Multi-targeted fibrosis tests
11605460 · 2023-03-14
Assignee
Inventors
Cpc classification
G16B40/00
PHYSICS
G01N2800/60
PHYSICS
G16H50/20
PHYSICS
G16H50/70
PHYSICS
G06N7/00
PHYSICS
G01N2800/085
PHYSICS
International classification
G16H50/20
PHYSICS
G06N7/00
PHYSICS
G16B40/00
PHYSICS
Abstract
Disclosed is a non-invasive method for assessing in a subject the presence and severity of a liver lesion, or the risk of death or liver-related events, including: 1) performing at least three binary logistic regressions on at least one variable, performed on the same variable(s) but each directed to a different single diagnostic target, thereby obtaining at least three scores; 2) combining the scores from step 1) in a multiple linear regression to obtain a new multi-targeted score; 3) optionally sorting the multi-targeted score obtained in step 2) in a classification of liver lesion stages or grades, thereby determining to which liver lesion stage or grade the subject belongs based on his/her multi-targeted score. Also disclosed is a single multi-targeted non-invasive test obtained by the combination of single-targeted non-invasive tests providing a unique score and a unique classification with improved accuracy compared to single-targeted diagnostic tests.
Claims
1. A method for treating an individual suffering from a liver fibrosis, comprising: determining in the individual the presence and severity of a liver fibrosis by: 1) performing at least 3 binary logistic regressions, or at least 3 other statistical analyses selected from linear discriminant analyses and multivariate analyses, on at least one variable, wherein the binary logistic regressions, or other statistical analyses, are performed on the same variable(s) but are each directed to a different single diagnostic target, thereby obtaining at least 3 scores; 2) combining the at least 3 scores obtained in step 1) in a multiple linear regression to obtain a new multi-targeted score, thereby determining the presence and severity of a liver fibrosis in the individual; and implementing an adapted patient care depending on the severity of the liver fibrosis, comprising: monitoring the individual by assessing the liver fibrosis severity at regular intervals, administering without delay at least one therapeutic agent to the individual, wherein said at least one therapeutic agent is an antifibrotic agent selected from the group consisting of simtuzumab, GR-MD-02, stem cell transplantation, Phyllanthus urinaria, Fuzheng Huayu, S-adenosyl-L-methionine, S-nitrosol-N-acetylcystein, silymarin, phosphatidylcholine, N-acetylcysteine, resveratrol, vitamin E, losartan, telmisartan, naltrexone, RF260330, sorafenib, imatinib mesylate, nilotinib, INT747, FG-3019, oltipraz, pirfenidone, halofuginone, polaorezin, gliotoxin, sulfasalazine, rimonabant, and combinations thereof, or wherein said at least one therapeutic agent is for treating the underlying cause responsible for the liver fibrosis, and/or starting a complication screening program for applying early prophylactic or curative treatment.
2. The method according to claim 1, wherein determining in the individual the presence and severity of a liver fibrosis further comprises a third step of sorting the multi-targeted score obtained in step 2) in a classification of liver fibrosis stages, thereby determining to which liver fibrosis stage the individual belongs based on his/her multi-targeted score.
3. The method according to claim 1, wherein determining in the individual the presence and severity of a liver fibrosis is carried out by: 1) performing at least 3 binary logistic regressions, or at least 3 other statistical analyses selected from linear discriminant analyses and multivariate analyses, on at least one variable, wherein the binary logistic regressions, or other statistical analyses, are performed on the same variable(s) but are each directed to a different single diagnostic target, thereby obtaining at least 3 scores; 1a) performing at least another binary logistic regression including the at least 3 scores obtained at step 1), wherein the diagnostic target of said binary logistic regression is a clinically relevant binary target, thereby identifying the significant single-targeted scores among those obtained by the binary logistic regressions, or other statistical analyses, of step 1), said significant single-targeted scores being independently associated with said clinically relevant binary diagnostic target; 1b) deriving a classification of liver fibrosis stages for each of the single-targeted binary logistic regressions, or other statistical analyses, found significant in step 1a); 1c) combining the classifications of step 1b) into a multi-targeted classification of liver fibrosis stages; and 2) combining the significant scores identified in step 1a) in a multiple linear regression to obtain a single multi-targeted score, thereby determining the presence and severity of a liver fibrosis in the individual.
4. The method according to claim 1, wherein step 1) comprises performing at least 3 binary logistic regressions.
5. The method according to claim 1, wherein step 1) comprises performing 4 binary logistic regressions, each targeting a different Metavir fibrosis stage corresponding to F1, F2, F3, and F4 stages.
6. The method according to claim 1, wherein step 1) comprises performing 7 binary logistic regressions, each with a different fibrosis target corresponding to Metavir fibrosis stages F≥1 (F≥1 vs. F0), F≥2 (F≥2 vs. F≤1), F≥3 (F≥3 vs. F≤2), F4 (F4 vs. F≤3), F1 vs. F0+F2+F3+F4, F2 vs. F0+F1+F3+F4, and F3 vs. F0+F1+F2+F4.
7. The method according to claim 1, wherein step 1) comprises performing 10 binary logistic regressions, each with a different fibrosis target corresponding to Metavir fibrosis stages F≥1 vs. F=0, F≥2 vs. F≤1, F≥3 vs. F≤2, F=4 vs. F≤3, F1 vs. F0+F2+F3+F4, F2 vs. F0+F1+F3+F4, F3 vs. F0+F1+F2+F4, F1+F2 vs. F0+F3+F4, F2+F3 vs. F0+F1+F4, and F1+F2+F3 vs. F0+F4.
8. The method according to claim 1, wherein the binary logistic regressions of step 1) are performed on at least one variable selected from biomarkers, clinical markers, qualitative markers, data obtained by a physical method of diagnosis, scores of fibrosis tests, descriptors of at least one image of the liver tissue of the individual previously obtained by an imaging method, and mathematical combinations thereof.
9. The method according to claim 8, wherein the binary logistic regressions of step 1) are performed on at least two descriptors of at least one image of the liver tissue of the individual previously obtained by an imaging method, said descriptors being selected from the group consisting of linearity percentage of the edges, mean of percentage of fibrosis around areas, area of stellar fibrosis among the total surface of the liver biopsy specimen, number of bridges, bridges thickness, mean area of porto-septal regions, bridges perimeter, ratio of bridges among the porto-septal areas, area of fibrosis in the bridges, fractal dimension of peri-sinusoidal fibrosis, perimeter of the organ, tissue or fragment thereof, fractal dimension of porto-septal fibrosis, ratio of peri-sinusoidal fibrosis among the whole fibrosis, length of the organ, tissue or fragment thereof, anfractuosity descriptors including native perimeter, smoothed perimeter and ratio between both perimeters, fractal dimension of fibrosis, interquartile range of total density, Arantius furrow thickness, mean native liver perimeter, mean total spleen perimeter, ratio spleen surface to liver surface, and mathematic combinations thereof.
10. The method according to claim 8, wherein the binary logistic regressions of step 1) are performed on at least one data obtained by a physical method of diagnosis, said physical method of diagnosis being an elastography method selected from Vibration Controlled Transient Elastography (VCTE) also known as Fibroscan, Acoustic Radiation Force Impulse (ARFI) imaging, supersonic shear imaging (SSI) elastometry, and NMR/MRI (nuclear magnetic resonance/magnetic resonance imaging) elastography.
11. The method according to claim 8, wherein the binary logistic regressions of step 1) are performed on at least one data obtained by a physical method of diagnosis, said physical method of diagnosis being a radiography method selected from X-ray, ultrasonography, computerized scanner, magnetic resonance imaging (MRI), functional magnetic resonance imaging, tomography, computed axial tomography, proton emission tomography (PET), single photon emission computed tomography, and tomodensitometry.
12. The method according to claim 8, wherein the binary logistic regressions of step 1) are performed on at least one score of fibrosis test obtained with a fibrosis test selected from APRI, FIB4, Fibrotest, ELF score, FibroMeter, Fibrospect, Hepascore, Zeng score, and NAFLD fibrosis score, wherein said fibrosis test comprises the combination in a simple mathematical function or a binary logistic regression of markers selected from biological markers and/or clinical markers.
13. The method according to claim 1, wherein the binary logistic regressions of step 1) correspond to a fibrosis test selected from the FibroMeter family of fibrosis tests and combinations thereof with Vibration Controlled Transient Elastography (VCTE) also known as Fibroscan.
14. The method according to claim 1, wherein the individual suffers from a liver condition selected from the group consisting of a liver impairment, a chronic liver disease, a hepatitis viral infection especially an infection caused by hepatitis B, C or D virus, a hepatotoxicity, a liver cancer, a steatosis, a non-alcoholic fatty liver disease (NAFLD), a non-alcoholic steato-hepatitis (NASH), an autoimmune disease, a metabolic liver disease, and a disease with secondary involvement of the liver.
Description
BRIEF DESCRIPTION OF THE DRAWINGS
(1)
(2)
(3)
(4)
EXAMPLES
(5) The present invention is further illustrated by the following examples.
Example 1: Multi-Targeted FibroMeter Constructed for Multi-Target Score (MFMs)
(6) Patients and Methods
(7) Populations
(8) A total of 2589 patients were initially included in the present study. The multi-target diagnostic test was developed using data from 1012 patients (derivation population), and an external validation was performed in 1577 patients (validation populations #1 to #5). Additional data were obtained with a validation population comprising 1220 patients suffering from chronic liver diseases with different etiologies (validation population #6). The overall population thus included 3809 patients.
(9) Derivation Population
(10) The derivation population included 1012 patients with CHC [4]. Thus, individual patient data were available from five centers, independent for study design, patient recruitment, blood marker determination and liver histology interpretation by an expert pathologist.
(11) Validation Populations
(12) Diagnostic populations—The validation population #1 included 641 patients with chronic hepatitis C (CHC) [5, 6]. The validation population #2 for chronic hepatitis B (CHB) was extracted from a previously published database [7] and included 152 patients all with chronic hepatitis (30.4% HBe Ag positive); inactive carriers of HBs Ag were excluded. The validation population #3 included 444 patients with CHC and (HIV) infection prospectively included from April 1997 to August 2007 if they had anti-HCV (hepatitis C virus) and anti-HIV (human immunodeficiency virus) antibodies, and HCV RNA in serum [8]. Population #4 comprised 225 patients with biopsy-proven nonalcoholic fatty liver disease (NAFLD) consecutively included in the study from January 2002 to March 2013 at Angers University Hospital and from September 2005 to July 2011 at Pessac University Hospital. NAFLD was defined as liver steatosis on liver biopsy after exclusion of concomitant steatosis-inducing drugs (such as corticosteroids, tamoxifen, amiodarone or methotrexate), excessive alcohol consumption (>210 g/week in men or >140 g/week in women), chronic hepatitis B or C infection, and histological evidence of other concomitant chronic liver disease (CLD). Patients were excluded if they had cirrhosis complications (ascites, variceal bleeding, systemic infection, or hepatocellular carcinoma). Population #5 included 115 patients with alcoholic liver disease (ALD) extracted from a database used in previously published works [9]. Population #6 included 1220 patients with different chronic liver disease (CLD) etiologies: CHC: 41.3%, NAFLD: 31.3%, alcohol, pure (ALD): 8.1% or mixed: 11.7%, CHB: 5.7%, co-infections (HIV/CHC, HIV/CHB, CHB/VHD, others): 1.2%, others combinations of previous etiologies: 0.7%. These patients were consecutively included between 2011 and 2016 in Angers and Pessac centers and represent a more recent population of clinical practice where liver biopsy is more often indicated when blood tests and VCTE are discordant. Therefore, this population was separately considered.
(13) Diagnostic Methods
(14) Histological Assessment
(15) Liver biopsies were performed using Menghini's technique with a 1.4-1.6 mm diameter needle. Biopsy specimens were fixed in a formalin-alcohol-acetic solution and embedded in paraffin; 5 μm thick sections were then cut and stained with hematoxylin-eosin-saffron. Liver fibrosis was evaluated according to Metavir fibrosis (F) stages [10] by two senior experts with a consensus reading in case of discordance in Angers and in the Fibrostar study [11] (part of validation population #1), and by a senior expert in other centers. These liver specimen findings served as a reference for the liver fibrosis evaluation by non-invasive tests.
(16) FibroMeter Variables
(17) Biological markers were those previously used in blood tests carried out to diagnose different lesions in chronic viral hepatitis [9, 12]. The following biological markers were included: platelets, aspartate aminotransferase, hyaluronate, urea, prothrombin index, alpha2-macroglobulin as used in FibroMeter.sup.V2G [4, 9] plus gamma-glutamyl transpeptidase (GGT) used in FibroMeter.sup.V3G [12] and alanine aminotransferase used in InflaMeter targeted for liver activity [13]. Clinical markers were also included (age and sex as used in FibroMeter.sup.V2G). Thus, 10 variables were available. The new tests were constructed by including hyaluronate (second generation as for FibroMeter.sup.V2G) or not (third generation as for FibroMeter.sup.V3G). Reference blood tests for comparison with the new test were FibroMeter.sup.V2G or FibroMeter.sup.V3G, targeted for significant fibrosis (F≥2), and CirrhoMeter.sup.V2G or CirrhoMeter.sup.V3G, targeted for cirrhosis, with previously calculated classifications [14, 15].
