The evaluation of liver fibrosis is a key step to manage

Transcription

1 GASTROENTEROLOGY 2011;140: CLINICAL LIVER, BILIARY TRACT Noninvasive Tests for Fibrosis and Liver Stiffness Predict 5-Year Outcomes of Patients With Chronic Hepatitis C JULIEN VERGNIOL,* JULIETTE FOUCHER,*, ERIC TERREBONNE,* PIERRE HENRI BERNARD, BRIGITTE LE BAIL, WASSIL MERROUCHE,* PATRICE COUZIGOU,* and VICTOR DE LEDINGHEN*, *Centre d Investigation de la Fibrose hépatique, Hôpital Haut-Lévêque, CHU Bordeaux, Pessac, France; Service d Hépato-Gastroentérologie, Hôpital Saint-André, CHU Bordeaux, Bordeaux, France; Service d Anatomie-Pathologique, Hôpital Pellegrin, CHU Bordeaux, Bordeaux, France; INSERM U1053, Université Segalen, Bordeaux, France BACKGROUND & AIMS: Liver stiffness can be measured noninvasively to assess liver fibrosis in patients with chronic hepatitis C. In patients with chronic liver diseases, level of fibrosis predicts liver-related complications and survival. We evaluated the abilities of liver stiffness, results from noninvasive tests for fibrosis, and liver biopsy analyses to predict overall survival or survival without liver-related death with a 5-year period. METH- ODS: In a consecutive cohort of 1457 patients with chronic hepatitis C, we assessed fibrosis and, on the same day, liver stiffness, performed noninvasive tests of fibrosis (FibroTest, the aspartate aminotransferase to platelet ratio index, FIB-4), and analyzed liver biopsy samples. We analyzed data on death, liver-related death, and liver transplantation collected during a 5-year follow-up period. RESULTS: At 5 years, 77 patients had died (39 liver-related deaths) and 16 patients had undergone liver transplantation. Overall survival was 91.7% and survival without liver-related death was 94.4%. Survival was significantly decreased among patients diagnosed with severe fibrosis, regardless of the noninvasive method of analysis. All methods were able to predict shorter survival times in this large population; liver stiffness and results of FibroTest had higher predictive values. Patient outcomes worsened as liver stiffness and FibroTest values increased. Prognostic values of stiffness (P.0001) and FibroTest results (P.0001) remained after they were adjusted for treatment response, patient age, and estimates of necroinflammatory grade. CONCLUSIONS: Noninvasive tests for liver fibrosis (measurement of liver stiffness or FibroTest) can predict 5-year survival of patients with chronic hepatitis C. These tools might help physicians determine prognosis at earlier stages and discuss specific treatments, such as liver transplantation. Keywords: Survival; Cirrhosis; FibroTest; FibroScan; Hepatitis C. The evaluation of liver fibrosis is a key step to manage a chronic liver disease and to assess its prognosis, as complications mainly occur in patients with advanced stages. Portal hypertension, ascites, or hepatocellular carcinoma are associated with a shorter survival. Early assessment of the risk of bad prognosis helps the physician to manage patients with cirrhosis and to make decisions about liver transplantation. Liver biopsy does not satisfy quality criteria as a surrogate end-point marker because of its complication rate, sampling error, intra- and interobserver variability, expense, and patient reluctance to undergo serial monitoring. 1 Different scores such as the Child Pugh score or the Model of End Stage Liver Disease are correlated with the prognosis of advanced cirrhosis and help the physicians to take care of cirrhotic patients. 2,3 In the past decade, noninvasive markers have been developed as an alternative for liver biopsy to evaluate the severity of chronic liver diseases. FibroTest (Biopredictive, Paris, France; Fibrosure-Labcorp, Burlington, VT), a surrogate marker, displays a significant correlation with survival. It has a 5-year prognostic value similar to that of liver biopsy for the prediction of cirrhosis decompensation and survival in patients with chronic hepatitis C virus (HCV), 4 chronic hepatitis B, 5 and alcoholic liver disease. 6 Aspartate aminotransferase to platelet ratio index (APRI), 7 and the FIB-4 score 8 have significant diagnostic performance for fibrosis and cirrhosis, and recently Nunes et al showed that hyaluronic acid, APRI, and FIB-4 were significantly associated with mortality in patients with chronic HCV infection. 9 Abbreviations used in this paper: APRI, aspartate aminotransferase to platelet ratio index; AUROC, area under the receiver operating characteristic; CI, confidence interval; HCV, hepatitis C virus; HIV, human immunodeficiency virus by the AGA Institute /$36.00 doi: /j.gastro

2 June 2011 PROGNOSIS VALUE OF LIVER STIFFNESS MEASUREMENT 1971 Liver stiffness measurement using M probe of FibroScan (Echosens, Paris, France) is a noninvasive method for the diagnosis of liver fibrosis. It has a high degree of accuracy and reproducibility in predicting bridging fibrosis and cirrhosis in patients with chronic liver diseases In chronic hepatitis C, its performance is highly significant, with area under the receiver operating characteristic (AUROC) of 0.83 and 0.95 for the diagnosis of significant fibrosis and cirrhosis, respectively. 14 There is a relationship between the value of liver stiffness and various complications of cirrhosis, such as esophageal varices, variceal bleeding, portal hypertension, ascites, and hepatocellular carcinoma. 