Primary biliary cirrhosis (PBC) is a slowly progressive

Transcription

1 AUTOIMMUNE, CHOLESTATIC AND BILIARY DISEASE Noninvasive Elastography-Based Assessment of Liver Fibrosis Progression and Prognosis in Primary Biliary Cirrhosis Christophe Corpechot, 1,2 Fabrice Carrat, 3 Armelle Poujol-Robert, 1 Farid Gaouar, 1 Dominique Wendum, 4 Olivier Chazouillères, 1,2 and Raoul Poupon 1,2 The development of liver fibrosis markers in primary biliary cirrhosis (PBC) is needed to facilitate the assessment of its progression and the effectiveness of new therapies. Here, we investigated the potential usefulness of transient elastography (TE) in the noninvasive evaluation of liver fibrosis stage and disease progression in PBC. We performed, first, a prospective performance analysis of TE for the diagnosis of METAVIR fibrosis stages in a diagnostic cohort of 103 patients and, second, a retrospective longitudinal analysis of repeated examinations in a monitoring cohort of 150 patients followed-up for up to 5 years. All patients were treated with ursodeoxycholic acid. Diagnostic thresholds of liver stiffness in discriminating fibrosis stages F1, F2, F3, and 5F4 were 7.1, 8.8, 10.7, and 16.9 kpa, respectively. TE showed high performance and was significantly superior to biochemical markers (e.g., aspartate aminotransferase [AST]/platelet ratio, FIB-4, hyaluronic acid, AST/alanine aminotransferase ratio, and Mayo score) in diagnosing significant fibrosis, severe fibrosis, or cirrhosis. Analysis of the monitoring cohort data set using generalized linear models showed the following: (1) an overall progression rate of kpa/year (P ) and (2) no significant progression in patients with F0-F1, F2, or F3 stages, but a significant increase ( kpa/year; P < ) in cirrhotic patients. A cut-off value of 2.1 kpa/year was associated with an 8.4-fold increased risk of liver decompensations, liver transplantations, or deaths (P < , Cox regression analysis). Conclusion: TE is one of the best current surrogate markers of liver fibrosis in PBC. Over a 5-year period, on-treatment liver stiffness appears stable in most noncirrhotic PBC patients, whereas it significantly increases in patients with cirrhosis. Progression of liver stiffness in PBC is predictive of poor outcome. (HEPATOLOGY 2012;56: ) Primary biliary cirrhosis (PBC) is a slowly progressive cholestatic disease. Despite ursodeoxycholic acid (UDCA) therapy, the disease remains active in a subset of patients and may progress toward cirrhosis and liver failure. New clinical trials aimed at assessing the benefit of therapeutic combinations with UDCA are therefore needed to further improve the overall prognosis of the disease. 1,2 PBC therapeutic trials are difficult to carry out, especially because of the slow course of the disease. The choice of selection criteria and primary endpoints is of critical importance. Because most patients are currently diagnosed and treated at an early stage, the use of hard endpoints, such as the occurrence of death or liver transplantation (LT), alone appears unsuitable. The use of additional endpoints, such as the occurrence of cirrhosis or of extensive fibrosis, seems undoubtedly more relevant in this condition. However, the necessity to perform repeated liver biopsies remains a serious limitation to the utilization of such Abbreviations: ACC, accuracy AIH, autoimmune hepatitis; ALP, alkaline phosphatase; ALT, alanine aminotransferase; APRI, aspartate aminotransferase/platelet ratio; AST, aspartate aminotransferase; AUROC, area under the receiver operating characteristic curve; BMI, body mass index; CI, confidence interval; ELF, European liver fibrosis test; HA, hyaluronic acid; HR, hazard ratio; IgG, immunoglobulin G; IQR, interquartile range; LD, liver decompensation; LSM, liver stiffness measurement; LT, liver transplantation; NLR, negative likelihood ratio; NPVs, negative predictive values; PBC, primary biliary cirrhosis; PLR, positive likelihood ratio; PPVs, positive predictive