A Prospective Analysis of the Prognostic Value of Biomarkers ( FibroTest ) in Patients with Chronic Hepatitis C

Background: FibroTest, a noninvasive method of measuring biomarkers of liver fibrosis, is an alternative to liver biopsy for determining the severity of chronic hepatitis C virus (HCV) infection. We compared the 5-year prognostic value of the FibroTest with biopsy staging for predicting cirrhosis decompensation and survival for patients with chronic HCV infection. Methods: Fibrosis stage was assessed on the same day by FibroTest and biopsy in a prospective cohort of 537 patients. Disease classification at baseline was 157 patients with severe fibrosis (FibroTest >0.58), 137 with moderate fibrosis (FibroTest 0.32–0.58), and 243 with no or minimal fibrosis (FibroTest <0.32). Results: In 64 untreated patients with severe fibrosis, survival without HCV complications was 73% [95% confidence interval (CI), 59%–086%; 13 complications], and survival without HCV-related death was 85% (95% CI, 73%–96%; 7 HCV deaths). Survival rates were higher in patients with moderate fibrosis, [99% (95% CI, 97%– 100%; 1 complication; P <0.001) and 100% (no HCV death; P <0.001) for with and without HCV-related complications, respectively], and in patients with minimal fibrosis [100% (no complication; P <0.001 vs severe) and 100% (no HCV death; P <0.001 vs severe), respectively]. FibroTest was a better predictor than biopsy staging for HCV complications, with area under the ROC curves (AUROC) 0.96 (95% CI, 0.93%–0.97%) vs 0.91 (95% CI, 0.85%–0.94%; P 0.01), respectively; it was also a better predictor for HCV deaths: AUROC 0.96 (95% CI, 0.93%–0.98%) vs 0.87 (95% CI, 0.70%–0.94%; P 0.046), respectively. The prognostic value of FibroTest was still significant (P <0.001) in multivariate analyses after taking into account histology, treatment, alcohol consumption, and HIV coinfection. Conclusion: The FibroTest measurement of HCV biomarkers has a 5-year prognostic value similar to that of liver biopsy. © 2006 American Association for Clinical Chemistry

Finding the best method to evaluate and manage patients infected with the hepatitis C virus (HCV) 6 continues to be a challenge (1,2 ).Liver biopsy for determining disease grade and stage has limitations (3)(4)(5) and risks (6 ); noninvasive alternatives to liver biopsy in patients infected with HCV (7 ) include 2 combinations of simple serum biochemical markers: FibroTest (FT) (Biopredictive) for the assessment of fibrosis, and ActiTest (AT) (Biopredictive) for the assessment of necroinflammatory activity (8 -16 ).With biopsy as the standard of reference, the diagnostic value of FT for a diagnosis of significant fibrosis (bridging fibrosis), as estimated by the area under the ROC curve (AUROC), is 0.73-0.86(8 ).AUROC values inside the range of other studies have been reported, but the usefulness of FT compared with biopsy was not

Patients and Methods patients
Study patients belonged to a prospective hospital-based cohort of 2865 patients with chronic HCV, seen at our institution from January 1997 to December 2002 (Fig. 1).Survival analysis included a core cohort of 537 patients who had been previously studied at baseline to assess the discordance rate between biopsy and FT (18 ).Inclusion criteria were 2 assessments of liver histology done Ͻ24 h apart, 1 of a liver biopsy specimen and the other with FT measured in serum.Patients had been referred by general practitioners, private specialists, or public general hospitals for the staging and treatment of HCV infection (18 ).
Most patients had no severe complications, and the disease had been discovered by the detection of HCV antibodies.Liver biopsy was indicated for all patients without a previous biopsy but was not performed in patients with contraindications or who refused.All patients received FT unless it was not available, the patient refused, or the hospital laboratory was unable to perform the FT the day of the biopsy.
Follow-up of core group patients was performed every 6 months, and biopsy or FT was repeated as considered necessary by the physician in charge.A group of 395 patients (validation group) who underwent a baseline FT either with no liver biopsy or with a Ͼ24-h lapse between liver biopsy and FT were analyzed separately to check the prognostic value of FT observed in the core group.
All procedures were followed in accordance with the current revision of the Helsinki declaration, and all participants gave informed consent.Biopsy was performed for routine management of chronic HCV infection and was not related to the study protocol.Consenting patients underwent FT testing if biochemistry unit personnel were available to perform the test and were blinded to the clinical data.The result of the FT was known after the results of the biopsy, and the clinical management at baseline was decided according to biopsy results before 2002.

liver biopsies
Liver biopsies were processed with standard techniques.A single pathologist (F.C.), who was unaware of the biochemical markers, evaluated the stage of fibrosis and grade of activity according to the METAVIR scoring system (22,23 ).Fibrosis was staged on a scale of 0 -4: F0, no fibrosis; F1, portal fibrosis without septa; F2, few septa; F3, numerous septa without cirrhosis; F4, cirrhosis.

