 (76.1) <0.001
Metavir F (%)
0 13 (3.5) 4 (2.9) 9 (3.8) 0.220
1 147 (39.4) 58 (41.7) 89 (38.0)
2 108 (29.0) 43 (30.9) 65 (27.8)
3 56 (15.0) 22 (15.8) 34 (14.5)
4 49 (13.1) 11 (7.9) 38 (16.2)
ALT (IU/L) 73 (48–120) 82 (53–121) 68 (45–115) 0.093
Prothrombin time (%) 95 (89–102) 95 (89–101) 96 (89–103) 0.593
Bilirubin (lmol/L) 10 (7–13) 9 (7–13) 10 (7–13) 0.376
Albumin (g/L) 44.0 (41.7–47.1) 44.7 (42.3–47.1) 44.0 (41.5–47.2) 0.168
Follow-up
Significant liver-related event (%)† 47 (12.6) 5 (3.6) 42 (17.9) <0.001
Death (all causes, %) 56 (15.0) 6 (4.3) 50 (21.4) <0.001
Liver-related death (%) 23 (6.2) 0 (0.0) 23 (9.8) <0.001
Quantitative variables are expressed as median with first and third quartiles in brackets and categorical variables as absolute number with percentage into brackets. SVR, sustained viral response.
* Either no or unsuccessful anti-viral therapy.
†Only thefirst significant liver-related event during follow-up was taken into account.
rate of events, followed by external validation. We thus chose to evaluate the prognostic significance of blood testsfibrosis classifications (Figure S1). This was relevant for two reasons. First,fibrosis classifications allow for an a prioristratification of patients. Second, fibrosis classifi- cations have been developed to estimate fibrosis stages and this original design would validate their ability to identify at-risk patients.
Figure 1b–d shows the cumulative incidence of first SLRE as a function of Fibrotest, FibroMeter or Cirrho- Meter fibrosis classifications. Unlike for Fibrotest, the increase in FibroMeter or CirrhoMeter classes was associ- ated with a progressively worsening prognosis. Moreover, patients included in the FM1 or CM1 classes (16.7% of
patients in each) had excellent prognoses with no SLRE during follow-up. FIB4 provided the highest Harrell C-index for the prediction offirst SLRE, but detailedfibro- sis classification is not available for this blood test. We thus evaluated the prognosis of three subgroups defined by its two recommended 1.45 and 3.25 diagnostic cut-offs.27 Interestingly, these three subgroups had significant differ- ent prognosis (P <0.001, Figure S3). However, 11 patients (4.6%) included in the subgroup with putative best progno- sis (i.e., FIB4<1.45) experienced SLRE during follow-up.
How to predict SLRE in clinical practice? .We have already shown in cross-sectional studies that combination offibrosis tests by binary logistic regression improves the
1.0
0.8
0.6
0.4
0.2
0.0
0 5 10 15
Follow-up (years)
20 P < 0.001
F0
F0 vs F1: P = 0.569 F1 vs F2: P = 0.086 F2 vs F3: P = 0.496 F3 vs F4: P < 0.001
F1 F2 F3 F4
Log rank test:
Metavir F stages:
Cumulative incidence of first significant liver-related event
(a)
0 5 10 15
Follow-up (years)
20 P < 0.001
FT1 vs FT2: P = 0.942 FT2 vs FT3: P = 0.629 FT3 vs FT4: P = 0.641 FT4 vs FT5: P = 0.054 FT5 vs FT6: P = 0.822 FT6 vs FT7: P = 0.728 FT7 vs FT8: P = 0.178 Log rank test:
1.0
0.8
0.6
0.4
0.2
0.0
Cumulative incidence of first significant liver-related event
FT1 FT2 FT3 FT4
FT5 FT6 FT7 FT8 Classes of Fibrotest classification:
(b)
FM1 vs FM2: P = 0.016 FM2 vs FM3: P = 0.553 FM3 vs FM4: P = 0.111 FM4 vs FM5: P = 0.470 FM5 vs FM6: P < 0.001 FM6 vs FM7: P = 0.001 Log rank test:
FM1 FM2 FM3 FM4
FM5 FM6 FM7 Classes of FibroMeter classification:
0 5 10 15
Follow-up (years)
20 P < 0.001 1.0
0.8
0.6
0.4
0.2
0.0
Cumulative incidence of first significant liver-related event
(c)
0 5 10 15
Follow-up (years)
20 P < 0.001
