Diagnostic and prognostic values of noninvasive biomarkers of fibrosis in patients with alcoholic liver disease

Authors

  • Sylvie Naveau,

    1. Assistance Publique–Hôpitaux de Paris (AP-HP), Hôpital Antoine Béclère, Service d'Hépato-gastroenterologie, Clamart, France
    2. University Paris-Sud, Faculté de médecine Paris-Sud, Institut Fédératif de Recherche 13, Clamart, France
    3. INSERM, Unite 764, Clamart, France
    Search for more papers by this author
  • Guillaume Gaudé,

    1. Assistance Publique–Hôpitaux de Paris (AP-HP), Hôpital Antoine Béclère, Service d'Hépato-gastroenterologie, Clamart, France
    Search for more papers by this author
  • Amani Asnacios,

    1. Assistance Publique–Hôpitaux de Paris (AP-HP), Hôpital Antoine Béclère, Service d'Hépato-gastroenterologie, Clamart, France
    Search for more papers by this author
  • Hélène Agostini,

    1. AP-HP, Unité de recherche clinique Paris-Sud, Clamart, France
    Search for more papers by this author
  • Annie Abella,

    1. AP-HP, Hôpital Antoine Béclère, Service de biochimie, Clamart, France
    Search for more papers by this author
  • Nadège Barri-Ova,

    1. Assistance Publique–Hôpitaux de Paris (AP-HP), Hôpital Antoine Béclère, Service d'Hépato-gastroenterologie, Clamart, France
    2. University Paris-Sud, Faculté de médecine Paris-Sud, Institut Fédératif de Recherche 13, Clamart, France
    Search for more papers by this author
  • Barbara Dauvois,

    1. Assistance Publique–Hôpitaux de Paris (AP-HP), Hôpital Antoine Béclère, Service d'Hépato-gastroenterologie, Clamart, France
    2. University Paris-Sud, Faculté de médecine Paris-Sud, Institut Fédératif de Recherche 13, Clamart, France
    Search for more papers by this author
  • Sophie Prévot,

    1. AP-HP, Hôpital Antoine Béclère, Service d'Anatomie Pathologique, Clamart, France
    Search for more papers by this author
  • Yen Ngo,

    1. AP-HP Groupe Hospitalier Pitié-Salpêtrière, Service of Hepato-Gastroenterology, Paris, France
    Search for more papers by this author
  • Mona Munteanu,

    1. Biopredictive, Paris, France
    Search for more papers by this author
    • Mona Munteanu is a full-time employee of Biopredictive.

  • Axel Balian,

    1. Assistance Publique–Hôpitaux de Paris (AP-HP), Hôpital Antoine Béclère, Service d'Hépato-gastroenterologie, Clamart, France
    Search for more papers by this author
  • Micheline Njiké-Nakseu,

    1. Assistance Publique–Hôpitaux de Paris (AP-HP), Hôpital Antoine Béclère, Service d'Hépato-gastroenterologie, Clamart, France
    Search for more papers by this author
  • Gabriel Perlemuter,

    1. Assistance Publique–Hôpitaux de Paris (AP-HP), Hôpital Antoine Béclère, Service d'Hépato-gastroenterologie, Clamart, France
    2. University Paris-Sud, Faculté de médecine Paris-Sud, Institut Fédératif de Recherche 13, Clamart, France
    3. INSERM, Unite 764, Clamart, France
    Search for more papers by this author
  • Thierry Poynard

    Corresponding author
    1. AP-HP Groupe Hospitalier Pitié-Salpêtrière, Service of Hepato-Gastroenterology, Paris, France
    • APHP Groupe Hospitalier Pitié-Salpêtrière, Service of Hepato-Gastroenterology, Groupe Hospitalier Pitie-Salpetriere, 47-83 Boulevard de l'Hopital, 75651 Paris Cedex 13, France
    Search for more papers by this author
    • fax: (33)14-2161427


  • Potential conflict of interest: Thierry Poynard is the inventor of the FibroTest with a capital interest in Biopredictive the company marketing FibroTest, the royalties belonging to the French public organization Assistance Publique Hôpitaux de Paris.