biochemical markers
FT is a noninvasive blood test that combines the quantitative results of 6 serum biochemical markers, [␣2-macroglobulin, haptoglobin, ␥-glutamyl transpeptidase, total bilirubin, apolipoprotein A1, and alanine aminotransferase (ALT)] with patient age and sex data in a patented artificial intelligence algorithm (24 ) to generate a measure of fibrosis and necroinflammatory activity in the liver (8 -16 ).This method provides a numerical quantitative estimate of liver fibrosis ranging from 0.00 to 1.00, corresponding to the METAVIR scoring system, which grades fibrosis from F0 (no fibrosis) to F4 (cirrhosis).An algorithm has been suggested that would classify patients into 3 groups: no or minimal fibrosis (FT 0 -0.31), moderate fibrosis (FT 0.31-0.58),and severe fibrosis (FT 0.58 -1.00) (8,18 ).
We measured ␥-glutamyl transpeptidase, ALT, AST, and total bilirubin with a Hitachi 917 analyzer and Roche Diagnostics reagents, and we measured ␣2-macroglobulin, apolipoprotein A1, and haptoglobin with a Dade Behring Nephelometer II.All tests were performed by personnel blinded to all patient characteristics, including biopsy results.All analytical studies, including intra-and interobserver and reproducibility studies, were performed independently of the present study, with CVs Ͻ10%, and have been reported previously (25)(26)(27)(28)(29)(30).

survival analysis
The 5-year survival without HCV-related cirrhosis or complications related to liver disease [defined as: liverrelated death, liver transplantation, or one of the following: decompensation, variceal bleeding, or hepatocellular carcinoma (HCC)] was the a priori main end-point used to compare the prognostic value of FT with histologic staging of biopsy samples.Decompensation was determined by the presence of ascites or hepatic encephalopathy or jaundice (total bilirubin, Ͼ51 mol/L).Ascites was deemed to be present when ascitic fluid was confirmed by paracentesis and/or abdominal imaging.HCC was diagnosed by histologic examination of liver tissue obtained by liver biopsy, or at autopsy, or if 1 or more hepatic space-occupying lesions observed at ultrasonography or computed tomography were shown to have vascular patterns typical of HCC by angiography, dualphase spiral tomography, or magnetic resonance imaging.Variceal bleeding was diagnosed on the basis of endoscopic findings in patients presenting with upper gastrointestinal hemorrhage.
To validate that the classification of HCV patients into 3 classes according to FT values, as previously recommended for disease management (8,18 ), was actually associated with mortality and morbidity, we used as 2 secondary endpoints survival without death related to HCV and the overall survival whatever the cause of death.We retrospectively compared FT with other indexes: the classical Child-Pugh score (31 ), the APRI index (32 ), and the Forns index (33 ).
The survival time was calculated from the date of FT to the endpoint date.This interval was censored at the time of last follow-up.For decompensated patients at baseline, only complications occurring during follow-up were taken into account.When several complications occurred, the first one was taken into account.Each year, for patients who had not been seen at our hospital in the previous 12 months, we found out whether they were living and, if not, the date and the cause of death.For patients who were still alive, we either interviewed the patients or obtained information through their physicians.For deceased patients who died outside our hospital, we obtained information about the date and cause of death from their physicians or family.If we could not obtain information on the patient, we sent a letter to the city of their birth to find out whether they were still living and, if not, the date of death.

statistical analysis
We used the 2 test for qualitative comparisons, the Mann-Whitney test for quantitative comparisons (34 ), time-dependent Kaplan-Meier analysis for survival curves, and the log-rank test and the Cox proportional hazard model for multivariate analysis (35 ).We checked the assumption of proportional hazards by plotting the scaled Schoenfeld residuals (36 ).
We compared patient survival with the survival expected in the French population, matched for age, sex, and follow-up period.The survival curve of the French population was calculated on the basis of age, sex, and follow-up period and conditional probabilities of death from official published census tables (35 ).For each patient, beginning from the date of FT assessment, we used the Ederer II method to calculate a yearly predicted cumulative survival rate from a person of the same age and sex having a similar period of follow-up, and we used the Z test for comparisons between actual and predicted survivals (36 ) to compare the prognostic values of FT and fibrosis staging.FT data, as well as other score values, were entered as continuous variables.We calculated AUROCs with an empirical nonparametric method according to DeLong et al. (37 ) or the binormal method if the sample size of the endpoint was Ͻ30 (38 ), and we compared results with the method of Zhou et al. (39 ).We used 2-sided statistical tests for all analyses; a P value of Յ0.05 was considered substantial.We used Number Cruncher Statistical Systems 2003 software for all analyses (34 ).
Data from patients who had advanced cirrhosis at baseline was excluded from sensitivity analysis, which including only data from Pugh class A patients.