CM1 vs CM2: P = 0.040 CM2 vs CM3: P = 0.829 CM3 vs CM4: P = 0.193 CM4 vs CM5: P = 0.066 CM5 vs CM6: P < 0.001 Log rank test:
1.0
0.8
0.6
0.4
0.2
0.0
Cumulative incidence of first significant liver-related event
CM1 CM2 CM3 CM4
CM5 CM6 Classes of CirrhoMeter classification:
(d)
Figure 1 | Cumulative incidence offirst significant liver-related events as a function of Metavirfibrosis (F) stages (a), or classes of Fibrotest (b), FibroMeterV2G(c), CirrhoMeterV2G(d) classifications. There was no significant difference among Metavir F0 to F3 stages (P=0.174) so that liver biopsy identified only two subgroups with different prognosis:
F0–3 vs. F4 (P<0.001). Diagnosticfibrosis classifications used in clinical practice to estimate the Metavirfibrosis stage from FibroMeterV2Gor CirrhoMeterV2Gresults are also prognostic scales: unlike for Fibrotest, the increase in FibroMeterV2Gor CirrhoMeterV2Gclasses was associated with a progressively worsening prognosis.
noninvasive diagnosis of liver fibrosis.11, 28 We used the same design in the present longitudinal study to deter- mine the most complementary combination of blood fibrosis tests for the prediction offirst SLRE.
In multivariate forward Cox model including the six blood fibrosis tests, age, sex, SVR and HCV genotype (one vs. others), independent predictors of first SLRE were CirrhoMeter (P <0.001), FibroMeter (P =0.008) and SVR (P =0.010). This result was not surprising as FibroMeter and CirrhoMeter, the first dedicated to sig- nificant fibrosis and the latter to cirrhosis, are clearly complementary with CirrhoMeter ‘stretching’ high FibroMeter values (Figure 2). The interaction between FibroMeter and CirrhoMeter was significant in the mul- tivariate analysis (P = 0.002). In addition, the cut-off between FM4 and FM5 classes clearly distinguished two groups: FM1–4 classes displayed a larger range of Fib- roMeter values than CirrhoMeter values and the opposite was observed in FM5–7 classes (Figure 2). We thus repeated the multivariate analysis separately in FM1–4 and FM5–7 groups. Multivariate backward Cox model including SVR, FibroMeter and CirrhoMeter showed FibroMeter (P =0.024) and SVR (P =0.040) as inde- pendent predictors of first SLRE in FM1–4 group. By contrast, CirrhoMeter (P <0.001) was the only indepen- dent predictor in FM5–7 group. These results suggested that FibroMeter better predicted first SLRE in the early stages of CHC, whereas CirrhoMeter was a better predic- tor in patients with advanced disease.
These results lead us to evaluate a combination of Fib- roMeter classification for patients included in the FM1–4 classes with CirrhoMeter classification for patients included in FM5–7 classes. This combination provided 10 intermediate classes (FM1–4 and CM1–6 if FM5–7). These 10 intermediate classes were grouped intofive final clas- ses owing to significantly different prognosis (FM1, FM2/
3, FM≥4 +CM≤4, FM≥4 +CM5, FM≥4 + CM6) defining the new FM/CM classification. The rates of patients included in the five classes of the FM/CM classification were, respectively from the first to the fifth: 16.7%, 41.5%, 27.2%, 10.0% and 4.6%.
FM/CM classification vs. liver biopsy. As with Metavir F staging, FM/CM classification split patients into five clas- ses but provided several advantages. First, the cumulative incidence offirst SLRE during follow-up was significantly different between adjacent classes of the FM/CM classifi- cation (Figure 3a) whereas there was no significant dif- ference between Metavir F0 to F3 stages (Figure 1a).