Abstract

FibroTest has been validated as a biomarker of fibrosis in patients with chronic viral hepatitis, with a similar prognostic value as biopsy. The aim of the study was to compare the diagnostic and prognostic values of FibroTest versus the recently patented biomarkers, FibrometerA, and Hepascore. A total of 218 consecutive patients with ALD and available liver biopsy examination were included. Biomarkers were compared using univariate area under the ROC curves (AUROC) and multivariate analysis (logistic regression and Cox). The median follow-up was 8.2 years. Eighty-five patients died, including 42 deaths related to liver complications. The diagnostic values of FibrometerA and Hepascore did not differ from that of FibroTest for advanced fibrosis (all AUROC = 0.83 ± 0.03) and cirrhosis (FibroTest and FibrometerA = 0.94 ± 0.02, Hepascore = 0.92 ± 0.02), and were significantly greater than those of nonpatented biomarkers (APRI, Forns, FIB4; P < 0.01). In multivariate analysis the most significant was FibroTest (P = 0.001), without independent diagnostic value for FibrometerA (P = 0.19), and Hepascore (P = 0.40). The prognostic values of FibroTest (AUROC for survival or non liver disease-related death = 0.79 ± 0.04), FibrometerA (0.80 ± 0.04), Hepascore (0.78 ± 0.04), did not differ from that of biopsy fibrosis staging (0.77 ± 0.04). In multivariate analysis the most significant were FibroTest (P = 0.004) and biopsy (P = 0.03), without independent prognostic values for FibrometerA (P = 0.41) and Hepascore (P = 0.28). In patients with alcoholic liver disease, FibrometerA and Hepascore did not improve the diagnostic and prognostic values of FibroTest. (HEPATOLOGY 2009;49:97-105.)

Progressive hepatic fibrosis with the development of cirrhosis is a feature in the majority of chronic liver disease cases. Liver biopsy is still recommended for most liver disease. It is an invasive procedure, however, which results in severe complications in about 0.5% of cases.1 In addition, sampling error is common and the biopsy specimen seems to be poorly reliable when its length is inferior to 25 mm.2 Moreover liver fibrosis is evaluated by histologic scores, which have interobserver variability.3 The assessment of liver fibrosis by noninvasive means is a major challenge that has stimulated the search for new approaches. A number of noninvasive tests have been commercialized and are being used increasingly in practice, especially in patients with chronic hepatitis C.

Little clinical data are available in patients with alcoholic liver disease (ALD). One of the first tests that we described in patients with alcoholic liver disease was the “PGA index,” which combines the prothrombin index (PI), gamma glutamyl transpeptidase (GGT) and apolipoprotein A1.4 We later modified it to include α-2 macroglobulin (the PGAA index).5 A second generation panel is the “FibroTest” that combines α-2 macroglobulin, haptoglobin, GGT, apolipoprotein A1, and total bilirubin corrected for age and gender, and which has shown high predictive values for significant fibrosis in patients with chronic hepatitis C, B, nonfatty liver disease,6, 7 and in patients with chronic ALD.8, 9 Hepascore combines bilirubin, GGT, hyaluronic acid, α-2 macroglobulin, age, and gender. It provides useful information regarding different fibrosis stages among hepatitis C patients10 and also for discriminating advanced fibrosis (F ≥ 2 according to the METAVIR scoring system) in patients with ALD.11 FibrometerA combines PI, α-2 macroglobulin, hyaluronic acid, and age with similar accuracy in patients with ALD.11, 12

The first aim of the study was to estimate if two newly patented biomarkers, FibrometerA and Hepascore, add diagnostic value in comparison or in combination with FibroTest.

The second aim was to compare the prognostic values of these scores for 10-year mortality with prognostic values of histologic features of liver biopsy. To be accepted as alternatives to liver biopsy, noninvasive biomarkers must show prognostic value based on hard clinical endpoints, such as liver-related mortality. Only FibroTest has shown a prognostic value similar to biopsy fibrosis stage in patients with chronic hepatitis C and B.13, 14

Abbreviations

AAH, acute alcoholic hepatitis; ALT, alanine aminotransferase; ALD, alcoholic liver disease; AST, aspartate aminotransferase; AUROC, area under the ROC curves; GGT, gamma glutamyl transpeptidase; PI, prothrombin index.

Patients and Methods

Study Population.

A total of 218 consecutive patients with heavy alcohol consumption and available liver biopsy examination and FibroTest results were retrospectively included for this analysis; however, they were initially prospectively included in a cohort of ALD for which one primary end-point was the identification of biomarkers.4, 5, 15, 16 The study group came from the original population (292 patients) that served for FibroTest validation in ALD.8 To be considered for inclusion, patients had to have consumed at least 50 g of alcohol per day over the previous year. Exclusion criteria included the presence of concomitant liver diseases, hepatitis surface antigen, antibodies to hepatitis C virus, HIV antibodies and immunosuppression, severe associated disease, unavailable serum, and unavailable biopsies. Patients also were excluded if the biopsy sample and serum were collected more than 1 month apart. Characteristics of included and nonincluded patients are given in Table 1. All patients gave informed consent for the use of data and serum for research purposes and agreed to participate in this noninterventional clinical research, which was approved by the local institutional review board, and was in accordance with the ethical guidelines of the Declaration of Helsinki.

