SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

The objective was to develop new blood tests to characterize different fibrosis parameters in viral and alcoholic chronic liver diseases. Measurements included 51 blood markers and Fibrotest, Fibrospect, ELFG, APRI, and Forns scores. The clinically significant fibrosis was evaluated via Metavir staging (F2-F4), and image analysis was used to determine the area of fibrosis. In an exploratory step in 383 patients with viral hepatitis, the area under the receiving operator characteristic (AUROC) curve for stages F2-F4 in a test termed the “Fibrometer” test combining platelets, prothrombin index, aspartate aminotransferase, α2-macroglobulin (A2M), hyaluronate, urea, and age was 0.883 compared with 0.808 for the Fibrotest (P = .01), 0.820 for the Forns test (P = .005), and 0.794 for the APRI test (P < 10−4). The Fibrometer AUROC curve was 0.892 in the validating step in 120 patients. The AUROC curve for stages F2-F4 in a test combining prothrombin index, A2M, hyaluronate, and age was 0.962 in 95 patients with alcoholic liver diseases. The area of fibrosis was estimated in viral hepatitis by testing for hyaluronate, γ-glutamyltransferase, bilirubin, platelets, and apolipoprotein A1 (aR2 = 0.645), and in alcoholic liver diseases by testing for hyaluronate, prothrombin index, A2M, and platelets (aR2 = 0.836). In conclusion, the pathological staging and area of liver fibrosis can be estimated using different combinations of blood markers in viral and alcoholic liver diseases. Whereas the Fibrometer has a high diagnostic accuracy for clinically significant fibrosis, blood tests for the area of liver fibrosis provide a quantitative estimation of the amount of fibrosis, which is especially useful in cirrhosis. (HEPATOLOGY 2005.)

The noninvasive diagnosis of liver fibrosis is an emerging issue for chronic liver disease (CLD).1 Several methods have been studied for the noninvasive diagnosis of hepatic fibrosis or cirrhosis: clinical2 or blood1 markers, signal analysis (ultrasound, magnetic resonance imaging, elastography3–6), and endoscopy.7 Although each method plays a role in the diagnosis of cirrhosis, only blood markers and elastography can be used for clinical application in the evaluation of fibrosis. Several blood tests have been developed with multivariate analysis using direct and/or indirect blood markers, and these have been used to evaluate clinically significant fibrosis (CSF).2, 8–14 However, these blood tests, as well as elastography, have been mainly designed to estimate the morphological pattern of fibrosis in hepatitis C virus (HCV)-related CLD based on semiquantitative histological staging, primarily the Metavir system.15 Moreover, these tests only had a single formula, whatever the cause of CLD.11, 16 In addition, the optical description of liver fibrosis relies on two distinct aspects: qualitative parameters (e.g., architectural structure) and quantitative parameters to determine the amount of fibrosis. Architectural disturbance is evaluated by histological staging, and the amount of fibrosis can be evaluated by the area of fibrosis (AOF).17

Although histological staging is the classical reference, it has several disadvantages—in particular, variability due to observational parameters18 and differences due to sampling19, 20 of liver specimens. Another problem stems from the difficulty of transforming staging into a binary variable when staging is evaluated by noninvasive tools. Indeed, whereas histological staging usually includes five stages from F0 to F4, the cutoff is fixed at F2 to define CSF as F2 or higher. Another limitation is the restriction of cirrhosis to one stage (F4), whereas the amount of fibrosis in cirrhosis is four times that of the other four stages.17 AOF is the only quantitative morphological method of determining the amount of liver fibrosis, and it has been suggested that it is superior to histological staging.17

The primary aim of this prospective study was to develop noninvasive blood tests to evaluate the different characteristics of liver fibrosis. The secondary aims were to adapt the tests to the cause of CLD (i.e., virus and/or alcohol) and to compare our blood test predicting CSF with previously published tests in viral-related CLD.

Patients and Methods

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Patients

Inclusion and Judgment Criteria.

A total of 598 patients with CLD due to alcohol, hepatitis B virus (HBV), or HCV registered at the Hepatogastroenterology Unit of the University Hospital in Angers, France, were prospectively enrolled in the study. Patients aged 18 to 74 years were included if they had consumed 50 g/d or more for the previous 5 years or if they tested positive for serum hepatitis B surface antigen or HCV RNA. Blood fasting samples were taken at entry and a percutaneous (1.6-mm–diameter needle, suction technique) liver biopsy was performed within 1 week. Patients were not included if they had other causes of liver disease or complicated cirrhosis or had received antifibrotic treatment within the previous 6 months.

The study protocol conformed to the Declaration of Helsinki and was approved by the local ethics committee. Informed consent to participate in the study was obtained from each patient. The main judgement criteria were CSF and AOF. The resulting score for CSF in viral CLD was determined using a test termed the “Fibrometer” test. The Fibrometer test was compared with the main published tests—Fibrotest, APRI, Fibrospect, ELFG, and Forns.8–12

Exploratory Population.

Four hundred seventy-eight consecutive patients with CLD were included from 1994 to 2002. CLD was due to virus (n = 383, CSF 55.6%) or alcohol (n = 95, CSF 69.5%). AOF was measured in 277 unselected patients (virus, n = 194, CSF 55.7%; alcohol, n = 83, CSF 65.9%).

Validating Population.