(18) Non-Invasive Tests
(19) A total of 19 variables (4 clinical markers and 15 biological markers) were used in 17 tests (14 blood tests, 1 elastometry technique and 2 combined). Eleven tests had been constructed in CHC populations and five in other CLD causes (two in NAFLD and one each in ALD, CHB or HIV-HCV).
(20) Blood tests—Fibrotest [16], Hepascore [17], Fib-4 [18] and APRI [19] were calculated according to published or patented formulas. FibroMeter.sup.V2G [20], CirrhoMeter.sup.V2G [3], FibroMeter.sup.V3G [12] and CirrhoMeter.sup.V3G [12] were constructed for Metavir fibrosis staging in CHC. FibroMeter/CirrhoMeter.sup.V2G differs from FibroMeter/CirrhoMeter.sup.V3G in that the hyaluronate included in the former is replaced by GGT in the latter. CirrhoMeters were constructed for cirrhosis diagnosis and include all of the FibroMeter biomarkers [3]. The Zeng score was constructed in CHB [21]. FibroMeter.sup.ALD2G (second generation) [13] and FibroMeter.sup.NAFLD [22] were constructed for Metavir fibrosis staging respectively in ALD and NAFLD. NAFLD fibrosis score was constructed for NASH-CRN (or Kleiner) fibrosis staging in NAFLD [23]. This body of tests provided at least one test specific to each etiology. All blood assays were performed in the same laboratories of each center except for the Fibrostar study (part of population #1) where they were centralized. Tests were used as raw data with no correction rules (e.g., expert systems).
(21) Liver elastometry—Vibration-controlled transient elastometry or VCTE (Fibroscan, Echosens, Paris, France) was performed by an experienced observer (>50 examinations before the study), blinded for patient data. Examination conditions were those recommended by the manufacturer [24]. VCTE examination was stopped when 10 valid measurements were recorded. Results (kPa) were expressed as the median and the interquartile range of all valid measurements.
(22) Test Construction
(23) The primary objective of the study was to construct multi-targeted FibroMeters displaying a significant increase in diagnostic performance when compared to mono-targeted tests of the FibroMeter family. In particular, the aim was to obtain a multi-targeted test with Obuchowski index and area under the receiver operating characteristics (AUROC) for cirrhosis significantly superior to those of FibroMeter, and with AUROC for cirrhosis superior or equal to that of CirrhoMeter.
(24) The second objective of the study was to obtain multi-targeted FibroMeters displaying an improved diagnostic performance when compared to other fibrosis tests not belonging to the FibroMeter family, in particular Fibrotest, Hepascore, Zeng score and VCTE (also known as Fibroscan).
(25) The construction of the multi-target classification system was performed in 3 successive steps.
(26) Step 1: Single-target test construction—The single tests correspond to binary logistic regressions on the markers of the FibroMeter family of tests, in which said markers are combined as single markers, or as ratios of markers, or as arithmetic combinations of markers. These tests were built using a conventional binary logistic regression approach using as many diagnostic targets as possible by the five Metavir F stages. These targets were: fibrosis (F≥1), significant fibrosis (F≥2), severe fibrosis (F≥3), and cirrhosis (F=4). Four single-target tests were thus obtained. Six additional targets were obtained by binary targets using two cut-offs: e.g., F1 or F1+F2 or F1+F2+3 vs. other stages. The 6 additional targets were: F1 vs. F0+F2+F3+F4, F2 vs. F0+F1+F3+F4, F3 vs. F0+F1+F2+F4, F1+F2 vs. F0+F3+F4, F2+F3 vs. F0+F1+F4 and F1+F2+F3 vs. F0+F4. In total, ten single-target tests could thus be obtained.
(27) Step 2: Single-target test selection—Previous mono-targeted tests were included in stepwise multiple linear regression targeted for the five Metavir stages. Metavir stages were normalized to 1, i.e., divided by 4, to obtain a score between 0 and 1. This new score was called multi-targeted FibroMeter (MFM).
(28) Step 3: Multi-target test classification—Briefly, the correspondence between the previous MFM score and Metavir stages was derived according to the published method [20]. This optional step resulted in a classification including 6 fibrosis classes: 0/1 (corresponding to Metavir F0/1, 1/2 (F1/2), 2 (F2±1), 3 (F3±1), 3/4 (F3/4) and 4 (F4).
(29) Statistics
(30) Accuracy—The diagnostic accuracy of each test score was expressed with two descriptors. The main descriptor was the Obuchowski index (OI) [25] to better take into account differences in fibrosis stage prevalence between populations and thus limit spectrum bias. This index is a multinomial version of the AUROC adapted to ordinal references such as pathological fibrosis staging. With N (=5: F0 to F4) categories of the gold standard outcome and AUROCst, it estimates the AUROC of diagnostic tests differentiating between categories s and t. The OI is a weighted average of the N(N−1)/2 (=10) different AUROCst corresponding to all the pair-wise comparisons between two of the N categories. Additionally, the OI was assessed using a penalty function proportional to the difference in fibrosis stages, i.e., a penalty of 1 when the difference between stages was one, 2 when the difference was two, and so on. The reference prevalence was standardized according to the largest series of CHC with liver biopsies [26] to facilitate comparisons between etiologies. Thus, the result can be interpreted as the probability that the non-invasive test will correctly rank two randomly chosen patients with different fibrosis stages.
(31) The second descriptor for the diagnostic accuracy of test score was the AUROC, i.e., the classical index for binary diagnostic targets.
(32) The overall accuracy of classification tests was assessed by the rate of well-classified patients according to Metavir F.
(33) Optimism bias—By definition, optimism bias maximizes performance in the population where tests are constructed: this affected FibroMeter, CirrhoMeter and MFM in the derivation population and VCTE in the validation population #1 for its fibrosis classification. Thus, external validation was performed outside these populations.
(34) Sample size calculation—The size of the main populations (derivation and validation #1) was that necessary to detect a significant difference between two tests for the diagnosis of cirrhosis. With an a risk of 0.05, a 13 risk of 0.05, a cirrhosis prevalence of 0.12, an AUROC correlation of 0.82 and bilateral testing, the required sample size was 659 patients for the following expected AUROC values for cirrhosis: first test: 0.92, second test: 0.90 [3].
(35) Miscellaneous—Quantitative variables were expressed as mean±standard deviation. Data were reported according to STARD [27] and Liver FibroSTARD statements [28], and analyzed on an intention to diagnose basis. The main statistical analyses were performed under the control of professional statisticians (SB, GH) using SPSS version 18.0 (IBM, Armonk, N.Y., USA) and SAS 9.2 (SAS Institute Inc., Cary, N.C., USA).
(36) Results
(37) Population Characteristics
(38) The main characteristics of the studied populations are depicted in Table 2.
(39) TABLE-US-00004 TABLE 2 Population characteristics. Validation Populations Derivation #1 #2 #3 #4 #5 #6 Etiology CHC CHC CHB HIV/CHC NAFLD Alcohol Miscellaneous Patients (n) 1012 641 152 444 225 115 1220 Male (%) 59.6 60.5 81.5 68.7 65.3 64.3 67.3 Age (years) 45.4 ± 12.5 51.4 ± 11.2 40.0 ± 11.3 40.5 ± 5.8 56.5 ± 12.0 50.8 ± 23.9 50.7 ± 13.3 Body mass NA 24.8 ± 4.0 NA NA 31.3 ± 5.0 23.9 ± 4.2 29.2 ± 6.3 index (kg/m.sup.2) Metavir (%): F0 4.3 3.7 15.1 5.9 25.3 11.3 10.1 F1 43.3 38.7 44.1 24.3 37.3 14.8 32.5 F2 27.0 25.4 25.7 38.5 16.9 14.8 25.0 F3 13.9 18.4 6.6 19.6 15.6 7.0 17.5 F4 11.4 13.7 8.6 13.7 4.9 52.2 14.8 Score 1.85 ± 1.08 2.00 ± 1.13 1.49 ± 1.10 2.11 ± 1.10 1.37 ± 1.16 2.74 ± 1.49 1.94 ± 1.22 Significant 52.3 57.6 40.8 69.8 37.3 73.9 57.4 fibrosis (%) Biopsy 21.2 ± 7.9 24.4 ± 8.7 21.6 ± 7.4 21 ± 10 30.8 ± 12.0 NA 27.6 ± 11.4 length (mm) NA: not available
(40) Diagnostic Performance
(41) Derivation Population
(42) Multi-FibroMeters were only compared to mono-targeted FibroMeters in this population of 1012 CHC since performance was optimized due to optimism bias for all these tests and not for others. Main diagnostic indices are reported in Table 3 (see below). These diagnostic indices were similar between Multi-FibroMeter.sup.V2G and FibroMeter.sup.V2G (diagnostic target: significant fibrosis) for significant fibrosis or CirrhoMeter.sup.V2G (diagnostic target: cirrhosis) for cirrhosis, especially accuracies were not significantly different (details not shown). AUROCs for all diagnostic targets and Obuchowski indexes are listed in Table 4 below. As expected, Multi-FibroMeter.sup.V2G ranked first for all diagnostic targets (Table 4). Pairwise comparisons are detailed in Table 5 below for cirrhosis AUROC since this is the main binary diagnostic target and for Obuchowski indexes in Table 6 below since this reflects overall performance. Cirrhosis AUROCs of Multi-FibroMeters were higher than FibroMeters and CirrhoMeters: this improvement was significant vs. FibroMeters but not vs. CirrhoMeters so the objective was reached. Obuchowski indexes of Multi-FibroMeters were significantly improved vs. FibroMeters (objective reached) and CirrhoMeters (beyond the objective).
(43) TABLE-US-00005 TABLE 4 AUROCs for all diagnostic targets and Obuchowski indices for Metavir fibrosis (F) stages of multi-targeted FibroMeters vs. published mono-targeted FibroMeters in the CHC derivation population (1012 patients). Obuchowski AUROC index F ≥ 1 F ≥ 2 F ≥ 3 F = 4 Value Rank FibroMeter.sup.V2G 0.854 0.853 0.884 0.907 0.843 3 CirrhoMeter.sup.V2G 0.825 0.811 0.874 0.919 0.819 5 Multi-FibroMeter.sup.V2G 0.862 0.856 0.897 0.929 0.853 1 FibroMeter.sup.V3G 0.852 0.851 0.880 0.893 0.838 4 CirrhoMeter.sup.V3G 0.821 0.814 0.874 0.911 0.818 6 Multi-FibroMeter.sup.V3G 0.861 0.855 0.892 0.919 0.850 2 Best result per diagnostic target is indicated in bold.
(44) TABLE-US-00006 TABLE 3 Diagnostic indices of blood tests for significant fibrosis or cirrhosis in the CHC derivation population (1012 patients). 95% confidence intervals in parentheses. Likelihood Predictive value ratio Kappa Sensitivity Specificity % Posi- Nega- Diagnostic Accuracy Test Cut-off.sup.a index.sup.b (%) (%) Positive Negative tive tive odds ratio (%) AUROC.sup.c Significant fibrosis: FibroMeter.sup.V2G 0.4115 0.560 80.4 75.6 78.3 77.8 3.29 0.26 12.7 78.1 0.853 (0.507-0.609) (77.0-83.8) (71.7-79.4) (74.8-81.8) (74.1-81.6) (9.4-17.3) (75.5-80.6) (0.830-0.876) Multi- 0.399 0.567 81.9 74.7 78.0 79.0 3.24 0.24 13.3 78.5 0.85 FibroMeter.sup.V2G (0.515-0.616) (78.6-85.1) (70.9-78.6) (74.6-81.5) (75.3-82.7) (9.9-18.3) (75.9-81.0) (0.834-0.879) Cirrhosis: CirrhoMeter.sup.V2G 0.442 0.602 54.8 97.9 76.8 94.4 25.86 0.46 56.0 93.0 0.919 (0.502-0.684) (45.7-63.9) (96.9-98.8) (67.7-86.0) (92.9-95.9) (30.3-111.2) (91.4-94.6) (0.893-0.945) Multi- 0.748 0.603 60.0 96.7 69.7 95.0 17.94 0.41 43.4 92.5 0.929 FibroMeter.sup.V2G (0.527-0.680) (51.0-69.0) (95.5-97.8) (60.6-78.7) (93.5-96.4) (27.4-74.8) (90.9-94.1) (0.910-0.949) AUROC: area under the receiver operating characteristic. .sup.aDiagnostic cut-offs of blood tests were fixed a posteriori in this derivation population (maximum Youden index (= maximum accuracy) for significant fibrosis and maximum accuracy for cirrhosis). .sup.bKappa index reflecting agreement with liver specimen (all p < 0.001). .sup.cAUROC is independent of diagnostic cut-off.