Results patients
Of the 2865 patients seen during the study period, 537 (19%) were included in the core group and 395 in the validation group (Table 1).Core group patients were younger, with lower body-mass index, more infections attributable to intravenous drug use, higher transaminases and ALT, more genotype 3, less fibrosis, and less

complications and survival
A total of 29 complications, including 2 liver transplantations, were observed during the 5-year follow-up, with a total of 20 deaths, including 9 deaths attributable to HCV (Table 2A).One patient treated with beta-blockers bled from esophageal varices but survived.

comparison with biopsy and other indexes
The ROC curves comparing the sensitivity and specificity of FT vs fibrosis at biopsy for complications, HCV-related death, and overall death are shown in Fig. 3. FT had a substantially higher prognostic value than biopsy for HCV complications and HCV-related deaths compared (Table 3).

Table 2. Death and survival analysis.
A. Causes of death and complications during the 5-year follow-up.The FT AUROC for survival without complications (0.96; 95% CI, 0.93%-0.97%)was also greater (all P Ͻ0.01) than the AUROCs of the other indexes (Pugh, APRI, and Forns; Table 3), and each of the 6 components of FT analyzed separately (platelets, prothrombin time, and albumin; see Table 1 in the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol52/issue10).FT had also better prognostic value than the ActiTest or histologic activity grade (data not shown).

variables associated with complications and survival (see table 2 in the online data supplement)
FT was the only variable significantly associated with both survival without HCV-related complications and overall survival.Fibrosis staging at biopsy was not associated with overall survival.Heavy alcohol consumption (Ͼ50 g/day) and absence of HCV treatment were associated with overall death.

test values on ft and biopsy for persons with complications and death
Nine patients died from HCV-related causes: all of them had been classified by FT at baseline as having cirrhosis; 7 of these patients had the same classification made by biopsy, but 2 patients had been classified as noncirrhotic by biopsy, one as stage F1 and the other F2.
Among the 20 patients who had nonlethal complications, 12 had been classified by both biopsy and FT as having cirrhosis at baseline, 2 were classified as stage F2 by biopsy and FT, 5 were classified as cirrhotic by FT but not by biopsy (2 F2 and 3 F3), and 1 was classified as F3 by FT and cirrhotic by biopsy.

sensitivity analyses
When the 10 patients with baseline severe cirrhosis were excluded, (5 of them died from HCV and 8 had complications), all the main results were the same (see Table 3 in the online Data Supplement).
occurred in the moderate group (n ϭ 98); and 22% complications (27 complications) and 17% HCV-related mortality (17 deaths) occurred in the severe group (n ϭ 168) (P Ͻ0.001 between the severe vs the 2 other groups for both complications and deaths).

