Diagnosis of liver fibrosis and cirrhosis
09585613 · 2017-03-07
Assignee
Inventors
Cpc classification
G16H50/20
PHYSICS
International classification
Abstract
This invention relates to method of diagnosing the presence and/or severity of a liver pathology and/or of monitoring the effectiveness of a curative treatment against a liver pathology in an individual, leading to a score, comprising the combination, of at least one blood test and of at least one data issued from a physical method of diagnosing liver fibrosis, said data being selected from the group consisting of medical imaging data and clinical measurements, said combination being performed through a mathematical function. This invention also relates to a method wherein the combination through a mathematical function, of at least one blood test and of at least one data issued from a physical method of diagnosing liver fibrosis, is performed at least twice and the at least two resulting scores are combined in an algorithm based on the diagnostic reliable intervals.
Claims
1. A microprocessor comprising a computer algorithm to perform a method of diagnosing the presence and/or severity of a liver pathology and/or of monitoring the effectiveness of a curative treatment against a liver pathology in an individual comprising: obtaining a blood test score of at least one blood test from an individual, wherein the blood test comprises obtaining a blood sample from the individual and analyzing a marker for liver pathology in the blood sample to obtain a score; obtaining a result of using at least one measuring device to practice a non-invasive physical method for diagnosing liver fibrosis, wherein the physical method is further defined as comprising medical imaging and/or clinical measurement and is further defined as elastometry; and performing a mathematical function to combine the blood test score with the result of the physical method for diagnosing liver function to obtain a second score useful for the diagnosis of the presence and/or severity of a liver pathology and/or of monitoring the effectiveness of a curative treatment against a liver pathology in the individual.
2. The microprocessor of claim 1, wherein the blood test is further defined as an Hepascore, Fibrotest, FibroMeter, Elf score, or Fibrospect blood test.
3. The microprocessor of claim 1, wherein the liver disease or condition is significant porto-septal fibrosis, severe porto-septal fibrosis, centrolobular fibrosis, cirrhosis, or persinusoidal fibrosis and of alcoholic or non-alcoholic origin.
4. The microprocessor of claim 1, wherein the mathematical function is a logistic regression.
5. The microprocessor of claim 1, wherein the mathematical function is a binary logistic regression.
6. The microprocessor of claim 1, wherein the individual is a patient with chronic Hepatitis C.
7. The microprocessor of claim 1, wherein performing a mathematical function to combine the blood test score with the result of the physical method for diagnosing liver function to obtain a second score is done at least twice to obtain at least two second scores, and the at least two second scores are then combined in an algorithm based on diagnostically reliable intervals.
8. The microprocessor of claim 1, wherein the method further comprises treating the individual for a liver pathology.
9. The microprocessor of claim 1, wherein elastometry is further defined as selected from the group consisting of Fibroscan, Acoustic Radiation Force Impulse imaging (ARFI imaging), supersonic elastometry, transient elastography (TE) and MRI stiffness.
Description
(1) The invention will be better understood in view of the following examples, which are read with consideration of the figures:
(2)
(3)
(4)
(5)
(6)
(7)
EXAMPLES
(8) The following examples may be read, when appropriate, with references to the figures, and shall not be considered as limiting in any way the scope of this invention.
Example 1
(9) Blood fibrosis tests and liver stiffness measured by ultrasonographic elastometry like Fibroscan are well correlated with the histological stages of fibrosis. In this study, we aimed to improve non-invasive diagnosis of liver fibrosis stages via a novel combination of blood tests and Fibroscan.
(10) Methods: 349 patients with chronic hepatitis C across three centres were included in the study. For each patient, a liver biopsy and the following fibrosis tests were done: Fibroscan (FS), Fibrotest, FibroMeter (FM, for significant fibrosis or cirrhosis), Hepascore, Fib4, and APRI. Reference for liver fibrosis was Metavir F staging. Fibrosis tests independently associated with significant fibrosis (F2) or cirrhosis (F4) were identified by stepwise binary logistic regression repeated on 1000 bootstrap samples of 349 patients.
Results: Prevalences of diagnostic targets were, significant fibrosis: 67.9%, cirrhosis: 11.8%. Multivariate analyses on the 1000 bootstrap samples indicated that FM and FS were the tests most frequently associated with significant fibrosis or cirrhosis. We thus implemented 2 new scores combining FS and FM by using binary logistic regression: F2-score for the diagnosis of significant fibrosis and F4-score for cirrhosis. F2-score provided reliable diagnosis of significant fibrosis, with predictive values90%, in 55.6% of patients. F4-score provided reliable diagnosis of cirrhosis, with predictive values95%, in 89.1% of patients. An algorithm combining F2-score and F4-score, as a function of their interval of highest diagnostic accuracy, produced a new diagnostic classification (% of patients): F0/1 (9.5%), F1/2 (17.2%), F21 (27.2%), F2/3 (33.2%), F31 (10.9%), and F4 (2.0%). According to liver biopsy results, this new classification provided 88.0% diagnostic accuracy, outperforming FM (67.6%, p<10.sup.3), FS (55.3%, p<10.sup.3) and Fibrotest (33.2%, p<10.sup.3) classifications. Furthermore, diagnostic accuracy of the new classification did not significantly differ over the 3 centres (92.9%, 85.7%, and 86.3%, p=0.20) or between patients with biopsies < or 25 mm (respectively: 87.2% versus 88.5%, p=0.72).
Conclusions: The non-invasive diagnosis of liver fibrosis in patients with chronic hepatitis C is improved by a combination of FibroMeter and Fibroscan. A new classification using the two scores derived from the test combination is much more accurate than single fibrosis tests and provides a non-invasive diagnosis in 100% of patients with 88% accuracy without any liver biopsy.
Patients
(11) The exploratory set included 349 patients. 132 patients from the 512 of the Fibrostar study were already included in the exploratory set. We thus removed these patients from the validation set which finally included 380 patients. The characteristics of both exploratory and validation sets are detailed in the Table 1 of Example 1. Among the 2 groups, 93.5% of liver biopsy were considered as reliable.
(12) Implementation of the New Classifications (Exploratory Set)
(13) New Scores Combining Blood Fibrosis Tests and LSE
(14) Significant fibrosisThe fibrosis tests most frequently selected by the stepwise binary logistic regression repeated on the 1000 bootstrap samples for the diagnosis of significant fibrosis were LSE and FibroMeter (Table 2 of Example 1). F2-index was implemented by including these 2 fibrosis tests as independent variables in a binary logistic regression performed in the whole population of the exploratory set. The regression score of F2-index, specifically designed for the diagnosis of significant fibrosis, was: 3.9066 FibroMeter+0.1870 LSE result2.8345. F2-index had a significantly higher AUROC than FibroMeter and LSE (Table 3 of Example 1).