16 Recently, Masuzaki et al showed the association between liver stiffness and risk of hepatocellular carcinoma development in patients with hepatitis C. 17 Another study by Fung et al showed that the risks of hepatocellular carcinoma development and liver-related mortality were significantly higher in patients with liver stiffness 10 kpa in a population of patients with HBeAg-negative chronic hepatitis B. 18 To our knowledge, there is no study evaluating a relationship between survival and liver stiffness in chronic hepatitis C. The aim of this prospective study was to evaluate 5-year prognostic value of liver stiffness, FibroTest, APRI, and FIB-4 for predicting survival and liver-related death in patients with chronic hepatitis C. Materials and Methods Patients From April 2003 to February 2009, we prospectively collected data on a large cohort of consecutive patients presenting to our center with chronic hepatitis C. At baseline, we noted the clinical, biological, endoscopic, and radiologic parameters, and any liver decompensation in the past. During the follow-up, we collected the natural history events, such as liver complications, liver transplantation, or death. Study patients belonged to a prospective hospital-based cohort of the Hepatology Unit of Haut-Lévêque Hospital (University Hospital of Bordeaux, Pessac, France). In this cohort study, we targeted follow-up consultations and examinations every 6 or 12 months or closer. The end of the follow-up was March 1, Final analysis of the data was done in February Therefore, all patients had at least 1-year follow-up. We included all consecutive patients aged older than 18 years and with chronic hepatitis C of any severity. The determination of chronic hepatitis C was made using standard diagnostic criteria: serological detection of hepatitis C antibodies and positive serum HCV-RNA by polymerase chain reaction for 6 months. Exclusion criteria were chronic hepatitis B virus infection and all other causes of chronic liver disease. Patients with human immunodeficiency virus (HIV) infection were included. The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki. Patients were enrolled after written informed consent was obtained. The study was registered at the ClinicalTrials.gov (Identifier: NCT ). Clinical and Biological Parameters For all patients, the following parameters were determined at the time of liver stiffness. Clinical parameters included weight, height, past history of ascites or bleeding varices, and hepatocellular carcinoma. Biological parameters included aspartate aminotransferase, alanine aminotransferase, -glutamyl-transpeptidase, total bilirubin, platelet count, prothrombin time, and different scores: FibroTest, APRI, and FIB-4 sample analysis was centralized in the laboratory of our hospital. Esophageal varices were examined by upper gastrointestinal endoscopy. Because ascites is a physical limitation to the technique since waves do not propagate through liquids, patients with ascites were excluded. FibroTest and ActiTest scores were computed on the Biopredictive Web site ( 19 APRI index was calculated as follows: aspartate transaminase ( upper limit of normal) 100/platelet count (giga/l). 20 FIB-4 score was calculated as follows: (age [in years] aspartate aminotransferase [U/L])/(platelet count [giga/l] alanine aminotransferase [U/L] ½ ). 8 Liver Stiffness Measurement Liver stiffness measurement (LSM) was performed with the patient lying in dorsal decubitus with the right arm in maximal abduction, on the right lobe of the liver, through intercostal spaces. The operator, assisted by a time-motion ultrasound image, located a liver portion at least 6-cm thick and free of large vascular structures. When the target area had been located, he pressed the probe button to commence the measurements. The measurement depth was between 25 and 65 mm using the M probe of FibroScan. Ten validated measurements were performed on each patient. Results were expressed in kilopascals (kpa). Only procedures with at least 10 validated measurements and an interquartile range 30% of the median value were considered reliable. 12,21 Measurement of liver stiffness was performed in our unit by specialized nurses. Liver Biopsies Patients with possible indication of treatment or with discordance between FibroTest and LSM 14 had a liver biopsy, which was processed with standard techniques. Two pathologists, unaware of the biochemical markers, evaluated the stage of fibrosis and grade of activity according to the METAVIR scoring system. 22 Fibrosis was staged on a scale of 0 4: F0, no fibrosis; F1, portal fibrosis without septa; F2, few septa; F3, numerous septa without cirrhosis; F4, cirrhosis.

3 1972 VERGNIOL ET AL GASTROENTEROLOGY Vol. 140, No. 7 Follow up Patients had a conventional treatment of their diseases during the follow-up. The overall 5-year survival (whatever the cause of death, and including liver transplantation) was the a priori main end point used to compare the prognostic value of the different methods. Survival time was calculated from the date of liver stiffness to the end-point date (March 1, 2009). This interval was censored at the time of last follow-up. Each year, for patients who had not been seen at our hospital in the previous 12 months, we found out whether they were living and, if not, the date and the cause of death. For patients who were still alive, we either interviewed the patients or obtained information through their physicians. For patients who died outside our hospital, we obtained information about the date and cause of death from their physicians or family. The survival without liver-related death (including death related to liver disease and liver transplantation) was also assessed. Statistical Analysis For the analysis, 2 groups were defined. The core group was the group of patients with all liver fibrosis scores available (FibroTest, APRI, FIB-4, liver stiffness, liver biopsy). The non-core group (whole group) was the group of patients with more than one liver fibrosis score, therefore, with the highest power for pair-wise comparisons. Tests were used as follow: 2 test for qualitative comparisons, Mann-Whitney test for quantitative comparisons, AUROC for the diagnostic and univariate prognostic analyses. 23 The prognostic AUROCs were estimated by the empirical (nonparametric) method of DeLong et al, equivalent to the Mann-Whitney statistic and compared using the paired method. 24 Comparisons between diagnostic AUROCs were performed using the Obuchowski method to take into account the risk of multiple testing and the spectrum effect The Obuchowski measure allows comparison of 2 biomarkers with a single test, avoiding appropriate correction for the type I error when comparing 2 biomarkers for different stages or grades. 