Table 1. Characteristics of Included Patients with Alcoholic Liver Disease
CharacteristicsIncluded (n = 218)Nonincluded (n = 74)
  1. Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; ENT, ears, nose, and throat; GGT, gamma glutamyl transpeptidase; NA, not available.

Age at biopsy in years, mean ± SE47 ± 0.749 ± 0.7
Male, n (%)170 (78)56 (78)
Biopsy  
 Fibrosis stage, n (%)  
  No fibrosis (F0)16/218 (7)5/58 (9)
  Nonseptal fibrosis (F1)65/218 (30)17/58 (29)
  Few septa (F2)48/218 (22)7/58 (12)
  Numerous septa (F3)22/218 (10)5/58 (9)
  Cirrhosis (F4)67/218 (31)24/58 (41)
 Necroinflammatory activity grade, n (%)  
  No acute alcoholic hepatitis155/218 (71)43/58 (74)
  Acute alcoholic hepatitis63/218 (29)15/58 (8)
Biopsy quality, mean ± SE  
 Sample size (mm)15 ± 0515 ± 08
 Number of fragments2.2 ± 0.12.3 ± 0.3
 Number of portal tracts14.4 ± 0.713.8 ± 1.5
Markers, mean ± SE  
 α2-macroglobulin (g/L)2.00 ± 0.051.93 ± 0.11
 ALT (IU)65 ± 567 ± 6
 AST (IU)95 ± 8120 ± 22
 Apolipoprotein A1 (g/L)1.40 ± 0.041.22 ± 0.11
 GGT (IU/l)319 ± 24449 ± 115
 Haptoglobin (g/l)1.35 ± 0.051.21 ± 0.12
 Total bilirubin (μmol/L)42 ± 532 ± 7
 Hyaluronic acid (μg/L)306 ± 28260 ± 58
 FibroTest (.00-1.00)0.51 ± 0.020.55 ± 0.07
 FibrometerA (0.00-1.00)0.59 ± 0.03NA
 Hepascore (0.00-1.00)0.67 ± 0.02NA
 FIB4 (0.00-1.00)0.66 ± 0.07NA
 Forns2.04 ± 0.07NA
 APRI1.84 ± 0.18NA
Deaths during follow-up  
 Liver-related deaths42NA
  Hemorrhage14NA
  Hepatocellular carcinoma7NA
  Decompensation21NA
Nonliver-related deaths43NA
  ENT carcinoma8NA
  Lung carcinoma4NA
  Other carcinoma4NA
  Myocardial infarction9NA
  Respiratory failure4NA
  Other6NA
  Unknown8NA
  Alcohol consumption during follow-up (%)  
   Abstinent46 (21)NA
   Not abstinent108 (50)NA
   Unknown64 (29)NA

Histological Analysis.

The biopsy samples were processed routinely and stained with hematoxylin-eosin-Safran, Masson's trichrome, and picrosirius red stains. A single pathologist, unaware of patient characteristics, analyzed the histologic features. A scoring system adapted from the METAVIR scoring system was used.17–19 According to this system, fibrosis was staged on a scale from F0 to F4; F0 = no fibrosis, F1 = fibrosis without septa, F2 = few septa, F3 = numerous septa without cirrhosis, and F4 = cirrhosis. Perivenular fibrosis was considered as stage F1 when septa were not present. Necroinflammatory activity was graded on a 2-grade scale according to the presence or absence of at least two alcoholic hepatitis features (necrosis, polynuclear infiltrate, Mallory bodies): A0 = absence and A1 = presence of acute alcoholic hepatitis (AAH) features. We validated the reproducibility of this classification in another study including 295 cases of ALD.19

Serum Fibrosis Markers.

The following parameters were assessed prospectively on fresh serum: alanine aminotransferase (ALT), aspartate aminotransferase (AST), GGT, bilirubin, PI (%), α-2 macroglobulin, and apolipoprotein A1. Haptoglobin and hyaluronic acid were determined from serum stored at −80°C. Hyaluronic acid was assessed using a radiometric assay (Pharmacia HA test, Pharmacia Diagnostic, Uppsala, Sweden). FibroTest (Biopredictive, Paris, France; FibroSURE, LabCorp, Burlington, NC) was carried out using published recommended pre-analytical and analytical procedures.20, 21 Formulas for calculating patented biomarkers were taken from USPTO patents (FibroTest issued patent 7,225,080, FibrometerA application number 20070178443, Hepascore application number 20070225919).

We retrospectively assessed three nonpatented scores: Forns,22 APRI,23 and FIB4,24 using formulas of original publications, as well as the Pugh prognostic score.25

Survival Analysis.

The 5-year and 10-year survivals or non liver-related death were the main endpoints used to compare the prognostic value of biomarkers with histologic staging of biopsy samples. The survival was calculated from the date that the biomarker was carried out to the endpoint date. This interval was censored at the time of last follow-up. For patients who had not been seen at our hospital in the previous 6 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 the 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

Diagnostic analysis.