values; ROC, receiver operating characteristic; TE, transient elastography; UDCA, ursodeoxycholic acid; ULN, upper limit of normal. From the 1 Service d Hepatologie, Centre de reference des Maladies Inflammatoires des Voies Biliaires, Hôpital Saint-Antoine, Assistance Publique Hôpitaux de Paris (APHP), Paris, France; 2 UMR_S938, Faculte demedecine Saint-Antoine, Universite Pierre et Marie Curie (UPMC), Paris, France; 3 Unite de Sante Publique, Hôpital Saint-Antoine, Assistance Publique Hôpitaux de Paris (APHP), Paris, France; and 4 Service d Anatomie et Cytologie Pathologiques, Hôpital Saint-Antoine, Assistance Publique Hôpitaux de Paris (APHP), Paris, France. Received October 18, 2011; accepted January 3,

2 HEPATOLOGY, Vol. 56, No. 1, 2012 CORPECHOT ET AL. 199 endpoints. Because there are still no reliable surrogate markers of liver fibrosis in PBC, there is a continued need for developing noninvasive methods of liver fibrosis assessment and monitoring. In this regard, transient elastography (TE) appears as one of the most promising methods. 3 Liver stiffness measurement (LSM) using TE has indeed been shown as a simple, reliable surrogate marker of liver fibrosis in various chronic liver diseases. 4 Though the effectiveness of this method is well established in patients with chronic hepatitis C, only preliminary data are, so far, available in PBC. 5-7 Additionally, the progression rate of liver stiffness, used as an evaluation of fibrosis course and prognosis in PBC, has never been assessed. Consequently, our aim was first to define the diagnostic performance of TE in a large cohort of patients with PBC and then to determine the time course of changes in liver stiffness as a noninvasive assessment of fibrosis progression and prognosis in a monitoring cohort of UDCA-treated patients followed-up for several years. Patients and Methods Patient Populations Diagnostic criteria. PBC was diagnosed in all patients based on the following classical criteria (i.e., if at least two of the three following conditions were fulfilled): (1) alkaline phosphatase (ALP) level above 1.5 the upper limit of normal (ULN) for at least 6 months; (2) presence of serum antimitochondrial antibody at a minimal titer of 1:40; and (3) histological evidence of inflammatory destructive cholangitis of interlobular bile ducts on biopsy. PBC/autoimmune hepatitis (AIH) overlap syndrome was diagnosed when at least two of the following criteria were recorded at any time until or during the study: (1) serum activity of alanine aminotransferase (ALT) at least 5 ULN; (2) serum immunoglobulin G (IgG) level at least 2 ULN or a positive test for anti smooth muscle antibodies; and (3) histological evidence of moderate or severe periportal or periseptal lymphocytic interface hepatitis. 8 Diagnostic cohort. One hundred and forty-six patients with PBC who underwent both TE and percutaneous liver biopsy between September 2004 and January 2010 were consecutively recruited from the Department of Hepatology of Saint-Antoine Hospital (Paris, France). Among them, 60 patients (41%) were previously included in an earlier study of TE published in The average time interval between biopsy and TE was months (range, 0-16). Biopsy failure or inadequate liver specimen (<8 mm) was recorded in 5 patients (3%). Failure of TE (i.e., zero valid shots) occurred in 6 patients (4%), whereas unreliable results (as defined below) were obtained in 32 further patients (22%). The 103 remaining patients composed the definitive diagnostic cohort. The main characteristics of the patients at the time of biopsy are shown in Table 1. Eleven patients (11%) were newly diagnosed and thus did not receive any specific treatment at the time of recruitment. The other patients (89%) were all previously treated with UDCA at a dose of mg/kg/day. Nineteen (18%) patients received additional corticosteroids (e.g., budesonide or prednisone) and/or mycophenolate mofetil because of established PBC-AIH overlap syndrome (n ¼ 8; 