Second, no SLRE occurred during the follow-up of patients included in thefirst class of the FM/CM classifi- cation whereas 1 F0 patient according to baseline liver biopsy had jaundice during follow-up. Third, patients included in the highest class of the FM/CM classification had a markedly worse prognosis (Figure 3a) than cir- rhotic patients as defined by liver biopsy (Figure 1a).
Finally, in multivariate stepwise forward Cox Model including Metavir F, FibroMeter, CirrhoMeter and FM/
CM classifications, FM/CM classification was selected at the first step of the multivariate analysis as independent predictor of SLRE.
Liver-related death
Survival without liver-related death was significantly dif- ferent between Metavir F4 and F3 stages (P <0.001), but not among F0 to F3 stages (P =0.456, Figure S4a).
All blood fibrosis tests had a higher C-index than liver biopsy (Table 2). CirrhoMeter was the only independent predictor of liver-related death among blood fibrosis tests. A multivariate stepwise forward Cox Model includ- ing Metavir F, FibroMeter, CirrhoMeter and FM/CM classifications showed that FM/CM classification was the
Test First significant liver-related event Liver-related death Metavir F 0.811 (0.751–0.868)* 0.849 (0.755–0.920)†
APRI 0.861 (0.813–0.902) 0.869 (0.793–0.932)
FIB4 0.879 (0.832–0.919) 0.911 (0.851–0.954)
Fibrotest 0.838 (0.776–0.897) 0.856 (0.765–0.932) Hepascore 0.809 (0.729–0.880) 0.894 (0.811–0.951) FibroMeterV2G 0.870 (0.812–0.922) 0.898 (0.842–0.945) CirrhoMeterV2G 0.835 (0.762–0.902) 0.884 (0.795–0.952)
* Comparison of Metavir F vs. blood tests: P=0.039 vs. APRI, P=0.002 vs. FIB4, P=0.005 vs. FibroMeterV2G,P>0.340 for other comparisons.
† Comparison of Metavir F vs. blood tests: P=0.011 vs. FIB4, P=0.078 vs. Fibro- MeterV2G,P>0.190 for other comparisons.
Table 2 | Harrell C-indexes of liver biopsy (Metavir F) and bloodfibrosis tests for the prediction of the first significant liver-related event or liver-related death
only independent predictor of liver-related death. Fig- ure 3b shows that survival without liver-related death worsened as a function of FM/CM classification. Detailed results are available in Supporting information.
Evaluation of liver prognosis in clinical practice To summarise, liver prognosis in CHC is best assessed by a combination of two complementary tests: the Fib- roMeter (targeted forfibrosis) in patients with mild dis- ease and CirrhoMeter (targeted for cirrhosis) in patients with advanced disease. Both tests provide a combined FM/CM classification which is the best independent pre- dictor of first SLRE as well as liver-related death. The incidences of 5- and 10-year SLRE or liver-related death as a function of FM/CM classification are presented in Figure 4 (details in Table S3).
DISCUSSION
Currently, the use of blood fibrosis tests for the evalua- tion offibrosis in CHC patients remains controversial.2, 3 This is due to the reported moderate accuracy of blood fibrosis tests for the diagnosis of intermediate stages of fibrosis, and the lack of perfect gold standard in the cross-sectional studies that evaluated their diagnostic accuracy. Basically, a more comprehensive approach to
validate the use of blood fibrosis test in clinical practice is the evaluation of their prognostic accuracy for clinical events in longitudinal studies. We focused the current work on this approach, and our study has several strengths: (i) a large number of patients included, (ii) only a single cause of chronic liver disease: CHC, (iii) a well-balanced repartition of fibrosis stages as described in a large reference cohort of CHC patients,29 (iv) two clinically relevant outcomes evaluated: SLRE (for which prediction by blood fibrosis tests is poorly evaluated in the literature) and liver-related death, (v) long-term fol- low-up (median: 9.5 years) and (vi) direct comparison of liver biopsy with six blood fibrosis tests including a new test specifically developed for cirrhosis.