The first outcome was advanced fibrosis, defined as stage F ≥ 2 or higher. The second outcome was the presence of cirrhosis (F = 4). Spearman's two-tailed test was used to assess correlations. Receiver operating characteristic (ROC) curves were constructed. Sensitivity, specificity, and positive and negative predictive values (PPV and NPV) were calculated using cut-offs described previously for each score. The overall diagnostic performances of scores were evaluated by the area under the ROC curves (AUROC) calculated with an empirical nonparametric method according to De Long et al.26 Direct comparisons used the Zhou et al. method.27

Indirect comparisons between sensitivities, specificities, AUROC and predictive values are hazardous due to variability factors, the two most important being the prevalence of stages defining advanced and nonadvanced fibrosis and the biopsy length.28, 29

Indirect comparisons used two AUROC corrections as described by Poynard et al.27, 28 One correction included the difference of mean stages observed between advanced fibrosis and nonadvanced fibrosis (DANA).28 The second correction included biopsy length. The AUROC is expressed relative to the best possible AUROC achieved by liver biopsy of a given length versus the true gold standard (almost entire liver). For a 15-mm length biopsy, the best AUROC for the diagnosis of advanced fibrosis is 0.92. Therefore, an observed AUROC for a biomarker equal to 0.86 will be transformed to an adjusted AUROC of 0.86/0.92 = 0.93.29

FibroTest, FibrometerA, and Hepascore were entered as continuous variables. For all analyses, two-sided statistical tests were used; a P value of 0.05 or less was considered significant. Multiple logistic regression analysis was carried out to test the statistical independence diagnostic value of these scores.

Prognostic Value.

We used Kaplan-Meier analysis for survival curves estimates, the log-rank test for comparisons, prognostic AUROC for univariate biomarkers and biopsy comparisons, and the Cox proportional hazard model for multivariate comparisons. Cumulative survival was expressed with standard error (SE) and their 95% confidence interval (CI).

Patient survival was compared with the survival expected in the French population, matched for age, gender, and follow-up period. The survival curve of the French population was calculated based on age, gender, and follow-up period and conditional probabilities of death from official, published census tables.30 For each patient, a yearly predicted cumulative survival rate was calculated from a subject of the same age and gender having a similar period of follow-up, beginning from the date of biomarker assessment, a method known as the Ederer II method, and the comparisons between actual and predicted survivals used the Z test.31

We used Number Cruncher Statistical Systems 2007 Software for all analyses.32

Results

Patient Characteristics.

The main features of patients are summarized in Table 1. The majority (78%) of patients were male and cirrhosis was present at biopsy in 31%.

Diagnostic Analysis

FibroTest versus FibrometerA and Hepascore.

Figure 1 shows the box-plot of fibrosis scores according to fibrosis stage. Significant correlations were found between fibrosis stages versus FibroTest (r = 0.71) (Fig. 1A), FibrometerA (r = 0.72) (Fig. 1B), and Hepascore (r = 0.71) (Fig. 1C). Correlation with fibrosis persisted after adjustment for the presence of acute alcoholic hepatitis (data not shown).

Figure 1.

Notched box plots showing the relationship between the stage of fibrosis and (A) FibroTest, (B) FibrometerA, and (C) Hepascore. The horizontal line inside each box represents the median and the width of each box the median ± 1.57 interquartile range/√n to assess the 95% level of significance between group medians. Failure of the shaded boxes to overlap signifies statistical significance (P < 0.05). The horizontal lines above and below each box encompass the interquartile range (from 25th to 75th percentile), and the vertical lines from the ends of the box encompass the adjacent values (upper: 75th percentile + 1.5 times interquartile range, lower 25th percentile − 1.5 times interquartile range).

For discriminating between advanced versus nonadvanced fibrosis, the FibroTest, FibrometerA, and Hepascore AUROC were similar at 0.83 ± 0.03 (Fig. 2A). For discriminating between cirrhosis versus non-cirrhosis stages, the FibroTest and the FibrometerA AUROC = 0.94 ± 0.02 and the Hepascore AUROC = 0.92 ± 0.02 (Fig. 2B). There were no significant differences between the AUROC of these three biomarkers (Table 2).

Figure 2.

ROC curves of FibroTest, FibrometerA, and Hepascore for the detection of (A) advanced (moderate to severe) fibrosis, and (B) cirrhosis. The diagonal line represents that achieved by chance alone (AUROC = 0.50); the ideal AUROC is 1.00. (A) For discriminating between advanced versus nonadvanced fibrosis, the FibroTest, FibrometerA, and Hepascore AUROC were similar = 0.83 ± 0.03. (B) For discriminating between cirrhosis versus non-cirrhosis stages, the FibroTest and the FibrometerA AUROC = 0.94 ± 0.02 and the Hepascore AUROC = 0.92 ± 0.02. There were no significant differences between the AUROC of these three biomarkers.