The primary aim of this study was to validate the performance of the Fibrometer test in a sample closer to the general HCV population21 and to compare it with recently published tests.11, 12 Thus, 120 patients with CLD due to HCV were selected between 2002 and 2004 with a liver specimen measuring 15 mm long or more and a CSF prevalence of less than 50%.

Methods

Clinical and Biological Data.

Clinical data recorded were age, sex, cause of CLD, and body mass index. Analyses of blood samples provided the following usual variables: hemoglobin, mean corpuscular volume, lymphocyte count, platelet (PLT) count, cholesterol, urea, creatinine, sodium, bilirubin, γ-glutamyltransferase (GGT), alkaline phosphatase, aspartate aminotransferase (AST), alanine aminotransferase (ALT), albumin, α1 and α2 globulins, β globulins, γ globulins, βγ block, prothrombin index (PI), apolipoprotein-A1 (apo-A1), and ferritin. Several automats were used for biochemical (Hitachi 917, Roche Diagnostics, Mannheim, Germany; Capillarys autoanalyzer, Sebia, Issy les Moulineaux, France) or hematological (Advia 120, Bayer Diagnostic, Frankfurt, Germany) usual variables. The method for PI (Neoplastin CI plus, Diagnostica Stago, Asnières, France) has been previously described.22 From these single variables, 17 ratio indexes including from 2 to 6 variables were calculated using classical ratios: AST/ALT and AST/PLT, called APRI,10 and 15 new ratios, including GAPRI (GGT/PLT) × 100 and HAPRI: (hyaluronic acid/PI) × 100. Haptoglobin was also determined to perform the Fibrotest8 according to recommendations23 in 244 recently registered patients with viral CLD from the exploratory population.

The following direct blood markers of fibrosis were obtained: α2-macroglobulin (A2M),24 PGA25 and PGAA26 scores, N-terminal peptide of type III procollagen, hyaluronic acid (HA), transforming growth factor β1, and laminin. Sera were kept at −80°C for assays as previously described for a maximum of 48 months.17 Briefly, haptoglobin, A2M, and apo-A1 levels were determined via laser immunonephelometry (BN2; Dade Behring, Marburg, Germany). The apo-A1 assay was standardized against international reference proteins SP1-01, and the A2M and haptoglobin assays were calibrated against international standards.27 However, HA was re-evaluated with a new method (Corgenix Inc. EIA; Biogenic SA, Mauguio, France); results of previous radio-immunoassay and new enzyme-linked binding protein methods for HA were well correlated (r = 0.987, P < 10−4).

In the validation population, sandwich immunoassay was used to quantify serum YKL-40 (Metra Biosystems/Quidel, San Diego, CA) and enzyme-linked immunosorbent assay for tissue inhibitor of metalloproteinase 1 and matrix metalloproteinase 2 (Human Biotrak; Amersham Biosciences, Buckinghamshire, UK). Overall, 51 blood variables were available.

Liver Histological Assessment.

Biopsy specimens were stained with hematoxylin-eosin-saffron and 0.1% picrosirius red solution. Fibrosis was staged by two independent pathologists, who were blinded for patient characteristics according to Metavir staging,15 and was also validated in alcoholic CLD28 with a final consensual judgment. AOF was measured using a Leica Quantimet Q570 image processor (Leica, Rueil-Malmaison, France) as previously described.17

Statistical Analysis.
Statistical tests.

Quantitative variables were expressed as the mean ± SD unless otherwise specified. Pearson (rp) and Spearman (rs) correlation coefficients were used as necessary. Agreement of quantitative variables was evaluated by the intraclass correlation coefficient (ric). Multiple linear regression analysis and binary logistic regression analysis were used. Forward stepwise logistic regression analysis was used to determine score probability ranging from 0 to 1, with a cutoff fixed at 0.5 for a unique dependent variable (e.g., CSF presence). Scores used were considered personal when β coefficients of score probability were provided by running multivariate analysis of composite variables in our population and native when coefficients were provided by a previous population, usually published. Scores presented are native unless otherwise specified. The prediction (or performance) of each model is expressed either by the diagnostic accuracy (DA), i.e., true positives and negatives, and by the area under the receiver operating characteristic (AUROC) curve in logistic regression, or by the adjusted R2 coefficient (aR2) in linear regression. AUROC curves were compared using the Hanley-McNeil method for paired data.29 Statistical software used were SPSS version 11.5.1 (SPSS Inc., Chicago, IL) and SAS version 8.02 (SAS Institute Inc., Cary, NC).

Sample size calculation.

The size of the exploratory population was determined to show a significant difference between the Fibrometer and the Fibrotest in viral CLD. With an α risk of 0.05, a β risk of 0.2, a CSF prevalence of 0.5, an AUROC curve correlation of 0.7, and a bilateral test, the sample size was 180 patients for the following hypothesis of AUROC curve: Fibrometer, 0.90; Fibrotest, 0.84.8, 30

Results

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

General Characteristics

CLD was secondary to HCV in 337 patients and to HBV in 46 patients in the exploratory population. The 95 patients with alcoholic CLD were significantly older and had more marked fibrosis than patients with viral CLD. The mean alcohol intake (g/d) was 94 ± 50 in 95 patients with alcoholism versus 17 ± 38 in 354 patients with viral CLD without alcoholism and 82 ± 81 in 29 patients with viral CLD and alcoholism. The AOF (expressed as a percentage) for stages F0-F1 versus F2-F4 was 7.3 ± 2.1 versus 12.8 ± 7.2, respectively (P < 10−4), in viral CLD and 7.8 ± 3.9 versus 26.5 ± 12.2, respectively (P < 10−4), in alcoholic CLD. The DA of the AOF for CSF was 73.3% (P < 10−4) in viral CLD and 84.1% (P < 10−4) in alcoholic CLD. The exploratory and validation populations were similar except for the degree of fibrosis, which was significantly less marked in the latter as expected (Table 1).