By using several statistical methods (Harrell C-index, comparison of Kaplan–Meier curves, multivariate Cox Model), we demonstrated that blood fibrosis tests are at least as accurate, and for some significantly more accu- rate than liver biopsy for the prediction of SLRE or liver-related death in CHC patients. We also identified a combination comprising a blood test targeted for the diagnosis of significant fibrosis (FibroMeterV2G) and another targeted for cirrhosis (CirrhoMeterV2G) as accu- rate to stratify CHC patients in subgroups having signifi- cant different liver-related prognosis. Finally, the present
1.00 Significant liver-related event during follow-up
no yes
1.00 0.80
0.80 0.60
0.60 0.40
0.40 0.20
0.20
FibroMeter
CirrhoMeter
F0/1 F1/2
F1/2 F0/1
F 1
F3 F2±1 F3±1 /4
F3±1 F2±1
F4
F4
F3/4
FM5-7 FM1-4
FibroMeter fibrosis classification
CirrhoMeter fibrosis classification
0.00 0.00 Figure 2 |Scatter plot of
FibroMeter2GV(dedicated to significantfibrosis diagnosis) and CirrhoMeter2GV(for cirrhosis diagnosis), with their correspondingfibrosis classifications. These two blood tests are
complementary with CirrhoMeterV2G‘stretching’ high FibroMeterV2Gvalues.
Two distinct groups can be distinguished: FM1–4classes displayed a larger range of FibroMeterV2Gvalues than CirrhoMeterV2Gvalues, and the opposite was observed in FM5–7classes.
prognostic study validates the use of blood fibrosis tests and proposes an unprecedented combination of blood tests for the assessment of liver prognosis in clinical practice.
One of the 13 F0 patients included in our study had jaundice after 5.5 years of follow-up. This surprising evo- lution may be explained by the conjunction of two draw- backs of fibrosis staging on liver biopsy: underestimation of liver fibrosis at inclusion because of sample bias and rapid progression offibrosis during follow-up than cannot be closely follow by repeated liver biopsy. Our study also demonstrates the poor discriminative ability of intermedi- ate pathological fibrosis stages for the medium-term pre- diction of liver prognosis. Indeed, despite the good discriminative ability of Metavir F staging according to the Harrell C-index, the cumulative incidence of first SLRE and survival without liver-related death were not different among F0 to F3 stages in our population, and only F4 patients had a significantly worse prognosis. This result is in accordance with those obtained in the HALT-C cohort where incidence of liver-related outcomes was significantly
higher in Ishak F6 stage vs. F5 and in F5 vs. F4, but not in F4 vs. F3.30
Our results demonstrate that blood fibrosis tests are globally at least as accurate as Metavir F staging for the prediction of liver prognosis in CHC patients. Indeed, none of the blood fibrosis tests had a significantly lower Harrell C-index than liver biopsy for the prediction of SLRE or liver-related death. In multivariate analyses including each blood test with Metavir F staging, almost all blood fibrosis tests were independent predictors of SLRE or liver-related death, suggesting that bloodfibrosis tests provide relevant prognostic information. Moreover, FibroMeter and CirrhoMeter were independent predic- tors of SLRE with no role for Metavir F, as was Cirrho- Meter for liver-related death.
Among the six bloodfibrosis tests evaluated, multivar- iate analysis identified FibroMeter and CirrhoMeter as the best combination for the prediction of first SLRE.
These two tests, the former specifically targeted for sig- nificant fibrosis and the latter for cirrhosis,22 are very complementary (Figure 2) and their association improves
1.0
0.8
0.6
0.4
0.2
0.0
0 5 10 15
Follow-up (years)
1.0
0.8
0.6
0.4
0.2
0.0
0 5 10 15 20
Follow-up (years)
Survival without liver-related death
20 P < 0.001
P < 0.001
FM/CM1
FM/CM1 vs FM/CM2: P = 0.089 FM/CM2 vs FM/CM3: P = 0.006 FM/CM3 vs FM/CM4: P = 0.036 FM/CM4 vs FM/CM5: P < 0.001
FM/CM1 vs FM/CM2: P = 0.568 FM/CM2 vs FM/CM3: P = 0.002 FM/CM3 vs FM/CM4: P = 0.662 FM/CM4 vs FM/CM5: P < 0.001 FM/CM2
FM/CM3 FM/CM4 FM/CM5
Log rank test:
Log rank test:
Classes of FM/CM classification:
Cumulative incidence of first significant liver-related event
FM/CM1 FM/CM2 FM/CM3 FM/CM4 FM/CM5
Classes of FM/CM classification:
(a)
(b)
Figure 3 |Cumulative incidence offirst significant liver-related events (a) and survival without liver-related death (b) as a function of the FM/CM classification. FM/CM classification identifiedfive patient subgroups with significant different prognosis.