(15) Severe fibrosisThe fibrosis tests most frequently selected by the 1000 bootstrap multivariate analyses were LSE and FibroMeter (Table 2 of Example 1). The regression score of F3-index including these 2 fibrosis tests and specifically designed for the diagnosis of severe fibrosis was: 3.3135 FibroMeter+0.1377 LSE result4.2485. F3-index had a higher AUROC than FM and LSE, but the difference was significant only with FibroMeter (Table 3 of Example 1).
CirrhosisThe fibrosis tests most frequently selected by the 1000 bootstrap multivariate analyses were also LSE and FibroMeter (Table 2 of Example 1). The regression score of F4-index including these 2 fibrosis tests and specifically designed for the diagnosis of cirrhosis was: 3.6128 FibroMeter+0.1484 LSE result6.4999. F4-index had a higher AUROC than FM and LSE, but the difference was significant only with FibroMeter (Table 3 of Example 1).
Intervals of Reliable Diagnosis
(16) Significant fibrosisF2-index included 32 (9.2%) patient in the 90% negative predictive value (NVP) interval and 161 (46.1%) patients in the 90% positive predictive value (PPV) interval (Table 4 of Example 1). Thus, F2-index allowed a reliable diagnosis of significant fibrosis with 90% accuracy in 55.3% of patients, versus 33.8% with LSE (p<10.sup.3) and 55.6 with FibroMeter (p=1.00). The indeterminate interval between F2-index values>0.248 and <0.784 was divided into two new intervals according to the statistical cut-off of 0.500. 90.2% of the patients included in the lower interval (>0.248-<0.500) had F1/2 stages according to liver biopsy results, and 96.8% of patients included in the higher interval (0.500-<0.784) had F1/2/3 stages (
(17) Severe fibrosisF3-index included 174 (49.9%) patients in the intervals of 90% predictive values for severe fibrosis (Table 4 of Example 1), versus 41.8% with FibroMeter (p<10.sup.3) and 46.4% with LSE (p=0.235). By dividing the intermediate interval of F3-index according to the statistical cut-off of 0.500, F3-index provided 4 IRD (F2, F21, F2, F3;
CirrhosisF4-index included 313 (89.7%) patients in the intervals of 95% predictive values for cirrhosis (Table 4 of Example 1), versus 65.9% with FibroMeter (p<10.sup.3) and 87.4% with LSE (p=0.096). Dividing the intermediate interval according to the cut-off 0.500 did not allow for distinguish two different groups. Finally, F4-index provided 3 IRD (F3, F2, and F4) which well classified 95.1% of patients (
New Classifications
(18) The first classification (classification A) was derived from both F2- and F3-indexes used with their IRD (Table 5 of Example 1). Classification A included 6 classes: F0/1, F1/2, F21, F2/3, F2, and F3. It provided 86.2% diagnostic accuracy in the exploratory set. The second classification (classification B) was derived from the IRD of F2- and F4 indexes (Table 5 of Example 1). Classification B included 6 classes (F0/1, F1/2, F21, F2/3, F2, F4) and provided 88.3% diagnostic accuracy (p=0.143 vs classification A). The third classification (classification C) was derived from the IRD for significant fibrosis of FibroMeter, and those for severe fibrosis of LSE (Table 5 of Example 1). Results of FibroMeter and LSE RDI were discordant in 2 patients which had thus undetermined diagnosis (Table 5 of Example 1). Classification C finally included 8 classes (F0/1, F1, F1/2, F2, F21, F2/3, F2, F3) and provided 84.0% diagnostic accuracy (p=0.229 vs classification A).
(19) Validation of the Classifications (Validation Set)
(20) Diagnostic accuracy of fibrosis tests classificationsThe rates of well classified patients by the new classifications A and B were not significantly different in the validation set (respectively: 84.2% vs 82.4%, p=0.149), but were significantly higher than those of FibroMeter, LSE and Fibrotest (Table 6 of Example 1). One patient had undetermined diagnosis with the classification C that provided 70.3% diagnostic accuracy. Among already published classifications, FibroMeter provided the highest diagnostic accuracy (69.7%, p<0.029 vs LSE and Fibrotest), and Fibrotest the lower (p<10.sup.3 vs others). Finally, according to their diagnostic accuracies in the validation set, the classifications were ordered as follow: A, B>C>FibroMeter>LSE>Fibrotest (Table 6 of Example 1).
(21) Influencing factorsIn the whole study population, we performed a stepwise binary logistic regression including age, sex, biopsy length, Metavir F, and IQR/median as independent variables. Misclassification by classification A was independently associated only with the ratio IQR/median. In the validation set, classification A provided 88.2% diagnostic accuracy in patients with IQR/median<0.21 versus 70.1% in patients with IQR/median0.21 (p=0.010). In the subgroup of patients with IQR/median<0.21, classification A had the highest diagnostic accuracy with p=0.007 versus classification B (85.5%), and p<10.sup.3 versus others.
(22) Management for antiviral therapy in clinical practiceAntiviral therapy was considered when FibroMeter classification was F2/3, LSE: F2, Fibrotest: F2, classifications A and B: F21, and classification C: F2. By using classification A, 12.1% of patients in the validation set were considered for antiviral therapy whereas they had no/mild fibrosis at liver biopsy (Table 7 of Example 1). On the other hand, 9.7% of patients had no treatment whereas they had significant fibrosis at liver biopsy. Finally, classification A provided the highest rate of patients well managed for antiviral therapy (78.2%, p<0.040 versus others classifications).
(23) TABLE-US-00003 TABLE 1 OF EXAMPLE 1 Patients characteristics at inclusion Set All Exploratory Validation p Patients (n) 729 349 380 Male sex (%) 61.3 60.2 62.4 0.531 Age (years) 51.7 11.2 52.1 11.2 51.3 11.2 0.347 Metavir F (%): <10.sup.3 0 4.0 1.4 6.3 1 37.7 30.7 44.2 2 25.8 35.5 16.8 3 17.6 20.6 14.7 4 15.0 11.7 17.9 0.020 Significant 58.3 67.9 49.5 <10.sup.3 fibrosis (%) Reliable biopsy (%) 93.5 92.6 94.2 0.391 LSE result (kPa) 10.0 7.9 9.9 8.1 10.1 7.7 0.755 IQR/median <0.21 66.9 66.2 67.6 0.700 (%) LSE: liver stiffness evaluation; kPa: kilopascal; IQR: interquartile range
(24) TABLE-US-00004 TABLE 2 OF EXAMPLE 1 Selection of candidate predictors at bootstrapped stepwise binary logistic regressions, as a function of diagnostic target Significant fibrosis Severe fibrosis Cirrhosis Fibrosis tests (Metavir F 2) (Metavir F 3) (Metavir F = 4) FibroMeter 920 903 610 FibroMeter F4 284 Fibrotest 113 173 88 Hepascore 216 74 172 Fib4 85 103 62 APRI 350 504 59 LSE 964 1000 993
Stepwise binary logistic regressions were performed on 1000 bootstrap samples of 349 subjects from the exploratory set. The table depicts the number of times any fibrosis test was selected across the 1000 multivariate analyses. For each diagnostic target, LSE and FibroMeter were the mostly selected variables.