27 This measure is a multinomial version of the AUROC. With n ( 5), categories of the gold standard outcome (histological fibrosis stage) and AUROCst, the estimate of the AUROC of diagnostic tests for differentiating between categories s and t, the Obuchowski measure, is a weighted average of the N(N 1)/2 ( 10) different AUROCst corresponding to all the pairwise comparisons between 2 of the N categories. Each pairwise comparison has been weighted to take into account the distance between activity grades (ie, the number of units on the ordinal scale). A penalty function proportional to the difference in METAVIR units between grades was defined: the penalty function was 0.25 when the difference between stages was 1, 0.50 when the difference was 2, 0.75 when the difference was 3, and 1 when the difference was 4. The Obuchowski measure can be interpreted as the probability that the noninvasive index will correctly rank 2 randomly chosen patient samples from different activity grades according to the weighting scheme, with a penalty for misclassifying patients. The overall Obuchowski measure is not equivalent to a usual AUROC curve, as the measurements are weighted according to the distance between stages. For survival analyses, time-dependent Kaplan-Meier analysis for survival curves, the log-rank test for univariate comparisons and the Cox proportional hazard model for multivariate analysis were used. The hazard ratio, together with its 95% confidence interval (CI) and the corresponding P value were calculated for assessment of relative risks using logistic regression. Because of the limited number of events, only overall 5-year survival was detailed in the Results section. Because of possible colinearity effects, a sensitivity analysis was performed excluding from FibroTest analyses, age, and ActiTest. We used liver stiffness values expressed in log values to perform statistical analysis and comparisons. We used 2-sided statistical tests for all analyses. Due to the number of biomarker comparisons (n 4), a P value.01 was considered as significant, between.01 and.05 as borderline,.05 as not significant. Statistical analyses were performed using Number Cruncher Statistical Systems software 28 and R software [library(nonbinroc)and library(rocr)]. 29 Modeling for Multivariate Analysis The predetermined factors associated with prognosis were fibrosis as the main factor, and 3 covariates, ie, HCV treatment (none, no sustained virologic response, and sustained virologic response), age, and necroinflammatory grade at baseline presumed by ActiTest. Model 1 aimed to assess the prognostic value of each estimate of fibrosis including biopsy, independent of treatment impact, and age of the patient. Model 2 aimed to estimate the independent additive prognostic value of each test when fibrosis is already estimated using biopsy. Model 3 aimed to estimate the independent additive prognostic value of each serum test when fibrosis is already estimated using stiffness. Model 4 aimed to estimate the independent additive prognostic value of stiffness and each other serum test when fibrosis is already estimated using FibroTest. Model 5 aimed to estimate the value of combining the 4 best independent predictors identified by previous Models: stiffness an FibroTest, adjusted on treatment, age, and ActiTest. All the variables have been introduced initially in the regression analysis as continuous variables for continuous data: age and biomarkers. If using the continuous quantitative value of the Cox regression model, the maximum likelihood procedure cannot reach convergence (no completion), the semi-quantitative value of the continuous data was used in the multivariate model. To

4 June 2011 PROGNOSIS VALUE OF LIVER STIFFNESS MEASUREMENT 1973 Figure 1. Chart-flow of the study population. prevent any risk of misleading conclusion related to the use of continuous data for FibroTest and categorical classes for liver stiffness, we performed the multivariate analysis twice, once including the highest number possible of continuous variables and another including the biomarkers with the same number of categories to prevent artificial differences in the significance of respective hazard ratio. The adjustment factors (age, ActiTest) were still expressed as continuous variables in both models. Results Characteristics of Patients A total of 1830 patient were recruited (Figure 1). Among these patients, 214 (11.7%) were excluded for not interpretable LSM: liver stiffness failure (no valid shot) in 57 patients (3.1%), 10 valid measurements in 22 patients (1.2%), interquartile range/liver stiffness 0.30 in 135 patients (7.4%). Among these remaining 1616 patients (88.3%), 159 patients (10.1%) were lost to followup. Characteristics of lost patients were similar to those of the 1457 other patients, except for mean age ( v y; P.0001), baseline FibroTest value ( vs ; P.02), and prevalence of liver stiffness 9.5 kpa (131/159 vs 1069/1457; P.01). For APRI and FIB-4, no statistical difference was observed between lost patients and patients with long-term follow-up. A total of 1457 patients had a complete follow-up. Median follow-up was 47.3 months (interquartile range, months). Clinical and biological characteristics of the patients are indicated in Table 1. A majority of them were male (53.4%), and mean age was years. A total of 140 patients (9.6%) were HIV-HCV co-infected. For each noninvasive test, we used the most usual published cut-off for severe fibrosis or cirrhosis (F3F4 or F4 according to METAVIR score). Therefore, 388 patients (26.6%) had liver stiffness 9.5 kpa, 283 patients (19.4%) had FibroTest 0.74, 148 patients (10.2%) had APRI 2, and 199 patients (13.7%) had FIB-4 score Liver biopsy was performed in 789 patients (54.2%) and showed severe fibrosis or cirrhosis in 260 cases (33%). Median liver biopsy samples length was 17 mm and median number of portal tract was 14. At baseline, 1069 patients had a liver stiffness inferior to 9.5 kpa. Three hundred eighty-eight, 257, 144, 84, 51, and 29 patients had a measurement 9.5, 12.5, 20, 30, 40, and 50 kpa, respectively. Among these 1457 patients, the core group included the 663 patients with all liver fibrosis scores available. The non-core group included the other 794 patients. Characteristics of these 2 groups are indicated in Table 1 and Supplementary Table 3.