(45) TABLE-US-00007 TABLE 5 Comparison of AUROCs for cirrhosis of multi-targeted FibroMeters and mono-targeted FibroMeters in the CHC derivation population (1012 patients in Table 3) by Delong test. FM.sup.2G CM.sup.2G MFMF.sup.2G FM.sup.3G CM.sup.3G MFMF.sup.3G FibroMeter.sup.V2G — 0.2316 8.10.sup.-6 0.0039 0.6764 0.0594 CirrhoMeter.sup.V2G — 0.2419 0.0280 0.0978 0.9864 Multi-FibroMeter.sup.V2G — 4.10.sup.-8 0.0459 0.0342 FibroMeter.sup.V3G — 0.0945 2.10.sup.−6 CirrhoMeter.sup.V3G — 0.3800 Multi-FibroMeter.sup.V3G — FM.sup.2G: FibroMeter.sup.V2G, CM.sup.2G: CirrhoMeter.sup.V2G, MFM.sup.2G: multi-targeted FibroMeter.sup.V2G, FM.sup.3G: FibroMeter.sup.V3G, CM.sup.3G: CirrhoMeter.sup.V3G, MFM.sup.3G: multi-targeted FibroMeter.sup.V3G Significant differences are indicated in bold.
(46) TABLE-US-00008 TABLE 6 Comparison of Obuchowski indices of multi-targeted FibroMeters and mono-targeted FibroMeters in the CHC derivation population (1012 patients in Table 3) by z test. FM.sup.2G CM.sup.2G MFMF.sup.2G FM.sup.3G CM.sup.3G MFMF.sup.3G FibroMeter.sup.V2G — 0.0005 1.10.sup.−5 0.0344 0.0003 0.0152 CirrhoMeter.sup.V2G — 4.10.sup.−7 0.0068 0.6657 1.10.sup.−5 Multi-FibroMeter.sup.V2G — 2.10.sup.−7 4.10.sup.−7 0.0574 FibroMeter.sup.V3G — 0.0017 1.10.sup.−6 CirrhoMeter.sup.V3G — 3.10.sup.−6 Multi-FibroMeter.sup.V3G — FM.sup.2G: FibroMeter.sup.V2G, CM.sup.2G: CirrhoMeter.sup.V2G, MFM.sup.2G: multi-targeted FibroMeter.sup.v2G, FM.sup.3G: FibroMeter.sup.V3G, CM.sup.3G: CirrhoMeter.sup.V3G, MFM.sup.3G: multi-targeted FibroMeter.sup.V3G Significant differences are indicated in bold.
(47) CHC Validation Population
(48) Multi-FibroMeters were compared to 10 other single tests in this population of 641 CHC where optimism bias was excluded (Table 7 below). Combined Elasto-FibroMeters were considered apart in this comparison due to optimism bias. Again, Multi-FibroMeter.sup.V2G ranked first for Obuchowski indexes.
(49) TABLE-US-00009 TABLE 7 AUROCs for all diagnostic targets and Obuchowski indices for Metavir fibrosis (F) stages of all tests in the CHC validation population (641 patients). p values of pair comparisons are reported in Tables 8 and 9. Obuchowski AUROC index F ≥ 1 F ≥ 2 F ≥ 3 F = 4 Value Rank FibroMeter.sup.V2G 0.827 0.812 0.830 0.863 0.797 2 CirrhoMeter.sup.V2G 0.783 0.785 0.816 0.858 0.770 5 Multi-FibroMeter.sup.V2G 0.822 0.808 0.838 0.880 0.798 1 FibroMeter.sup.V3G 0.819 0.798 0.816 0.844 0.785 4 CirrhoMeter.sup.V3G 0.769 0.771 0.796 0.840 0.756 7 Multi-FibroMeter.sup.V3G 0.818 0.804 0.826 0.868 0.792 3 APRI 0.769 0.751 0.768 0.814 0.742 10 Fib4 0.757 0.762 0.773 0.802 0.741 11 Fibrotest 0.797 0.769 0.800 0.822 0.762 6 Hepascore 0.750 0.776 0.804 0.849 0.752 9 Zeng score 0.740 0.757 0.791 0.810 0.734 12 VCTE 0.704 0.788 0.839 0.897 0.754 8 Elasto-FibroMeter.sup.V2G 0.795 0.843 0.878 0.922 0.812 ND .sup.a Elasto-FibroMeter.sup.V3G 0.795 0.842 0.877 0.922 0.812 ND .sup.a VCTE: vibration controlled transient elastography (by Fibroscan). .sup.a ND: not done due to optimism bias.
(50) TABLE-US-00010 TABLE 8 Comparison of AUROCs for cirrhosis of all test pairs in the CHC validation population (641 patients, Table 7) by Delong test. FM.sup.2G CM.sup.2G MFMF.sup.2G FM.sup.3G CM.sup.3G MFMF.sup.3G APRI Fib4 FT HS Zeng VCTE EFM.sup.2G EFM.sup.3G FM.sup.2G — 0.773 0.038 0.012 0.204 0.625 0.005 0.004 0.010 0.381 0.005 0.087 <0.001 <0.001 CM.sup.2G — 0.059 0.431 0.017 0.510 0.045 0.007 0.128 0.671 0.031 0.103 <0.001 <0.001 MFMF.sup.2G — <0.001 0.003 0.081 <0.001 <0.001 <0.001 0.025 <0.001 0.380 <0.001 <0.001 FM.sup.3G — 0.824 0.004 0.102 0.031 0.148 0.779 0.097 0.017 <0.001 <0.001 CM.sup.3G — 0.036 0 248 0.037 0.446 0.710 0.216 0.029 <0.001 <0.001 MFMF.sup.3G — 0.004 <0.001 0.008 0.342 0.010 0.190 <0.001 <0.001 APRI — 0.503 0.779 0.169 0.869 <0.001 <0.001 <0.001 Fib4 — 0.504 0.110 0.782 0.001 <0.001 <0.001 Fibrotest — 0.101 0.598 0.001 <0.001 <0.001 Hepascore — 0.016 0.019 <0.001 <0.001 Zeng — <0.001 <0.001 <0.001 VCTE — 0.024 0.028 EFM.sup.2G a — 0.856 EFM.sup.3G a — FM.sup.2G: FibroMeter.sup.V2G, CM.sup.2G: CirrhoMeter.sup.V2G, MFM.sup.2G: multi-targeted FibroMeter.sup.V2G, FM.sup.3G: FibroMeter.sup.V3G, CM.sup.3G: CirrhoMeter.sup.V2G, MFM.sup.3G: multi-targeted FibroMeterV.sup.3G, FT: Fibrotest, HS: Hepascore, VCTE: vibration controlled transient elastography (by Fibroscan), EFM.sup.2G: Elasto-FibroMeterV.sup.2G, EFM.sup.3G: Elasto-FibroMeterV.sup.3G. Significant differences are shown in bold. .sup.a Optimism bias
(51) Pairwise comparisons for cirrhosis AUROCs are detailed in Table 8 hereinabove. AUROCs of Multi-FibroMeters were significantly improved vs. FibroMeters or CirrhoMeters (borderline significance between multi-FibroMeter.sup.V2G and CirrhoMeter.sup.V2G) which was beyond the objective and reinforced the results observed in the derivation population. In addition, AUROCs of Multi-FibroMeters were significantly superior to all other single blood tests (except between Multi-FibroMeter.sup.V3G and Hepascore), but not vs. VCTE. Considering significant improvements brought by Multi-FibroMeters, it should be underlined that Multi-FibroMeter.sup.V2G had new advantages of significant superiority vs. Fibrotest (p<0.001) or Hepascore (p=0.025) which was not the case previously for CirrhoMeter.sup.V2G (p=0.128 or p=0.671, respectively). The new advantages were more marked for Multi-FibroMeter.sup.V3G since the differences became significant vs. APRI, Fibrotest and Zeng score whereas the AUROCs for cirrhosis of these last 3 tests were not significantly different with FibroMeter.sup.V2G and even CirrhoMeter.sup.V3G. Concerning VCTE, AUROCs for cirrhosis of Multi-FibroMeter.sup.V3G became not significantly different from that of VCTE whereas this latter was significantly higher than those of FibroMeter.sup.V3G or CirrhoMeter.sup.V3G. In other words, Multi-FibroMeter.sup.V3G deleted the superiority of VCTE over its corresponding mono-targeted tests.
(52) Pairwise comparisons for Obuchowski indexes are detailed in Table 9 below. Obuchowski indexes of Multi-FibroMeters were significantly improved vs. FibroMeters or CirrhoMeters (except between Multi-FibroMeter.sup.V2G and FibroMeter.sup.V2G). Obuchowski indexes of Multi-FibroMeters were significantly higher than all those of other blood tests. This was a new advantage mainly between multi-FibroMeter.sup.V3G and Hepascore. There was also the occurrence of significant superiority of Multi-FibroMeters vs. all single blood tests at the difference of CirrhoMeters but this improvement had less clinical interest since CirrhoMeters are only used for cirrhosis diagnosis. Concerning comparison between Multi-FibroMeters and VCTE, the differences remained not significant as for FibroMeters or CirrhoMeters. Concerning comparison between Multi-FibroMeter.sup.V3G and Elasto-FibroMeters, despite an optimism bias favoring Elasto-FibroMeters, the differences became not significant contrary to FibroMeter.sup.V3G or CirrhoMeter.sup.V3G. In other words, Multi-FibroMeter.sup.V3G deleted the superiority of Elasto-FibroMeters over corresponding mono-targeted blood tests.
(53) TABLE-US-00011 TABLE 9 Comparison of Obuchowski indices of all test pairs in the CHC validation population (641 patients, Table 7) by z test. FM.sup.2G CM.sup.2G MFMF.sup.2G FM.sup.3G CM.sup.3G MFMF.sup.3G APRI Fib4 FT HS Zeng VCTE EFM.sup.2G EFM.sup.3G FM.sup.2G — 0.003 0.995 0.002 <0.001 0.237 <0.001 <0.001 0.002 0.004 <0.001 0.096 0.244 0.241 CM.sup.2G — 0.004 0.148 0.003 0.032 0.146 0.073 0.562 0.306 0.081 0.576 0.012 0.011 MFMF.sup.2G — <0.001 <0.001 0.040 0.001 <0.001 0.001 0.004 <0.001 0.090 0.233 0.229 FM.sup.3G — 0.001 0.035 0.013 0.002 0.039 0.053 0.004 0.242 0.041 0.036 CM.sup.3G — <0.001 0.494 0.336 0.660 0.853 0.322 0.962 0.001 <0.001 MFMF.sup.3G — 0.004 <0.001 0.008 0.022 0.002 0.158 0.118 0.109 APRI — 0.958 0.312 0.678 0.719 0.678 <0.001 <0.001 Fib4 — 0.242 0.668 0.722 0.680 <0.001 <0.001 Fibrotest — 0.487 0.117 0.747 <0.001 <0.001 Hepascore — 0.307 0.914 <0.001 <0.001 Zeng — 0.393 <0.001 <0.001 VCTE — <0.001 <0.001 EFM.sup.2G a — 0.884 EFM.sup.3G a — FM.sup.2G: FibroMeter.sup.V2G, CM.sup.2G: CirrhoMeter.sup.V2G, MFM.sup.2G: multi-targeted FibroMeter.sup.V2G, FM.sup.3G: FibroMeter.sup.V3G, CM.sup.3G: CirrhoMeter.sup.V2G, MFM.sup.3G: multi-targeted FibroMeterV.sup.3G, FT: Fibrotest, HS: Hepascore, VCTE: vibration controlled transient elastography (by Fibroscan), EFM.sup.2G: Elasto-FibroMeter.sup.V2G, EFM.sup.3G: Elasto-FibroMeter.sup.V3G. Significant differences are shown in bold. .sup.a Optimism bias
(54) Non-CHC Validation Populations
(55) AUROC for cirrhosis and. Obuchowski indices were compared in 11 to 17 fibrosis tests in 4 other etiologies in Table 10 below. Multi-FibroMeters had higher Obuchowski indices than corresponding mono-targeted blood tests (except in ALD). As there was a few variations of diagnostic indices between all etiologies for most tests (i.e., no significant difference of Obuchowski indices compared to those of CHC validation population, results not shown), etiologies were pooled resulting in a non-CHC population of 935 patients in Table 11 below.