Table 2. Accuracy of Biomarkers for the Diagnosis of Advanced Fibrosis and Cirrhosis in Patients with Alcoholic Liver Disease
ScoresAdvanced Fibrosis F0F1 versus F2F3F4Cirrhosis F0F1F2F3 versus F4
AUROC95% CIAUROC95% CI
  • Abbreviations: AUROC, area under the ROC curves; CI, confidence interval.

  • *

    FibroTest AUROC greater than that of Forns (P < 0.0001), APRI (P < 0.0001), and FIB4 (P = 0.0007) scores.

  • Forns score had a negative diagnostic value, explained by the decrease of GGT in cirrhotic patients and the absence of significant differences according to fibrosis stage of platelets.

Patented biomarkers (n = 218)    
 FibroTest*0.830.77–0.880.940.90–0.96
 FibrometerA0.830.77–0.870.940.90–0.97
 Hepascore0.830.77–0.880.920.87–0.97
Nonpatented scores    
 Forns0.380.30–0.460.380.27–0.47
 APRI0.590.51–0.670.670.59–0.75
 FIB40.700.62–0.760.800.72–0.86

In logistic regression including the three scores, FibroTest was the only score independently associated with advanced fibrosis (P < 0.001). Both FibroTest (P < 0.0001) and FibrometerA (P < 0.01) were associated independently with cirrhosis (Table 3).

Table 3. Multivariate Analysis of Biomarkers Accuracy for the Diagnosis of Advanced Fibrosis or Cirrhosis in Patients with Alcoholic Liver Disease
MarkersAdvanced Fibrosis ORCirrhosis 95% CIPOR95% ClP
  1. Abbreviations: CI, confidence interval; OR, odds ratio.

FibroTest20.04.0–100.60.000362545.0–8700<0.0001
Fibrometer A3.40.9–12.80.0777.44.6–13050.003
Hepascore2.30.5–11.80.310.310.01–6.70.46

For discriminating advanced fibrosis the AUROC of a score combining FibroTest and FibrometerA was 0.85, not significantly different than the AUROC of FibroTest (0.83) or FibrometerA alone (0.83). For discriminating cirrhosis, the AUROC of scores combining FibroTest and FibrometerA and the AUROC of scores combining FibroTest and PI were identical (0.95). They were not significantly different from the AUROC of FibroTest and FibrometerA alone. The comparisons of the five AUROC for discriminating advanced fibrosis and cirrhosis were not significantly different (Table 2).

Although AUROC were not significantly different, the simultaneous use of two scores improved positive and negative predictive values (Supporting Table 1). In particular, the concomitant presence of FibroTest ≥0.70 and PI <70% improved the PPV for significant fibrosis to 100% and for cirrhosis to 89%. However, this combination decreased the number of selected patients more than the other combinations (21%). Only one combination reached 100% Sp and therefore 100% PPV: FibroTest > = 0.70 and PI < 70%. Only FibroTest (alone) reached 100% Se and therefore 100% NPV: FibroTest <0.30.

The proportion of patients classified using the cutoffs for sensitivity 90% (Se90) and specificity 90% (Sp90) have been assessed. For FibroTest and advanced fibrosis, the Se90 = 0.23 concerned 27% (59/218) and Sp90 = 0.64 concerned 39% (85/218), that is 66% of patients classified; for cirrhosis Se90 = 0.56 concerned 57% (124/218) and Sp90 = 0.78 concerned 31% (67/218), that is 88% classified. For FibroMeterA and advanced fibrosis, Se90 = 0.11 concerned 21% (45/218) and Sp90 = 0.95 concerned 41% (89/218), that is 62% classified; for cirrhosis Se90 = 0.92 concerned 57% (125/218) and Sp90 = 0.997 concerned 33% (72/218), that is 90% classified. For Hepascore and advanced fibrosis, Se90 = 0.25 concerned 20% (44/218) and Sp90 = 0.94 concerned 40% (88/218), that is 60% classified; for cirrhosis, Se90 = 0.97 concerned 63% (138/218) and Sp90 = 0.99 concerned 31% (68/218), that is 94% classified.

Other Scores.

The diagnostic values of FibroTest, FibrometerA and Hepascore were significantly greater than those of nonpatented scores, APRI (P < 0.0001), Forns (P < 0.0001), and FIB4 (P = 0.0007) both for the diagnosis of advanced fibrosis or cirrhosis (Table 2).

Prognostic Analysis.