5 1974 VERGNIOL ET AL GASTROENTEROLOGY Vol. 140, No. 7 Table 1. Characteristics of the 1457 Patients and Comparison of Core and Non-Core Groups Characteristics Whole group (n 1457) Core group (n 663) Non-core group (n 794) P value Clinical characteristics Males, n (%) 778 (53.4) 371 (55.9) 407 (51.2) NS Age (y), mean SE NS Body mass index, mean SE Tobacco use (pack-year), median (range) 2 (0 20) 5 (0 20) 0 (0 20) NS Alcohol use, (drink/wk), median (range) 0 (0 7) 0 (0 7) 0 (0 7) NS HIV co-infection, n (%) 140 (9.6) 65 (9.8) 75 (9.4) NS HCV genotypes, n (%) Genotype 1/4 849 (58.3)/125 (8.6) 438 (66)/69 (10) 411 (52)/56 (7).001 Genotype 2/3 192 (13.2)/165 (11.3) 65 (10)/61 (9) 127 (16)/104 (13).001 FibroScan value (kpa), median (IQR) 6.7 (5.1 10) 6.9 (5.2 10) 6.6 ( ) NS Fibrosis according to liver biopsy, n (%) 789 (54.2).0001 F0 29 (3.7) 25 (3.8) 4 (0.5) F1 220 (27.9) 194 (29.3) 26 (3.3) F2 280 (35.5) 244 (36.8) 36 (4.5) F3 118 (15) 100 (15.1) 18 (2.3) F4 142 (18) 100 (15.1) 42 (5.3) Esophageal varices, n (%) NS No endoscopy 1086 (74.5) 490 (45.1) 596 (54.9) Varices stage 0 or (23.1) 152 (22.9) 167 (21.0) Varices stage 2 or 3 52 (3.6) 21 (3.2) 31 (3.9) Past history of ascites 20 (1.4) 4 (0.6) 16 (2.0).02 Hepatocellular carcinoma at inclusion 22 (1.5) 9 (1.4) 13 (1.6) NS Child Pugh score, n (%).002 A 1405 (96.4) 649 (97.9) 756 (95.2) B 46 (3.2) 12 (1.8) 34 (4.3) C 6 (0.4) 2 (0.3) 4 (0.5) Liver stiffness measurement (kpa),n(%) (26.6) 176 (26.5) 212 (26.7) NS (9.9) 57 (8.6) 87 (11.0) NS (5.8) 29 (4.4) 55 (6.9) (3.5) 17 (2.6) 34 (4.3) NS (2) 6 (0.9) 23 (2.9).007 Biological characteristics, median (IQR) Platelet count (G/L) 209 ( ) 206 ( ) 211 ( ) NS Prothrombin time (%) 100 (92 100) 100 (94 100) 100 (91 100).005 Bilirubin ( mol/l) 10 (8 13) 10 (8 13) 10 (8 14).005 GGT (IU/L) 55 (29 123) 60 (32 135) 51 (27 113).006 AST (IU/L) 43 (31 66) 45 (32 67) 41 (30 66).03 ALT (IU/L) 60 (39 99) 65 (41 05) 57 (37 95).006 FIB ( ) 1.35 ( ) 1.32 ( ) NS APRI 0.51 ( ) 0.53 ( ) 0.50 ( ) NS ActiTest 0.41 ( ) 0.43 ( ) 0.39 ( ).004 FibroTest 0.40 ( ) 0.45 ( ) 0.37 ( ).008 AST, aspartate aminotransferase; ALT, alanine aminotransferase; GGT, -glutamyl transpeptidase; IQR, interquartile range; SE, standard error. Accuracy of Biomarkers for Fibrosis Staging As indicated in Table 2, liver stiffness and FibroTest had comparable diagnostic accuracy for the diagnosis of fibrosis stage, but higher accuracy in comparison with other biomarkers, when biopsy was taken as the reference. In the core group (n 663), 46 patients had all noninvasive test values (APRI, FIB-4, FibroTest, liver stiffness) indicating stages F3F4 fibrosis. Among them, 40 patients had F3F4 fibrosis at liver biopsy. In the core group (n 663), 113 patients had FibroTest and liver stiffness indicating F3F4 fibrosis. Among them, 97 patients had F3F4 fibrosis at liver biopsy. Survival A total of 1457 patients were analyzed after collecting all the data of the living and dead patients in February At the beginning of the study, 22 patients had a hepatocellular carcinoma. The overall number of death/transplantation was 93 (6.4%) of 1457 patients, 53 liver-related and 40 not liver-related deaths. Among these 93 patients, 29 patients (31.2%) had liver stiffness measurement 14.5 kpa (cut-off value for the diagnosis of cirrhosis) at inclusion. At inclusion, 27 dead or transplanted patients had Child B or C cirrhosis (Supplementary Table 5).

6 June 2011 PROGNOSIS VALUE OF LIVER STIFFNESS MEASUREMENT 1975 Table 2. Overall Accuracy (waurocs) of Biomarkers for the Diagnosis of Fibrosis Stages: 663 Patients Without Missing Data for All Pair-Wise Comparisons wauroc, median (SE) Comparison: Z test (P significance) LSM FibroTest FIB-4 LSM (0.007) FibroTest (0.007) 1.80 (.07) FIB (0.007) 5.03 (.0002) 3.14 (.002) APRI (0.007) 6.90 (.0001) 5.25 (.0001) 2.60 (.009) NOTE. The Obuchowski measure estimates the overall accuracy (wauroc) that is the weighted mean of the 10 pairwise comparisons stage by stage (F0 vs F1, F0 vs F2, F0 vs F3, F0 vs F4, F1 vs F2, F1 vs F3, F1 vs F4, F2 vs F3, F2 vs F4, and F3 vs F4). LSM, liver stiffness measurement; SE, standard error; wauroc, weighted area under the receiver operating characteristic. During follow-up, 703 patients (48%) did not receive any treatment for HCV infection, 199 patients (14%) had sustained virological response after treatment, and 555 (38%) had failure of treatment (nonresponse or relapse). Among patients with liver stiffness 9.5 kpa, 133 patients did not receive any treatment (49 deaths in this group), 65 patients were sustained virological responders (2 deaths in this group), and 190 patients had treatment failure (22 deaths in this group). Among patients with liver stiffness 9.5 kpa, 570 patients did not receive any treatment (18 deaths in this group), 134 patients were sustained virological responders (no deaths in this group), and 365 patients had treatment failure (2 deaths in this group). Details of the causes of deaths are given in Table 3.Inthe overall population (1457 patients), 5-year overall survival was (95% CI: ) and the 5-year survival without liver-related death or transplantation was (95% CI: ). In the core group (n 663), the number of deaths/ transplantations was 27 (4.1%), 17 liver-related deaths and 10 not liver-related deaths. The 5-year overall survival was (95% CI: ). The 5-year survival without liver-related death was (95% CI: ). In patients with missing biomarkers (n 794), the overall survival (0.890; 95% CI: ) was lower than in the core group (log-rank test 11.8; P.0006) as well as for liver-related death (0.925; 95% CI: ; log-rank test 6.0; P.01). Table 3. Causes of Death/Transplantation During 5 Years Follow up Liver-related deaths (whole group/core group) Deaths unrelated to liver disease (whole group/core group) n n Deaths 55/17 Deaths 38/10 Hepatocellular carcinoma 21/9 Non-liver cancer 14/3 Hepatic failure 9/1 Cardiovascular disease 8/2 Variceal bleeding 2/1 HIV 3/1 Hepatorenal syndrome 3/0 Unknown 13/4 Infection 4/1 Liver transplantation 16/5 HIV, human immunodeficiency virus. The survival probability of patients classified according to liver stiffness, FibroTest, APRI, FIB-4, and liver biopsy is detailed in Figure 2A. Liver stiffness and FibroTest values were divided into 6 different classes. As indicated in Figures 2B and C, overall survival was decreased according to liver stiffness and FibroTest values. For example, the 5-year overall survival was 96% in patients with liver stiffness 9.5 kpa, and 47% in patients with liver stiffness 40 kpa, and 97% for FibroTest 