(25) TABLE-US-00005 TABLE 3 AUROC of FibroMeter, LSE and their synchronous combination as a function of diagnostic target and patient group Diagnostic Set target Fibrosis test Exploratory Validation p All Metavir FibroMeter 0.806 0.026 0.839 0.022 0.333 0.813 0.017 F 2 LSE 0.785 0.026 0.828 0.022 0.207 0.791 0.017 F 2-index 0.835 0.023 0.875 0.019 0.180 0.846 0.015 FibroMeter vs LSE 0.513 0.685 0.301 FibroMeter vs F 2- 0.027 0.0020 0.0002 index LSE vs F 2-index 0.024 0.0086 0.0002 Metavir FibroMeter 0.776 0.025 0.880 0.020 0.0012 0.829 0.016 F 3 LSE 0.816 0.025 0.881 0.019 0.038 0.847 0.016 F 3-index 0.830 0.022 0.918 0.017 0.0016 0.875 0.014 FibroMeter vs LSE 0.163 0.993 0.324 FibroMeter vs F 3- <10.sup.4 0.0002 <10.sup.4 index LSE vs F 3-index 0.458 0.014 0.019 Metavir FibroMeter 0.814 0.031 0.897 0.021 0.027 0.861 0.018 F = 4 LSE 0.878 0.032 0.927 0.017 0.176 0.905 0.017 F4-index 0.890 0.028 0.947 0.014 0.069 0.921 0.015 FibroMeter vs LSE 0.059 0.193 0.026 FibroMeter vs F4- 0.0004 0.0002 <10.sup.4 index LSE vs F4-index 0.511 0.120 0.133
(26) TABLE-US-00006 TABLE 4 OF EXAMPLE 1 Rate of patients included in the intervals of reliable diagnosis defined by the 90% negative (NPV) and positive (PPV) predictive values for significant fibrosis (Metavir F 2) and 95% predictive values for cirrhosis (Metavir F = 4), as a function of patient group and fibrosis test. Metavir F 2 Metavir F 3 Metavir F = 4 Fibrosis NPV PPV NPV + PPV NPV PPV NPV + PPV NPV PPV PPV NPV + PPV Set test 90% 90% 90% 90% 90% 90% 95% 90% 95% 95% Exploratory FibroMeter 3.2 52.4 55.6 41.8 0.0 41.8 65.9 0.0 0.0 65.9 (90.9) (89.6) (89.7) (89.7) () (89.7) (94.8) () () (94.8) Fibroscan 1.1 32.7 33.8 43.3 3.2 46.4 86.0 2.6 1.4 87.4 (100.0) (90.4) (90.7) (90.1) (90.9) (90.1) (94.7) (88.9) (100.0) (94.8) F 2-index .sup.a 9.2 46.1 55.3 44.7 5.2 49.9 87.7 3.2 2.0 89.7 (90.6) (90.1) (90.2) (89.7) (88.9) (89.7) (94.8) (90.9) (100.0) (94.9) Validation FibroMeter 1.2 47.6 48.8 47.0 0.0 47.0 64.2 0.0 0.0 64.2 (100.0) (72.0) (72.7) (94.2) () (94.2) (97.2) () () (97.2) Fibroscan 0.9 37.3 38.2 44.5 2.1 46.7 83.3 2.1 1.8 85.2 (100.0) (76.4) (77.0) (93.2) (100.0) (93.5) (93.1) (100.0) (100.0) (93.2) F 3-index .sup.a 7.6 41.5 49.1 51.2 7.3 58.5 85.2 2.4 2.1 87.3 (100.0) (82.5) (85.2) (95.3) (100.0) (95.9) (93.6) (100.0) (100.0) (93.8) All FibroMeter 2.2 50.1 52.3 44.3 0.0 44.3 65.1 0.0 0.0 65.1 (93.3) (81.5) (82.0) (92.0) () (92.0) (93.9) () () (95.9) Fibroscan 1.0 34.9 35.9 43.9 2.7 46.5 84.7 2.4 1.6 86.3 (100.0) (83.1) (83.6) (91.6) (94.4 (91.8) (93.9) (93.8) (700.0) (94.0) F4-index .sup.a 8.4 43.9 52.3 47.9 6.2 54.1 86.5 2.8 2.1 88.5 (94.7) (86.6) (87.9) (92.6) (95.2) (92.9) (94.2) (94.7) (100.0) (94.3) Cut-offs for NPV 90% and PPV 90% were calculated in the exploratory set and tested in the validation set and the whole population. Significant fibrosis. Cut-offs for NPV 90%: FibroMeter: 0.110, Fibroscan: 3.2, F 2-index: 0.248; cut-offs for PPV 90%: FibroMeter: 0.608, Fibroscan: 9.2, F 2-index: 0.784. Severe fibrosis. Cut-offs for NPV 90%: FibroMeter: 0.554, Fibroscan: 6.8, F 3-index: 0.220; cut-offs for PPV 90%: Fibroscan: 32.3, F 3-index: 0.870. Cirrhosis. Cut-offs for NPV 95%: FibroMeter: 0.757, Fibroscan: 14.5, F4-index: 0.244; cut-offs for PPV 90%: Fibroscan: 34.1, F4-index: 0.817; Cut-offs for PPV 95%: Fibroscan: 35.6, F4-index: 0.896. .sup.a SF-index for significant fibrosis, X-index for severe fibrosis, and C-index for cirrhosis.
(27) TABLE-US-00007 TABLE 5 OF EXAMPLE 1 Implementation of 3 new classifications for the non invasive diagnosis of fibrosis, derived from the interpretation of the interval of reliable diagnosis of several fibrosis tests (F 2- and F 3 indexes, F 3- and F 4 indexes, FibroMeter and Fibroscan). Reliable intervals of F 2-index F0/1 F1/2 F2 1 F 2 Reliable F 2 F0/1 F1/2 F1/2 intervals (29/32) (55/61) (50/63) of F 3-index F2 1 F2 1 F2/3 (32/32) (65/86) F 2 F 2 (54/57) F 3 F3/4 (16/18) Reliable F 3 F0/1 F1/2 F2 1 F2/3 intervals (29/32) (55/61) (92/95) (90/118) of F4-index F 2 F 2 (35/36) F4 F4 (7/7) Reliable F 2 F0/1 F1/2 F1/2 F2 intervals of (9/9) (68/74) (21/23) (23/45) LSE for F 3 F2 1 F1 F1/2 F2 1 F2/3 1/1 (23/26) (8/9) (43/48) F 2 F2 F2/3 F 2 (4/13) (6/9) (77/80) F 3 F 3 (10/10) The new classifications are depicted in italic (into brackets: rate of well classified patients in each class of the new classification according to liver biopsy results). Grey cells correspond to discordant results.