The median follow-up of the 218 included patients was 8.2 years (range = 5 days to 11.8 years). Four patients were lost to follow-up. Eighty-five patients died. Forty-two deaths were attributable to liver disease. The details of the causes of death are given in Table 1. During the 5-year follow-up, only two patients were lost to follow-up and the cause of death was unknown in three cases. At 5- and 10-years, the cumulative global survival, regardless of the cause of death, were respectively 79 ± 3% (95% CI = 73-84) and 62 ± 4% (95% CI = 55-70) (Fig. 3A,B for FibroTest, and Supporting Table 2 and Supporting Figure S3B for others). These global survival rates were significantly lower than those of matched controls (Tables 4,5).

Figure 3.

(A) Survival curves or non–liver related death according to the three categories of baseline FibroTest. The upper curve corresponds to the FibroTest value <0.32 (n = 81; no or minimal fibrosis, 92.0%; 95% CI, 84.9-99.0), the intermediate curve to FibroTest value >0.32 and <0.58 (n = 43; moderate fibrosis, 87.5%; 95% CI, 75.5-99.5) and the lower curve to FibroTest >0.58 (n = 94; severe fibrosis, 62.6%; 95% CI, 52.2-73.1; P < 0.0001 versus the two other survivals). (B) Overall survival curves according to the three categories of baseline FibroTest. The upper curve corresponds to the FibroTest value <0.32 (n = 81; no or minimal fibrosis, 71.4%; 95% CI, 60.7-82.2), the intermediate curve to FibroTest value >0.32 and <0.58 (n = 43; moderate fibrosis, 69.8%; 95% CI, 55.2-84.4) and the lower curve to FibroTest >0.58 (n = 94; severe fibrosis, 42.4%; 95% CI, 31.1-53.6; P < 0.0001 versus the two other survivals).

Table 4. Five-Year Survival According to Baseline FibroTest Values in Patients with Alcoholic Liver Disease
Baseline ValuesnLiver-Related DeathSurvival or Nonliver-Related DeathDeathOverall SurvivalOverall Survival in Paired Controls
  • *

    Survival of the severe fibrosis group was significantly lower than the two other groups (P < 0.0001).

  • Overall survival of patients with alcoholic liver disease was significantly lower than that of paired controls in patients with no or minimal fibrosis (P < 0.05), moderate fibrosis (P < 0.01), and severe fibrosis (P < 0.0001).

FibroTest value
 (0.00–0.31) No or minimal fibrosis81198.7 (96.0–100)988.9 (82.0–95.7)98.4 (97.8–98.9)
 (0.32–0.58) Moderate fibrosis43392.1 (83.5–100)783.4 (72.1–94.6)98.3 (97.8–98.8)
 (0.59–1.00) Severe fibrosis*942868.3 (58.5–78.0)3958.4 (48.4–68.4)97.1 (96.4–97.8)
All2183284.5 (79.5–89.4)5574.7 (68.9–80.5)97.8 (97.4–98.2)
Table 5. Ten-Year Survival According to Baseline FibroTest in Patients with Alcoholic Liver Disease
Baseline ValuesnLiver-Related DeathSurvival or Nonliver-Related DeathDeathOverall SurvivalOverall Survival in Paired Controls
  • *

    Survival of the severe fibrosis group was significantly lower than the two other groups (P < 0.0001).

  • Overall survival of patients with alcoholic liver disease was significantly lower than that of paired controls, in patients with no or minimal fibrosis, moderate fibrosis, and severe fibrosis (all P < 0.0001).

FibroTest value
 (0.00–0.31) No or minimal fibrosis81592.0 (84.9–99.0)2171.4 (60.7–82.2)97.7 (97.0–98.4)
 (0.32–0.58) Moderate fibrosis43487.5 (75.5–99.5)1269.8 (55.2–84.4)97.8 (97.1–98.5)
 (0.59–1.00) Severe fibrosis*943262.6 (52.2–73.1)5042.4 (31.1–53.6)96.4 (95.5–97.3)
All2184178.5 (72.4–84.6)8358.8 (51.7–65.9)97.2 (96.7–97.6)

At 5 and 10 years, the cumulative survival or non–liver related death was, respectively, 84 ± 3% (95% CI = 79-88) and 78 ± 3% (95% CI = 72-84). The survival outcomes of patients classified according to FibroTest are detailed in Figure 3.

The ROC curves comparing the sensitivity and specificity of biomarkers and biopsy for overall death and for liver-related deaths are shown in Fig. 4A and B, respectively. Prognostic values of FibroTest, FibrometerA, and Hepascore were not significantly different from biopsy fibrosis staging for either overall survival or survival or non liver-related death (Table 6). Patented biomarkers had significantly higher prognostic AUROC than Pugh prognostic scores or nonpatented markers (Table 6). FibroTest (P < 0.05) and fibrosis staging at biopsy (P < 0.02) were associated with overall survival in a multivariate proportional hazard model (Table 7). FibrometerA and Hepascore did not have independent prognostic values. FibroTest was strongly associated with survival or non–liver related death (P = 0.004), followed by fibrosis staging at biopsy (P = 0.03). FibrometerA and Hepascore no longer had more independent prognostic values. If we considered all unknown causes of death as being either attributable to liver disease or not, the results did not differ (data not shown).

