7 févr. 2012 LiverCenter Non Invasive Biomarkers in chronic hepatitis C and B DU 2016 Thierry Poynard + AP-HP Groupe Hospitalier Pitié Salpêtrière, UPMC Liver Center, Université Paris 6, INSERM U680, Biopredictive France
7 févr. 2012
LiverCenter
Non Invasive Biomarkers in chronic hepatitis C and B
DU 2016
Thierry Poynard +
AP-HP Groupe Hospitalier Pitié Salpêtrière, UPMC Liver Center, Université Paris 6, INSERM U680, Biopredictive France
Serum Biomarker Imaging Biomarker
FibroTest FibroMax
Choice Hepatologist
Epidemiologist GP
FibroScan Aixplorer
2
Biomarkers of liver injury in chronic hepatitis
• Unmet need
• Historic
• Methods and based evidence
• Guidelines in practice
2
FRANCE FIBROSIS ICEBERG
Biopsy: 2% (8 000 /yr) FibroTest: 10% (50 000/yr) Imaging: 10% (50 000/yr)
No-estimate: 78%
0.25 Million Chronic Hepatitis C 0.25 Million Chronic Hepatitis B
USA FIBROSIS ICEBERG
Biopsy: 1% (50 000 /yr) FibroSure: 1% (50 000 /yr) Imaging: 1% (50 000 /yr)
No-estimate: 97%
3 Millions Chronic Hepatitis C 1 Million Chronic Hepatitis B
"META-analysis of histological data in VIRal hepatitis"
Hepatology 1996
METAVIR
THE METAVIR cooperative group. Inter- and intra-observer variation, Hepatology 1995
7 févr. 20127 févr. 2012
Viral necrosis Activity
Fibrosis Steatosis
Alcohol Ash
Nash
Liver Injury
8
7 févr. 2012
FibroMAX: HCV-HBV-ALD-NAFLD
ActiTest
FibroTest SteatoTest
AshTest
NashTest
FibroMAX
7
2
Biomarkers of liver injury in chronic hepatitis
• Unmet need
• Historic
• Methods and based evidence
• Guidelines in practice
2
7 févr. 2012
Fibrosis biomarkers: 24 years history
SJG 2008
n=100
n=1 million
7 févr. 2012
Haptoglobin
Alpha2Macroglobulin
Apolipoprotein A1
Total Bilirubin
Gamma GT
In Situ In Serum: FibroTest
Imbert-Bismut, Lancet 2001
Liver Injury
Activated Stellate CellsFibrotic Matrix
7 févr. 2012
Rational of FibroTest:
• Alpha 2 macroglobulin: key protein for Collagenase metabolism
• Apolipoprotein A1 key protein for Collagen trapping
• Haptoglobin: key protein for binding Free Hemoglobin oxidant
• Total Bilirubin: specific marker of severe late Fibrosis
• Gamma Glutamyl Transpeptidase: sensitive marker of early Fibrosis
• No transaminases: to prevent inflammatory necrosis confusion (ActiTest)
• Proteomic has blindly proved the major diagnostic value of
• Apolipoprotein A1, A2M
• HaptoglobinParadis Cell Mol Biol 1996, Paradis Hepatology 1996, Mathurin Hepatology 1996, Imbert Bismut 2001, Langlois 2006, Watanabe 2009, Ho 2010
2
Biomarkers of liver injury in chronic hepatitis
• Unmet need
• Historic
• Methods and based evidence
• Guidelines in practice
2
7 févr. 2012
Period 1: 1991-2004 Optimistic
Looking for a fibrosis biomarker with accuracy > 90%
7 févr. 2012
Biopsy =
Gold Standard
Biopsy=
0% False Positive0% False Negative
7 févr. 2012
Liver Injury
Serum biomarker Imaging biomarker
F4
F1
F0
Fibrotic Liver Disease
F2
F3
Hemorrhage Liver failure Cancer
FibroTest OK AUROC >80%
FibroTest OK FibroScan OKAUROC >80%
«Gray Zone»: Biopsy
Imbert Bismut 2001, Castera 2005
7 févr. 2012
Period 2: 2005-2009: Sceptic
Standard statistical methods were inappropriate
Period 3: 2010-2015
New methods
19
7 févr. 2012
• Sampling error Bedossa 2003
• Inter-observers variability Rousselet 2005
• Discordance studies Poynard 2004, Halfon 2006
• Prognostic studies Ngo 2006, Vergniol 2011