Table 1. Main Characteristics of Patients
 Exploratory PopulationValidating Population
WholeAlcoholVirusP Value*
  • Abbreviation: ULN, upper limit of normal.

  • *

    Alcohol versus virus.

  • Validation versus exploratory virus: P = .02.

  • Validation versus exploratory virus: P < 10−4.

No. of patients47895383120
Age (yrs)44.8 ± 12.549.8 ± 11.243.5 ± 12.5<10−444.1 ± 12.2
Sex (% male)64.971.663.2.1362.5
Metavir F stage (%):
 F05.912.64.2 16.7
 F135.617.940.1 35.8
 F225.216.827.3<10−423.3
 F311.911.611.9 11.7
 F421.441.116.4 12.5
CSF (%)58.569.555.6.0147.5
Metavir F score (mean)2.1 ± 1.32.5 ± 1.52.0 ± 1.2<10−41.7 ± 1.2
AOF (%)13.2 ± 10.020.1 ± 13.510.3 ± 6.2<10−4
Liver specimen length (mm)18.2 ± 6.418.4 ± 6.018.1 ± 6.5.6820.7 ± 4.9
ALT and AST < ULN (%)22.630.520.7.0418.3

Prediction of CSF

Viral CLD.
Exploratory population.

CSF was predicted in viral CLD by logistic regression using 7 different variables. The AUROC curve of derived score probability (Fibrometer test) was 0.883 ± 0.019 (95% CI: 0.846-0.921), and the DA was 82.1%.

The regression function was −0.007 PLT (G/L) − 0.049 PI (%) + 0.012 AST (UI/L) + 0.005 A2M (mg/dL) + 0.021 HA (μg/L) − 0.270 urea (mmol/L) + 0.027 age (years) + 3.718. The distribution of the 17.9% of misclassified patients via blood Fibrometer was: F0, 0; F1, 7.1; F2, 9.8; F3, 1.0; F4, 0%. The percentage of misclassified patients in each F stage was: F1, 17.4; F2, 38.7; F3, 9.1. The box plots of Fibrometer in relation to the Metavir F stages are shown in Fig. 1. The AUROC curve and ric are presented in Table 2, and the diagnostic indexes are shown in Table 3. The agreement between the Fibrotest and Fibrometer test was good (ric = 0.79, P < 10−4). Correlations with Metavir stages were: Fibrometer, rs = 0.745, P <10−4; Fibrotest, rs = 0.65, P < 10−4. The Fibrometer test AUROC curve was significantly higher than that of the Fibrotest (P = .01), Forns score (P = .005), and APRI (P < 10−4) (Fig. 2, Table 2). In logistic regression including the four tests, only the Fibrometer test (P < 10−4) independently predicted CSF.

thumbnail image

Figure 1. Probability score of CSF as determined by blood test as a function of Metavir F stages in (A) viral and (B) alcoholic CLD. Box plots represent median, quartiles, and extremes. CSF, clinically significant fibrosis.

Download figure to PowerPoint

Table 2. Comparison of AUROC Curve of Fibrometer Test and Main Published Tests As a Function of Population and Source of Algorithm
TestExploratory PopulationValidation Population
PersonalNativericPersonalNativeric
  • NOTE. ric = intraclass correlation coefficient.

  • *

    Corresponds to the algorithm from the exploratory population.

  • P = .44 (personal vs. native).

  • P = .03 (vs. Fibrometer).

  • §

    P = .01 (vs. Fibrometer).

  • P = .11 (vs. personal Fibrometer).

  • P = .59 (vs. native Fibrometer).

  • #

    P = .61 (personal vs. native Fibrotest).

  • **

    P < 10−4 (vs. Fibrometer).

  • ††

    P = .05 (vs. native Fibrometer).

  • ‡‡

    P = .52 (vs. native Fibrometer) or P = .22 (vs. personal Fibrometer).

  • §§

    P = .11 (vs. native Fibrometer) or P = .04 (vs. personal Fibrometer).

  • ∥∥

    P = .005 (vs. Fibrometer).

  • ¶¶

    P = .39 (vs. native Fibrometer).

Fibrometer0.883 ± 0.0190.907 ± 0.0270.892 ± 0.029*,0.88, P < 10−4
Fibrotest80.820 ± 0.0260.808 ± 0.027§0.95, P < 10−40.857 ± 0.0360.871 ± 0.034,#0.89, P < 10−4
APRI100.794 ± 0.028**0.822 ± 0.037††
Fibrospect120.869 ± 0.034‡‡
ELFG110.834 ± 0.037§§
Forns90.820 ± 0.030∥∥0.864 ± 0.059¶¶
Table 3. Diagnostic Indexes of Blood Tests for Clinically Significant Fibrosis As a Function of Cause of Chronic Liver Disease
 SensitivitySpecificityPositive Predictive ValueNegative Predictive ValueDiagnostic Accuracy
  1. NOTE. All values are expressed as % (95% CI).