the prediction of first SLRE: FibroMeter is more predic- tive at lowfibrosis levels and CirrhoMeter at high fibro- sis levels since it amplifies the test scale like a two-speed gearbox. Finally, similar to the Child–Pugh score or hepatic venous pressure gradient measurement in cir- rhotic patients, CirrhoMeter can stratify high FibroMeter values to improve the identification of subgroups with significantly different prognoses. Because it requires a very large population to robustly determine a posteriori prognostic subgroups, we decided to stratify patients according to the a priori fibrosis classifications or diag- nostic cut-offs that are currently used in clinical practice.
In this setting, the combination of FibroMeter and Cir- rhoMeterfibrosis classifications in the FM/CM classifica- tion has several advantages compared to Metavir F staging. First, whereas Metavir F staging identified only two groups of patients with different prognoses (F0–3 vs.
F4), FM/CM classification identified five significantly dif- ferent groups (Figure 3a). Second, contrary to Metavir F staging, no SLRE occurred in the lower class of FM/CM classification, defining thus a subgroup with excellent prognosis. Third, the highest class of FM/CM identified a subgroup of very high-risk patients that was not identi- fied by Metavir F staging.
FIB4, an inexpensive and easy-to-calculate bloodfibro- sis test, showed very good prognostic accuracy in our study. It was not selected as an independent predictor of
first SLRE by multivariate analysis, probably because it was masked by the high complementary role of FibroMe- ter and CirrhoMeter for this endpoint. By using its two recommended diagnostic cut-offs, FIB4 identified three subgroups of patients whose prognosis was significantly different. However, 11 patients from the best-prognosis subgroup experienced SLRE during their follow-up whereas, by comparison, no patients included in thefirst class of FibroMeter or CirrhoMeter classification had SLRE. This could be due to an inadequate choice for FIB4 thresholds. Indeed, no SLRE occurred during follow-up of patients included in thefirst three deciles of FIB4 (see Fig- ure S5a). This suggests that FIB4 is also able to determine a subgroup of patients with excellent prognosis, and FIB4 thresholds should be refined for this purpose.
Two recent studies have suggested that annual surveil- lance by a noninvasive test of liver fibrosis is cost effec- tive compared to the current surveillance with liver biopsy in CHC patients.31, 32 As our study results but using a different approach, these works provide argu- ments in favour of noninvasive testing instead of liver biopsy in clinical practice. We may suppose thatfibrosis tests with the best prognostic accuracy will be the most cost effective. Furthermore, the FM/CM classification has the advantage of using two tests (FibroMeter and Cirrho- Meter) which are simultaneously calculated with no additional costs because they include the same markers
FibroMeter fibrosis classification
CirrhoMeter fibrosis classification F0/1F F1/2
1
F3 F2±1 F3±1 /4 F4
F3±1
1 0.0 0.0 0.0 0.0
0.0 0.0
0.0
7.0 7.5
0.7 6.0
4.5 15.7
18.2 29.8
76.5 94.1 39.1 59.4 0.0
FM/CM classification:
5-years SLRE (%) 10-years SLRE (%) 5-years liver death (%) 10-years liver death (%)
2 3 4 5
F4 F3/4
Figure 4 |Evaluation of liver-related prognosis from the results of the FibroMeterV2Gand CirrhoMeterV2Gdiagnostic classifications. In patients with mild disease (i.e. included in the threefirst classes of the FibroMeterV2G
classification), the FibroMeterV2Gclassification is used to identify two prognostic subgroups. In patients with advanced disease (i.e. included in the four last classes of the FibroMeterV2Gclassification), the CirrhoMeterV2G classification identify three additional prognostic subgroups. Incidence of liver-related events progressively increases as a function of thesefive prognostic subgroups which define the prognostic FM/CM classification.