8%) or suboptimal biochemical response to UDCA with significant histological activity grade on previous biopsy (n ¼ 11; 10%). Monitoring cohort. The monitoring cohort consisted of 150 patients with PBC who underwent at least 2 TEs with a minimal time interval of 6 months between October 2004 and March All the patients from this cohort were followed up in Saint- Antoine Hospital. Their main characteristics, as recorded at the time of first TE, are shown in Table 1. Among these patients, 52 (35%) had previously participated in the diagnostic cohort. Twenty-one patients (14%) were previously diagnosed with a PBC-AIH overlap syndrome. Fifty-six patients (37%) were receiving corticosteroids (e.g., budesonide or prednisone) and/or mycophenolate mofetil in association with UDCA (combined therapies) either at the beginning or during the study because of features of PBC-AIH overlap syndrome (n ¼ 21; 14%) or histologically active disease with incomplete biochemical response to UDCA (n ¼ 35; 23%). It was decided to maintain these patients in the study and to analyze the effect of the combined treatments retrospectively. In total, Address reprint requests to: Christophe Corpechot, M.D., Service d Hepatologie, Centre de reference des Maladies Inflammatoires des Voies Biliaires, Hôpital Saint-Antoine, Assistance Publique Hôpitaux de Paris (APHP), 184 rue du Faubourg Saint-Antoine, Paris, France. christophe.corpechot@sat.aphp.fr; fax: Copyright VC 2012 by the American Association for the Study of Liver Diseases. View this article online at wileyonlinelibrary.com. DOI /hep Potential conflict of interest: Nothing to report. Additional Supporting Information may be found in the online version of this article.

3 200 CORPECHOT ET AL. HEPATOLOGY, July 2012 Table 1. Characteristics at Entry of the Diagnostic and Monitoring Cohorts Variable Diagnostic Cohort Monitoring Cohort Population size (%) 103 (100) 150 (100) Time from diagnosis (years) (0-27.2) (0-17.5) Women (n, %) 87 (84) 134 (89) Age (years) ( ) ( ) BMI (kg.m 2 ) ( ) ( ) Total bilirubin (mg/dl) ( ) ( ) ALP ( ULN) ( ) ( ) GGT ( ULN) ( ) ( ) AST ( ULN) ( ) ( ) ALT ( ULN) ( ) ( ) Albumin (g/l) ( ) ( ) Prothrombin index (%) (52-142) (34-142) Platelet count (10 9 /L) (62-489) (58-443) IgG (g/l) ( ) ( ) IgM (g/l) ( ) ( ) Serum HA (lg/l) (10-347) (10-430) APRI ( ) ( ) FIB ( ) ( ) AST/ALT ratio ( ) ( ) Mayo score ( ) ( ) LSM (kpa) ( ) ( ) Biopsy length (mm) (8-40) Fibrosis stage (%) F0-F1 51 (50)* 83 (55) F2 22 (21)* 30 (20) F3 15 (14.5)* 25 (17) F4 15 (14.5)* 12 (8) Activity grade (%) A0 19 (18.5) A1 54 (52.5) A2 28 (27) A3 2 (2) PBC-AIH overlap syndrome (%) 8 (8) 21 (14) Abbreviations: GGT, gamma-glutamyl transferase; IgM, immunoglobulin M. *According to biopsy. According to TE. 507 validated measures of TE were recorded during 395 patient-years (average follow-up, years; range, ). The average number of TEs per patient was (range, 2-8). The average time interval between 2 consecutive TEs was years (range, ). Biochemical response to UDCA was estimated using the Paris I and II criteria 9,10 at the time of first TE in the patients who were previously treated for more than 6 months (n ¼ 129). Among them, 50 (39%) patients were previously used for establishing the Paris I criteria, including 37 (29%) who were additionally used for defining the Paris II criteria. Evidence of suboptimal biochemical response was present in 34 (27%) patients according to the Paris I criteria and in 71 (55%) patients according to the Paris II criteria. LSM. Measurement of liver stiffness by TE was performed using a FibroScan (EchoSens, Paris, France), as previously described. 11 As recommended by the manufacturer, only procedures with 10 valid measurements, a success rate greater than 60%, and an interquartile range (IQR)/median ratio of less than 30% were considered as reliable and eligible for the study. Serum Fibrosis Markers. Four surrogate blood indices of liver fibrosis were assessed at baseline: aspartate aminotransferase (AST)/platelet ratio (APRI), FIB- 4, hyaluronic acid (HA), and AST/ALT ratio. APRI and FIB-4 were calculated according to published formulas. 