Virus80.5 (74.4–86.6)84.1 (77.9–90.3)86.3 (80.8–91.7)77.6 (70.8–84.5)82.1 (77.7–86.5)
Alcohol91.8 (84.9–98.7)92.6 (82.7–100)96.6 (91.9–100)83.3 (70.0–96.7)92.0 (86.4–97.7)
thumbnail image

Figure 2. AUROC curve for CSF as determined by the Fibrometer test versus the Fibrotest (P = .01), APRI (P < 10−4), and Forns (P = .005) scores in the exploratory population.

Download figure to PowerPoint

Validating population.

The AUROC curve of the Fibrometer test variables from the score probabilities of the validating and exploratory populations were similar: 0.907 ± 0.027 versus 0.892 ± 0.029, respectively (P = .44). The respective DAs were 83.3% and 77.5%. The AUROC curve and ric of the Fibrometer test and of the main published tests (APRI, Fibrospect, ELFG, Fibrotest) as a function of the population and the source of the algorithm are presented in Table 2 and Fig. 3. Briefly, whatever the source of the score probability, the Fibrometer test had the highest AUROC curve. By plotting the DA against the AUROC curve, Fig. 4 shows the improvement of the Fibrometer test and Fibrotest compared with their composite variables.

thumbnail image

Figure 3. AUROC curve for CSF as determined by main blood tests in the validating population. Probability scores used are native unless unavailable (Table 2).

Download figure to PowerPoint

thumbnail image

Figure 4. Plots of diagnostic accuracy against the AUROC curve for CSF as determined by the Fibrometer test and Fibrotest and their composite variables obtained via logistic regression analysis in the validation population. A2M, α2-macroglobulin; HA, hyaluronic acid; PI, prothrombin index; GGT, γ-glutamyltransferase; B, bilirubin; H, haptoglobin; ApoA1, apolipoprotein A1; AUROC, area under the receiver operating characteristic.

Download figure to PowerPoint

Alcoholic CLD.

CSF was predicted in alcoholic CLD according to four different variables with an AUROC curve score probability of 0.962 ± 0.018 (95% CI: 0.926-0.998) and a DA of 92.0%. The regression function was: −0.169 PI (%) + 0.015 A2M (mg/dL) + 0.032 HA (μg/L) − 0.140 age (years) + 16.541. The box plots of score probability against Metavir F stages are shown in Fig. 1, and diagnostic indexes are shown in Table 3. The distribution of the 8.0% of misclassified patients was: F0, 2.3; F1, 0; F2, 4.5; F3, 0; F4, 1.1. The percentage of misclassified patients in each F stage was: F0, 16.7; F2, 26.7; F4, 2.7.

Prediction of AOF

The AOF was estimated in viral CLD according to five variables with an aR2 of 0.645 and in alcoholic CLD according to four variables with an aR2 of 0.836. The respective regression functions were: 0.015 HA (μg/L) + 0.091 bilirubin (μmol/L) − 1.666 apo-A1 (g/L) − 0.034 GGT (UI/L) + 3.037 GAPRI [(GGT/PLT).100] + 9.491 and 0.090 HA (μg/L) + 0.028 A2M (mg/dL) − 0.009 PTL (G/L) − 0.017 HAPRI [(HA/PI) × 100] + 2.166. The correlations between estimated and measured AOF are depicted in Fig. 5.

thumbnail image

Figure 5. Correlation between the AOF measured via image analysis and estimated via blood test as a function of Metavir F stages in (A) viral (rp = 0.809) and (B) alcoholic (rp = 0.922) CLD. AOF, area of fibrosis.

Download figure to PowerPoint

Performance in Different Settings

Normal Aminotransferases.

The AUROC curve for CSF was similar in patients with and without normal ALT and AST both in viral and alcoholic CLD (data not shown).

Effect of Liver Specimen Size.

The Fibrometer AUROC curve was 0.864 ± 0.030 versus 0.892 ± 0.026 in the smallest (12.9 ± 3.2 mm) and largest (23.2 ± 4.7 mm) specimens, respectively, divided according to the median length (18 mm). Specimen length was not correlated with AOF (rp = 0.02, P = .75) or AOF as estimated by the blood test (rp = −0.01, P = .89). The predictions of AOF via blood test were similar in the smallest (aR2 = 0.661) and largest (aR2 = 0.578) specimens.

Effect of the Cause of CLD.

The Fibrometer AUROC curve was 0.868 ± 0.056 in HBV patients versus 0.883 ± 0.021 in HCV patients (CSF prevalence: 69.0% vs. 54.2%, respectively; P = .07). In combined alcoholic and viral CLD, cause of CLD was an independent predictor of CSF or AOF (Table 4). Score probabilities obtained for one cause were tested for the other cause (e.g., the tests developed for viral CLD were applied to alcoholic CLD) and in combined causes (Table 4). Briefly, when a test designed for a specific cause was applied to another cause or combined causes, its results (AUROC or aR2) were poorer than when it was applied for its specific cause.

Table 4. Comparison of Blood Tests As a Function of Cause
Original Design of TestVirusAlcoholVirus + Alcohol
  • *

    According to PLT, HA, PI, A2M, urea, and cause.

  • According to PLT, HA, PI, urea, and cause.