(56) Pairwise comparisons for cirrhosis AUROCs are detailed in Table 12 below. AUROCs of Multi-FibroMeters were significantly improved vs. FibroMeters but not vs. CirrhoMeters which fitted with objectives. AUROCs of Multi-FibroMeters were significantly superior to several single blood tests: APRI, Fib4 and Fibrotest (except for multi-FibroMeter.sup.V3G), this last difference being a new advantage of multi-FibroMeter.sup.V2G vs. FibroMeter.sup.V2G. Considering Multi-FibroMeter.sup.V3G, the significant inferiority observed between Hepascore and the corresponding FibroMeter.sup.V3G was deleted to become non-significant.
(57) Pairwise comparisons for Obuchowski indexes are detailed in Table 13 below. Obuchowski indexes were significantly improved vs. FibroMeters or CirrhoMeters. Obuchowski indexes of Multi-FibroMeters were significantly superior to all other single blood tests (except with Hepascore). Other new (minor) advantages were the significant superiority of Multi-FibroMeter.sup.V3G over APRI, Fibrotest or Zeng score at the difference of CirrhoMeters.
(58) Comparisons with the 3 tests including VCTE were performed in a subset of 376 patients (Table 14 below).
(59) TABLE-US-00012 TABLE 10 AUROC for cirrhosis and Obuchowski indices of all tests in the CHB (n = 152), HIV/CHC (n = 444), NAFLD (n = 224) and ALD (n = 115) validation populations. CHB HIV/CHC NAFLD ALD AUROC AUROC AUROC AUROC F = 4 Obuchowski F = 4 Obuchowski F = 4 Obuchowski F = 4 Obuchowski FibroMeter.sup.V2G 0.918 0.789 0.785 0.760 0.836 0.773 0.903 0.758 CirrhoMeter.sup.V2G 0940 0.768 0.832 0.737 0.857 0.750 0.900 9.772 Multi-FibroMeter.sup.V2G 0.942 0.802 0.823 0.766 0.850 0.783 0.905 0.770 FibroMeter.sup.V3G 0.909 0.781 0.758 0.749 0.793 0.749 0.819 0.715 CirrhoMeter.sup.V3G 0.940 0.761 0.809 0.727 0.898 0.723 0.849 0,738 Multi-FibroMeter.sup.V3G 0.942 0.793 0.794 0.756 0.803 0.759 0.847 0.728 APRI 0.810 0.727 0.678 0.712 0.679 0.680 0.527 0.532 Fib4 0.890 0.731 0.743 0.699 0.691 0.691 0.707 0.625 Fibrotest 0.887 0.767 0.793 0.733 0.697 0.670 — — Hepascore 0.912 0.781 0.819 0.723 0.920 0.780 0.920 0.780 Zeng score 0.921 0.783 0.790 0.711 0.920 0.785 0.871 0.772 VCTE 0.906 0.746 — — 0.951 0.808 — — Elasto-FibroMeter.sup.V2G 0.951 0.815 — — 0.960 0.846 — — Elasto-FibroMeter.sup.V3G 0.947 0.812 — — 0.953 0.840 — — FibroMeter.sup.NAFLD — — — — 0.819 0.714 — — NAFLD fibrosis score — — — — 0.775 0.673 — — FibroMeter.sup.ALD2G 0.915 0.758 0.830 0.728 0.949 0.803 .sup. 0.929 .sup.a .sup. 0.794 .sup.a VCTE: vibration controlled transient elastography (by Fibroscan) .sup.a Optimism bias
(60) TABLE-US-00013 TABLE 11 AUROCs for all diagnostic targets and Obuchowski indices Metavir fibrosis (F) stages of 12 blood tests in the non-CHC validation populations (935 patients). Obuchowski AUROC index F ≥ 1 F ≥ 2 F ≥ 3 F = 4 Value Rank FibroMeter.sup.V2G 0.797 0.829 0.849 0.874 0.780 2 CirrhoMeter.sup.V2G 0.748 0.792 0.855 0.892 0.754 7 Multi-FibroMeter.sup.V2G 0.793 0.829 0.859 0.895 0.786 1 FibroMeter.sup.V3G 0.774 0.814 0.827 0.838 0.763 6 CirrhoMeter.sup.V3G 0.718 0.770 0.829 0.862 0.731 9 Multi-FibroMeter.sup.V3G 0.779 0.817 0.837 0.862 0.773 3 APRI 0.733 0.729 0.710 0.676 0.712 11 Fib4 0.684 0.734 0.767 0.788 0.694 12 Fibrotest 0.731 0.764 0.763 0.809 0.729 10 Hepascore 0.789 0.819 0.849 0.902 0.772 4 Zeng score 0.738 0.789 0.829 0.876 0.741 8 FibroMeter.sup.ALD2G 0.769 0.817 0.872 0.912 0.771 5
(61) TABLE-US-00014 TABLE 12 Comparison of AUROC for cirrhosis of 12 blood test pairs in the non-CHC validation populations (935 patients in Table 11) by z test. FM.sup.2G CM.sup.2G MFMF.sup.2G FM.sup.3G CM.sup.3G MFMF.sup.3G APRI Fib4 FT HS Zeng FMA FM.sup.2G — 0.0008 7.10.sup.−6 0.0015 0.8941 0.9017 1.10.sup.−6 0.0004 0.1142 0.0868 0.8031 0.0140 CM.sup.2G — 0.8683 0.0001 0.0022 0.0153 1.10.sup.−7 2.10.sup.−6 0.0079 0.7685 0.1311 0.5144 MFMF.sup.2G — 1.10.sup.−7 0.0253 0.0005 2.10.sup.−9 4.10.sup.−7 0.0033 0.8201 0.0793 0.4726 FM.sup.3G — 0.0492 0.0012 0.0002 0.0276 0.8840 0.0099 0.3337 0.0016 CM.sup.3G — 0.9363 2.10.sup.−5 0.0007 0.2013 0.2955 0.7893 0.0531 MFMF.sup.3G — 3.10.sup.−7 0.0002 0.1475 0.2356 0.7913 0.0473 APRI — 0.0222 0.0035 1.10.sup.−6 0.0002 1.10.sup.−7 Fib4 — 0.1938 9.10.sup.−5 0.0074 2.10.sup.−5 Fibrotest — 0.0088 0.2941 0.0076 Hepascore — 0.0624 0.3011 Zeng — 0.0207 FMA — FM.sup.2G: FibroMeter.sup.V2G, CM.sup.2G: CirrhoMeter.sup.V2G, MFM.sup.2G: multi-targeted FibroMeter.sup.V2G, FM.sup.3G: FibroMeter.sup.V3G, CM.sup.3G: CirrhoMeter.sup.V3G, MFM.sup.3G: multi-targeted FibroMeter.sup.V3G, FT: Fibrotest, HS: Hepascore, FMA: FibroMeter.sup.ALD2G. Significant differences are shown in bold.
(62) TABLE-US-00015 TABLE 13 Comparison of Obuchowski indices of 12 blood test pairs in the non-CHC validation population (935 patients in Table 11) by z test. FM.sup.2G CM.sup.2G MFMF.sup.2G FM.sup.3G CM.sup.3G MFMF.sup.3G APRI Fib4 FT HS Zeng FMA FM.sup.2G — 4.10.sup.−6 0.0250 8.10.sup.−8 2.10.sup.−11 0.0576 2.10.sup.−9 6.10.sup.−15 8.10.sup.−9 0.3588 0.0001 0.3211 CM.sup.2G — 2.10.sup.−7 0.1723 2.10.sup.−9 0.0040 0.0017 7.10.sup.−7 0.0304 0.094 0.3256 0.0360 MFMF.sup.2G — 8.10.sup.−12 2.10.sup.−16 7.10.sup.−7 2.10.sup.−11 0 2.10.sup.−10 0.1430 7.10.sup.−6 0.1159 FM.sup.3G — 1.10.sup.−6 0.009 7.10.sup.−6 9.10.sup.−11 5.10.sup.−6 0.4128 0.0268 0.4538 CM.sup.3G — 8.10.sup.−12 0.1495 0.0009 0.8425 0.0013 0.4686 0.0001 MFMF.sup.3G — 2.10.sup.−8 4.10.sup.−14 2.10.sup.−6 0.9348 0.0036 0.8732 APRI — 0.0837 0.2373 5.10.sup.−6 0.0468 0.0001 Fib4 — 0.0132 2.10.sup.−7 0.0003 5.10.sup.−7 Fibrotest — 1.10.sup.−6 0.2858 0.0012 Hepascore — 0.0036 0.9319 Zeng — 0.0115 FMA — FM.sup.2G: FibroMeter.sup.V2G, CM.sup.2G: CirrhoMeter.sup.V2G, MFM.sup.2G: multi-targeted FibroMeter.sup.V2G, FM.sup.3G: FibroMeter.sup.V3G, CM.sup.3G: CirrhoMeter.sup.V3G, MFM.sup.3G: multi-targeted FibroMeter.sup.V3G, FT: Fibrotest, HS: Hepascore, FMA: FibroMeter.sup.ALD2G. Significant differences are shown in bold.
(63) TABLE-US-00016 TABLE 14 AUROCs for all diagnostic targets and Obuchowski indices of blood tests (FibroMeter family), VCTE and FibroMeter + VCTE combined tests in the non-CHC validation populations (376 patients). Obuchowski AUROC index F ≥ 1 F ≥ 2 F ≥ 3 F = 4 Value Rank FibroMeter.sup.V2G 0.744 0.852 0.833 0.884 0.789 4 CirrhoMeter.sup.V2G 0.688 0.809 0.829 0.902 0.764 6 Multi-FibroMeter.sup.V2G 0.731 0.859 0.854 0.904 0.797 3 FibroMeter.sup.V3G 0.712 0.836 0.829 0.862 0.771 7 CirrhoMeter.sup.V3G 0.661 0.796 0.821 0.880 0.747 9 Multi-FibroMeter.sup.V3G 0.718 0.847 0.840 0.880 0.783 5 VCTE 0.705 0.794 0.861 0.880 0.766 8 Elasto-FibroMeter.sup.V2G 0.772 0.881 0.915 0.940 0.833 1 Elasto-FibroMeter.sup.V3G 0.765 0.878 0.913 0.935 0.829 2
(64) Comparisons of the Multi-FibroMeters to VCTE were also performed in the combined validation populations #1 to #6 (1746 patients). AUROC for significant fibrosis (F≥2) and for cirrhosis (F=4), Obuchowski index and rate of correctly classified patients were compared. The Multi-FibroMeter.sup.V2G displayed the best results in terms of Obuchowski index (OI=0.777) and rate of correctly classified patients (83%). VCTE displayed an Obuchowski index of 0.755 and a rate of correctly classified patients of 80%. The Multi-FibroMeter.sup.V3G also displayed better results than VCTE in terms of Obuchowski index (OI=0.759) and rate of correctly classified patients (82.7%). AUROC for significant fibrosis (F≥2) was 0.786 for VCTE vs. 0.817 for MFM.sup.V2G and 0.804 for MFM.sup.V3G. AUROC for cirrhosis (F=4) were equivalent between the MFM.sup.V2G (0.885) or MFM.sup.V3G (0.860) and VCTE (0.898).
(65) Overall Population
(66) As shown in Table 15 below, the diagnostic performance of the Multi-FibroMeters was also evaluated in the overall population (3809 patients) since there was no optimism bias in statistical comparisons within the FibroMeter family. The MFM.sup.V2G displayed the best results in terms of AUROC for significant fibrosis, AUROC for severe fibrosis, AUROC for cirrhosis, and Obuchowski index. The MFM.sup.V2G also displayed a very high rate of correctly classified patients, only second to that of the MFM.sup.V3G.
(67) TABLE-US-00017 TABLE 15 Diagnostic performance in the overall population (3809 patients). Obuchowski AUROC index Classification F ≥ 1 F ≥ 2 F ≥ 3 F = 4 Value Rank Rate Rank FibroMeter.sup.V2G 0.788 0.832 0.849 0.878 0.791 2 82.1 3 CirrhoMeter.sup.V2G 0.747 0.800 0.846 0.897 0.769 5 81.8 4 Multi-FibroMeter.sup.V2G 0.778 0.833 0.863 0.906 0.795 1 86.0 2 FibroMeter.sup.V3G 0.767 0.823 0.837 0.855 0.776 4 79.5 6 CirrhoMeter.sup.V3G 0.722 0.790 0.835 0.879 0.754 6 80.8 5 Multi-FibroMeter.sup.V3G 0.764 0.823 0.849 0.886 0.782 3 86.1 1 The best result per diagnostic target is indicated in bold.