but with different coefficients. To our knowledge, pub- lished cost-effectiveness studies have evaluated only a single noninvasive test (Fibrotest31 or Fibroscan32) against liver biopsy, and none has yet compared different noninvasive tests between them. Such article will help to define the best strategy for clinical practice, and if a strategy that uses the FM/CM classification dominates the others.
Results from the present study were obtained in a sin- gle-centre population of CHC patients and should be independently validated in a large multicenter cohort. In addition, our study has two main limitations. Thefirst is the lack of elastometry evaluation. Indeed, as for blood fibrosis tests, liver stiffness measured by Fibroscan appears to be a prognostic factor in CHC patients.33–35 However, Fibroscan became available in 2004 and thus most of our patients had no liver stiffness measurement at the time of their inclusion. The second limitation is the absence of repeated measurements for bloodfibrosis tests during follow-up. Noninvasive tests offibrosis can be eas- ily repeated but the prognostic significance of their evolu- tion during patient follow-up remains poorly evaluated.
In the upcoming area of new anti-viral treatments, evalua- tion offibrosis and even more cirrhosis regression will be particularly relevant in patients who will achieve SVR.
In conclusion, blood fibrosis tests are also prognostic tests at least as accurate as liver biopsy for the predic- tion of liver prognosis in patients with CHC. A combi- nation of two complementary blood tests, one targeted for fibrosis (FibroMeter) and the other for cirrhosis (CirrhoMeter), is more accurate than liver biopsy and appears as a simple and accurate method for the base- line assessment of prognosis in clinical practice.
AUTHORSHIP
Guarantor of the article:Paul Cales, MD.
Author contributions: Jer^ome Boursier: study concept and design, analysis and interpretation of data, drafting of the manuscript, statistical analysis and study supervi- sion. Charlene Brochard, Sophie Michalak, Yves Gallois, Isabelle Fouchard-Hubert, Frederic Oberti and Marie- Christine Rousselet: acquisition of data. Sandrine Bertrais: statistical analysis. Paul Cales: study concept and design, drafting of the manuscript, critical revision of the manuscript for important intellectual content, obtained funding and study supervision. All authors approved thefinal version of the manuscript.
ACKNOWLEDGEMENTS
We thank the following contributors: Gilles Hunault, Pascal Veillon, Gwena€elle Soulard; and Kevin L. Erwin (for English proofreading).
Declaration of personal interests: Paul Cales is inventor of the patented blood test FibroMeter.
Declaration of funding interests: This study was funded in part by grant from the University Hospital of Angers.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article:
Data S1.Liver-related death –supplementary results.
Figure S1.Diagnostic fibrosis stage classifications used in clinical practice for the interpretation of bloodfibrosis tests results. These classifications give an estimated Meta- vir F stage from the bloodfibrosis test result.23, 24 (a) Fi- brotest classification; (b) FibroMeter2GV classification; (c) CirrhoMeter2GVclassification.
Figure S2.Cumulative incidence of first significant liver-related event (a) and survival without liver-related death (b) during follow-up.
Figure S3.Cumulative incidence of first significant liver-related event (a) and survival without liver-related death (b) as a function of the three subgroups defined by the two recommended diagnostic cut-offs for FIB4 (<1.45 and>3.25).
Figure S4.Survival without liver-related death as a func- tion of Metavirfibrosis (F) stages (a), or classes of Fibrotest (b), FibroMeter (c), CirrhoMeter (d) classifications.
Figure S5.Cumulative incidence of first significant liver-related event as a function of deciles of FIB4 (a), Fib- roMeter2GV (b), CirrhoMeter2GV (c), Fibrotest (d), Hepa- score (e) or APRI (f).
Table S1.Paired comparisons of Harrell C-indexes of bloodfibrosis tests and liver biopsy (Metavir F) for the pre- diction offirst significant liver-related event or liver-related death.