Clinically significant fibrosis (AUROC curve) in CLD due to:
 Virus (Fibrometer test)0.883 ± 0.0190.916 ± 0.0300.883 ± 0.017
 Alcohol0.769 ± 0.0270.962 ± 0.0180.819 ± 0.021
 Virus or alcohol*0.871 ± 0.0200.939 ± 0.0250.890 ± 0.016
AOF (aR2) in CLD due to:
 Virus0.6530.3460.496
 Alcohol0.4210.8380.593
 Virus or alcohol0.5000.8170.773

Discussion

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Methods.

The present study is original because it measures the AOF in addition to histological staging, takes into account the cause of CLD, and includes a large number of blood variables (n = 51). The reference method—liver biopsy—may vary due to sampling19, 20 and observer errors.18 The results of our blood tests for CSF were better in the largest specimens; thus test results could be even better with recent criteria for specimen length (≥20 mm).31, 32 It has been suggested that variability is greater for AOF evaluation than it is for histological staging.31 However, blood tests for AOF were not sensitive to specimen length in the present study, whereas we18 and others32 have shown that histological staging is sensitive to specimen length. In addition, inter- and intraobserver agreement is excellent for AOF evaluation in our unit.17

The quantitative variable AOF can be used to analyze blood tests via linear regression, thus providing a direct linear estimation of the pathological characteristic based on its reference unit (expressed as a percentage). Moreover, logistic regression analysis provides a CSF score probability. Score probability alone could be used for histological staging, but this would be an indirect and nonlinear score with no reference unit. Thus, compared with histological staging, the advantage of estimations derived from image analysis is that they provide precise and easily interpreted results. On the other hand, the determination of an F stage instead of CSF following logistic regression such as that suggested in the Fibrotest or elastography6 has not been statistically validated, because this would require other methods such as discriminant analysis. Blood test results might also depend on the prevalence of CSF. However, Fibrometer AUROC curves were identical between patients with the largest specimens in the exploratory population (0.892 ± 0.026) and the validating population (0.892 ± 0.029), with similar specimen lengths but a significantly lower Metavir F score in the latter.

Although there was less observer variability for histological staging when a consensual reading was performed by two experts,18 blood test results were not perfect (e.g., the DA for the Fibrometer test was 83.3% in the validation population). This might be due to preanalytical or analytical variability in blood variables or variability in the pathological reference. Recent studies with the Fibrotest have suggested that most errors are due to the histological staging itself.16, 33 It should be noted that the plots of blood test results (data not shown) and of interobserver agreement in relation to F stages18 all had the same V-shape with a nadir at the F2 stage, suggesting that the difficulty in distinguishing F2 from F1 or F3 stages via histological staging was the main cause of misclassification by blood tests. The reproducibilty of the blood tests will mainly depend on variations in analytical conditions that should be taken into account when extrapolating results or when external validation studies are conducted.

New Tests.

The following independent variables were included in the different tests: PLT, PI, AST, A2M, HA, urea, age, bilirubin, GGT, albumin, and apo-A1. Most of these variables reflect liver function or are known fibrosis markers. For example, of the first seven variables of the Fibrometer test for the prediction of CSF in viral CLD, PLT,34 A2M,24, 26 HA,14 AST,10 and age34 are known fibrosis markers in HCV-related hepatitis. The independent role of PI, which expresses prothrombin time as a percentage, confirms that it is a useful marker of liver fibrosis.22, 35 The role of urea has not been reported, but the urea synthesis rate is decreased in patients who have cirrhosis.36 Our tests included direct and indirect fibrosis markers. Other authors included either indirect, easily accessible markers8–10, 13 or direct markers11, 12, 14 for their putative specificity. We suggest that the combination of both markers might increase certain advantages and limit other disadvantages. Thus, direct markers are more specific but can be altered by other phenomena, especially during connective tissue damage. However, an indirect marker such as PI is organ-specific. Finally, the combination of markers slightly increases sensitivity (Fig. 4) and increases specificity by limiting errors due to isolated variations in biomarkers for diseases other than liver fibrosis.

The use of blood tests to estimate the AOF is new. Measurement of the AOF via image analysis is limited to clinical research. Unlike histological stagings, the AOF provides precise quantification of the extensive variations in fibrosis during cirrhosis. Indeed, because cirrhosis corresponds to only one15 or two37 stages, the range of blood test results for CSF in patients with cirrhosis is very limited (Fig. 1). Thus, in viral-related CLD, the interquartile range of blood tests for AOF was 7.2%-10.4% in F0-F3 versus 11.2%-19.0% in F4, whereas the corresponding ranges for the Fibrometer test for CSF were 0.03-0.99 and 0.99-1.00, respectively. Finally, AOF estimation via blood test is the only statistically validated quantitative test for the noninvasive diagnosis of fibrosis. However, the aim of AOF evaluation is not to distinguish mild F stages due to overlap of AOF values in these stages.

The exact meaning of these tests requires further investigation. Indeed, their relationship with fibrosis is complex, because the blood markers interact with the dynamics and the cause of fibrosis.17, 22, 38 For example, HA is a marker of fibrogenesis, and its blood level depends on the persistence of the cause of CLD, especially on alcohol intake.38 This could explain why the predictive models included different variables depending on the cause of CLD.

Similar performances of the Fibrometer test have been observed in the evaluation of HBV- and HCV-related CLD as with the Fibrotest.39 As in the present study, better results were observed for alcohol-related CLD than for viral-related CLD with the ELFG test11 but not the Fibrotest.16 This higher performance could be due to the higher CSF or F4 prevalence in alcohol-related CLD.16

Comparison With Other Tests.