0.75, and 0% in patients with FibroTest Prediction of Survival In the core group, a multivariate analysis of liver stiffness, biomarkers, and liver biopsy accuracy for the prognosis of patients was calculated in 4 models. As indicated in Table 4, overall survival and survival without liver-related deaths were significantly associated with liver stiffness and FibroTest, whatever age and treatment (Model 1), with an additive prognostic value when fibrosis stage estimated using liver biopsy and necroinflammatory grade presumed using ActiTest were taken into account (Model 2). The prognostic values of FIB-4 and APRI scores were also significant but with lower hazard ratio than stiffness and FibroTest for Model 1, borderline significant for Model 2, and not significant for Model 3 and Model 4 (Supplementary Table 1 and 6). The combination of liver stiffness and FibroTest (Model 4) had a highly significant prognostic accuracy with an AUROC between (95% CI: ) in the core group, and (95% CI: ; P.37) in the non-core group. For the prognostic value for 5-year mortality for the whole group of patients, Model 5 had the best prognosis value compared to liver stiffness and FibroTest. AUROC of Model 5 was (95% CI: ) vs (95% CI: ) for liver stiffness and (95% CI: ) for FibroTest. There was no significant difference between liver stiffness and FibroTest AUROCs (P.61) and borderline differences between Model 5 and liver stiffness (P.06) and between liver stiffness and FibroTest (P.02) (Supplementary Tables 2 and 4). The estimated individual risk estimated by model 5 in the core group was Exp( 2.67*ActiTest 2.63E-02*Age

7 1976 VERGNIOL ET AL GASTROENTEROLOGY Vol. 140, No. 7 Figure 2. Overall survival probability according to liver stiffness, biomarkers, and liver biopsy. (A) Overall survival according to published cut-offs for the diagnosis of severe fibrosis or cirrhosis. (B) Overall survival according to different cut-offs of liver stiffness (kpa). Overall survival (95% confidence interval) using liver stiffness: 9.5 kpa: 96% (94% 98%); 9.5 kpa: 77% (72% 82%); 20 kpa: 66% (61% 71%); 30 kpa: 57% (50% 64%); 40 kpa: 47% (37% 57%); 50 kpa: 42% (29% 55%). (C) Overall survival according to different cut-offs of FibroTest. Overall survival (95% confidence interval) using FibroTest: 0.75: 97% (96% 98%); 0.75 and 0.80: 87% (78% 95%); 0.80 and 0.85: 84% (75% 94%); 0.85 and 0.90: 75% (61% 90%); 0.90 and 0.95: 69% (54% 84%); 0.95 to 1: 0% (0% 22%). 4.00*FibroTest 1.12* LSM in 3 classes (0: LSM 7, 1: 7 LSM 12.5, 2: LSM 12.5) 1.68*treatment in 3 classes (0: no treatment, 1: nonresponder, 2: sustained virological response). In the core group the corresponding prognostic AUROC for survival was (95% CI: ; P.0001). In the whole group, the corresponding prognostic AUROC for survival was (95% CI: ; P.0001). As obesity has been recognized as an independent predictor of prognosis in chronic liver disease, in a sensitivity analysis we added body mass index as another possible prognostic factor in model 5; there was not significant prognostic value (hazard ratio 0.98; 95% CI: ). As HIV infection has been recognized as an independent predictor of prognosis in chronic liver disease, in a sensitivity analysis we added HIV infection as another possible

8 June 2011 PROGNOSIS VALUE OF LIVER STIFFNESS MEASUREMENT 1977 Table 4. Prognostic Value of Biopsy and Biomarkers for Predicting Overall Survival in the Core Group (663 Patients) Multivariate analysis Model 1 Model 2 Model 3 Model 4 Estimate Univariate analysis Independent value of each estimate Additive value vs biopsy and inflammation estimate Additive value vs LSM Additive value vs FibroTest Prognostic aim None Treatment and age Biopsy, ActiTest, treatment, and age LSM, ActiTest, treatment, and age FibroTest, ActiTest, treatment, and age Covariate adjusted AUROC (95% CI) HR (95% CI) P value HR (95% CI) P value HR (95% CI) P value HR (95% CI) P value HR (95% CI) P value Stage METAVIR biopsy 0.76 ( ) 2.6 ( ) ( ).0001 NA NA NA NA NA NA LSM 0.82 ( ) 57 (20 161) (15 127) ( ).001 NA NA 3.1 ( ).002 FibroTest 0.80 ( ) 110 (16 752) (15 549) ( ) (5 569).002 NA NA APRI 0.66 ( ) 1.5 ( ) ( ) a ( ) a ( ) a ( ) a.11 FIB ( ) 4.4 ( ) a ( ) a ( ) a ( ) a ( ) a.02 ActiTest 0.56 ( ) 2.3 ( ) ( ).03 NA NA NA NA NA NA Treatment 0.71 ( ) 0.2 ( ).0002 NA NA NA NA NA NA NA NA Age 0.72 ( ) 1.07 ( ).0001 NA NA NA NA NA NA NA NA AUROC, area under the receiver operating characteristic; CI, confidence interval; HR, hazard ratio; LSM, liver stiffness measurement; NA, not applicable. a Semi-quantitative value used in the multivariate model when the continuous quantitative value was not applicable. prognostic factor in model 5: there was not significant prognostic value: hazard ratio 0.65 (95% CI: ; P.37). As hepatocellular carcinoma was present at baseline in 22 patients, in a sensitivity analysis we excluded these patients in model 5. All factors were still significant: treatment, hazard ratio (95% CI: ); LSM, hazard ratio (95% CI: ; P.0001); FibroTest, hazard ratio (95% CI: ; P.0001); age, hazard ratio (95% CI: ; P.01); and ActiTest, hazard ratio (95% CI, ; P.03). The overall performance of the best combination of tests (Model 5) in the core group and in the whole group is indicated in Table 5. In the core group, the combination of tests had an AUROC of 0.91 for the overall prognosis and this AUROC was 0.89 in the whole group. Discussion Until now, prognosis markers evaluated for use in chronic hepatitis C have been the histologic fibrosis staging of biopsy specimens, 30,31 FibroTest, 4 hyaluronic acid, 32 APRI, FIB-4, 9 and the Child Pugh score for patients with cirrhosis. To our knowledge, our study is the first showing that liver stiffness has a prognostic value for overall survival and survival without liver-related death in patients with chronic HCV infection. The present study independently validated the prognostic value of FibroTest and showed that liver stiffness and FibroTest can similarly predict survival. The second main finding, that liver stiffness and FibroTest had better prognosis values than liver biopsy, the classical gold standard for staging, was expected as the prognosis gets worse as these quantitative fibrosis estimates increased. This cannot be the case for the METAVIR cirrhosis as there is a unique stage F4 for cirrhosis. A better prognostic value of FibroTest vs biopsy had been already observed by Ngo et al. 4 The third main finding is the ranking of noninvasive methods for prognosis purpose. From the results, liver stiffness and FibroTest had better prognostic values than FIB-4 and APRI score. We have not evaluated other biomarkers, such as Fibrometer, ELF, and Hepascore. As necrosis and Table 5. Overall Performance of the Best Combination of Tests: Liver Stiffness, Fibrotest and Actitest, Adjusted on Age and Treatment Response Model 5: Combination of tests, with the highest performance value Core group (663 patients, 27 deaths) Whole group (1457 patients, 89 deaths) Hazard ratio (95% CI) P value Hazard ratio (95% CI) P value LSM 3.1 ( ) a ( ) a.0001 FibroTest 55 ( ) (14 255).0001 ActiTest 0.07 ( ) ( ).002 Treatment 0.19 ( ) ( ).0001 Age 1.03 ( ) ( ).002 AUROC of combination ( ) ( ).0001 AUROC, area under the receiver operating characteristic; CI, confidence interval; LSM, liver stiffness measurement. a Semi-quantitative value used in the multivariate model when the continuous quantitative value was not applicable.