• Spectrum effect Poynard 2007, Lambert 2008
• Exceeding limits of biopsy Metha 2009
• Biopsy has a gray zone Poynard 2012
• Direct meta-analyses Poynard 201522
8 Key methodological issues:Biopsy is no more a perfect gold standard
Sampling error:AUROCs (F1 vs F2) of Biopsy vs Whole Liver according to length
Bedossa Hepatology 2003
AUROC 15 mm = 0.82 AUROC 25 mm = 0.89
«We showed that with 25-mm long biopsy specimens, only 75% were scored correctly»
7 févr. 2012
F0 F1 F2 F3 F4
Inter-Observers variability:Biopsy has lower inter-observers concordance for intermediate stages
Rousselet, Hepatology 2005
22
7 févr. 2012
Discordances studies: independent endpoints
• 537 prospective cases hepatitis C
• 154 (29%) discordances FibroTest/Biopsy
• Error attributable
• To FibroTest: 2%
• To Biopsy: 18%
25
Poynard Clin Chem 2004, Halfon AJG 2006
F4.1
F1
F0
F2
F3
7 Stages Presumed by Biomarkers
Decompensated
F4.2
F4.3
Varices
FibroTest
0.48
0.74
0.85
0.95
TE
7.1
9.5
20
50
CHC Poynard J Hepatol 2014 CHB Poynard J Hepatol 2014
12.5
0.58
F4.1
F1
F0
F2
F3
7 Stages Presumed by Biomarkers
Decompensated
F4.2
F4.3
Varices
FibroTest
0.48
0.74
0.85
0.95
TE
7.1
9.5
20
50
CHC Poynard J Hepatol 2014 CHB Poynard J Hepatol 2014
12.5
0.58
25
7 févr. 2012
• Sampling error Bedossa 2003
• Inter-observers variability Rousselet 2005
• Discordance studies Poynard 2004, Halfon 2006
• Prognostic studies Ngo 2006, Vergniol 2011
• Spectrum effect Poynard 2007, Lambert 2008
• Exceeding limits of biopsy Metha 2009
• Biopsy has a gray zone Poynard 2012
29
3/7 key methodological issues not well understoodBiopsy is no more a perfect gold standard
F4
F1
F0
Fibrotic Liver Disease
F2
F3
DANA=4
DANA=Difference between Advanced and non-advanced fibrosis stages
Obuchowski measure=AUROCs Pair-wise comparison between all stages
Black and White Spectrum
FibroTest AUROC=0.98
F4
F1
F0
Fibrotic Liver Disease
F2
F3
DANA=1
DANA=Difference between Advanced and non-advanced fibrosis stages
Obuchowski measure=AUROCs Pair-wise comparison between all stages
Gray Spectrum
FibroTest AUROC=0.67
F4
F1
F0
Fibrotic Liver Disease
F2
F3
DANA=2.5
DANA=Difference between Advanced and non-advanced fibrosis stages
Obuchowski measure=AUROCs Pair-wise comparison between all stages
FibroTest AUROC=0.85
Standard Spectrum
7 févr. 2012
Hazardous AUROC Scores for defining test performance:
AUROC Score* Biopsy (length) FibroTest (Spectrum)
0.90-1.00 Excellent 100mm F1 vs F2 F0 vs F4
0.80-0.90 Good 25 mm F1 vs F2 F01 vs F234
0.70-0.80 Fair 5 mm F1 vs F2 F0 vs F2
0.60-0.70 Poor 5 mm F0 vs F1 F1 vs F2
0.50-0.60 Fail
*Sebastiani CCLM 2011, Bedossa Hepatology 2003, Poynard Clin Chem 2007
7 févr. 2012
Hazardous AUROC Scores for defining test performance:
AUROC Score* FibroTest Spectrum DANA
0.90-1.00 Excellent 50% F0 vs 50% F4 4
0.80-0.90 Good F01 vs F234 20% each stage 2.5
0.70-0.80 Fair 50% F0 vs 50% F2 2
0.60-0.70 Poor 50% F1 vs 50% F2 1
0.50-0.60 Fail
Poynard Clin Chem 2007, Lambert Clin Chem 2008,
7 févr. 2012
Hazardous Tables due to Spectrum Effect (October 2012)
Ochi Hepatology 2012
Real-time tissue elastography cut-off values by stage in the training set were 2.47 for F1, 2.67 for F2, 3.02 for F3, and 3.36 for F4. Usingthese cut-off values, the diagnostic accuracy of hepatic fibrosis in the validation set was 82.6%-96.0% in all stages.