(28) TABLE-US-00008 TABLE 6 OF EXAMPLE 1 Diagnostic accuracies (% of well classified patients) of several fibrosis tests classifications as a function of patient group Set Explor- Valida- atory tion p All Classification Classification A 86.2 84.2 0.516 85.3 Classification B 88.3 82.4 0.038 85.4 Classification C 84.0 70.3 <10.sup.3 77.3 FibroMeter 67.6 69.7 0.575 68.7 Fibroscan .sup.a 54.4 63.3 0.024 58.7 Fibroscan .sup.b 45.0 59.0 <10.sup.3 51.8 Fibroscan .sup.c 46.1 59.0 10.sup.3 52.4 Fibroscan .sup.d 52.7 63.9 0.004 58.1 Fibrotest 33.5 43.9 0.005 38.8 p Classification A vs classification B 0.143 0.146 1.000 Classification A vs classification C 0.229 <10.sup.3 <10.sup.3 Classification A vs FibroMeter <10.sup.3 <10.sup.3 <10.sup.3 Classification A vs Fibroscan .sup.a <10.sup.3 <10.sup.3 <10.sup.3 Classification A vs Fibrotest <10.sup.3 <10.sup.3 <10.sup.3 Classification B vs classification C 0.032 <10.sup.3 <10.sup.3 Classification B vs FibroMeter <10.sup.3 <10.sup.3 <10.sup.3 Classification B vs Fibroscan .sup.a <10.sup.3 <10.sup.3 <10.sup.3 Classification B vs Fibrotest <10.sup.3 <10.sup.3 <10.sup.3 Classification C vs FibroMeter <10.sup.3 0.720 <10.sup.3 Classification C vs Fibroscan .sup.a <10.sup.3 0.049 <10.sup.3 Classification C vs Fibrotest <10.sup.3 <10.sup.3 <10.sup.3 FibroMeter vs Fibroscan .sup.a <10.sup.3 0.029 <10.sup.3 FibroMeter vs Fibrotest <10.sup.3 <10.sup.3 <10.sup.3 Fibroscan .sup.a vs Fibrotest <10.sup.3 <10.sup.3 <10.sup.3 .sup.a 6 classes (de Ledinghen, GCB 2008); .sup.b 4 classes (Ziol 2005), .sup.c 4 classes (Stebbing 2009 + 9.6 kPa pour F 3), .sup.d 3 classes (Stebbing 2009)
(29) TABLE-US-00009 TABLE 7 OF EXAMPLE 1 Management of patient for antiviral therapy according to the results of fibrosis tests classifications (rates of patients in the validation population, %) Liver biopsy result Management according Metavir F0/1 Metavir F 2 classification result .sup.a No treatment Treatment No treatment Treatment Well managed Classification A 41.5 12.1 9.7 36.7 78.2 Classification B 27.0 26.7 4.2 42.1 69.1 Classification C 33.9 19.7 7.3 39.1 73.0 FibroMeter 38.3 12.2 12.8 36.7 75.0 Fibroscan (VDL) 42.5 10.8 16.9 29.8 72.3 Fibroscan (Ziol) 42.5 10.8 16.9 29.8 72.3 Fibroscan (Steb 4 cl) 41.9 11.4 16.3 30.4 72.3 Fibroscan (Steb 3 cl) 41.9 11.4 16.3 30.4 72.3 Fibrotest 30.3 20.3 7.5 41.9 72.2 .sup.a Indication for antiviral therapy: Classifications A and B: F2 1; Classification C: F2; FibroMeter: F2/3; Fibroscan VDL: F2; Fibroscan Ziol and Stebbing 4 classes: F2; Fibroscan Stebbing 3cl: F2/3; Fibrotest: F2
Example 2
Patients
(30) 390 patients with chronic liver disease (CLD) hospitalized for a percutaneous liver biopsy at the University Hospitals of Angers and Bordeaux (France) were enrolled. 194 patients were included from April 2004 to June 2007 at the Angers site (group A, exploratory set), and 196 from September 2003 to April 2007 at the Bordeaux site (group B, validation set). Patients with the following cirrhosis complications were not included: ascites, variceal bleeding, systemic infection, and hepatocellular carcinoma. The non-invasive assessment of liver fibrosis by blood fibrosis tests and LSE was performed within one week prior to liver biopsy.
(31) Methods
(32) Histological Liver Fibrosis Assessment
(33) Percutaneous liver biopsy was performed using Menghini's technique with a 1.4-1.6 mm diameter needle. In each site, liver fibrosis was evaluated by a senior pathologist specialized in hepatology according to Metavir staging (with a consensus reading in Angers). Significant fibrosis was defined by Metavir stages F2. Liver fibrosis evaluation was considered as reliable when biopsy length was 15 mm and/or portal tract number8 (17).
(34) Fibrosis Blood Tests
(35) The following blood tests were calculated according to published formulas or patents: APRI, FIB-4, Fibrotest, Hepascore, and FibroMeter (FM). Cause-specific formulas were used for FibroMeter (9, 18, 19). All blood assays were performed in the same laboratories of each site. The inter-laboratory reproducibility was excellent for these tests (20).
(36) Liver Stiffness Evaluation
(37) LSE (FibroScan, EchoSens, Paris, France) was performed by an experienced observer (>50 LSE before the study), blinded for patient data. LSE conditions were those recommended by the manufacturer, as detailed elsewhere (21, 22). LSE was stopped when 10 valid measurements were recorded. The LSE result was expressed in kPa and corresponded to the median of all valid measurements performed within the LSE. Inter-quartile range (kPa) was defined as previously described (21).
(38) Statistical Analysis
(39) Quantitative variables were expressed as meanstandard deviation, unless otherwise specified. When necessary, diagnostic cut-off values of fibrosis tests were calculated according to the highest Youden index (sensitivity+specificity1). This technique allows maximizing the diagnostic accuracy with equilibrium between a high sensitivity and a high specificity by selecting an appropriate diagnostic cut-off. The diagnostic cut-off is here the values of blood test or LSE that distinguishes the patients as having or not the diagnostic target (significant fibrosis or cirrhosis).
(40) Accuracy of fibrosis testsThe performance of fibrosis tests was mainly expressed as the area under the receiver operating characteristic curve (AUROC). The reliable individual diagnosis was determined either by the traditional negative (NPV) and positive (PPV) predictive values, or by the recently described method of reliable diagnosis intervals (18) (see Appendix for precise definitions). AUROCs were compared by the Delong test (23).