Discussion
Until now, the prognostic markers validated for use in chronic HCV have been the histologic fibrosis staging of biopsy specimens, the Pugh score for patients with cirrhosis (19 -21, 40 ), and hyaluronic acid (41 ).
Our results indicate that FT has a better 5-year prognostic value than biopsy estimates, regardless of the treatment and risk factors.Even after exclusion of decompensated cirrhotic patients, the prognostic value of FT was at least similar to biopsy.In the multivariate model, biopsy was more predictive than FT for complications but not for overall survival.Because of the applicability, cost, and risks of biopsy, prognostic studies with FT and other noninvasive methods should be the first-line approach.
Although this prognostic study was not specifically designed to validate FT as a true surrogate endpoint of HCV chronic hepatitis severity (42 ), we observed that FT fulfilled almost all of the 13 criteria of a surrogate endpoint biomarker (43 ), including specificity and detection limit for fibrosis, with lϽ5% false positives or false negatives (8,9,18,26 ).FT was indicative of response to virologic treatment.The virologic treatment response that we observed, with FT improvement and cirrhosis reversal, has been previously observed (9,11 ); intra-and interobserver variability of FT has been studied; preanalytical and analytical recommendations have been issued (8,26,44 ); and security algorithms have been designed for detection of false positives or false negatives (18 ).Serial monitoring of FT is possible, as has been demonstrated in several studies (10,11,45 ).
In comparison, liver biopsy does not satisfy quality criteria as a surrogate endpoint marker because of its complication rate, sampling error, intra-and interobserver variability, expense, and patient reluctance to undergo serial monitoring.The results of our analysis of hard clinical endpoints during this 5-year follow-up demonstrate a higher failure rate of biopsy than FT (18 ).When a false negative was defined as a patient without baseline cirrhosis (FT or biopsy) who suffered HCV-related complication or death during the 5 years, there were 7 false negatives (true cirrhosis not detected at baseline) attributable to biopsy in this study and only 1 to FT.
We used a quantitative estimate for FT and a discrete estimate for biopsy, but a comparison of AUROCs by use of discrete FT scores vs histology (all in 5 classes from F0 to F4) for the main endpoint (HCV complications) was still significant in favor of FT: 0.96 (95% CI, 0.94%-0.97%)vs 0.91 (95% CI, 0.85%-0.95%),P ϭ 0.003, respectively.Validation studies of biomarkers vs biopsy as surrogate markers should be performed by independent groups.
Our study was not conclusive for the use of FT score to determine the need for treatment, but the results supported a previously suggested treatment algorithm based on FT values (18 ).Because patients with low FT scores were unlikely to develop complications, decisions not to treat such patients were unlikely to be associated with clinical decompensation, at least over a relatively short follow-up period.Patients with moderate FT scores were at low risk, with a 1% risk at 5 years.Although availability of FT test results could influence therapeutic decisions, during follow-up (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004), most patients were managed according to biopsy results, and therapeutic decisions were never related to the FT results from 1997 to 2002.Interestingly, for nontreated patients with severe FT Ͼ0.58, overall survival was significantly lower than controls (P ϭ 0.02), whereas for sustained virologic responders with FTϾ 0.58, overall survival was similar to that of controls.Thus, FT permitted identification of patients with a high mortality risk.Because the Pugh index was validated in patients with cirrhosis, a group that represented Ͻ20% of the included patients, the prognostic superiority of FT over Pugh was expected.The superiority of FT over the Forns and APRI indexes was also expected, because these indexes have lower values for the diagnosis of advanced fibrosis as assessed by AUROCs (13,46 ).
One weakness of the present study was that the included population was not a random, community-based population.Baseline comparisons between the included and nonincluded populations (18 ), however, have not identified obvious causes of bias: data were very similar for variables associated with a risk of bias in prognostic studies.
The patients of the core group were more likely to be responders to HCV treatment, with better response variables: younger, lower body mass index, less steatosis, and more genotype 3. Patients in the core group were more often new patients with an indication of first liver biopsy, which could explain more simultaneous biopsy and FT than in patients of the validation group.
Compared with the nonincluded population, the core group or the validation group likely included more patients who returned more frequently or had other features suggestive of more aggressive disease.The survival in the core group was higher than in a community-based retrospective cohort (47 ), however, and lower than in another prospective tertiary hospital cohort (40 ).These 2 studies differed from ours in design, including only untreated patients and looking at follow-up from the time of HCV detection.Although the rate of outcomes in the patients with severe FT represents nearly a 20% cumulative incidence, the low event rate is also a limitation of this study.Our core group, however, had the advantage of including treated and untreated patients with a wide range of fibrosis severity and risk factors, without the exclusion of heavy drinkers or HIV-coinfected patients.After we excluded decompensated patients, fewer events were included, but the main results were similar to those for the entire prospective cohort.Another advantage of this study was that it allowed the observation of similar prognostic values of FT in the validation population of 395 patients.
As previously reported (47 ), deaths unrelated to HCV, including 3 suicides, were as frequent as HCV-related death.The overall survival of patients with nonsevere fibrosis at baseline was close to that of paired controls in the general population.In patients with severe fibrosis, overall survival was 5% lower than that of the control population, a rate that should be confirmed with longer follow-up.
In conclusion, the FT has prognostic values at least similar to that of liver biopsy, with the advantage of being noninvasive.
This work was supported by research grants from the Association pour la Recherche sur le Cancer and from the Association pour la Recherche sur les Maladies He ´patiques Virales.The FibroTest patent belongs to public hospital organization Assistance Publique Hopitaux de Paris and is licensed by Biopredictive, a start-up company of Paris University.One of the authors (T.P.) is the inventor of FibroTest and is a consultant for and has a capital interest in Biopredictive.T.P. was responsible as a scientist for the design, execution, analysis, and reporting of this study and did not receive any specific honorarium for this study.Another author (M.M.) is a full-time employee of Biopredictive and participated in the analysis of the study.All her work was supervised and checked by T.P. and by Y.N. and V.R., who have no conflict of interest.

Table 1 .
(18 )cteristics of included patients.thanthevalidationgroup.Comparisons between the patients of core group and the 2328 nonincluded patients were previously published(18 ).Pugh class B. At baseline in the severe group, 12 patients had signs of portal hypertension: 3 with esophageal varices, 2 with gastropathy, and 8 with signs of portal hypertension by ultrasound.During follow-up, 8 patients were treated with beta-blockers for portal hypertension.The median follow-up of the cohort was 57 months (range, 1-93 months).Details of the number of patients at risk at each interval are given in Fig.1.

Table 3 .
Comparison of AUROC for survival endpoints determined from biomarkers and biopsy specimen histology