The area under the receiver operating characteristic curve of elastic ratio better correlated than serum fibrosis markers in both early and advanced fibrosis stages.
Conclusion: Real-time tissue elastography is useful in evaluating hepatic fibrosis and PH in patients with NAFLD. (HEPATOLOGY 2012;1271-1278)
32
7 févr. 2012
• Sampling error Bedossa 2003
• Inter-observers variability Rousselet 2005
• Discordance studies Poynard 2004, Halfon 2006
• Prognostic studies Ngo 2006, Vergniol 2011
• Spectrum effect Poynard 2007, Lambert 2008
• Exceeding limits of biopsy Metha 2009
• Biopsy has a gray zone Poynard 2012
29
3/7 key methodological issues not well understoodBiopsy is no more a perfect gold standard
Using 25 mm liver biopsy a perfect market cannot be validated
Black shading represents the set of conditions under which the AUROC values exceed what has already been observed
Metha J Hepatol 2009
34
7 févr. 2012
Exceeding limits of biopsy: >90% accuracy is impossible for advanced fibrosis
35
«Comparison of 8 diagnostic algorithms for liver fibrosis in hepatitis C: New algorithms are more precise and entirely non-invasive».
Boursier et al, Hepatology 2012
7 févr. 2012
Misleading presentation using biopsy as Gold-Standard
Boursier Hepatology 2012
Mathematically impossible with biopsy as «Gold Standard
36
7 févr. 2012
• Sampling error Bedossa 2003
• Inter-observers variability Rousselet 2005
• Discordance studies Poynard 2004, Halfon 2006
• Prognostic studies Ngo 2006, Vergniol 2011
• Spectrum effect Poynard 2007, Lambert 2008
• Exceeding limits of biopsy Metha 2009
• Biopsy has a gray zone Poynard 2012
29
3/7 key methodological issues not well understoodBiopsy is no more a perfect gold standard
37
7 févr. 2012
Review of tests by Gebo, Hepatology 2002
« These panels of tests may have the greatest value in predicting fibrosis or cirrhosis »
« Biochemical tests were best at predicting no or minimal fibrosis, or at predicting advanced fibrosis/cirrhosis, and were poor at predicting intermediate levels of fibrosis »
37
FibroTest/FibroSure has a Gray Zone
Biopsy has a Gray Zone
41
7 févr. 2012
Review of fibrosis tests by Nguyen, Hepatology 2011
41
7 févr. 2012
Liver Biopsy Analysis Has a Low Level of Performance for Diagnosis of IntermediateStages of Fibrosis
The gray anatomy of 27,869 virtual biopsies and 6,500 patients
Poynard Clin Gastro Hepatol 2012 Poynard, BMC 2005, J Hepatol 2011
7 févr. 2012
The gray zone of liver biopsy: 27,864 virtual biopsies
Area Fibrosis (Log)
25 mm Liver Biopsies
Poynard Clin Gastro Hepatol 2012
7 févr. 2012
The gray zone of liver biopsy: 27,864 virtual biopsies
Poynard Clin Gastro Hepatol 2012
Area Fibrosis (Log)
25 mm Liver Biopsies
Lower gray zone of FibroTest relative to biopsy
Lower gray zone F2vsF1 for FibroTest vs Biopsy
58% lower F2vsF1 vs F1vsF0 41% lower F2vsF1 vs F4vsF3.