(41) Synchronous combination of fibrosis testsCombinations of blood tests and LSE were studied in 3 populations: group A, B, and A+B. In each population, we performed a forward binary logistic regression using significant fibrosis determined on liver biopsy as the dependent variable, and blood fibrosis tests and LSE results as independent variables. Then, by using the regression score provided by the multivariate analysis, we implemented a new fibrosis test for the diagnosis of significant fibrosis. The same methodology was used for the diagnosis of cirrhosis.
(42) Sample sizeSample size was determined to show a significant difference for the diagnosis of significant fibrosis between FM and synchronous combination in the exploratory population. With risk: 0.05, risk: 0.20, significant fibrosis prevalence: 0.70, AUROC correlation: 0.70, and a bilateral test, the sample size was 159 patients for the following hypothesis of AUROC: FM: 0.84, synchronous combination: 0.90. The software programs used for statistical analyses were SPSS for Windows, version 11.5.1 (SPSS Inc., Chicago, Ill., USA) and SAS 9.1 (SAS Institute Inc., Cary, N.C., USA).
(43) Results
(44) Patients
(45) The characteristics of the 390 patients are summarized in Table 1 of Example 2. Mean age of patients was 52.4 years, 67.9% were male, and 74.4% had significant fibrosis. 89.5% of patients had a liver biopsy considered as reliable. Liver Stiffness Evaluation failure occurred in 12 patients (overall failure rate: 3.1%). Among the 390 patients included, 332 had all 5 blood tests and LSE available.
(46) TABLE-US-00010 TABLE 1 OF EXAMPLE 2 Patient characteristics at inclusion. Group All A B (n = 390) (n = 194) (n = 196) p .sup.a Age (years) 52.4 13.4 50.8 12.7 53.9 14.0 0.03 Male sex (%) 67.9 68.0 67.9 0.97 Cause of liver <10.sup.3 disease (%) Virus 48.7 54.1 43.4 Alcohol 27.2 26.3 28.1 NAFLD 4.9 9.8 0.0 Other 19.2 9.8 28.6 Metavir fibrosis <10.sup.3 stage (%) F0 7.2 4.1 10.2 F1 18.5 19.6 17.3 F2 23.1 26.3 19.9 F3 20.3 27.3 13.3 F4 31.0 22.7 39.3 <10.sup.3 Significant 74.4 76.3 72.4 0.39 fibrosis (%) Reliable biopsy (%) 89.5 95.3 82.6 <10.sup.3 IQR/LSE result <0.21 59.4 58.5 60.3 0.73 (%) IQR: interquartile range (kiloPascal) .sup.a By t-test or .sup.2 between the groups A and B
Diagnosis of Significant Fibrosis
Accuracy of Blood Tests and LSE (Table 2 of Example 2)
(47) LSE AUROC was significantly higher than that of Hepascore, FIB-4, and APRI for the diagnosis of significant fibrosis, and was not significantly different from FibroMeter and Fibrotest AUROCs.
(48) TABLE-US-00011 TABLE 2 OF EXAMPLE 2 AUROCs of blood tests and liver stiffness evaluation (LSE) as a function of diagnostic target, in the 332 patients having all 5 blood tests and LSE available. Significant fibrosis Cirrhosis AUROC: FibroMeter (FM) 0.836 0.834 Fibrotest (FT) 0.826 0.813 Hepascore (HS) 0.799 0.806 FIB-4 0.787 0.793 APRI 0.762 0.691 LSE 0.858 0.915 Comparison (p) .sup.a: FM vs FT 0.622 0.326 FM vs HS 0.074 0.101 FM vs FIB-4 0.030 0.078 FM vs APRI 0.004 <10.sup.3 FM vs LSE 0.417 <10.sup.3 FT vs HS 0.195 0.786 FT vs FIB-4 0.119 0.416 FT vs APRI 0.022 <10.sup.3 FT vs LSE 0.257 <10.sup.3 HS vs FIB-4 0.700 0.663 HS vs APRI 0.264 <10.sup.3 HS vs LSE 0.046 <10.sup.3 FIB-4 vs APRI 0.302 <10.sup.3 FIB-4 vs LSE 0.016 <10.sup.3 APRI vs LSE 0.003 <10.sup.3 .sup.a By Delong test
Synchronous Combination
(49) Combination of non-invasive tests (Table 3 of Example 2)In each of the three populations tested, significant fibrosis defined by liver biopsy was independently diagnosed by FibroMeter at the first step and Liver Stiffness Evaluation at the second step. The regression score provided by the binary logistic regression performed in group A (exploratory set) was: 3.6224.FM+0.4408.LSE result3.9850. This score was used to implement a diagnostic synchronous combination of FibroMeter and Liver Stiffness Evaluation called significant fibrosis-index (SF-index). This new fibrosis test was then evaluated in the validation sets: group B (Bordeaux center) and the pooled group A+B.
(50) TABLE-US-00012 TABLE 3 OF EXAMPLE 2 Fibrosis tests independently associated with significant fibrosis or cirrhosis defined by liver biopsy, as a function of patient group (A: Angers, B: Bordeaux). Significant fibrosis Cirrhosis Patient Independent Diagnostic Independent Diagnostic Group variables .sup.a p accuracy (%) .sup.b variables .sup.a p accuracy (%) .sup.b A 1. FibroMeter <10.sup.3 82.0 1. LSE <10.sup.3 89.7 2. LSE <10.sup.3 87.6 2. FibroMeter 0.031 88.7 B 1. FibroMeter <10.sup.3 78.2 1. LSE <10.sup.3 82.4 2. LSE 0.012 80.3 2. FibroMeter 0.017 83.0 All 1. FibroMeter <10.sup.3 80.6 1. LSE <10.sup.3 85.1 2. LSE <10.sup.3 85.3 2. FibroMeter 10.sup.3 86.1 LSE: liver stiffness evaluation; .sup.a Variables independently associated with significant fibrosis or cirrhosis with increasing order of step (the first step is the most accurate variable); .sup.b Cumulative diagnostic accuracy for the second step
Performance of SF-index (Table 4 of Example 2)SF-index AUROCs were not significantly different between groups A and B. SF-index AUROC was significantly higher than that of FibroMeter (FM) or Liver Stiffness Evaluation (LSE) in the whole population.