Table S2.Evaluation of bloodfibrosis tests against Meta- vir F staging for the prediction of liver-related outcomes.
Each score was separately introduced with Metavir F stage in a multivariate stepwise forward Cox model adjusted on age, sex, sustained viral response and HCV genotype (1 vs.
others).
Table S3.Cumulative incidence (%) of first significant liver-related event (SLRE) or liver-related death as a function of thefive classes of the FM/CM classification.
REFERENCES
1. Cales P, Veillon P, Konate A,et al.
Reproducibility of blood tests of liver fibrosis in clinical practice.Clin Biochem2008;41: 10–8.
2. EASL. Clinical practice guidelines:
management of hepatitis C virus infection.J Hepatol2011;55: 245–64.
3. Ghany M, Strader DB, Thomas DL, Seeff LB. Diagnosis, management, and treatment of hepatitis C: an update.
Hepatology2009;49: 1335–74.
4. Bedossa P, Carrat F. Liver biopsy: the best, not the gold standard.J Hepatol 2009;50: 1–3.
5. Mayo MJ, Parkes J, Adams-Huet B, et al.Prediction of clinical outcomes in primary biliary cirrhosis by serum enhanced liverfibrosis assay.
Hepatology2008;48: 1549–57.
6. Naveau S, Gaude G, Asnacios A,et al.
Diagnostic and prognostic values of noninvasive biomarkers offibrosis in patients with alcoholic liver disease.
Hepatology2009;49: 97–105.
7. Parkes J, Roderick P, Harris S,et al.
Enhanced liverfibrosis test can predict clinical outcomes in patients with chronic liver disease.Gut2010;59:
1245–51.
8. Ngo Y, Munteanu M, Messous D,et al.
A prospective analysis of the prognostic value of biomarkers (FibroTest) in patients with chronic hepatitis C.Clin Chem2006;52: 1887–96.
9. Nunes D, Fleming C, Offner G,et al.
Noninvasive markers of liverfibrosis are highly predictive of liver-related death in a cohort of HCV-infected individuals with and without HIV infection.Am J Gastroenterol2010;105:
1346–53.
10. Yu ML, Lin SM, Lee CM,et al.A simple noninvasive index for predicting long-term outcome of chronic hepatitis C after interferon-based therapy.
Hepatology2006;44: 1086–97.
11. Boursier J, De Ledinghen V, Zarski J, et al.A new combination of blood test and Fibroscan for accurate non-invasive diagnosis of liverfibrosis stages in chronic hepatitis C.Am J Gastroenterol 2011;106: 1255–63.
12. Castera L, Sebastiani G, Le Bail B, de Ledinghen V, Couzigou P, Alberti A.
Prospective comparison of two algorithms combining non-invasive methods for staging liverfibrosis in chronic hepatitis C.J Hepatol2010;52:
191–8.
13. Sebastiani G, Halfon P, Castera L,et al.
SAFE biopsy: a validated method for large-scale staging of liverfibrosis in
chronic hepatitis C.Hepatology2009;
49: 1821–7.
14. Cales P, de Ledinghen V, Halfon P, et al.Evaluating the accuracy and increasing the reliable diagnosis rate of blood tests for liverfibrosis in chronic hepatitis C.Liver Int2008;28: 1352–62.
15. D’Amico G, Garcia-Tsao G, Pagliaro L.
Natural history and prognostic indicators of survival in cirrhosis: a systematic review of 118 studies.J Hepatol2006;44: 217–31.
16. D’Amico G, Villanueva C, Burroughs AK,et al.Clinical stages of cirrhosis: a multicenter cohort study of 1858 patients.Hepatology2010;52: 329A.
17. Wai CT, Greenson JK, Fontana RJ,et al.
A simple noninvasive index can predict both significantfibrosis and cirrhosis in patients with chronic hepatitis C.
Hepatology2003;38: 518–26.
18. Sterling RK, Lissen E, Clumeck N,et al.
Development of a simple noninvasive index to predict significantfibrosis in patients with HIV/HCV coinfection.