Biopsyn=27,864
Fibrotestn=6500
Poynard Clin Gastro Hepatol 2012
7 févr. 2012
Biopsy is no more a perfect gold standard
FibroTest and Elastography have similar performance
2006: Approval Markers French Health Authorities HCV2011: Guidelines EASL 2011
7 févr. 2012
Period 2: 2005-2009: Sceptic
Standard statistical methods were inappropriate
Period 3: 2010-2015
New methods
(c) BioPredictive 2008 - All Rights Reserved - No reproduction without written permission
Benefit/Risk must be evaluated for each change in the formula:
High Risk False Positive Negative
5/954 (0.52%)
High Risk False Positive Negative
38/7494 (0.51%)
FibroTest Global Quality Estimates
High Risk False Positive Negative 3349/345,695 (0.97%)
High Risk False Positive Negative
491/24,872 (1.97%)
FibroScan (Roulot et al 2008) >7.1 kPa= 12.6%: False Positives ?
Poynard BMC Gastro 2011, Roulot J Hepatol 2008
(c) BioPredictive 2008 - All Rights Reserved - No reproduction without written permission
One Test, One formula
360,000 FibroTest for Quality Control
Risk of False positive/negative of FibroTest
• Tertiary center: 1.97%
• HIV co-infection: 1.77%
• Sub-Saharan origin: 2.61%
7 févr. 2012
Which Fibrometer for patients with Hepatitis C ? Too many variants = Risk of false positive
FibroMeter Variant Year Components
FM-1G 2005 PLT, PI, AST, A2M, HA, Urea, Age
FM-2G V* 2008 + Gender
FM-3G 2008 Switch GGT/HA
FM-3G+ (CirrhoMeter) 2009 New formula for cirrhosis
FM-HICV 2010 AST, A2M, PI
CSF-Index 2011 Combined with LSM
SF-Index 2011 Combined with LSM
C-Index 2011 Combined with LSM
*ONLY one ( FM-2G V) is approved by Haute Autorité de Santé
PLT: platelet counts, PI prothrombin index, AST aspartate amino transferase, A2M alpha2 macroglobulin, HA hyaluronic acid
7 févr. 2012
7 févr. 2012
Biopsy vs Serum marker Main advantages/disadvantages
Serum Marker FibroTest
Less accurate for intermediate stages
No grey zone relatively to biopsy
Fibrosis only ActiTest/SteatoTest
Delays result proprietary tests 1-48h
False positive/hemolysis/inflammation/Gilbert
Yes but 0.97% (3349/345695; 0.94-1.00)
Nguyen Hepatology, 2011 Poynard BMC Gastro 2011
7 févr. 2012
Period 3: 2010-2015
Welcome in a world without perfect Gold Standard
7 févr. 2012
Gold Standard
25 mm Biopsy 0%False PositiveFalse Negative
7 févr. 2012
Truth in the Absence of
Gold Standard
25 mm Biopsy 25%False PositiveFalse Negative
7 févr. 2012
Area of fibrosis estimated by biopsy according to its length (mm) in subjects scoring METAVIR F0 (no fibrosis) on large surgical section.
Area of fibrosis >5.3%: 16.3% false positives 20mm biopsy for diagnosis of advanced fibrosis >16.5%: 0.3% false positives 20mm biopsy for diagnosis of cirrhosis.
Cirrhosis
Advanced fibrosis
Poynard J Hepatol 2012
7 févr. 2012
Poynard J Hepatol 2011
Truth
FibroTest FibroScan
5-30 mm Biopsy
ALT
7 févr. 2012
Distribution of 1893 subjects according to the 16 possible combinations of the 4 tests' results: presumed advanced fibrosis (present=1) or not (=0)
16 combinations of 4 tests results Number of subjects
FibroTest LSM ALT Biopsy Observed Expected by model
0 0 0 0 621 615.5
0 0 0 1 186 191.1
...