(51) TABLE-US-00013 TABLE 4 OF EXAMPLE 2 AUROCs of synchronous combinations (FM + LSE index). Significant fibrosis Cirrhosis Patient group All A B All A B AUROC: FibroMeter 0.834 0.839 0.843 0.835 0.822 0.839 LSE 0.867 0.889 0.850 0.923 0.931 0.922 FM + LSE index .sup.a 0.892 0.917 0.874 0.917 0.923 0.913 Comparison (p) .sup.b: FM vs LSE 0.162 0.150 0.839 <10.sup.3 10.sup.3 0.004 FM vs FM + LSE index <10.sup.3 <10.sup.3 0.210 <10.sup.3 <10.sup.3 <10.sup.3 LSE vs FM + LSE index 0.011 0.081 0.042 0.458 0.463 0.445 Comparison with those of FibroMeter (FM) and liver stiffness evaluation (LSE), as a function of diagnostic target and patient group (A: Angers, B: Bordeaux). .sup.a SF-index for significant fibrosis, C-index for cirrhosis .sup.b By Delong test
As shown on Table 4 of Example 2, SF-index inherited of the lowest misclassification rate provided by each single test in each fibrosis stage: the blood test in F0/1 stages, and LSE in F2 stages (see also
(52) Discordances between LSE and FMDiscordances between fibrosis tests for the diagnostic target were calculated according to the diagnostic cut-off determined by the highest Youden index. FM and LSE were concordant in 279 (73.0%) patients of whom 88.9% were correctly classified according to liver biopsy (F1: 77.0%, F2: 94.3%). FM and LSE were discordant in the 103 (27.0%) remaining patients of whom 68 (66.0%) were correctly classified by SF-index according to liver biopsy results (Table 5 of Example 2). Finally, SF-index correctly classified 316 (82.7%) patients and improved correct classification (i.e., discordances between FM and LSE resolved by SF-index) in 33 (8.6%) patients.
(53) Moreover, the SF-index resolved 66% of discordant cases between the blood test and LSE (Table 5 of Example 2).
(54) TABLE-US-00014 TABLE 5 OF EXAMPLE 2 Discordances. Impact of FM + LSE Patients (n) according to Classification by fibrosis tests .sup.a index on classification diagnostic target studied FM + LSE index .sup.b FM and LSE .sup.c by FM and LSE F 2 F4 Correct Both incorrect Favorable 0 0 Discordant 68 54 Both correct Neutral 248 275 Incorrect Both incorrect 31 28 Discordant Unfavorable 35 25 Both correct 0 0 Net improvement 33 .sup.d (8.6%) 29 .sup.e (7.6%) Impact of FM + LSE index on discordances between FibroMeter (FM) and liver stiffness evaluation (LSE) for the diagnosis of significant fibrosis or cirrhosis in the whole population. .sup.a Respective diagnostic cut-off values used for significant fibrosis or cirrhosis, according to the highest Youden index: FM: 0.538 and 0.873; LSE: 6.9 and 13.0 kiloPascals; FM + LSE index: 0.753 (SF-index) and 0.216 (C-index) .sup.b Classification by SF-index for significant fibrosis or C-index for cirrhosis expressed as correct or incorrect according to liver biopsy. .sup.c Classification of both tests based on liver biopsy. Discordant means than one test is correct and the other one is incorrect. .sup.d Favorable (68) unfavorable (35) effect = improvement (33) .sup.e Favorable (54) unfavorable (25) effect = improvement (29)
Methods Reliably Classifying 100% of Patients
(55) New sequential algorithmSF-index included significantly more patients than FM or LSE in the classical intervals of 90% predictive values (see Appendix for precise definition), especially in the 90% NPV interval (Table 6 of Example 2). By using SF-index with 90% predictive values in 81.7% of patients and liver biopsy required in the remaining 18.3% of patients, a correct diagnosis of significant fibrosis based on liver biopsy was obtained in 91.9% of patients (Table 6 of Example 2). This two-step sequential algorithm was called Angers SF-algorithm (
(56) Reliable diagnosis intervals of SF-indexWith this recently described method (18), accuracy is made 90% in the interval(s) between the previous intervals of 90% predictive values by changing the diagnostic target. The interest is to offer a reliable diagnosis for all patients. In the indeterminate interval determined by the 90% predictive values of SF-index, the proportion of Metavir fibrosis stages was F0: 20.0%, F1: 40.0%, and F2: 32.9% according to LIVER BIOPSY (
(57) Comparison of algorithms (Table 7 of Example 2)We compared the Angers SF-algorithm to those previously published in Bordeaux (24) and in Padova (16). The population tested was the 332 patients having Fibrotest, FibroMeter, APRI, and LSE available. The Padova algorithm had significantly higher accuracy (95.2%) compared to other algorithms due to a significantly higher rate of required LB. The Angers algorithm had a significantly lower rate of required liver biopsy compared to other algorithms. Thus, Angers SF-algorithm had the best compromise between high correct classification and low liver biopsy requirement, reflected by a much lower liver biopsy/accuracy ratio.
(58) Diagnosis of Cirrhosis
(59) Accuracy of Blood Tests and LSE (Table 2 of Example 2)
(60) LSE had a significantly higher AUROC than the blood tests for the diagnosis of cirrhosis.
(61) Synchronous Combination
(62) Combination of non-invasive tests (Table 3 of Example 2)The most accurate combination of fibrosis tests for the diagnosis of cirrhosis was LSE+FM. The regression score provided by the binary logistic regression performed in the group A (exploratory set) was: 0.1162.LSE result+1.9714.FM4.6616. This score was used to implement a diagnostic synchronous combination of LSE and FM called cirrhosis-index (C-index). This new fibrosis test was then evaluated in the validation sets: group B (Bordeaux center) and the pooled group A+B.
(63) Performance of C-index (Table 4 of Example 2)C-index AUROCs were not significantly different between groups A and B. In each group tested, C-index had a significantly higher AUROC than FM, but the difference with the LSE AUROC was not significant.
(64) Discordances between LSE and FMFM and LSE were concordant in 303 (79.3%) patients of whom 90.8% were correctly classified according to LIVER BIOPSY (F3: 94.7%, F4: 82.1%). FM and LSE were discordant in the 79 (20.7%) remaining patients of whom 54 (68.4%) were correctly classified by C-index according to LIVER BIOPSY results (Table 5 of Example 2). Finally, C-index correctly classified 329 (86.1%) patients and improved correct classification (i.e., discordances between FM and LSE resolved by C-index) in 29 (7.6%) patients.