9 1978 VERGNIOL ET AL GASTROENTEROLOGY Vol. 140, No. 7 inflammation have been associated with higher immune response and better response to HCV treatment, 33 biomarkers including transaminases in their fibrosis score (FIB-4, APRI, Fibrometer) should have a disadvantage in term of prognostic value. Indeed, in the present study ActiTest was an independent prognostic factor of survival. Because ascites is a physical limitation to the technique because elastic waves do not propagate through liquids, patients with ascites at baseline were excluded. The prognosis of those patients is known to be bad, and our results may have been underestimated compared to the reality. LSM failure is also related to overweight or obesity. The new XL probe is now used to assess liver stiffness in those patients, with a good feasibility and good correlations with M probe. 34 Therefore, the usefulness of XL probe for liver stiffness as a prognosis factor in obese and overweight patients should be evaluated. In our study, we decided to use well-established and validated liver stiffness cut-offs for severe fibrosis and cirrhosis. To test clinically relevant cut-offs, we chose the previously published values for severe fibrosis and cirrhosis, 14 doing the hypothesis these stages of fibrosis should be associated with a worse prognosis. In order to have a sensitive test, we decided to evaluate a cut-off for severe fibrosis instead of a cut-off for cirrhosis. Therefore, we hypothesized that all patients with cirrhosis were included in the analysis. We also studied the prognosis value of liver stiffness 20 kpa, known to be associated with liver-related complications. 16 Our results show that in the group of patients with cirrhosis, an increasing liver stiffness value is associated with a worse prognosis, introducing for the first time the concept of noninvasive prediction of survival in cirrhotic patients with liver stiffness. The evolution of liver stiffness or FibroTest values over time is extremely important. However, this point was not evaluated in our study and needs further evaluation. In our study, we decided not to test the Child Pugh score as a prognosis factor. Indeed, the prognosis value of Child Pugh score is well-established in cirrhotic patients. In our study, patients included exhibit chronic hepatitis C with various degrees of fibrosis, and few of them (17.4%) display cirrhosis at liver biopsy. Natural history of HCV patients is related to the amount of fibrosis, we therefore chose to test only noninvasive techniques for the diagnosis of fibrosis. A large number of patients were treated during the follow-up. Few data exist showing that liver stiffness may decrease under treatment, 35 suggesting it may be correlated with improved survival, as it is now well-described that sustained virological response ameliorates long-term outcomes. 36,37 In our study, we found that liver stiffness and FibroTest were accurate prognostic factors independent of treatment and its response. A few patients, were lost to follow-up. Most of them had all baseline data collected. Their follow-up was exactly the same as all the other previously known patients and most of the liver-related complications were treated in our center. Therefore, we conclude that data collected during this prospective monocentric follow-up were absolutely valuable to creating a strong survival database. Death and liver transplantation are sudden and objective criteria that can be precisely assessed. Because a large majority of the patients were treated in our center, and almost all the events occurred in the southwest area of France, this allows us a sharp evaluation of survival using the Kaplan- Meier method. Our weakness of the present study was that the included population was not a random, communitybased population. During the 5 years of follow-up, 53 liver-related and 40 unrelated deaths occurred in a population of 1457 patients. This incidence is comparable to data extracted from the natural history of viral cirrhosis. 38 Our models have not been validated in an independent model, therefore, there is a risk of overfitting and a decrease in the prognostic ability would probably occur in a new sample. However, for FibroTest, similar prognostic values have been already observed for patients with chronic hepatitis C, 4 chronic hepatitis B, 5 and alcoholic liver disease. 6 In this prospective study, we confirmed the prognostic value of liver stiffness and FibroTest on survival. This information is of major importance, helping us to sharpen our various tools for the follow-up of our patients. Liver stiffness, as a good predictive factor of survival, may help the physician to evaluate earlier the severity of those diseases, to decide with stronger arguments of a liver transplantation or a portosystemic shunt, and to evaluate more precisely the surgical risk of our cirrhotic patients. It may also be used as a complement to the score foie to determine the emergency of a liver transplantation in a cirrhotic patient. LSM and/or FibroTest could replace liver biopsy for the evaluation of the disease, whatever the stage of the disease. In our clinical practice, liver stiffness and FibroTest are already a strong daily parameter to give information to patients and their family on the severity of the liver disease. Using those predictive data, it might help us in the future to make our patients and their families aware of the importance of total alcoholic abstinence or on serious antiviral therapy compliance. Supplementary Table 1-6. Supplementary Material Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at and at doi: /j.gastro References 1. Bedossa P, Dargere D, Paradis V. Sampling variability of liver fibrosis in chronic hepatitis C. Hepatology 2003;38: Pugh RN, Murray-Lyon IM, Dawson JL, et al. Transection of the oesophagus for bleeding oesophageal varices. Br J Surg 1973; 60: Kamath PS, Kim WR. The model for end-stage liver disease (MELD). Hepatology 2007;45:

10 June 2011 PROGNOSIS VALUE OF LIVER STIFFNESS MEASUREMENT Ngo Y, Munteanu M, Messous D, et al. A prospective analysis of the prognostic value of biomarkers (FibroTest) in patients with chronic hepatitis C. Clin Chem 2006;52: Ngo Y, Benhamou Y, Thibault V, et al. An accurate definition of the status of inactive hepatitis B virus carrier by a combination of biomarkers (FibroTest-ActiTest) and viral load. PLoS One 2008; 3:e 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: Wai CT, Greenson JK, Fontana RJ, et al. A simple noninvasive index can predict both significant fibrosis and cirrhosis in patients with chronic hepatitis C. Hepatology 2003;38: Sterling RK, Lissen E, Clumeck N, et al. Development of a simple noninvasive index to predict significant fibrosis in patients with HIV/HCV coinfection. Hepatology 2006;43: Nunes D, Fleming C, Offner G, et al. Noninvasive markers of liver fibrosis are highly predictive of liver-related death in a cohort of HCV-infected individuals with and without HIV infection. Am J Gastroenterol 2010;105: de Lédinghen V, Vergniol J. Transient elastography for the diagnosis of liver fibrosis. Expert Rev Med Devices 2010;7: Wong VW, Vergniol J, Wong GL, et al. Diagnosis of fibrosis and cirrhosis using liver stiffness measurement in nonalcoholic fatty liver disease. Hepatology 2010;51: Fraquelli M, Rigamonti C, Casazza G, et al. Reproducibility of transient elastography in the evaluation of liver fibrosis in patients with chronic liver disease. Gut 2007;56: de Lédinghen V, Douvin C, Kettaneh A, et al. Diagnosis of hepatic fibrosis and cirrhosis by transient elastography in HIV/hepatitis C virus-coinfected patients. J Acquir Immune Defic Syndr 2006;41: Castera L, Vergniol J, Foucher J, et al. Prospective comparison of transient elastography, Fibrotest, APRI, and liver biopsy for the assessment of fibrosis in chronic hepatitis C. Gastroenterology 2005;128: Friedrich-Rust M, Ong MF, Martens S, et al. Performance of transient elastography for the staging of liver fibrosis: a metaanalysis. Gastroenterology 2008;134: Foucher J, Chanteloup E, Vergniol J, et al. Diagnosis of cirrhosis by transient elastography (FibroScan): a prospective study. Gut 2006;55: Masuzaki R, Tateishi R, Yoshida H, et al. Prospective risk assessment for hepatocellular carcinoma development in patients with chronic hepatitis C by transient elastography. Hepatology 2009;49: Fung J, Lai CL, Seto WK, et al. Prognostic significance of liver stiffness for hepatocellular carcinoma and mortality in HBeAgnegative chronic hepatitis B. J Viral Hepat 2010 Jul 26. [Epub ahead of print]. 19. Imbert-Bismut F, Ratziu V, Pieroni L, et al. Biochemical markers of liver fibrosis in patients with hepatitis C virus infection: a prospective study. Lancet 2001;357: Forns X, Ampurdanes S, Llovet JM, et al. Identification of chronic hepatitis C patients without hepatic fibrosis by a simple predictive model. Hepatology 2002;36: Castera L, Foucher J, Bernard PH, et al. Pitfalls of liver stiffness measurement: a 5-year prospective study of 13,369 examinations. Hepatology 2010;51: The French METAVIR Cooperative Study Group. Intraobserver and interobserver variations in liver biopsy interpretation in patients with chronic hepatitis C. Hepatology 1994;20: Zhou X, Obuchowski N, McClish D. Statistical methods in diagnostic medicine. 1st ed. New York: John Wiley & Sons, DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44: Lambert J, Halfon P, Penaranda G, et al. How to measure the diagnostic accuracy of noninvasive liver fibrosis indices: the area under the ROC curve revisited. Clin Chem 2008;54: Obuchowski NA, Goske MJ, Applegate KE. Assessing physicians accuracy in diagnosing paediatric patients with acute abdominal pain: measuring accuracy for multiple diseases. Stat Med 2001; 20: Hillis SL, Obuchowski NA, Schartz KM, et al. A comparison of the Dorfman-Berbaum-Metz and Obuchowski-Rockette methods for receiver operating characteristic (ROC) data. Stat Med 2005;24: Hintze J. NCSS 2003 user guide. Kaysville, UT: Number Cruncher Statistical Systems, R Development Core Team R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing, Available at: Accessed XXX. 30. Ikeda K, Saitoh S, Suzuki Y, et al. Disease progression and hepatocellular carcinogenesis in patients with chronic viral hepatitis: a prospective observation of 2215 patients. J Hepatol 1998;28: Niederau C, Lange S, Heintges T, et al. Prognosis of chronic hepatitis C: results of a large, prospective cohort study. Hepatology 1998;28: Guechot J, Serfaty L, Bonnand AM, et al. Prognostic value of serum hyaluronan in patients with compensated HCV cirrhosis. J Hepatol 2000;32: McHutchison JG, Lawitz EJ, Shiffman ML, et al. Peginterferon alfa-2b or alfa-2a with ribavirin for treatment of hepatitis C infection. N Engl J Med 2009;361: de Lédinghen V, Vergniol J, Foucher J, et al. Feasibility of liver transient elastography with FibroScan using a new probe for obese patients. Liver Int 2010;30: Vergniol J, Foucher J, Castera L, et al. Changes of non-invasive markers and FibroScan values during HCV treatment. J Viral Hepat 2009;16: Cardoso AC, Moucari R, Figueiredo-Mendes C, et al. Impact of peginterferon and ribavirin therapy on hepatocellular carcinoma: incidence and survival in hepatitis C patients with advanced fibrosis. J Hepatol 2010;52: Bruno S, Stroffolini T, Colombo M, et al. Sustained virological response to interferon-alpha is associated with improved outcome in HCV-related cirrhosis: a retrospective study. Hepatology 2007;45: Marcellin P, Pequignot F, Delarocque-Astagneau E, et al. Mortality related to chronic hepatitis B and chronic hepatitis C in France: evidence for the role of HIV coinfection and alcohol consumption. J Hepatol 2008;48: Received November 28, Accepted February 18, Reprint requests Address requests for reprints to: Victor de Lédinghen, MD, PhD, Centre d Investigation de la Fibrose hépatique, Service d Hépato- Gastroentérologie, Hôpital Haut-Lévêque, Pessac, France. victor.deledinghen@chu-bordeaux.fr; fax: Conflicts of interest Victor de Lédinghen received grants from Schering-Plough and Roche. The remaining authors disclose no conflicts. We disclosed to study participants potential investigator conflicts of interest.