(68) Fibrosis Staging
(69) Classifications of FibroMeters [20], CirrhoMeters [15] and Multi-FibroMeters included 6 to 7 fibrosis classes reflecting Metavir staging. The new classes developed for Multi-FibroMeters were: F0/1, F1/2, F2±1, F3±1, F3/4 and F4. The rate of correctly classified patients ranked in the same order for the 6 tests as a function of the 3 populations: the derivation population (1012 CHC patients), the validation population #1 (676 CHC patients) and the combined validation populations #2 to #5 (936 non-CHC patients) (Table 16 below). These rates were significantly higher (p<0.001) in Multi-FibroMeters vs. corresponding FibroMeter.sup.V2/3G or CirrhoMeter.sup.V2/3G in the 3 populations. These rates were not significantly different between both Multi-FibroMeters.sup.V2/3G.
(70) As shown in
(71) TABLE-US-00018 TABLE 16 Rate of correctly classified patients by fibrosis stagings of multi-targeted FibroMeters vs. published mono-targeted FibroMeters in the 3 main populations. CHC derivation CHC validation Non-CHC validation (1012 patients) (676 patients) .sup.a (935 patient) F0 F1 F2 F3 F4 All F Rank All F Rank All F Rmk FibroMete.sup.V2G 56.8 92.2 88.9 80.0 87.7 81.6 3 83.6 3 77.9 3 CirrhoMetet.sup.V2G 43.2 89.7 91.1 82.9 93.0 81.5 4 82.5 4 76.3 4 Multi- 50.0 92.9 96.3 97.9 90.4 92.3 .sup.b 2 88.0 .sup.b 2 81.3 .sup.b 2 FibroMeter.sup.V2G FibroMeter.sup.V3G 43.2 91.3 94.1 83.6 73.7 86.9 6 81.4 5 69.4 6 CirrhoMeter.sup.V3G 45.5 95.0 90.0 77.1 79.8 87.3 5 81.1 6 75.7 5 Multi- 50.0 94.1 97.8 97.1 87.7 92.8 .sup.b c 1 88.5 .sup.b c 1 81.4 .sup.b c 1 FibroMeter.sup.V3G p .sup.d 0.266 <0.001 <0.001 <0.001 <0.001 <0.001 — <0.001 — <0.001 — Best result per diagnostic target is indicated in bold. .sup.a more patients were available with these 6 tests than in the core population .sup.b p < 0.001 vs. corresponding FibroMeter.sup.v2/3G or CirrhoMeter.sup.v2/3G by paired Wilcoxon test .sup.c vs. Multi-FibroMeter.sup.v2G by paired Wilcoxon test: p = 0.443 in CHC derivation, p = 0.439 in CHC validation, p = 1 in non-CHC validation populations .sup.d by paired Cochran test
(72) Advantages of the Multi-Targeted FibroMeter (MFMs)
(73) The primary objective of the study was to construct multi-targeted FibroMeters displaying a significant increase in diagnostic performance when compared to mono-targeted tests of the FibroMeter family. The accuracy between Multi-FibroMeters and FibroMeters or CirrhoMeters was thus compared through the assessment of five judgement criteria: 1) whether the AUROC for cirrhosis of the MFM was superior to that of the FibroMeter, 2) whether the Obuchowski index of the MFM was superior to that of the FibroMeter, 3) whether the AUROC for significant fibrosis of the MFM was equal or superior to that of the FibroMeter, 4) whether the rate of correctly classified patients (also called “classification metric”) of the MFM was superior to that of the FibroMeter, and 5) whether the AUROC for cirrhosis of the MFM was equal or superior to that of the CirrhoMeter.
(74) The second objective of the study was to construct multi-targeted FibroMeters, as obtained for the primary objective, displaying an improved diagnostic performance when compared to other fibrosis tests not belonging to the FibroMeter family, in particular Fibrotest, Hepascore, Zeng score and VCTE. The accuracy between Multi-FibroMeters and said fibrosis tests was thus compared through the assessment of three, in some cases four, judgement criteria: 1) whether the AUROC for cirrhosis of the MFM was superior to that of Fibrotest, Hepascore, and Zeng score; and equivalent to that of VCTE, 2) whether the Obuchowski index of the MFM was superior to that of the other fibrosis tests, 3) whether the AUROC for significant fibrosis of the MFM was superior to that of the other fibrosis tests, and 4) whether the rate of correctly classified patients (also called “classification metric”) of the MFM was superior to that of the Fibrotest and of VCTE.
(75) Table 17 below presents a summary of the diagnostic performance of both MFM.sup.V2G and MFM.sup.V3G when assessed as described above, through comparison with mono-targeted tests of the FibroMeter family and with Fibrotest, Hepascore, Zeng score and VCTE.
(76) TABLE-US-00019 TABLE 17 Diagnostic performance of Multi-FibroMeters when compared to the indicated tests in combined populations of maximum size.sup.a. Criteria fulfilled by Multi-FibroMeter Judgment criteria Test compared V2G V3G Primary objective: AUROC cirrhosis > FibroMeter
Obuchowski index >
.sup.b
AUROC significant F ≥ Yes Yes Classification metric >
AUROC cirrhosis ≥ CirrhoMeter
.sup.b Secondary objectives: AUROC cirrhosis > Fibrotest Yes
Obuchowski index > Yes Yes AUROC significant F > Yes Yes Classification metric > Yes Yes AUROC cirrhosis > Hepascore
No .sup.c Obuchowski index > Yes No .sup.c AUROC significant F > Yes Yes AUROC cirrhosis > Zeng score
No .sup.c Obuchowski index > Yes Yes AUROC significant F > Yes Yes AUROC cirrhosis ≈ VCTE Yes No .sup.d Obuchowski index >
No .sup.c AUROC significant F > Yes No .sup.c Classification metric >
VCTE: vibration controlled transient elastography (by Fibroscan), F: fibrosis. Results indicated in bold depict a significant difference. Results indicated in italics depict a statistical advantage of the Multi-FibroMeter over the test compared, in comparison with the FibroMeter or CirrhoMeter compared to the same test. .sup.aCombined population: overall population for primary objective (3809 patients) and combination of populations #1 to #6 for secondary objectives (2796 patients except for Fibrotest: 1461 patients and VCTE: 1746 patients) to avoid optimism bias in comparisons. “Yes” and “no” indicate whether the criterion was reached or not with the following precision: .sup.b Borderline significance .sup.c Non-significant superior value of Multi-FibroMeter .sup.d Significant inferior value of Multi-FibroMeter
(77) The primary objective was fulfilled, with both MFM.sup.V2G and MFM.sup.V3G displaying a significant increase in diagnostic performance when compared to the corresponding FibroMeter. Thus, all of five judgement criteria were positively met by the Multi-FibroMeters. In particular, AUROCs for cirrhosis of Multi-FibroMeters were significantly increased when compared to the corresponding FibroMeter. It should be noted that for cirrhosis diagnosis the most relevant comparator of Multi-FibroMeter is FibroMeter and not CirrhoMeter since FibroMeter (like other blood tests) is the classical test used whatever the target diagnostic. In other words, the objective was that Multi-FibroMeters added the diagnostic performance for cirrhosis of CirrhoMeter to FibroMeter. Considering discrimination of Metavir fibrosis stages, the performance of Multi-FibroMeters, evaluated by Obuchowski index, was significantly increased compared to FibroMeter. Regarding fibrosis classification reflecting Metavir stages, i.e., the rate of correctly classified patients, Multi-FibroMeters had significantly higher accuracy than FibroMeters.
(78) Concerning the secondary objective, the judgement criteria were all positively met by the MFM.sup.V2G when compared to the other fibrosis tests, and only partially met by the MFM.sup.V3G. It should be noted that the usual reference for non-invasive diagnosis of cirrhosis is VCTE. The results showed that AUROC for cirrhosis of VCTE and Multi-FibroMeter.sup.V2G were equivalent. AUROC for significant fibrosis and Obuchowski index were significantly increased in Multi-FibroMeter.sup.V2G. This last result was confirmed by the rate of correctly classified patients.
(79) In conclusion, using multi-targeted FibroMeters significantly improves the fibrosis staging accuracy compared to classical single-target blood tests or VCTE (also known as Fibroscan), especially when the underlying cause of the liver lesion is chronic hepatitis C.
(80) For the diagnosis of cirrhosis, Multi-targeted FibroMeters are even matching VCTE, usually considered as the reference for non-invasive diagnosis of cirrhosis.
(81) With the use of a single non-invasive test, the multi-targeted FibroMeter, it is thus now possible to accurately diagnose either significant fibrosis or cirrhosis. Multi-targeted FibroMeters thus provide unique non-invasive tests for the accurate diagnostic of the presence and severity of fibrosis, including cirrhosis.
(82) Importantly, the present diagnostic method, i.e., the construction of a multi-targeted diagnostic test, can be applied to any non-invasive diagnostic test based on a semi-quantitative (ordinal) reference, e.g., a severity score in radiology.
REFERENCES
(83) 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 Chou R, Wasson N. Blood tests to diagnose fibrosis or cirrhosis in patients with chronic hepatitis C virus infection: a systematic review. Annals of internal medicine 2013; 158:807-820. 3 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. 4 Cales P, de Ledinghen V, Halfon P, Bacq Y, Leroy V, Boursier J, et al. Evaluating the accuracy and increasing the reliable diagnosis rate of blood tests for liver fibrosis in chronic hepatitis C. Liver Int 2008; 28:1352-1362. 5 Boursier J, de Ledinghen V, Zarski J P, Fouchard-Hubert I, Gallois Y, Oberti F, et al. Comparison of eight diagnostic algorithms for liver fibrosis in hepatitis C: new algorithms are more precise and entirely noninvasive. Hepatology 2012; 55:58-67. 6 Boursier J, de Ledinghen V, Zarski J P, Rousselet M C, Sturm N, Foucher J, et al. A new combination of blood test and fibroscan for accurate non-invasive diagnosis of liver fibrosis stages in chronic hepatitis C. Am J Gastroenterol 2011; 106:1255-1263. 7 Leroy V, Sturm N, Faure P, Trocme C, Marlu A, Hilleret M N, et al. Prospective evaluation of FibroTest®, FibroMeter®, and HepaScore® for staging liver fibrosis in chronic hepatitis B: comparison with hepatitis C. J Hepatol 2014; 61:28-34. 8 Calès P, HaIfon P, Batisse D, Carrat F, Perré P, Penaranda G, et al. Comparison of liver fibrosis blood tests developed for HCV with new specific tests in HIV/HCV co-infection J Hepatol 2010; 52:238-244. 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 Intraobserver and interobserver variations in liver biopsy interpretation in patients with chronic hepatitis C. The French METAVIR Cooperative Study Group. Hepatology 1994; 20:15-20. 11 Zarski J P, Sturm N, Guechot J, Paris A, Zafrani E S, Asselah T, et al. Comparison of nine blood tests and transient elastography for liver fibrosis in chronic hepatitis C: The ANRS HCEP-23 study. J Hepatol 2012; 56:55-62. 12 Cales P, Boursier J, Bertrais S, Oberti F, Gallois Y, Fouchard-Hubert I, et al. Optimization and robustness of blood tests for liver fibrosis and cirrhosis. Clin Biochem 2010; 43:1315-1322. 13 Cales P, Boursier J, Oberti F, Hubert I, Gallois Y, Rousselet M C, et al. FibroMeters: a family of blood tests for liver fibrosis. Gastroenterol Clin Biol 2008; 32:40-51. 14 Boursier J, Bertrais S, Oberti F, Gallois Y, Fouchard-Hubert I, Rousselet M C, et al. Comparison of accuracy of fibrosis degree classifications by liver biopsy and non-invasive tests in chronic hepatitis C. BMC Gastroenterol 2011; 11:132. 15 Cales P, Boursier J, Oberti F, Bardou D, Zarski J P, de Ledinghen V. Cirrhosis Diagnosis and Liver Fibrosis Staging: Transient Elastometry Versus Cirrhosis Blood Test. J Clin Gastroenterol 2015; 49:512-519. 16 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. 17 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. 18 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. 19 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. 20 Leroy V, Halfon P, Bacq Y, Boursier J, Rousselet M C, Bourliere M, et al. Diagnostic accuracy, reproducibility and robustness of fibrosis blood tests in chronic hepatitis C: a meta-analysis with individual data. Clin Biochem 2008; 41:1368-1376. 21 Zeng M D, Lu L G, Mao Y M, Qiu D K, Li J Q, Wan M B, et al. Prediction of significant fibrosis in HBeAg-positive patients with chronic hepatitis B by a noninvasive model. Hepatology 2005; 42:1437-1445. 22 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. 23 Angulo P, Hui J M, Marchesini G, Bugianesi E, George J, Farrell G C, et al. The NAFLD fibrosis score: a noninvasive system that identifies liver fibrosis in patients with NAFLD. Hepatology 2007; 45:846-854. 24 Castera L, Forns X, Alberti A. Non-invasive evaluation of liver fibrosis using transient elastography. J Hepatol 2008; 48:835-847. 25 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. 26 Thein H H, Yi Q, Dore G J, Krahn M D. Estimation of stage-specific fibrosis progression rates in chronic hepatitis C virus infection: a meta-analysis and meta-regression. Hepatology 2008; 48:418-431. 27 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. 28 Boursier J, de Ledinghen V, Poynard T, Guechot J, Carrat F, Leroy V, et al., An extension of STARD statements for reporting diagnostic accuracy studies on liver fibrosis tests: The Liver-FibroSTARD standards. J Hepatol 2014. 29 Cales P, Boursier J, Ducancelle A, Oberti F, Hubert I, Hunault G, et al., Improved fibrosis staging by elastometry and blood test in chronic hepatitis C. Liver Int 2014; 34:907-917. 30 Boursier J, Zarski J P, de Ledinghen V, Rousselet M C, Sturm N, Lebail B, et al., Determination of reliability criteria for liver stiffness evaluation by transient elastography. Hepatology 2013; 57:1182-1191.