12,13 Serum HA concentration was measured by enzyme-linked immunosorbent assay (Corgenix, Broomfield, CO). The Mayo risk score for PBC 14 was additionally tested as a potential surrogate for liver fibrosis. Liver Histological Assessment. Liver tissue specimens were obtained by percutaneous biopsy. Specimens of less than 8 mm in total length were considered as noneligible. Biopsies were analyzed by one experienced pathologist blinded to the results of TE and clinical data. Histological stage was determined according to Ludwig s classification. 15 Liver fibrosis and necroinflammatory

4 HEPATOLOGY, Vol. 56, No. 1, 2012 CORPECHOT ET AL. 201 activity were evaluated according to the METAVIR scoring system. 16 Ductopenia was defined as a ratio of portal tracts without apparent interlobular bile duct over the total number of portal tracts below 50%. Data Analysis Performance analysis. Comparisons of LSMs, according to the different histological fibrosis stages and necroinflammatory activity grades, were performed using Kruskal-Wallis test. The diagnostic performance of TE was determined in terms of sensitivity, specificity, positive (PPV) and negative (NPV) predictive values, positive (PLR) and negative (NLR) likelihood ratio, accuracy (ACC), and area under the receiver operating characteristic curve (AUROC). Optimal stiffness thresholds for discrimination between fibrosis categories were determined at the maximum total sensitivity and specificity. The stability of the estimates was checked by a bootstrap method 17 using 100 internal random replications. Comparison of paired AUROCs and 95% confidence intervals (CIs) was performed using the nonparametric Delong test. 18 Repeated-measures analysis. With time intervals between LSMs being unequal both between and within patients, a generalized linear mixed model assuming two random effects (a 0 ¼ intercept in kpa, i.e., mean LSM at baseline; b 0 ¼ slope in kpa/year, i.e., mean change in LSM per year) for each fibrosis category (i.e., F0-F1, F2, F3, and F4) was applied to assess the progression rate of liver stiffness and its 95% CI for each fibrosis stage. Regression coefficients were assessed using the maximum likelihood estimation method based on an unstructured covariance model. The influence of clinical and biochemical variables was considered as a fixed effect and was assessed by introducing these covariates into the model according to the following equation: a ¼ a 0 þ R(a Vn V n ); b ¼ b 0 þ R(b Vn V n ), where V 1 V n are the covariates, a 0 and b 0 are the baseline coefficients, and a Vn and b Vn are the covariate-related coefficients. It was assumed that the effects of each covariate were equally weighted among all fibrosis categories. Wald s test was used to determine the significance of variables effect. All variables found to be individually linked to LSM at baseline or to the progression rate of liver stiffness were tested together in a multivariate model, and a backward stepwise procedure was used to select the independent prognostic variables. Survival analysis. Outcome, as of June 2011, was recorded as either death, LT, or liver decompensation (LD; namely, ascites, variceal bleeding, hepatic encephalopathy, hepatocellular carcinoma, doubling of total serum bilirubin level above 6 mg/dl, or minimal criteria for LT) versus right censored. A single Cox proportional hazards model was used to identify the variables recorded at the time of last LSM individually associated with a poor outcome, including the mean variation in LSM per unit of time (DLSM/qt). This variable was computed by averaging the variations in LSM calculated from all couples of sequential measures available per patient. Because LSM and DLSM/qt were associated with a significant risk of poor outcome, a receiver operating characteristic (ROC) analysis was performed to determine the