Figure 4.

Discordances observed between baseline Fibrotest, Fibrometer, and biopsy in 42 patients who died from a liver complication.

Table 6. Accuracy of Biomarkers and Biopsy for the Prognosis of Patients with Alcoholic Liver Disease
MarkersSurvival or Nonliver-Related DeathOverall Death
AUROC95% CIAUROCs95% CI
  • Abbreviations: AUROC, area under the ROC curves; CI, confidence interval.

  • *

    FibroTest has higher AUROCs than Pugh (P = 0.02; P = 0.02), FIB4 (P = 0.004; P = 0.20), APRI (P = 0.005; P = 0.005), and Forns (P < 0.0001; P < 0.0001) for liver-related death and overall survival respectively.

  • FibrometerA has higher AUROCs than Pugh (P = 0.003; P = 0.009), FIB4 (P = 0.003; P = 0.17), APRI (P = 0.001; P = 0.004), and Forns (P < 0.0001; P < 0.0001).

  • Hepascore has higher AUROCs than Pugh (P = 0.01; P = 0.006), FIB4 (P = 0.008; P = 0.16), APRI (P = 0.006; P = 0.006), and Forns (P < 0.0001; P < 0.0001).

FibroTest*0.790.68–0.860.690.61–0.76
FibrometerA0.800.71–0.870.690.61–0.76
Hepascore0.770.69–0.850.690.62–0.76
Fibrosis staging at biopsy0.770.70–0.830.690.61–0.76
Pugh0.690.58–0.770.620.54–0.68
FIB40.650.54–0.740.640.55–0.71
APRI0.600.50–0690.560.48–0.64
Forns0.400.30–0.490.430.35–0.51
Table 7. Multivariate Analysis of Biomarkers and Biopsy Accuracy for the Prognosis of Patients with Alcoholic Liver Disease
MarkersLiver-Related DeathOverall Death
RR95% CIPRR95% CIP
  • Abbreviations: CI, confidence interval; RR, risk ratio.

  • *

    Risk ratio for unknown abstinence versus nonabstinent was also significant = 0.24 (P < 0.001).

  • Risk ratio for unknown abstinence versus nonabstinent was also significant = 0.06 (P < 0.001).

FibroTest23.23.2–167.30.0023.71.2–11.70.02
Fibrosis staging at biopsy1.50.98–2.30.061.41.05–1.70.02
FibrometerA2.00.3–12.10.431.30.5–3.50.65
Hepascore0.30.04–2.80.340.90.2–3.10.81
Abstinent (versus nonabstinent)0.2*0.08–0.50.0010.40.22–0.710.002

The discordances between biomarkers and biopsy for persons with liver-related death are described in Figure 4. There were two false positives for FibroTest, and one False positive and one false negative for Fibrometer and for biopsy.

Acute Alcoholic Hepatitis.

The diagnostic value of AshTest was AUROC = 0.75 (95% CI = 0.67-0.82).8, 9 The 1-, 5-, and 10-year prognostic values of alcoholic hepatitis were significant in univariate analysis when present at biopsy or presumed with AshTest but not in multivariate analysis when the presence of cirrhosis at biopsy or presumed with FibroTest was entered into the model (Supporting Table 4).

Discussion

The first main finding of our study is the absence of significant differences between the diagnostic values of FibroTest, FibrometerA, and Hepascore in patients with ALD, both for the diagnosis of advanced fibrosis and for the diagnosis of cirrhosis. FibroTest had the best independent diagnostic value in multivariate analysis, and adding FibrometerA or Hepascore to FibroTest did not significantly improve the AUROC. Combinations marginally improved the predictive values of these tests alone.

The second main finding was that biomarkers had similar prognostic values as biopsy, the classical gold standard, at 5-year and 10-year follow-ups. FibroTest had the best prognostic value in multivariate analysis, without clinically significant improvement when combined with the other biomarkers.

Diagnostic Values.

Our results were similar to previous direct comparisons in patients with hepatitis C, in which no significant differences between these patented biomarkers were found.33, 34

Indirect comparisons between sensitivities, specificities, AUROC and predictive values are hazardous due to variability factors, the two most important being the prevalences of stages defining advanced and nonadvanced fibrosis and the biopsy length (28, 29).