Example 2: Multi-Targeted FibroMeter Constructed for Multi-Target Classification (MFMc)
(84) Patients and Methods
(85) Populations
(86) A total of 3901 patients were included in the present study: the multi-target diagnostic algorithm was developed using data from 1012 patients (derivation population), and an external validation was performed in 1330 patients (validation populations #1, #2 and #3). The prognostic relevance of the fibrosis classification resulting from this new diagnostic system was also assessed in a prospective cohort of 1559 patients (validation population #4).
(87) Derivation Population
(88) The derivation population included 1012 patients with chronic hepatitis C (CHC) (5). Thus, individual patient data were available from five centers, independent for study design, patient recruitment, biological marker determination and liver histology interpretation by an expert pathologist.
(89) Validation Populations
(90) Diagnostic populations—The validation population #1 included 676 patients with CHC (6, 7). The validation population #2 included 450 patients with CHC and HIV infection prospectively included from April 1997 to August 2007 if they had anti-HCV (hepatitis C virus) and anti-HIV (human immunodeficiency virus) antibodies, and HCV RNA in serum (8). The validation population #3 for chronic hepatitis B (CHB) was extracted from a previously published database (9) and included 204 patients all with chronic hepatitis (30.4% HBe Ag positive); inactive carriers of HBs Ag were excluded.
(91) Prognostic population—All subjects over 18 years of age who were received for consultation or hospitalized for a chronic liver disease in the Department of Hepatology at the University Hospital of Angers from January 2005 to December 2009 were invited to join a study cohort (validation population #4), whatever the severity or etiology of their disease (viral hepatitis, alcoholic liver disease, non-alcoholic fatty liver disease (NAFLD), other causes). The resulting 1559 patients were then followed until death, liver transplantation or Jan. 1, 2011. The study was approved by an Institutional Review Board (AC-2012-1507) and informed consent was obtained from all patients.
(92) Diagnostic Methods
(93) Histological Assessment
(94) Liver biopsies were performed using Menghini's technique with a 1.4-1.6 mm diameter needle. Biopsy specimens were fixed in a formalin-alcohol-acetic solution and embedded in paraffin; 5 μm thick sections were then cut and stained with hematoxylin-eosin-saffron. Liver fibrosis was evaluated according to Metavir fibrosis (F) stages (10) by two senior experts with a consensus reading in case of discordance in Angers and in the Fibrostar study (11) (part of validation population #1), and by a senior expert in other centers. The area of porto-septal fibrosis was centrally measured by automated morphometry as recently described (12) in the validation population #1.
(95) FibroMeter Variables
(96) Biological markers were those previously used in various blood tests carried out to diagnose different lesions in chronic viral hepatitis (13, 14). The following biological markers were included: platelets, aspartate aminotransferase (AST), hyaluronate, urea, prothrombin index, alpha2-macroglobulin as used in FibroMeter.sup.V2G (5, 13) plus gamma-glutamyl transpeptidase (GGT) (used in FibroMeter.sup.V3G (14) and QuantiMeterV targeted for area of fibrosis (13)), bilirubin (used in QuantiMeterV) and alanine aminotransferase (ALT) (used in InflaMeter targeted for liver activity (15)). Clinical markers were also included (age and sex as used in FibroMeter.sup.V2G). Thus, with the addition of the AST/ALT ratio, 12 variables were available. Reference blood tests for comparison with the new test were FibroMeter.sup.V2G, targeted for significant fibrosis (F≥2), and CirrhoMeter.sup.V2G, targeted for cirrhosis, with previously calculated classifications (
(97) Liver Elastometry
(98) Vibration-controlled transient elastometry or (Fibroscan, Echosens, Paris, France) was performed by an experienced observer (>50 examinations before the study), blinded for patient data. Examination conditions were those recommended by the manufacturer (16). VCTE examination was stopped when 10 valid measurements were recorded. Results (kPa) were expressed as the median and the interquartile range (IQR) of all valid measurements. The 6-class fibrosis classification recently developed in CHC was used here for VCTE (
(99) Test Construction
(100) The construction of the multi-target classification system was performed in four progressive steps, summarized in
(101) Step 1: Single-target test construction—These tests were built using a conventional binary logistic regression (BLR) approach, using as many diagnostic targets as possible by the five Metavir F stages. These targets were: fibrosis (F≥1), significant fibrosis (F≥2), severe fibrosis (F≥3), and cirrhosis (F=4). Four single-target tests were thus obtained, called FMF≥1, FMF≥2, FMF≥3 and FMF=4, respectively.
(102) Step 2: Single-target test selection—Significant fibrosis was independently predicted by the FMF≥2 test (p<0.001) and the FMF=4 test (p<0.001) with a significant one-way interaction (p=0.001), whereas cirrhosis was independently diagnosed by the FMF≥1 test (p<0.003) and the FMF=4 test (p=0.038). Thus, three of the independent single-target tests were considered relevant for multi-target staging.
(103) Step 3: Single-target test classifications—The test scores (range: 0 to 1) were transformed into fibrosis classifications including several classes of predicted F stages according to a previously described segmentation method (17). Three classifications for FMF≥1, FMF≥2 and FMF=4 tests were thus obtained. Here, “class” refers to fibrosis classification (staging) by non-invasive tests.
(104) Step 4: Multi-target test classification—Briefly, each of the most accurate parts of the three retained test classifications (
(105) Step 5: this optional step is a multiple linear regression with the Metavir reference as dependent variable (or diagnostic target) on multi-target test classification. The score obtained can been normalized either before the regression being applied to the normalized Metavir score or after the regression normalization being applied to the regression score. If necessary, the final score is fully normalized (range 0 to 1) by bounding the extreme values (0 and 1).
(106) Statistics
(107) Quantitative variables were expressed as mean±standard deviation. The discriminative ability of each test was expressed as the area under the receiver operating characteristic curve (AUROC) and the overall accuracy as assessed by the rate of well-classified patients according to Metavir F. In classification calculations, test classes were used with their median value, e.g., 1.5 for F1/2. By definition, optimism bias maximizes performance in the population where test classifications are constructed: this affected FibroMeter.sup.V2G, CirrhoMeter.sup.V2G and MFM in the derivation population and VCTE in the validation population #1. Data were reported according to STARD (18) and Liver FibroSTARD statements (19), and analyzed on an intention to diagnose basis. Survival curves were estimated by the Kaplan-Meier method and were compared using the log-rank test. The main statistical analyses were performed under the control of professional statisticians (SB, GH) using SPSS version 18.0 (IBM, Armonk, N.Y., USA) and SAS 9.2 (SAS Institute Inc., Cary, N.C., USA).
(108) Results
(109) Population Characteristics
(110) The main characteristics of the studied populations are depicted in Table 18 below. In the prognostic population, median follow-up was 2.8 years (IQR: 1.7-3.9). During follow-up, there were 262 deaths (16.8%), of which 115 (7.4%) were liver-related.
(111) TABLE-US-00020 TABLE 18 Characteristics of the populations. Population Validation Derivation #1 #2 #3 #4 Patients (n) 1012 676 450 204 1559 Male (%) 59.6 60.1 68.9 77.0 68.9 Age (years) 45.4 ± 21.2 51.6 ± 11.2 40.5 ± 5.8 39.6 ± 12.1 54.6 ± 14.9 Cause (%): Virus 100 (HCV) 100 (HCV) 100 (HCV/HIV) 100 (HBV) 30.5 Alcohol — — — — 41.2 NAFLD — — — — 20.0 Other — — — — 8.3 Metavir (%): F0 4.3 4.0 5.8 14.7 F0/1 .sup.a: 15.1 F1 43.4 37.6 24.7 44.1 F1: 3.4, F1/2: 29.4 F2 27.0 25.7 36.4 26.5 F2 ± 1: 10.8 F3 12.9 18.2 19.6 5.9 F3 ± 1: 19.8, F3/4: 13.5 F4 11.4 14.5 13.6 8.8 F4: 8.0 Significant 52.3 58.4 69.6 41.2 52.1 .sup.a fibrosis (%) Biopsy length 21.2 ± 7.9 24.3 ± 9.0 NA 22.8 ± 7.9 — (mm) NA: not available. .sup.a According to FibroMeter.sup.V2G classification
(112) Multi-Target Test Characteristics (Derivation Population)
(113) Test Accuracy
(114) Single-target test accuracy—The discriminative ability of the new single-target tests (FMF≥1, FMF≥2, FMF≥3, FMF=4) compared to previously published tests (FibroMeter.sup.V2G, CirrhoMeter.sup.V2G) can be summarized as follows. First, the highest AUROCs were observed with the new tests. Second, for each of the new single-target tests, the highest AUROC was observed at the diagnostic target for which the test was constructed, as expected.
(115) Fibrosis classification accuracy—Table 19 below shows the overall fibrosis classification accuracy (as assessed by correct classification rate) of published tests (FibroMeter.sup.V2G: 87.6%, CirrhoMeter.sup.V2G: 87.5%) compared to the new multi-target test (MFM: 92.7%, p<0.001) in the derivation population. The accuracy was only fair in Metavir F0 for all tests. The gain in Metavir F1 for the new MFM was only moderate as the published tests already have high accuracy in this stage. In contrast, the MFM provided substantial gains in Metavir F2 and especially in F3, where it increased accuracy by 16.3% and 22.8% (p<0.001), respectively in the derivation population and validation population #1, compared to CirrhoMeter.sup.V2G.
(116) MFM increased accuracy in most fibrosis classes, e.g., in F4 class: MFM: 96.0%, CirrhoMeter.sup.V2G: 88.0%, FibroMeter.sup.V2G: 79.2% (details in Table 20 below). The comparison of classical diagnostic indices for a single diagnostic target was performed between MFM and FibroMeter.sup.V2G for severe fibrosis (Table 21 below); overall accuracies were 83.0% and 80.4%, respectively, p<0.010.