optimal cutoffs of these parameters for the prediction of adverse events. Survival rates without LT and LD were estimated using Kaplan-Meier s method. Their comparison, according to the LSM and DLSM/qt cut-off values, was performed using the log-rank test. Cox s multiple regression analysis, using a forward selection procedure, was performed to identify independent prognostic factors. A two-side P value of less than 5% was considered significant. Statistical analyses were carried out using JMP 8.02 and SAS 9.2 systems (SAS Institute Inc., Cary, NC). Results Diagnostic Performance Ability in differentiating fibrosis stages. The values of LSM recorded from the diagnostic cohort ranged from 3.6 to 69.1 kpa (median, 7.8; average, ). There was a significant association between LSM and histological fibrosis stage (P < ), but there was no patent correlation with necroinflammatory activity grade (Fig. 1). No significant association with the presence of ductopenia was found. The optimal stiffness thresholds for the diagnosis of fibrosis stage F1, F2, F3, and ¼F4 were 7.1, 8.8, 10.7, and 16.9 kpa, respectively. The performance profile of TE and its statistics are shown in Table 2. LSM was found to be of high performance for the diagnosis of severe liver fibrosis (F3) or cirrhosis (F4), with both sensitivity and specificity 90%. On the other hand, it showed rather poor sensitivity (despite high PPV and specificity) for the detection of mild (F1) or significant (F2) liver fibrosis. Because histological staging of low-grade liver fibrosis is poorly reliable, F0 and F1 stages were grouped as a single category (F0-F1). Using this four-stage classification, 76 patients (74%) were correctly classified by TE. This percentage reached 80% or above in patients with F0-F1 or F3- F4 stages (93% in cirrhotic patients), but fell to 45% in patients with intermediate fibrosis (F2), in which probabilities of under- and overstaging were 32% and 23%, respectively (Supporting Fig. 1).

5 202 CORPECHOT ET AL. HEPATOLOGY, July 2012 Fig. 1. Distribution of LSM according to histological fibrosis stage (A) and necroinflammatory activity grade (B). Factors associated with TE failure or unreliable results. Failure of TE and unreliable LSM were recorded in 6 (4%) and 32 (22%) of the 146 patients, respectively, of the initial cohort. They were significantly associated with lower platelet count (P ¼ 0.02) or prothrombin index (P ¼ 0.03) and greater fibrosis stage (P ¼ 0.04), but not with age, gender, body mass index (BMI), or serum level of liver enzymes. Unreliable LSMs were mostly the result of an unsatisfactory IQR/median ratio (i.e., >0.3). Comparison with serum fibrosis markers.. The ability of TE in predicting significant fibrosis (F2), severe fibrosis ( F3), or cirrhosis (F4) was compared to those of APRI, FIB-4, HA, and AST/ALT ratio (which are all unrelated to cholestasis-related biochemistry and are reported to be of potential interest in PBC ), as well as that of the Mayo risk score. Each of these markers was associated with histological fibrosis stage. The ROC curves in differentiating the three diagnostic targets (i.e., F2, F3, and ¼F4) are shown in Fig. 2. The AUROCs from LSMs were significantly greater than those of the four serum indices and the Mayo risk score for the prediction of fibrosis F2, F3, or ¼F4 (Supporting Table 1). When they were combined with LSM in a multiple regression model, the serum markers did not improve diagnostic accuracy. Table 2. Performance Profile of TE in Differentiating Liver Fibrosis Stages Stage No. Cutoff Se Sp PPV NPV PLR NLR ACC AUROC Diagnostic cohort F þ F F ¼F Bootstrap statistic ( ) ( ) þ ( ) ( ) ( ) ( ) ( ) F1 9, ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) F2 5, ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) F3 3, ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ¼F4 1, ( ) Cutoffs are expressed in kpa and bootstrap estimates as mean 6 SD (95% CI). Abbreviations: Se, sensitivity; Sp, specificity.