11 1979.e1 VERGNIOL ET AL GASTROENTEROLOGY Vol. 140, No. 7 Supplementary Table 1. Multivariate Analysis of Biomarkers and Biopsy Accuracy for the Prognosis of Patients in the Whole Group Multivariate analysis Model 1 Model 2 Model 3 Model 4 Estimate Univariate Independent value of each estimate Additive value vs biopsy and inflammation estimate Additive value vs LSM Additive value vs FibroTest Prognostic aim None Treatment and age Biopsy, ActiTest, Treatment, and age LSM, ActiTest, Treatment, and age FibroTest, ActiTest, Treatment, and age Covariate adjusted Maximum, n HR (95% CI) P value HR (95% CI) P value HR (95% CI) P value HR (95% CI) P value HR (95% CI) P value Overall survival Liver biopsy ( ) ( ).0001 NA NA NA NA NA NA LSM (30 81) (27 79) ( ) a.0002 NA NA 17 (7.7 37).0001 FibroTest (70 626) (78 651) ( ) (14 255).0001 NA NA APRI ( ) a ( ) a ( ) a ( ) a ( ) a.0005 FIB ( ) a ( ) a ( ) a ( ) a ( ) a.0002 ActiTest ( ) (5.3 26).0001 NA NA NA NA NA NA Treatment ( ).0001 NA NA NA NA NA NA NA NA Age ( ).0001 NA NA NA NA NA NA NA NA Survival without LRD Liver biopsy ( ) ( ).0001 NA NA NA NA NA NA LSM (66 309) (7.4-27) a ( ) NA NA 7.0 (3.3-15).0001 FibroTest ( ) ( ) ( ) ( ).0001 NA NA APRI ( ) a ( ) a ( ) a ( ) a ( ) a.0002 FIB (6.0 15) a (6.3 16) a ( ) a ( ) a ( ) a ActiTest (5.0 41) (7.9 59).0001 NA NA NA NA NA NA Treatment ( ).009 NA NA NA NA NA NA NA NA Age ( ).007 NA NA NA NA NA NA NA NA CI, confidence interval; HR, hazard ratio; LRD, liver-related death; LSM, liver stiffness measurement; NA, not applicable. a Semi-quantitative value used in the multivariate model when the continuous quantitative value was not applicable. Supplementary Table 2. Sensitivity Analysis of Model 5 Excluding Variables With Possible Colinearity Effect (Age and ActiTest) in the Whole Group (n 1414) Model 5 modified: Combination of the 2 tests, with the highest prognostic value, without age and ActiTest Hazard ratio (95% CI) Significance P value Liver stiffness measurement 13 (4.6 58).0001 FibroTest 16 (5 58).0001 Treatment 0.27 ( ).0001 AUROC ( ).0001 AUROC, area under the receiver operating characteristic; CI, confidence interval.

12 June 2011 PROGNOSIS VALUE OF LIVER STIFFNESS MEASUREMENT 1979.e2 Supplementary Table 3. Characteristics of Core Group (No Missing Biomarker) in Comparison to Non-Core Group Characteristics Non-core group Core group P value n Liver related death, n (%) 38 (4.8) 17 (2.6%).03 Any death, n (%) 66 (8.6) 27 (4.1%).0005 Duration follow-up, (mos), 45.8 (10.5) 48.3 (10.7).0001 mean (SE) Liver biopsy F4 (%) Liver stiffness F4 (%) FibroTest F4 (%) Concordance LSM/FibroTest, (SE) 0.46 (0.02) 0.34 (0.03).01 LSM, liver stiffness measurement. Supplementary Table 5. Details of Discordances Between Estimates Among Dead Patients Stage F1 or F2 Fibrosis at Baseline Liver Biopsy Liver stiffness (kpa) FibroTest Biopsy fibrosis stage Cause of death Heart disease HIV infection Unknown Hepatocellular carcinoma 5 a Hepatocellular carcinoma Liver transplantation with cirrhosis 7 b Variceal bleeding Lung cancer Unknown a This patient had a new liver biopsy 1 year later with F4 fibrosis. b This patient had a new liver biopsy 4 years later with F2 fibrosis. Supplementary Table 4. Sensitivity Analysis of Model 5 Using FibroTest and LSM As Semi-Quantitative Variables Supplementary Model 5: Combination of tests, with the highest prognostic value Core group, 663 patients, 27 deaths Whole group, 1457 patients, 89 deaths Hazard ratio (95% CI) P value Hazard ratio (95% CI) P value LSM a 3.3 ( ) ( ).0001 FibroTest a 3.2 ( ) ( ).0001 ActiTest 0.09 ( ) ( ).009 Treatment 0.19 ( ) ( ).0001 Age 1.03 ( ) ( ).002 CI, confidence interval; LSM, liver stiffness measurement. a LSM and FibroTest were expressed as categorical variables in 3 classes.