(117) TABLE-US-00021 TABLE 19 Classification accuracy (rate of correctly-classified patients, %) of published single-target tests and the new multi-target test (MFM) as a function of Metavir fibrosis (F) stages in the derivation population and the validation populations #1, #2 and #3. Test Population/F n FibroMeter.sup.V2G CirrhoMeter.sup.V2G MFM p .sup.a Derivation: F0 44 56.8 43.2 54.5 0.212 F1 439 92.3 89.7 92.5 0.010 F2 273 89.0 91.2 96.3 <0.001 F3 141 80.1 83.0 99.3 <0.001 F4 115 87.7 93.0 91.3 0.174 Overall 1012 87.6 87.5 92.7 <0.001 Validation #1: F0 27 29.6 37.0 25.9 0.247 F1 254 85.0 85.8 87.0 0.562 F2 174 91.4 89.7 95.4 0.048 F3 123 80.5 74.8 97.6 <0.001 F4 98 84.7 83.7 83.7 0.895 Overall 676 83.6 82.5 88.2 <0.001 Validation #2: F0 26 15.4 19.2 23.1 0.549 F1 111 78.4 73.9 79.3 0.161 F2 164 84.8 86.6 92.7 0.004 F3 88 83.0 85.2 96.6 0.001 F4 61 80.3 85.2 82.0 0.417 Overall 450 78.2 79.1 84.7 <0.001 Validation #3: Overall .sup.b 204 81.4 76.5 82.8 0.021 MFM: multi-target FibroMeter, n: number of patients. Bold figures indicate the highest accuracy per stage and population. Underlined accuracies show a noteworthy improvement brought about by MFM compared to the previously published CirrhoMeter.sup.V2G test. .sup.a by paired Cochran test between all tests .sup.b no result per F stage due to small sample size
(118) TABLE-US-00022 TABLE 20 Classification accuracy (rate of correctly-classified patients) of published single- target tests and new multi-target test (MFM) as a function of their specific fibrosis classes in the derivation population and the validation population #1. Classes Derivation population Validation population #1 .sup.a FM2G CM2G MFM FM2G CM2G MFM DA DA DA DA DA DA n (%) n (%) n (%) n (%) n (%) n (%) F0/1 152 92.8 126 88.1 125 96.0 40 90.0 58 77.6 42 90.5 F1 50 80.0 — — — — 19 78.9 — — — — F1/2 380 88.4 405 89.6 56 85.7 231 83.5 240 80.8 24 75.0 F2 ± 1 126 91.3 152 95.4 531 95.7 118 88.1 117 91.5 371 91.9 F3 ± 1 203 86.2 203 80.8 224 86.6 187 84.0 182 81.9 197 84.3 F3/4 76 78.9 76 78.9 51 86.3 68 75.0 51 74.5 26 76.9 F4 25 79.2 50 88.0 25 96.0 13 69.2 28 89.3 16 81.3 Overall 1012 87.6 1012 87.5 1012 92.7 676 83.6 676 82.5 676 88.2 FM2G: FibroMeter.sup.V2G, CM2G: CirrhoMeter.sup.V2G, MFM: multi-target FibroMeter, n: number of patients, DA: diagnostic accuracy. Bold figures indicate the highest accuracy per class and population. .sup.a No results in the validation populations #2 and #3 due to small sample sizes
(119) TABLE-US-00023 TABLE 21 Comparison of classical diagnostic indices between FibroMeter.sup.V2G and the multi-target test (MFM) for severe fibrosis (Metavir F ≥ 3). Sensitivity Specificity PPV NPV Accuracy Test Cut-off (%) (%) (%) (%) (%) LR+ LR− Fibro 0.6275 .sup.a 83.6 79.4 57.8 93.5 80.4 4.05 0.21 Meter.sup.V2G MFM ≥F3 ± 1 .sup.b 75.0 85.7 64.0 91.0 83.0 5.25 0.29 PPV: positive predictive value, NPV: negative predictive value, LR: likelihood ratio, MFM: multi-target FibroMeter .sup.a Maximum Youden index .sup.b i.e., between classes F2 ± 1 and F3 ± 1
(120) Cirrhosis
(121) Cirrhosis diagnosis—Cirrhosis is an important diagnostic target. Fibrosis classification by MFM compared favorably to the other tests, especially with CirrhoMeter.sup.V2G: the sensitivity for cirrhosis of fibrosis classes including F4 was 91.3% vs. 93.0%, respectively; the positive predictive value (PPV) for cirrhosis of the F4 class was 96.0% vs. 88.0% respectively.
(122) Cirrhosis classification—Areas of porto-septal fibrosis (median (IQR)) were, in Metavir staging: F3: 2.7% (2.2), F4: 5.2% (6.5); and in MFM classes: F3±1: 2.3% (3.9), F3/4: 3.2% (3.9), F4: 7.3% (4.3). Thus, MFM was able to distinguish early (F3/4) and definitive (F4) cirrhosis.
(123) Classification Precision and Refinement
(124) Precision evaluates the capability of a fibrosis test classification to precisely reflect Metavir F stage. The mean F scores varied from 1.84±1.08 to 2.13±0.84 among test classifications (p<0.001). This showed that the classification precision had differed from one test to another. Therefore, the precision was comprehensively evaluated using four criteria: agreement, difference and linearity of test classification with Metavir F staging, and dispersion of Metavir F stages within test classes. Briefly, MFM classification had satisfactory precision criteria among the new tests (details in Table 22 below).
(125) TABLE-US-00024 TABLE 22 Fibrosis classification precision: agreement, exactness, dispersion and linearity. Derivation population (1012 patients). Metavir FMF ≥ 1/ Simplified F FM ≥ 1 FM ≥ 2 FM2G FMF ≥ 2 FM = 4 CM2G MFM MFM General characteristics: Class number 5 4 6 7 6 6 6 6 6 F score (mean ± SD) 1.84 ± 1.08 1.99 ± 0.70 1.91 ± 0.93 1.90 ± 0.97 2.07 ± 0.80 2.05 ± 0.86 2.02 ± 0.97 2.13 ± 0.84 1.82 ± 1.08 p vs Metavir .sup.a — <0.001 0.014 0.037 <0.001 <0.001 <0.001 <0.001 0.267 Agreement with Metavir F: Weighted kappa — 0.471 0.600 0.664 0.529 0.534 0.641 0.563 0.703 Intra-class correlation — 0.671 0.775 0.806 0.746 0.746 0.804 0.780 0.826 coefficient Exactness (F difference with Metavir): Absolute difference .sup.b — 0.73 ± 0.55 0.68 ± 0.55 0.65 ± 0.51 0.70 ± 0.55 0.70 ± 0.57 0.66 ± 0.54 0.68 ± 0.54 0.65 ± 0.52 Raw difference — 0.14 ± 0.91 0.07 ± 0.87 0.05 ± 0.83 0.22 ± 0.86 0.20 ± 0.88 0.18 ± 0.84 0.29 ± 0.82 0.03 ± 0.83 Dispersion (mean number of F 1 2.83 ± 0.38 2.63 ± 0.53 2.25 ± 0.58 2.75 ± 0.49 2.70 ± 0.51 2.30 ± 0.56 2.72 ± 0.50 1.84 ± 0.36 stages/fibrosis class) Linearity (correlation .sup.c with) Metavir F — 0.554 0.640 0.680 0.623 0.612 0.676 0.661 0.703 Porto-septal fibrosis area .sup.d 0.550 0.238 0.288 0.326 0.197 0.329 0.356 0.226 0.354 FM2G: FibroMeter.sup.V2G, CM2G: CirrhoMeter.sup.V2G, FMF: single-target test, MFM: multi-target FibroMeter, n: number of patients. Best results between non-invasive tests are depicted in bold (Metavir F is excluded) .sup.a Paired t test for F score between blood test and Metavir .sup.b Absolute difference in F score between test classification and Metavir stage (mean ± SD), i.e., deletion of minus sign in negative difference .sup.c Pearson correlation .sup.d Results obtained in validation population #1 (676 patients)
(126) However, the MFM classification had two imprecise classes including three F stages (i.e., large dispersion). Therefore, a simplified MFM classification having a maximum of two F stages per fibrosis class was developed (
(127) Multi-Target Test Validation
(128) Classification Accuracy in Validation Populations.
(129) Comparison between blood tests—As expected, due to loss of optimism bias, there was an accuracy decrease (from −4.0% to −5.0%) in fibrosis classifications of FibroMeter.sup.V2G, CirrhoMeter.sup.V2G and MFM in the CHC validation population #1 compared to the derivation population (Table 19). However, the overall accuracy of MFM was still significantly higher than those of FibroMeter.sup.V2G or CirrhoMeter.sup.V2G in validation populations #1 (CHC), #2 (HIV/CHC) and #3 (CHB).
(130) Comparison with VCTE—VCTEs were available in 647 patients from population #1 and 152 patients from population #3. MFM accuracy was not significantly different from VCTE accuracy (Table 24 below). Other diagnostic indices were close between MFM and VCTE, especially for cirrhosis diagnosis despite an optimism bias in favor of VCTE. For example, in population #1, the sensitivities for cirrhosis of fibrosis classes including F4 were 86.0% and 81.7%, respectively for MFM and VCTE; the PPVs for cirrhosis of the F4 class were 80.0% and 76.7%, respectively.
(131) Validation of Fibrosis Classes
(132) Diagnostic population—The MFM fibrosis classification was validated by good correlations with other liver fibrosis descriptors, namely histological Metavir F, porto-septal fibrosis area, and liver stiffness measured by VCTE. More importantly, these liver fibrosis descriptors were significantly different between adjacent fibrosis classes of the MFM test.
(133) TABLE-US-00025 TABLE 23 Simplified classification of multi-target test (MFM). The cut-offs of this classification and those of the exhaustive classification are different; thus, two new classes of the simplified classification (F1 and F2/3) lie across two classes of the exhaustive classification. x denotes the predominant F stages per fibrosis class. Fine dashed lines delineate the employed parts of the single-target tests (left column); coarse dashed lines delineate fibrosis classes of the multi-target test (right columns). Derivation population (1012 patients). Single-target test Metavir F stage Multi-target test Test name Cut-off 0 1 2 3 4 Class Accuracy (%) FMF ≥ 1 0 to ≤0.92 x x — — — F0/1 89.7 >0.92 to MFM .sup.a — x — — — Fl 73.4 FMF ≥ 2 MFM .sup.a to ≤0.28 >0.28 to ≤0.758 — x x — — F1/2 86.6 >0.758 to MFM .sup.a — — x x — F2/3 70.7 FMF = 4 MFM .sup.a to ≤0.135 >0.135 to ≤0.71 — — — x x F3/4 65.8 >0.71 to 1 — — — — x F4 86.0 Accuracy (%) — 72.7 90.4 71.4 70.9 78.3 — 80.4 .sup.b x denotes predominant Metavir F stage .sup.a MFM cut-offs (see FIG. 2B in main text). .sup.b Overall accuracy in the whole population
(134) TABLE-US-00026 TABLE 24 Overall accuracy (OA in %) of blood tests and VCTE (Fibroscan) classifications in two validation populations. MFM FibroMeter.sup.V2G CirrhoMeter.sup.V2G VCTE Population OA OA p.sup.a OA p.sup.a OA p.sup.a p.sup.b #1 88.6 84.1 <0.001 83.0 <0.001 87.8 0.691 0.009 #3 80.9 80.3 1 74.3 0.031 80.9 1 0.121 MFM: multi-target FibroMeter, VCTE: vibration-controlled transient elastography Bold figures indicate significant differences. .sup.aComparison vs. MFM by paired McNemar test .sup.bComparison of VCTE vs. CirrhoMeter.sup.V2G by paired McNemar test
(135) Prognostic Population
(136) Population Characteristics
(137) All subjects over 18 years of age who were received for consultation or hospitalized for a chronic liver disease in the Department of Hepatology at the University Hospital of Angers from January 2005 to December 2009 were invited to join a study cohort, whatever the severity or etiology of their disease (viral hepatitis, alcoholic liver disease, non-alcoholic fatty liver disease (NAFLD), other causes). The resulting 1559 patients were then followed until death, liver transplantation or Jan. 1, 2011. The study was approved by an Institutional Review Board (AC-2012-1507) and informed consent was obtained from all patients.
(138) Results
(139) The MFMc fibrosis classification was validated for prognostic ability of liver-related death (p<0.001 by log rank test).
(140) The MFMc classification offered good prognostic discrimination, especially between four fibrosis classes: F2±1, F3±1, F3/4 and F4. The prognostic discrimination between the F3/4 and F4 classes was improved compared to FibroMeter.sup.V2G (
(141) These results will raise the question as to whether a simplified or exhaustive classification should be used. An exhaustive classification (up to three F per class) has the apparent advantage of better accuracy compared to a simplified classification (up to two F per class). However, the latter offers better precision and prognostication. Thus, a simplified classification seems sufficient for clinical practice. The lack of interest of an exhaustive classification can be attributed to the sources of misclassification by histological staging (sample size and observer reading). This is reinforced by the better prognostication by non-invasive tests than by histological staging (21). Finally, prognostication is significantly altered only by F2±1 or even F2/3 class, and thus the minimal classification can be described into four classes: F0/1 (non-significant fibrosis), F2/3 (significant fibrosis), F3/4 (early cirrhosis) and F4 (definitive cirrhosis).