6 HEPATOLOGY, Vol. 56, No. 1, 2012 CORPECHOT ET AL. 203 Fig. 2. ROC curves of LSM, APRI, FIB-4, HA, Mayo risk score, and AST/ALT ratio in the prediction of significant fibrosis, severe fibrosis, and cirrhosis. Time Course of Liver Stiffness Study of the entire cohort. Fibrosis stages, as determined at baseline using TE, were distributed as follows: F0-F1, n ¼ 83 (55%); F2, n ¼ 30 (20%); F3, n ¼ 25 (17%); and F4, n ¼ 12 (8%). The patients were followed-up by TE for an average of years (range, ). The follow-up period did not differ according to the baseline fibrosis stages (F0-F1: years; F2: ; F3: ; F4: ; P ¼ 0.11). Figure 3A shows the time course of individual repeated LSMs on the entire cohort. Two generalized linear models with random effects were used to analyze both the global and fibrosis stage-specific trends in the progression of liver stiffness. The first (i.e., the global model) was a simple dual random effects model designed to analyze the whole data without differentiating fibrosis stages. The second (i.e., the four-stage model) allowed us to assess liver stiffness progression for each fibrosis category assuming two random effects per category (a f ¼ intercept in kpa; b f ¼ slope in kpa/year). The parameters estimated by these two models are shown in Table 3 (first part). The global model (blinded of fibrosis stages) showed a significant increase in the progression rate of liver stiffness, estimated as b global ¼ kpa/year (95% CI: kpa/year; P ¼ 0.02), suggesting a significant global progression of liver fibrosis. The four-stage model showed no significant changes in LSM among patients with F0-F1 (b F0F1 ¼ kpa/year; P ¼ 0.09), F2 (b F2 ¼ kpa/year; P ¼ 0.93), or F3 (b F3 ¼ kpa/year; P ¼ 0.66) stages, but a significant progression of liver stiffness in patients with F4 stage (b F4 ¼ kpa/year; P < ). The correlation between liver stiffness and time is shown graphically in Fig. 3B. Study of patients under UDCA alone. The analysis was replicated separately in the subpopulation of patients who received UDCA alone (n ¼ 94; 63%). The distribution of fibrosis stages, as determined by TE at baseline, in this subgroup did not differ from that of the entire cohort. The global model still showed a significant progression rate of liver stiffness on all patients ( kpa/year; 95% CI: ; P ¼ 0.03), whereas the four-stage model confirmed a significant progression of LSM in F4 (b F4 ¼ kpa/year; P ¼ ), but not in F0-F1 (b F0F1 ¼ kpa/year; P ¼ 0.08), F2 (b F2 ¼ kpa/year; P ¼ 0.98), or F3 (b F3 ¼ kpa/year; P ¼ 0.74) patients (Table 3, second part). Factors associated with liver stiffness progression. A single covariate analysis of factors associated with either LSM at baseline or the progression rate of LSM over time was performed. Three parameters were significantly linked to LSM at baseline, namely, Mayo risk score (P < ) as well as serum HA (P ¼ 0.004) and albumin (P ¼ 0.01, inverse correlation) levels.