The new tests presented in this study have several advantages, in particular adaptability and accuracy. Compared with our previous test for CSF,2 the present tests are more accurate because they include more variables and have been adapted to the cause of CLD (i.e., the test formula varies depending on the cause). To our knowledge, our test results for viral CLD (using the Fibrometer test) are also better than those reported in other published tests (Table 5). This was statistically demonstrated compared with the Fibrotest, Forns, and APRI scores in our exploratory population. This was also suggested in our validation population, in which the AUROC curve of other tests was similar to the published AUROC curves (Table 5). However, the difference between the Fibrometer test and other tests was not statistically significant (except for APRI), but the sample size of this validation population was not designed for such a comparison. The results of other tests could have been underestimated in the present study due to analitycal variability. However, our present results are within the range of original results (Table 5).

Table 5. Published Blood Tests of Clinically Significant Fibrosis in HCV-Related Chronic Liver Disease
StudyVariablesPublished AUROC*Present Study*
  • *

    The second figure corresponds to validation set.

  • Diagnostic accuracy in alcoholic and viral chronic liver disease.

  • Judgment criteria: Ishak stage ≥3 here and Metavir F stage ≥2 in all other studies.

Oberti et al.2PI, HA, A2M0.78
Imbert-Bismut et al.8A2M, apo-A1, GGT, bili, haptoglobin0.836–0.8700.808–0.871
Forns et al.9Age, GGT, cholesterol, PLT0.86–0.810.820
Wai et al.10AST/PLT0.80–0.880.794–0.822
Rosenberg et al.11Age, HA, PIIIP, TIMP10.770.834
Patel et al.12TIMP1, HA, A2M0.83–0.820.869
Sud et al.13AST, age, HOMA, alcohol, cholesterol0.84–0.77
Leroy et al.14PIIIP, MMP10.82
Present study (Fibrometer)PI, HA, A2M, AST, urea, PLT, age0.883–0.907

The good results of our tests can be attributed to the numerous direct and indirect markers. Indeed, most previous studies have limited the inclusion of markers to one of the two categories. Rosenberg et al. stated that the inclusion of PI and PLTs did not increase the performance of their direct markers.11 However, they did not include the most accurate single marker (A2M), which is included in several other tests.

The importance of cause in the diagnostic accuracy is suggested by three factors. First, the performance of one test designed for a single cause of CLD was always better in the corresponding population than when it was applied to the other single cause or to combined causes (Table 4). Second, as expected, the AUROC curve of the test designed for combined causes was less accurate when the cause of CLD was removed (data not shown) from the model, because this variable had an independent role in the test performance. Third, when the test designed for combined viral and alcoholic causes was applied to a single cause, results were always worse than results of the test designed for a single cause (Table 4).

The other attractive noninvasive tool for evaluating fibrosis is transient elastography (FibroScan). Its AUROC curve for CSF was 0.79 in mixed causes5 and 0.83 in HCV CLD.6 However, the intention-to-diagnosis principle3 was not applied, because 10% of included patients were excluded from the final analysis.

Application to Clinical Practice.

The limitations of these blood tests are their analytical variability due to assay or laboratory techniques and also to the indications. Indeed, these tests were performed in adults with compensated alcoholic or viral CLD before any specific treatment. Other situations must be investigated. Other causes of variations in blood parameters not related to CLD must be taken into account. This is especially important when a high score is obtained because of a single or a few abnormal variables, because there is a correlation between the percentage of abnormal values and score probability (data not shown). Although AOF evaluation is not used in clinical practice at present, this could change if it became easily available and if easy-to-interpret blood tests were used to obtain the AOF.

Although increasing attention has been paid to blood tests for CSF, their interpretation is not always simple. Indeed, results produce a probability score varying from 0 to 1 that can be applied in three ways. First, it can be applied as a qualitative variable (i.e., indicating the absence [P < .5] or presence [P ≥ .5] of CSF). Although this is a simplistic interpretation, the probability score was statistically designed for this purpose. Second, the probability score can be used as a semiquantitative variable by giving the correspondence between the score and the five Metavir fibrosis stages,33 which is interesting from a clinical point of view. However, the statistical validity of this conversion is debatable. Third, the probability score can be used as it is (i.e., as a quantitative variable, such as a fibrosis meter ranging from 0 [lack of fibrosis] to 1 [cirrhosis]). It is more reliable to use the probability score as a crude measure. Thus, for example, a P value of .48 means no CSF, whereas the fibrosis stage is probably F1. However, this probability score is close to the probability threshold of .5; thus, a P value of .52 (which is close to .48) means that CSF and stage F2 are probable. This example shows that the crude measure (P = .48) is a better indication of the patient's result than a qualitative interpretation such as no CSF or F1. However, a CSF probability score is a fibrosis meter with no unit of reference and with a nonlinear relationship to the F stage (Fig. 1). In addition, this fibrosis meter is limited in F4 patients where the AOF is a better indicator. Indeed, the extent of fibrosis in cirrhosis might be evaluated by a blood test for AOF which is an accurate meter with a real unit (fibrosis %).

Finally, we suggest that fibrosis can be determined by two independent noninvasive tests (two blood tests or a blood test + elastography). If the results agree, liver biopsy is not needed.