Hepatology2006;43: 1317–25.
19. Adams LA, Bulsara M, Rossi E,et al.
Hepascore: an accurate validated predictor of liverfibrosis in chronic hepatitis C infection.Clin Chem2005;
51: 1867–73.
20. Castera L, Vergniol J, Foucher J,et al.
Prospective comparison of transient elastography, Fibrotest, APRI, and liver biopsy for the assessment offibrosis in chronic hepatitis C.Gastroenterology 2005;128: 343–50.
21. Leroy V, Halfon P, Bacq Y,et al.
Diagnostic accuracy, reproducibility and robustness offibrosis blood tests in chronic hepatitis C: a meta-analysis with individual data.Clin Biochem 2008;41: 1368–76.
22. Boursier J, Bacq Y, Halfon P,et al.
Improved diagnostic accuracy of blood tests for severe fibrosis and cirrhosis in chronic hepatitis C.Eur J
Gastroenterol Hepatol 2009;21: 28–38.
23. Boursier J, Bertrais S, Oberti F,et al.
Comparison of accuracy offibrosis degree classifications by liver biopsy and non invasive tests in chronic hepatitis C.
BMC Gastroenterol2011;30: 132.
24. Poynard T, Imbert-Bismut F, Munteanu M,et al.Overview of the diagnostic value of biochemical markers of liver fibrosis (FibroTest, HCV FibroSure) and necrosis (ActiTest) in patients with chronic hepatitis C.Comp Hepatol 2004;3: 8.
25. Harrell FE Jr, Lee KL, Mark DB.
Multivariable prognostic models: issues
in developing models, evaluating assumptions and adequacy, and measuring and reducing errors.Stat Med1996;15: 361–87.
26. Empana JP, Tafflet M, Escolano S,et al.
Predicting CHD risk in France: a pooled analysis of the D.E.S.I.R., Three City, PRIME, and SU.VI.MAX studies.
Eur J Cardiovasc Prev Rehabil2011;18:
175–85.
27. Vallet-Pichard A, Mallet V, Nalpas B, et al.FIB64: an inexpensive and accurate marker offibrosis in HCV infection.
Comparison with liver biopsy and Fibrotest.Hepatology2007;46: 32–6.
28. Cales P, Boursier J, Ducancelle A,et al.
Improvedfibrosis staging by elastometry and blood test in chronic hepatitis C.Liver Int2013; doi: 10.
1111/liv.12327 [Epub ahead of print].
29. Thein HH, Yi Q, Dore GJ, Krahn MD.
Estimation of stage-specificfibrosis progression rates in chronic hepatitis C virus infection: a meta-analysis and meta-regression.Hepatology2008;48:
418–31.
30. Everhart JE, Wright EC, Goodman ZD, et al.Prognostic value of Ishakfibrosis stage:findings from the hepatitis C antiviral long-term treatment against cirrhosis trial.Hepatology2010;51:
585–94.
31. Liu S, Schwarzinger M, Carrat F, Goldhaber-Fiebert JD. Cost effectiveness offibrosis assessment prior to
treatment for chronic hepatitis C patients.PLoS ONE2011;6: e26783.
32. Canavan C, Eisenburg J, Meng L, Corey K, Hur C. Ultrasound elastography for fibrosis surveillance is cost effective in patients with chronic hepatitis C virus in the UK.Dig Dis Sci2013;58: 2691–704.
33. Masuzaki R, Tateishi R, Yoshida H, et al.Prospective risk assessment for hepatocellular carcinoma development in patients with chronic hepatitis C by transient elastography.Hepatology2009;
49: 1954–61.
34. Merchante N, Riveri-Juarez A, Tellee F, et al.Liver stiffness predicts clinical outcome in human immunodeficiency virus/hepatitis C virus co-infected patients with compensated liver cirrhosis.Hepatology2012;56: 228–38.
35. Vergniol J, Foucher J, Terrebonne E, et al.Noninvasive tests forfibrosis and liver stiffness predict 5-year outcomes of patients with chronic hepatitis C.
Gastroenterology2011;140: 1970–9.