The prevalence of each fibrosis stage according to the METAVIR scoring system was different between our study and Calés' study11: F0 = 7%, F1 = 30%, F2 = 22%, F3 = 10%, F4 = 31%; and F0 = 13%, F1 = 18%, F2 = 17%, F3 = 12%, F4 = 41%, respectively. The difference between the mean fibrosis values of advanced fibrosis minus the fibrosis stage of nonadvanced fibrosis (DANA) was 2.33 in the present study and 2.76 in the study by Calés.12 Poynard et al. recently showed the major impact of the DANA in the variability of the observed AUROC.28 The observed AUROC of FibroTest was 0.83 in the present study, equivalent to 0.86 after standardization using a DANA of 2.5 (same prevalence for all fibrosis stages), and 0.96 before and 0.93 after standardization in they study by Calés. When standardized, the difference between AUROC was 0.07 not 0.13.

One advantage of liver biopsy versus specific markers of fibrosis is the ability to diagnose alcoholic hepatitis, enabling patients to receive corticosteroid therapy. One advantage of FibroTest versus the other markers is that it has been validated together with AshTest, a patented specific biomarker of severe alcoholic hepatitis. In the present study presumed alcoholic hepatitis using AshTest had the same significant prognostic value as biopsy. This prognostic value was not found in multivariate analysis, but this was expected as all patients had been treated with corticosteroids, which had a significant impact on survival.33

When we compared the diagnostic value of the three patented biomarkers in the same patients, their diagnostic performances were quite similar. However, in logistic regression including the three scores, FibroTest was the only score to be associated independently with significant fibrosis. The absence of statistical independence of Hepascore can be explained by the fact that this score shares several parameters with FibroTest and FibrometerA.

The diagnostic value of FibroTest and FibrometerA for the diagnosis of alcoholic cirrhosis was almost perfect, with AUROC reaching 0.94.

It is not surprising, therefore, that diagnostic performances of scores combining biomarkers were not better than FibroTest used alone. Because liver biopsy is an imperfect gold standard, it is very difficult to speculate about predictive values, as it is possible that false negatives/positives of biomarkers could actually be false negatives/positives of liver biopsy. The best combination was FibroTest ≥0.70 and PI <70%. If the biopsy was always accurate, the diagnosis of significant fibrosis could be made with 100% certainty and the diagnosis of cirrhosis with 89% certainty when FibroTest was ≥0.70 and PI <70%.

Combinations did not improve negative predictive values, and FibroTest used alone had the best negative predictive value. Again, if biopsy was always accurate, cirrhosis could be ruled out with 100% certainty when FibroTest was lower than 0.30. This situation was observed in 35% of included patients.

FibroTest, FibrometerA, and Hepascore had significantly higher diagnostic values than the nonpatented scores, APRI, Forns, and Fib4. Greater accuracy of FibroTest compared to APRI and Forns has already been observed in patients with chronic hepatitis C and mixed liver diseases.36–38 In patients with ALD, AST/ALT ratio were different than in HCV patients, platelets can decrease according to alcohol consumption whatever the fibrosis stage, and GGT values decrease in patients with severe cirrhosis.39 It is important that these validated patented tests are covered by healthcare insurance programs.

Prognostic Analysis.

This study has shown that FibroTest, when used in patients with chronic ALD, can predict liver-related mortality at least as well as liver biopsy stage, as has already been observed in patients with chronic hepatitis C12 and B.13 The prognostic values of FibroTest, FibrometerA, Hepascore, and biopsy were similar and better than the Pugh prognostic score and those of the nonpatented scores.

These results confirm a posteriori that these markers have at baseline a high value for the diagnosis of advanced fibrosis. Liver biopsy is not a perfect gold standard and the diagnostic studies of fibrosis markers must take into account all the factors associated with the risk of biopsy failure. Because biomarkers and biopsy similarly predict survival, biomarkers seem to have the same rate of error as liver biopsy for baseline diagnosis.

The presence of baseline histologic signs of acute alcoholic hepatitis was not an independent prognostic factor. Similarly the AshTest, a validated biomarker of severe acute alcoholic hepatitis, also was not associated independently with liver disease mortality. Further studies with more patients and longer follow-up are needed to exclude the prognostic value of histologic or biologic alcoholic hepatitis independent of fibrosis, as these two liver injuries are highly associated.

Another finding of this study was that it showed that FibroTest, FibrometerA, and biopsy fibrosis staging also were useful in determining overall survival; their prognostic values were not unexpectedly, however, better for survival or nonliver-related death; this result was expected according to the numerous nonliver-related causes of death attributable to excessive alcohol consumption.

In patients with alcoholic liver disease, FibrometerA or Hepascore do not have better diagnostic or prognostic performances than FibroTest alone, or than combinations between these biomarkers. In multivariate analyses FibroTest was the most informative biomarker. As in patients with chronic viral hepatitis C and B, the prognostic value of FibroTest was similar to that of biopsy.

Ancillary