In conclusion, our results suggest that the different characteristics of liver fibrosis can be accurately estimated by combining a few blood markers in viral and alcoholic CLD.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

We would like to thank Dermot O'Toole, Daniel Chappard, Bruno Vielle, Yves Tourmen, Gwénaëlle Soulard, Frédéric Moal, Alain Godon, Elisabeth Mathieu, and Dale Roche for their contribution.

References

  1. Top of page
  2. Abstract
  3. Patients and Methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  • 1
    Afdhal NH, Nunes D. Evaluation of liver fibrosis: a concise review. Am J Gastroenterol 2004; 99: 11601174.
    Direct Link:
  • 2
    Oberti F, Valsesia E, Pilette C, Rousselet MC, Bedossa P, Aubé C, et al. Noninvasive diagnostic of hepatic fibrosis or cirrhosis. Gastroenterology 1997; 113: 16091616.
  • 3
    Aubé C, Winkfield B, Oberti F, Vuillemin E, Rousselet MCR, Caron C, et al. New US-Doppler signs improve the non-invasive diagnosis of cirrhosis or severe liver fibrosis. Eur J Gastroenterol Hepatol 2004; 16: 743751.
  • 4
    Aubé C, Racineux PX, Lebigot J, Oberti F, Croquet V, Argaud C, et al. Diagnosis and quantification of hepatic fibrosis with diffusion weighted MR imaging: preliminary results [in French]. J Radiol 2004; 85: 301306.
  • 5
    Ziol M, Handra-Luca A, Kettaneh A, Christidis C, Mal F, Kazemi F, et al. Noninvasive assessment of liver fibrosis by measurement of stiffness in patients with chronic hepatitis C. HEPATOLOGY 2005; 41: 4854.
  • 6
    Castéra L, Vergniol J, Foucher J, Le Bail B, Chanteloup E, Haaser M, et al. Prospective comparison of transient elastography, Fibrotest, APRI, and liver biopsy for the assessment of fibrosis in chronic hepatitis C. Gastroenterology 2005; 128: 343350.
  • 7
    Oberti F, Burtin P, Maïga M, Valsesia E, Pilette C, Calès P. Gastroesophageal endoscopic signs of cirrhosis: independent diagnostic accuracy, interassociation, and relationship to etiology and hepatic dysfunction. Gastrointest Endosc 1998; 48: 148157.
  • 8
    Imbert-Bismut F, Ratziu V, Pieroni L, Charlotte F, Benhamou Y, Poynard T, for the MULTIVIRC group. Biochemical markers of liver fibrosis in patients with hepatitis C virus infection: a prospective study. Lancet 2001; 37: 10691075.
  • 9
    Forns X, Ampurdanes S, Llovet JM, Aponte J, Quinto L, Martinez-Bauer E, et al. Identification of chronic hepatitis C patients without hepatic fibrosis by a simple predictive model. HEPATOLOGY 2002; 36: 986992.
  • 10
    Wai CT, Greenson JK, Fontana RJ, Kalbfleisch JD, Marrero JA, Conjeevaram HS, et al. A simple noninvasive index can predict both significant fibrosis and cirrhosis in patients with chronic hepatitis C. HEPATOLOGY 2003; 38: 518526.
  • 11
    Rosenberg WM, Voelker M, Thiel R, Becka M, Burt A, Schuppan D, et al.; The European Liver Fibrosis Group. Serum markers detect the presence of liver fibrosis: a cohort study. Gastroenterology 2004; 127: 17041713.
  • 12
    Patel K, Gordon SC, Jacobson I, Hezode C, Oh E, Smith KM, et al. Evaluation of a panel of non-invasive serum markers to differentiate mild from moderate-to-advanced liver fibrosis in chronic hepatitis C patients. J Hepatol 2004; 41: 935942.
  • 13
    Sud A, Hui JM, Farrell GC, Bandara P, Kench JG, Fung C, et al. Improved prediction of fibrosis in chronic hepatitis C using measures of insulin resistance in a probability index. HEPATOLOGY 2004; 39: 12391247.
  • 14
    Leroy V, Monier F, Bottari S, Trocme C, Sturm N, Hilleret MN, et al. Circulating matrix metalloproteinases 1, 2, 9 and their inhibitors TIMP-1 and TIMP-2 as serum markers of liver fibrosis in patients with chronic hepatitis C: comparison with PIIINP and hyaluronic acid. Am J Gastroenterol 2004; 99: 271279.
    Direct Link:
  • 15
    The French METAVIR Cooperative Study Group. Intraobserver and interobserver variations in liver biopsy interpretation in patients with chronic hepatitis C. HEPATOLOGY 1994; 20: 1520.
  • 16
    Naveau S, Raynard B, Ratziu V, Abella A, Imbert-Bismut F, Messous D, et al. Biomarkers for the prediction of liver fibrosis in patients with chronic alcoholic liver disease. Clin Gastroenterol Hepatol 2005; 3: 167174.
  • 17
    Pilette C, Rousselet MC, Bedossa P, Chappard D, Oberti F, Rifflet H, et al. Histopathological evaluation of liver fibrosis: quantitative image analysis vs semi-quantitative scores. J Hepatol 1998; 28: 439446.
  • 18
    Rousselet MC, Michalak S, Dupré F, Croué A, Bedossa P, Saint-André JP, et al.; Hepatitis Network 49. Sources of variability in histological scoring of chronic viral hepatitis. HEPATOLOGY 2005; 41: 257264.
  • 19