REFERENCES
(142) 1. Oberti F, Valsesia E, Pilette C, et al. Noninvasive diagnosis of hepatic fibrosis or cirrhosis. Gastroenterology 1997; 113:1609-16. 2. Chou R, Wasson N. Blood tests to diagnose fibrosis or cirrhosis in patients with chronic hepatitis C virus infection: a systematic review. Annals of internal medicine 2013; 158:807-20. 3. Boursier J, Bertrais S, Oberti F, et al. Comparison of accuracy of fibrosis degree classifications by liver biopsy and non-invasive tests in chronic hepatitis C. BMC Gastroenterol 2011; 11:132. 4. Cales P, Boursier J, Oberti F, et al. Cirrhosis Diagnosis and Liver Fibrosis Staging: Transient Elastometry Versus Cirrhosis Blood Test. Journal of clinical gastroenterology 2014. 5. Cales P, de Ledinghen V, Halfon P, et al. Evaluating the accuracy and increasing the reliable diagnosis rate of blood tests for liver fibrosis in chronic hepatitis C. Liver Int 2008; 28:1352-62. 6. Boursier J, de Ledinghen V, Zarski J P, et al. Comparison of eight diagnostic algorithms for liver fibrosis in hepatitis C: new algorithms are more precise and entirely noninvasive. Hepatology 2012; 55:58-67. 7. Boursier J, de Ledinghen V, Zarski J P, et al. A new combination of blood test and fibroscan for accurate non-invasive diagnosis of liver fibrosis stages in chronic hepatitis C. Am J Gastroenterol 2011; 106:1255-63. 8. Calès P, Halfon P, Batisse D, et al. Comparison of liver fibrosis blood tests developed for HCV with new specific tests in HIV/HCV co-infection J Hepatol 2010; 52:238-44. 9. Leroy V, Sturm N, Faure P, et al. Prospective evaluation of FibroTest®, FibroMeter®, and HepaScore® for staging liver fibrosis in chronic hepatitis B: comparison with hepatitis C. J Hepatol 2014; 61:28-34. 10. Intraobserver and interobserver variations in liver biopsy interpretation in patients with chronic hepatitis C. The French METAVIR Cooperative Study Group. Hepatology 1994; 20:15-20. 11. Zarski J P, Sturm N, Guechot J, et al. Comparison of nine blood tests and transient elastography for liver fibrosis in chronic hepatitis C: The ANRS HCEP-23 study. J Hepatol 2012; 56:55-62. 12. Sandrini J, Boursier J, Chaigneau J, et al. Quantification of portal-bridging fibrosis area more accurately reflects fibrosis stage and liver stiffness than whole fibrosis or perisinusoidal fibrosis areas in chronic hepatitis C. Mod Pathol 2014; 27:1035-45. 13. Cales P, Oberti F, Michalak S, et al. A novel panel of blood markers to assess the degree of liver fibrosis. Hepatology 2005; 42:1373-81. 14. Cales P, Boursier J, Bertrais S, et al. Optimization and robustness of blood tests for liver fibrosis and cirrhosis. Clin Biochem 2010; 43:1315-22. 15. Cales P, Boursier J, Oberti F, et al. FibroMeters: a family of blood tests for liver fibrosis. Gastroenterol Clin Biol 2008; 32:40-51. 16. Castera L, Forns X, Alberti A. Non-invasive evaluation of liver fibrosis using transient elastography. J Hepatol 2008; 48:835-47. 17. Leroy V, Halfon P, Bacq Y, et al. Diagnostic accuracy, reproducibility and robustness of fibrosis blood tests in chronic hepatitis C: a meta-analysis with individual data. Clin Biochem 2008; 41:1368-76. 18. Bossuyt P M, Reitsma J B, Bruns D E, et al. The STARD statement for reporting studies of diagnostic accuracy: explanation and elaboration. Clin Chem 2003; 49:7-18. 19. Boursier J, de Ledinghen V, Poynard T, et al. An extension of STARD statements for reporting diagnostic accuracy studies on liver fibrosis tests: The Liver-FibroSTARD standards. J Hepatol 2014. 20. Boursier J, Brochard C, Bertrais S, et al. Combination of blood tests for significant fibrosis and cirrhosis improves the assessment of liver-prognosis in chronic hepatitis C. Alimentary pharmacology & therapeutics 2014; 40:178-88. 21. Naveau S, Gaude G, Asnacios A, et al. Diagnostic and prognostic values of noninvasive biomarkers of fibrosis in patients with alcoholic liver disease. Hepatology 2009; 49:97-105. 22. Cales P, Boursier J, Ducancelle A, et al. Improved fibrosis staging by elastometry and blood test in chronic hepatitis C. Liver international 2014; 34:907-17. 23. Boursier J, Zarski J P, de Ledinghen V, et al. Determination of reliability criteria for liver stiffness evaluation by transient elastography. Hepatology 2013; 57:1182-91.
Example 3: Construction of the Multi-Targeted Classification in the MFMc
(143) The objective was to select and combine the most accurate parts of the three retained test classifications (
(144) Practically, the analysis was first started with the early F stages (
(145) The same calculation was then repeated to determine the best cut-off of FMF≥2 score (Table 2,
(146) The same calculations were then repeated to compare the FMF≥1/FMF≥2 classification to the FMF=4 classification (Tables 27 and 28,
(147) The relationship between the 3 scores included and their 3 respective parts retained is shown in
(148) TABLE-US-00027 TABLE 25 Comparison of correctly classified patients (%) between FMF ≥ 1 and FMF ≥ 2 scores as a function of growing FMF ≥ 1 cut-off. Correctly classified patients (%) by FMF ≥ 1 FMF ≥ 1 FMF ≥ 2 Both Cut-off < .sup.a ≥ .sup.b p < .sup.a ≥ .sup.b p < + ≥ .sup.c 0.7 100 91.5 <0.001 100 90.5 <0.001 90.6% 0.8 100 91.2 <0.001 100 90.2 <0.001 90.6% 0.85 98.7 91 <0.001 98.7 89.9 <0.001 90.6% 0.9 92.9 91.4 0.526 91 90.5 0.8969 90.9% 0.95 93.5 90.9 0.188 89.9 90.9 0.613 91.6% 0.96 93.9 90.6 0.054 90.7 90.6 0.929 91.6% 0.97 94.7 89.7 0.006 91.5 90.1 0.438 91.8% 0.98 94.3 89.6 0.007 91.3 90.1 0.511 91.9% 0.99 94.1 88.9 0.003 91.2 90 0.492 92.1% .sup.d 0.995 93.9 88.1 0.001 91.6 89.1 0.168 92.0% 0.997 93.8 87.6 0.001 91.5 89 0.208 92.1% 0.999 92 90.3 0.391 91 89.5 0.491 91.4% 0.9995 91.7 91 0.751 90.9 89.6 0.565 91.3% 0.9998 91.4 92.6 0.53 90.8 89.4 0.605 91.1% .sup.a Correctly classified patients (%) below the FMF ≥ 1 cut-off. .sup.b Correctly classified patients (%) beyond the FMF ≥ 1 cut-off. .sup.c Sum of correctly classified patients (%) below the FMF ≥ 1 cut-off by FMF ≥ 1 plus correctly classified patients (%) by FMF ≥ 2 beyond the FMF ≥ 1 cut-off. .sup.d Maximum rate deteimining the cut-off choice.
(149) TABLE-US-00028 TABLE 26 Comparison of correctly classified patients (%) between FMF ≥ 1 and FMF ≥ 2 scores as a function of growing FMF ≥ 2 cut-off. Correctly classified patients (%) by FMF ≥ 2 FMF ≥ 1 FMF ≥ 2 Both Cut-off < .sup.a ≥ .sup.b p < .sup.a ≥ .sup.b p <+≥ .sup.c 0.1 100 91.2 <0.001 100 90.2 <0.001 90.6% 0.2 93.9 91 0.142 90.9 90.5 0.873 91.2% 0.21 94.4 90.9 0.06 90.7 90.6 0.981 91.4% 0.22 94.3 90.8 0.058 89.6 91.2 0.277 91.9% 0.23 93.5 91 0.177 87.9 91.5 0.116 92.0% 0.25 93.8 90.8 0.096 88.3 91.5 0.154 92.1% 0.26 93.6 90.8 0.128 88.3 91.5 0.136 92.1% 0.27 93.9 90.7 0.074 88.7 91.4 0.215 92.1% .sup.d 0.28 93.7 90.7 0.091 88.7 91.4 0.184 92.1% 0.29 93.3 90.8 0.165 88.5 91.5 0.149 92.1% 0.3 93.6 90.7 0.101 89 91.4 0.231 92.1% 0.35 94 90.1 0.021 90.4 90.7 0.85 92.0% 0.4 94.2 89.7 0.009 90.9 90.4 0.797 92.0% 0.5 94.5 88.7 0.001 91.7 89.5 0.221 92.0% 0.6 95.1 86.7 <0.001 92.7 87.6 0.009 92.0% 0.7 94.4 86.1 <0.001 92.5 87 0.008 91.9% 0.8 93.2 87.3 0.008 91.7 87.6 0.068 91.7% 0.9 92 89.5 0.313 90.8 89.5 0.592 91.6% .sup.a Correctly classified patients (%) below the FMF ≥ 2 cut-off. .sup.b Correctly classified patients (%) beyond the FMF ≥ 2 cut-off. .sup.c Sum of correctly classified patients (%) below the FMF ≥ 2cut-off by FMF ≥ 1 plus correctly classified patients (%) by FMF ≥ 2 beyond the FMF ≥ 2 cut-off. .sup.d Maximum rate determining the cut-off choice.
(150) TABLE-US-00029 TABLE 27 Comparison of correctly classified patients (%) between FMF ≥ 1/FMF ≥ 2 classification and FMF = 4 score as a function of growing FMF ≥ 2 cut-off. Correctly classified patients (%) by FMF ≥ 2 FMF ≥ 1/FMF ≥ 2 FMF = 4 Both Cut-off < .sup.a ≥ .sup.b p < .sup.a ≥ .sup.b p < + ≥ .sup.c 0.2 93.9 91.6 92.9 90.8 91.4% 0.3 93.6 91.4 0.213 91.4 91.1 0.873 91.9% 0.33 94.1 91 0.065 91.6 91 0.721 92.1% 0.35 94.3 90.7 0.033 91.9 90.7 0.508 92.1% 0.37 94.1 90.79 0.049 91.8 90.79 0.563 92.1% 0.38 94.2 90.65 0.033 92 90.65 0.448 92.1% 0.39 94.3 90.54 0.024 92.1 90.54 0.369 92.1% 0.4 94.4 90.4 0.016 92.1 90.6 0.408 92.2% 0.5 94.7 89.5 0.002 92.5 89.9 0.139 92.3% 0.55 95 88.5 <0.001 92.7 89.4 0.071 92.5% 0.6 95.3 87.6 <0.001 92.7 89.1 0.043 92.7% 0.63 95.3 87.2 <0.001 92.7 88.9 0.049 92.8% 0.64 95.2 87.2 <0.001 92.6 89 0.059 92.8% 0.65 95.1 87.1 <0.001 92.4 89.2 0.091 92.9% 0.66 95.1 86.9 <0.001 92.5 89 0.072 92.9% 0.67 95.2 86.6 <0.001 92.4 89.1 0.088 93.0% .sup.d 0.68 95.1 86.8 <0.001 92.4 89 0.078 92.9% 0.69 94.7 87.3 <0.001 92.4 89 0.079 92.7% 0.7 94.7 87 <0.001 92.5 88.7 0.045 92.6% 0.8 93.8 87.6 0.005 92.4 88 0.045 92.2% 0.9 92.6 89.5 0.212 91.6 89.5 0.413 92.1% 0.95 92.5 88.7 0.237 91.1 91.5 0.906 92.4% 0.97 92.5 86.3 0.136 91.4 89 0.54 92.2% .sup.a Correctly classified patient (%) below the FMF ≥ 2 cut-off. .sup.b Correctly classified patients (%) beyond the FMF ≥ 2 cut-off. .sup.c Sum of correctly classified patients (%) below the FMF ≥ 2 cut-off by FMF ≥ 1/FMF ≥ 2 plus correctly classified patients (%) by FMF = 4 beyond the FMF ≥ 2 cut-off. .sup.d Maximum rate determining the cut-off choice.
(151) TABLE-US-00030 TABLE 28 Comparison of correctly classified patients (%) between FMF ≥ 1/FMF ≥ 2 classification and FMF = 4 score as a function of growing FMF = 4 cut-off Correctly classified patients (%) by FMF = 4 FMF ≥ 1/FMF ≥ 2 FMF = 4 Both Cut-off < .sup.a ≥ .sup.b p < .sup.a ≥ .sup.b p < + ≥ .sup.c 0.005 93.7 91.8 0.373 88.7 91.7 0.267 92.0% 0.006 94.8 91.5 0.182 90.2 91.5 0.583 92.1% 0.007 94.1 91.6 0.073 89.5 91.7 0.343 92.2% 0.01 93.3 91.6 0.35 89.8 91.8 0.342 92.2% 0.02 94.3 90.3 0.016 91.2 91.2 0.982 92.6% 0.03 94.9 88.9 0.001 92.5 89.7 0.127 92.5% 0.04 94.6 88.3 0.001 92.8 88.8 0.037 92.3% 0.05 94.7 87.1 0.001 93.1 87.6 0.008 92.3% 0.08 94.8 85.2 0.001 92.9 87 0.008 92.6% 0.085 94.9 84.7 <0.001 92.8 86.9 0.009 92.7% .sup.d 0.09 94.8 84.6 <0.001 92.8 86.8 0.009 92.7% 0.095 94.7 84.4 <0.001 92.7 86.8 0.011 92.7% 0.1 94.6 84.5 <0.001 92.8 86.5 0.008 92.6% 0.15 93.3 87 0.015 92 88 0.122 92.3% 0.2 92.8 88.5 0.115 91.4 90.4 0.704 92.4% .sup.a Correctly classified patient (%) below the FMF = 4 cut-off. .sup.b Correctly classified patients (%) beyond the FMF = 4 cut-off. .sup.c Sum of correctly classified patients (%) below the FMF = 4 cut-off by FMF ≥ 1/FMF ≥ 2 plus correctly classified patients (%) by FMF = 4 beyond the FMF = 4 cut-off. .sup.d Maximum rate determining the cut-off choice.