1 1 1 1 276 277.0
Poynard, J Hepatol 2011
7 févr. 2012
FibroTest Se LSM Se Biopsie Se
Performance for Cirrhosis: Sensitivity
The standard cutoffs: 0.74 FibroTest, 14.5 kPa Stiffness
Poynard, J Hepatol 2011
7 févr. 2012
FibroTest Sp LSM Sp Biopsy Sp
Performance for Cirrhosis: Specificity
The standard cutoffs: for cirrhosis 0.74 for FibroTest, and 14.5 kilo-Pascal for stiffness (LSM)
Poynard, J Hepatol 2011
7 févr. 2012
SWE Fibrotest 1 TE-M Fibrotest 2 TE-XL FibroTest 3
Poynard, J Hepatol 2013
Performances for diagnosis of Cirrhosis (HCV, HBV, NAFLD, ALD) of FibroTest, and Elastography: Transient M-XL probes and Share Wave
Latent Class Model: Best model for FibroTest with TE-XL or SWE (Likelihood ratio test 5.5, 6.9)
n = 322 simultaneous reliable tests
2
Biomarkers of liver injury in chronic hepatitis
• Unmet need
• Historic
• Methods and based evidence
• Guidelines in practice
2
Serum Biomarker Imaging Biomarker
FibroTest FibroMax
Choice Hepatologist
Epidemiologist GP
FibroScan SWE Aixplorer
Competitors
65
• AFEF, HAS
• EASL, AASLD
• WHO
Guidelines: HCV and HBV
7 févr. 20127 févr. 2012
FibroTest APRI FIB4TE
Imbert-Bismuth Lancet 2001
Sandrin Ultra Med Biol
2003
Wai Hepatology
2003
Sterling Hepatology
2006
• Applicability
• Variability
• False positive if activity
• False negative for cirrhosis
Fibroscan limitations
7 févr. 2012
Pitfalls of Fibroscan
3.1% Failures and Unreliable results 15.8%
23 sept. 2014
Performances for cirrhosis diagnosis
FibroTest Fibrosure Transient elastography
AUROC* 0.86 (0.71-0.92) 0.94 (0.93-0.95)
Applicability >95% 80 %
Afdhal, JVH Nov 2013 Chou, Ann Int Med 2013
Not in intention to diagnose
7 févr. 2012
Oliveri WJG 2008
7 févr. 2012
Choice of FibroScan Cutoffs
Castera 2005, Ketanneh 2007 Roulot 2008
For F2: 7.1 or 8.8 kPa ? Patients: false negatives ? Low negative predictive value
Healthy volunteers: 7.1 kPa 12.6% false positives ?
For screening 7.1 kPa ?
For patients 8.8 kPa ?
No rationale for changing cutoff according to liver disease
F2 8.8 kPa F4 14.5 kPa
F4 0.73
F2 0.48
Poynard PlosOne 2008
75
7 févr. 2012
Elasto-FibroTest® 1289 patients with CHC and 604 healthy volunteers
• For the diagnosis of cirrhosis Elasto-FibroTest has significantly higher performances than FibroTest or Fibroscan alone.
• For the diagnosis of advanced fibrosis (F234) no improvement in performance has been observed vs FibroTest alone, when a method without gold standard was used.