    Regev A, Berho M, Jeffers LJ, Milikowski C, Molina EG, Pyrsopoulos NT, et al. Sampling error and intraobserver variation in liver biopsy in patients with chronic HCV infection. Am J Gastroenterol 2002; 97: 26142618.
    Direct Link:
  • 20
    Siddique I, El-Naga HA, Madda JP, Memon A, Hasan F. Sampling variability on percutaneous liver biopsy in patients with chronic hepatitis C virus infection. Scand J Gastroenterol 2003; 38: 427432.
  • 21
    Guyader D, Lefeuvre C, Jacquelinet S, Prat M, Baudouard Y, Turlin B, et al. Epidemiology of hepatitis C virus infection in 1,304 HCV positive patients: variations according to the origin of transmission and year of diagnosis [in French]. Gastroenterol Clin Biol 1998; 22: 375380.
  • 22
    Croquet V, Vuillemin E, Ternisien C, Pilette C, Oberti F, Gallois Y, et al. Prothrombin index is an indirect marker of severe liver fibrosis. Eur J Gastroenterol Hepatol 2002; 14: 11331141.
  • 23
    Imbert-Bismut F, Messous D, Thibaut V, Myers RB, Piton A, Thabut D, et al. Intra-laboratory analytical variability of biochemical markers of fibrosis (Fibrotest) and activity (Actitest) and reference ranges in healthy blood donors. Clin Chem Lab Med 2004; 42: 323333.
  • 24
    Tiggelman AM, Boers W, Moorman AF, de Boer PA, Van der Loos CM, Rotmans JP, et al. Localization of alpha 2-macroglobulin protein and messenger RNA in rat liver fibrosis: evidence for the synthesis of alpha 2-macroglobulin within Schistosoma mansoni egg granulomas. HEPATOLOGY 1996; 23: 12601267.
  • 25
    Poynard T, Aubert A, Bedossa P, Abella A, Naveau S, Paraf F, et al. A simple biological index for detection of alcoholic liver disease in drinkers. Gastroenterology 1991; 100: 13971402.
  • 26
    Naveau S, Poynard T, Benattar C, Bedossa P, Chaput JC. Alpha-2-macroglobulin and hepatic fibrosis: diagnostic interest. Dig Dis Sci 1994; 11: 24262432.
  • 27
    Dati F, Schumann G, Thomas L, Aguzzi F, Baudner S, Bienvenu J, et al. Consensus of a group of professional societies and diagnostic companies on guidelines for interim reference ranges for 14 proteins in serum based on the standardization against the IFCC/BCR/CAP Reference Material (CRM 470). International Federation of Clinical Chemistry. Community Bureau of Reference of the Commission of the European Communities. College of American Pathologists. Eur J Clin Chem Clin Biochem 1996; 34: 517520.
  • 28
    Michalak S, Rousselet MC, Bedossa P, Pilette C, Chappard D, Oberti F, et al. Respective roles of porto-septal fibrosis and centrilobular fibrosis in alcoholic liver disease. J Pathol 2003; 201: 5562.
  • 29
    Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983; 148: 839843.
  • 30
    Calès P, Oberti F, Rousselet MC, Gallois Y. Can we avoid liver biopsy to assess the degree of clinically significant fibrosis? [In French]. Gastroenterol Clin Biol 2002; 26: A72.
  • 31
    Bedossa P, Dargere D, Paradis V. Sampling variability of liver fibrosis in chronic hepatitis C. HEPATOLOGY 2003; 38: 14491457.
  • 32
    Colloredo G, Guido M, Sonzogni A, Leandro G. Impact of liver biopsy size on histological evaluation of chronic viral hepatitis: the smaller the sample, the milder the disease. J Hepatol 2003; 39: 239244.
  • 33
    Poynard T, Munteanu M, Imbert-Bismut F, Charlotte F, Thabut D, Le Calvez S, et al. Prospective analysis of discordant results between biochemical markers and biopsy in patients with chronic hepatitis C. Clin Chem 2004; 50: 13441355.
  • 34
    Poynard T, Bedossa P. Age and platelet count: a simple index for predicting the presence of histological lesions in patients with antibodies to hepatitis C virus. J Viral Hepat 1997; 4: 199208.
  • 35
    Tran A, Hastier P, Barjoan EM, Demuth N, Pradier C, Saint-Paul MC, et al. Non invasive prediction of severe fibrosis in patients with alcoholic liver disease. Gastroenterol Clin Biol 2000; 24: 626630.
  • 36
    Shangraw RE, Jahoor F. Effect of liver disease and transplantation on urea synthesis in humans: relationship to acid-base status. Am J Physiol 1999; 276: G1145G1152.
  • 37
    Ishak K, Baptista A, Bianchi L, Callea F, De Groote J, Gudat F, et al. Histological grading and staging of chronic hepatitis. J Hepatol 1995; 22: 696699.
  • 38
    Parés A, Deulofeu R, Giménez A, Caballeria L, Bruguera M, Caballeria J, et al. Serum hyaluronate reflects hepatic fibrogenesis in alcoholic liver disease and is useful as a marker of fibrosis. HEPATOLOGY 1996; 24: 13991403.
  • 39
    Myers RP, Tainturier MH, Ratziu V, Piton A, Thibault V, Imbert-Bismut F, et al. Prediction of liver histological lesions with biochemical markers in patients with chronic hepatitis B. J Hepatol 2003; 39: 222230.