(65) Methods Reliably Classifying 100% of Patients
(66) New sequential algorithm (Table 6 of Example 2)The C-index included significantly more patients than FM or LSE in the classical intervals of 90% predictive values. By using C-index with 90% predictive values in 90.6% of patients and liver biopsy required in the remaining 9.4% of patients, a correct diagnosis of cirrhosis based on liver biopsy was obtained in 91.1% of patients (Table 6 of Example 2). This two-step sequential algorithm was called Angers C-algorithm (
(67) Reliable diagnosis intervals of C-indexIn the indeterminate interval determined by the 90% predictive values of C-index, the proportion of Metavir fibrosis stages was F2: 11.1%, F3: 22.2%, and F4: 58.3% according to liver (
(68) TABLE-US-00015 TABLE 6 OF EXAMPLE 2 New sequential algorithm. Rates of patients included and correctly classified by fibrosis tests in the intervals of 90% predictive values for the diagnosis of significant fibrosis or cirrhosis in the whole population, as a function of fibrosis test. Rate (%) of patients included in the intervals Diagnostic defined by 90% predictive values Accuracy (%) target Fibrosis test 90% NPV Indeterminate .sup.a 90% PPV Fibrosis test .sup.b Algorithm .sup.c Significant FibroMeter 0.3 36.4 63.4 57.3 93.7 fibrosis LSE 0.5 28.8 70.7 64.1 92.9 (F 2) SF-index 8.1 18.3 73.6 73.6 91.9 Cirrhosis FibroMeter 44.2 42.1 13.6 52.1 94.2 (F4) LSE 68.3 12.6 19.1 78.8 91.4 C-index 70.4 9.4 20.2 81.7 91.1 .sup.a Proportion of patients for whom diagnosis remains uncertain (NPV and PPV < 90%), thus requiring a liver biopsy. Comparison of patient rates by McNemar test. Significant fibrosis: LSE vs FibroMeter: p = 0.006, SF-index vs FibroMeter or LSE: p < 10.sup.3; cirrhosis: FibroMeter vs C-index or LSE: p < 10.sup.3, C-index vs LSE: p = 0.02. .sup.b Rate of patients correctly classified by the intervals of 90% (negative and positive) predictive values, among the whole population. Comparison of patient rates by McNemar test. Significant fibrosis: LSE vs FibroMeter: p = 0.005, SF-index vs FibroMeter or LSE: p < 10.sup.3; cirrhosis: FibroMeter vs C-index or LSE: p < 10.sup.3, C-index vs LSE: p = 0.007. .sup.c Algorithm is defined by a two-step procedure: the fibrosis test is initially used with the interval of 90% predictive values, and a liver biopsy is subsequently required for patients included in the interval of indeterminate diagnosis. Thus, algorithm accuracy is calculated as the sum of patients correctly classified by the fibrosis test in the whole population (4.sup.th result column) and liver biopsy requirement (2.sup.nd result column) where accuracy is 100% by definition. Comparison of rates by McNemar test between FibroMeter and C-index for cirrhosis: p = 0.04, others: p: NS.
(69) Comparison of sequential algorithms (Table 7 of Example 2)The Bordeaux algorithm had significantly higher accuracy for cirrhosis compared to other algorithms. However, Angers C-algorithm had a significantly lower rate of required liver biopsy compared to other algorithms. Thus, as for significant fibrosis, Angers C-algorithm had the best compromise between high correct classification and low liver biopsy requirement, reflected by a much lower liver biopsy/accuracy ratio.
(70) TABLE-US-00016 TABLE 7 OF EXAMPLE 2 Comparison of accuracies and liver biopsy (LB) requirements between sequential algorithms of Angers (present study), Bordeaux (24), and Padova (16), for the diagnosis of significant fibrosis or cirrhosis. Algorithm accuracy (%) Diagnostic Blood test All LB/accuracy target Algorithm accuracy (%) .sup.a LB (%) .sup.b causes .sup.c Virus Other ratio .sup.d Significant Angers SF 89.8 20.2 91.9 92.2 91.5 0.22 fibrosis Bordeaux 86.5 28.6 90.4 88.8 92.2 0.33 Padova 91.1 46.1 95.2 95.0 95.4 0.51 Cirrhosis Angers C 90.0 9.3 91.0 93.9 87.6 0.10 Bordeaux 92.3 25.3 94.3 94.4 94.1 0.27 Padova 81.1 20.5 84.9 86.0 83.7 0.25 Population tested is the 332 patients having FibroMeter, Fibrotest, APRI and LSE available together. Grey cells indicate the most important results. .sup.a Accuracy (%) of blood tests included in patients without liver biopsy whose proportion can be deduced from the following column. Paired comparison was not possible. .sup.b Rate (%) of liver biopsy required by the algorithm. Comparison of rates by McNemar test. Significant fibrosis: Angers vs Bordeaux: p = 0.02, Padova vs Angers or Bordeaux: p < 10.sup.3; cirrhosis: Angers vs Bordeaux or Padova: p < 10.sup.3; Bordeaux vs Padova: p = 0.129. .sup.c Comparison of patient rates by McNemar test. Significant fibrosis: Padova vs Angers: p = 0.02, or Bordeaux: p = 0.007; Angers vs Bordeaux: p = 0.50; cirrhosis: Bordeaux vs Angers: p = 0.04, or Padova: p < 10.sup.3; Angers vs Padova: p = 0.007. .sup.d Ratio: rate of required liver biopsy (2.sup.nd result column)/blood test accuracy (1.sup.st result column).