67
Poynard, CRHG 2012
8-week $ 63,000 12-Week $ 94,500 24-week $ 189,000
$ 1125 per pill
FibroTest used to identify cirrhosis at baseline or for follow-up
Number of Studies Quality Consistency Precision Strength of
evidence
32 Fair High High High
Chou, Ann Int Med 2013
FibroTest is the most validated test in Chronic Hepatitis C: Strength-of-evidence domains and overall ratings
Results:
Area under the ROC curves:
Significant fibrosis 0.84 (0.78 – 0.88) Cirrhosis 0.87 (0.85 – 0.90)
Am J Gastro 2013
APRI has lower performance than Fibrotest
due to its high variability
1. Analytical variability: ULN-AST definitions
2. Interaction with non-fibrosis features
• Activity and AST
• Steatosis and AST
Impact on performance: Obuchowski measure
Change in AUROCs between fibrosis stages
ULN 26 IU/L ULN >=30 IU/L
APRI 0,862 0,820
FibroTest 0,867 0,867
Significance 0,30 <0,0001
High variability of APRI associated with ULN definitions:
Impact on diagnosis performance
• Range of AST-ULN in controls: 26-49 IU/L
• According gender, BMI and cholesterol
• Fibrosis prevalence in CHC: spectrum effect
• Clinically significant fibrosis (F2F3F4): 35-69%
• Cirrhosis (F4): 11-32%
Higher Diagnostic and Prognostic performance of FibroTest versus APRI in CHC
Poynard, Ann Int Med 2013
5 years prognostic value in chronic hepatitis B FibroTest better biomarker
de Ledinghen APT 2013
LSM Not in intention to diagnose
10 year Prognostic value FibroTest versus TE n= 272445% CHC, 24% CHB, NAFLD 10%, ALD 6%
Biomarker
Analytic variability
Risk false positive due to
ActivityFasting Applicability Investment Cost
FibroTest < 7% no (ActiTest) no >95% 0 38€-200$
FibroScan Inter Observer yes yes 80 % 60,000€ 200,000$ 38€-350$
APRIHazard of AST
Upper Limit Normal
yes no ? 0 7€-40 $
Castera Hepatology 2010, Afdhal JVH 2013, Chou Ann Int Med 2013, Poynard BMC Gastro 2011, Poynard Ann Int Med 2013
Pro and Con of the three «Standard of Care» biomarkers
2
Biomarkers of liver injury in chronic hepatitis
• Unmet need
• Historic
• Methods and based evidence
• Guidelines in practice
2
2016
Direct comparisons between APRI, FIB4, FibroTest and TE n=185 comparisons: 99 Fibrosis n=12,725 / 86 Cirrhosis n=10,929
Houot APT 2016
FT better than TE
AUROC + 0.06
Fibrosis
Cirrhosis
FT = TE
AUROC + 0.00
13
Three reasons to assess fibrosis stages in 2016
•Expensive IFN-free regimen, priority to severe fibrosis
•Cirrhosis could need longer treatment
•Viral cure is not fibrosis cure
Cohen, Science 2013, Afdahl NEJM 2014, Poynard J Hepatol 2013
FibroTest similar to biopsy for estimating fibrosis progression
Progression to cirrhosis in 2472 patients
Biopsy FibroTest
Poynard et al, J Hepatol 2012
Response is associated with longer treatment in cirrhosis stage* vs non-cirrhosis in experienced Genotype 1, treated by Sofosbuvir-Ledispasvir:
80
85
90
95
100
12 weeks 24 weeks
Cirrhosis No Cirrhosis
Afdahl NEJM 2014*cirrhosis stage defined by biopsy or FibroTest
n=22 n=22n=87 n=87
23 sept. 2014
Survival without liver complications
n = 933NS
SVR n=43 HCC1 CholangiocarcinomaAll F4 before SVR2 F2 after
Poynard, J Hepatol 2013
Cirrhosis regression in SVR n=24/43 ( 56%)
FibroTest = 0.74
Poynard, J Hepatol 2013
Cirrhosis Occurrence in SVR n=15/128 (12%)
FibroTest = 0.74
Poynard, J Hepatol 2013
Fibrosis Progression in 13 HBV Sustained Virological Responders with occurrence of HepatoCellular Carcinoma at 10 years
FibroTest = 0.74 = F4
Poynard, J Hepatol 2014
F1: Subsaharan female, BMI 37 kg/m2
F4
F1
F0
France: 12,000,000 at Risk100%
5%
Death 15,000/year0.1%
Biomarker10% F2
F3
n= 1,016,557 (100 %) ActiTest
Poynard BmjOpen 2015
Fibrosis density Birth-Year and Gender
Other
USA
France
n= 470,762
Serum Biomarker Imaging Biomarker
FibroTest FibroMax
Choice Hepatologist
Epidemiologist GP
FibroScan Aixplorer