REFERENCES
(71) 1. Oberti F, Valsesia E, Pilette C, Rousselet M C, Bedossa P, Aube C, et al. Noninvasive diagnosis of hepatic fibrosis or cirrhosis. Gastroenterology 1997; 113:1609-1616. 2. 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; 14:1133-1141. 3. Afdhal N H, Nunes D. Evaluation of liver fibrosis: a concise review. Am J Gastroenterol 2004; 99:1160-1174. 4. Sebastiani G, Alberti A. Non invasive fibrosis biomarkers reduce but not substitute the need for liver biopsy. World J Gastroenterol 2006; 12:3682-3694. 5. Wai C T, Greenson J K, Fontana R J, Kalbfleisch J D, Marrero J A, Conjeevaram H S, et al. A simple noninvasive index can predict both significant fibrosis and cirrhosis in patients with chronic hepatitis C. Hepatology 2003; 38:518-526. 6. Sterling R K, Lissen E, Clumeck N, Sola R, Correa M C, Montaner J, et al. Development of a simple noninvasive index to predict significant fibrosis in patients with HIV/HCV coinfection. Hepatology 2006; 43:1317-1325. 7. Imbert-Bismut F, Ratziu V, Pieroni L, Charlotte F, Benhamou Y, Poynard T. Biochemical markers of liver fibrosis in patients with hepatitis C virus infection: a prospective study. Lancet 2001; 357:1069-1075. 8. Rosenberg W M, Voelker M, Thiel R, Becka M, Burt A, Schuppan D, et al. Serum markers detect the presence of liver fibrosis: a cohort study. Gastroenterology 2004; 127:1704-1713. 9. Cales P, Oberti F, Michalak S, Hubert-Fouchard I, Rousselet M C, Konate A, et al. A novel panel of blood markers to assess the degree of liver fibrosis. Hepatology 2005; 42:1373-1381. 10. Patel K, Gordon S C, Jacobson I, Hezode C, Oh E, Smith K M, 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; 41:935-942. 11. Adams L A, Bulsara M, Rossi E, DeBoer B, Speers D, George J, et al. Hepascore: an accurate validated predictor of liver fibrosis in chronic hepatitis C infection. Clin Chem 2005; 51:1867-1873. 12. Sandrin L, Fourquet B, Hasquenoph J M, Yon S, Fournier C, Mal F, et al. Transient elastography: a new noninvasive method for assessment of hepatic fibrosis. Ultrasound Med Biol 2003; 29:1705-1713. 13. Bourliere M, Penaranda G, Ouzan D, Renou C, Botta-Fridlund D, Tran A, et al. Optimized stepwise combination algorithms of non-invasive liver fibrosis scores including Hepascore in hepatitis C virus patients. Aliment Pharmacol Ther 2008; 28:458-467. 14. Bourliere M, Penaranda G, Renou C, Botta-Fridlund D, Tran A, Portal I, et al. Validation and comparison of indexes for fibrosis and cirrhosis prediction in chronic hepatitis C patients: proposal for a pragmatic approach classification without liver biopsy. J Viral Hepat 2006; 13:659-670. 15. Sebastiani G, Vario A, Guido M, Alberti A. Sequential algorithms combining non-invasive markers and biopsy for the assessment of liver fibrosis in chronic hepatitis B. World J Gastroenterol 2007; 13:525-531. 16. Sebastiani G, Vario A, Guido M, Noventa F, Plebani M, Pistis R, et al. Stepwise combination algorithms of non-invasive markers to diagnose significant fibrosis in chronic hepatitis C. J Hepatol 2006; 44:686-693. 17. Nousbaum J B, Cadranel J F, Bonnemaison G, Bourliere M, Chiche L, Chor H, et al. Clinical practice guidelines on the use of liver biopsy. Gastroenterol Clin Biol 2002; 26:848-878. 18. Cales P, De Ledinghen V, Halfon P, Bacq Y, Leroy V, Boursier J, et al. Evaluating accuracy and increasing the reliable diagnosis rate of blood tests for liver fibrosis in chronic hepatitis C. Liver Int 2008; 28:1352-1362. 19. Cales P, Laine F, Boursier J, Deugnier Y, Moal V, Oberti F, et al. Comparison of blood tests for liver fibrosis specific or not to NAFLD. J Hepatol 2009; 50:165-173. 20. Cales P, Veillon P, Konate A, Mathieu E, Ternisien C, Chevailler A, et al. Reproducibility of blood tests of liver fibrosis in clinical practice. Clin Biochem 2008; 41:10-18. 21. Boursier J, Konate A, Gorea G, Reaud S, Quemener E, Oberti F, et al. Reproducibility of liver stiffness measurement by ultrasonographic elastometry. Clin Gastroenterol Hepatol 2008; 6:1263-1269. 22. Boursier J, Konate A, Guilluy M, Gorea G, Sawadogo A, Quemener E, et al. Learning curve and interobserver reproducibility evaluation of liver stiffness measurement by transient elastography. Eur J Gastroenterol Hepatol 2008; 20:693-701. 23. DeLong E R, DeLong D M, Clarke-Pearson D L. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988; 44:837-845. 24. Castera 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; 128:343-350. 25. Lucidarme D, Foucher J, Le Bail B, Vergniol J, Castera L, Duburque C, et al. Factors of Accuracy of Transient Elastography (Fibroscan) for the Diagnosis of Liver Fibrosis in Chronic Hepatitis C. Hepatology 2009; in press. 26. Bossuyt P M, Reitsma J B, Bruns D E, Gatsonis C A, Glasziou P P, Irwig L M, et al. The STARD statement for reporting studies of diagnostic accuracy: explanation and elaboration. Clin Chem 2003; 49:7-18. 27. Halfon P, Bacq Y, De Muret A, Penaranda G, Bourliere M, Ouzan D, et al. Comparison of test performance profile for blood tests of liver fibrosis in chronic hepatitis C. J Hepatol 2007; 46:395-402. 28. Boursier J, Bacq Y, Halfon P, Leroy V, De Ledinghen V, De Muret A, et al. Improved diagnostic accuracy of blood tests for severe fibrosis and cirrhosis in chronic hepatitis C. Eur J Gastroenterol Hepatol 2009; 21:28-38. 29. Castera L. Use of elastometry (FibroScan) for the non-invasive staging of liver fibrosis. Gastroenterol Clin Biol 2007; 31:524-530. 30. Fraquelli M, Rigamonti C, Casazza G, Conte D, Donato M F, Ronchi G, et al. Reproducibility of transient elastography in the evaluation of liver fibrosis in patients with chronic liver disease. Gut 2007; 56:968-973. 31. Friedrich-Rust M, Ong M F, Martens S, Sarrazin C, Bojunga J, Zeuzem S, et al. Performance of transient elastography for the staging of liver fibrosis: a meta-analysis. Gastroenterology 2008; 134:960-974. 32. Cales P, Boursier J, Rousselet M C, Michalak S, Oberti F, Gallois Y, et al. Comparison of reproducibility of histology, blood tests and Fibroscan for liver fibrosis. Hepatology 2007; 46:834A. 33. Cales P, Boursier J, De Ledinghen V, Halfon P, Bacq Y, Leroy V, et al. Evaluation and improvement of a reliable diagnosis of cirrhosis by blood tests. Gastroenterol Clin Biol 2008; 32:1050-1060. 34. Castera L, Sebastiani G, Le Bail B, De Ledinghen V, Couzigou P, Alberti A. Prospective comparison of two algorithms combining non invasive tests for staging of liver fibrosis in chronic hepatitis C. Hepatology 2007; 46:320A. 35. Sebastiani G, Halfon P, Castera L, Mangia A, Di Marco V, Pirisi M, et al. Large-scale multicenter comparison of three algorithms combining serum non-invasive markers for liver fibrosis in chronic hepatitis C. J Hepatol 2008; 48:S282. 36. Poynard T, Halfon P, Castera L, Munteanu M, Imbert-Bismut F, Ratziu V, et al. Standardization of ROC curve areas for diagnostic evaluation of liver fibrosis markers based on prevalences of fibrosis stages. Clin Chem 2007; 53:1615-1622. 37. Lambert J, Halfon P, Penaranda G, Bedossa P, Cacoub P, Carrat F. How to measure the diagnostic accuracy of noninvasive liver fibrosis indices: the area under the ROC curve revisited. Clin Chem 2008; 54:1372-1378. 38. Obuchowski N A. An ROC-type measure of diagnostic accuracy when the gold standard is continuous-scale. Stat Med 2006; 25:481-493.