7 204 CORPECHOT ET AL. HEPATOLOGY, July 2012 Fig. 3. Time course of LSM as observed from the whole individual data (A) or estimated by the four-stage model as a function of baseline fibrosis stage (B). In (B), trends in liver stiffness progression are shown as linear regression straight lines (black lines) with 95% CIs (gray areas). Seven parameters were significantly linked to the progression rate of LSM, including serum HA level (P < ), FIB-4 (P ¼ ), APRI (P ¼ 0.002), platelet count (P ¼ 0.003, inverse correlation), Mayo risk score (P ¼ 0.02), AST/ALT ratio (P ¼ 0.03), and serum ALT activity (P ¼ 0.03, inverse correlation). The full results are available in Supporting Table 2. All the variables showing a P value 0.1 were selected for multivariate analysis. Baseline serum HA level was the only independent parameter associated with LSM at baseline (P ¼ 0.006) and the progression rate of LSM over time (P < ). Prognostic significance of liver stiffness progression. The patients were followed-up for an average of years (up to 5.1 years) after last LSM. Over this extended follow-up, 11 (7%) adverse events (e.g., 2 deaths, including 1 nonliver related, 4 LTs, and 5 LDs) occurred. Putative prognostic factors, including baseline LSM and variation in LSM over time (DLSM/qt), were studied in uni- and multivariate analyses (Table 4). The three independent predictors of adverse outcome were DLSM/qt (HR: 1.3; 95% CI: ; P < ), serum total bilirubin level (hazard ratio [HR]: 1.1; 95% CI: ; P ¼ ), and biochemical response to treatment according to the Paris I criteria (HR: 2.9; 95% CI: ; P ¼ 0.04). Optimal thresholds of 2.1 kpa/year (95% CI: ) for DLSM/qt and 9.6 kpa (95% CI: ) for baseline LSM were associated with an 8.4 (95% CI: ; P < ) and 5.1 (95% CI:

8 HEPATOLOGY, Vol. 56, No. 1, 2012 CORPECHOT ET AL. 205 Table 3. Parameters of the Global and Four-Stage Models as Assessed From All the Patients and Those Treated With UDCA Alone LSM at Baseline (a, kpa) Progression Rate (b, kpa/year) Estimate SD Test P Value Estimate SD Test P Value All patients (n ¼ 150) Global model Any stage < Four-stage model F0-F < F < F < F < < Patients Treated With UDCA Alone (n ¼ 95) Global model Any stage < Four-stage model F0-F < F < F < F < Abbreviation: SD, standard deviation ; P < ) times increased risk of adverse outcome, respectively. All events but one (1 death from metastatic breast cancer) occurred in patients with TE evidence of cirrhosis (n ¼ 8) or severe fibrosis (n ¼ 2) at baseline and a DLSM/qt >2.1 kpa/year. Comparisons of survival rates according to baseline LSM and DLSM/qt are shown in Fig. 4. Discussion Only a minority of the patients included in the present study participated in our previous TE analysis including both patients with PBC and primary sclerosing cholangitis. 5 So, the present study was not only PBC specific, as compared to the previous one, but was also largely composed of new patients. Of the 146 patients initially enrolled, 32 (22%) were excluded because of unreliable TE results. This significant proportion, mainly related to an IQR/median ratio >30%, was slightly higher, but overall similar than that (17%) reported in a series of 13,369 examinations carried out in 7,261 individuals with various chronic liver diseases. 24 In the latter study, the main reason for unreliable results was obesity, whereas in our experience, it mostly depended on liver fibrosis stage. In an intent-to-diagnose perspective, 25 considering that LSM was uninterpretable in approximately 1 in 5 cases, corrected AUROCs for the prediction of significant fibrosis, severe fibrosis, and cirrhosis can be estimated at 0.83, 0.86, and 0.89, respectively. There are currently no serum surrogate markers of liver fibrosis routinely recommended in PBC. Methods based on pharmacokinetics of substances removed exclusively or principally by the liver (i.e., sulfobromophthalein or indocyanine green) were abandoned. 26 Table 4. Uni- and Multivariate Analyses of Prognostic Factors Associated With Adverse Outcome Over the Extended Follow-up Factors HR (95% CI) Khi-2 P Value Univariate analysis Female gender 0.89 ( ) Age 1.01 ( ) BMI 0.97 ( ) Total bilirubin 1.11 ( ) 32.9 < ALP 1.38 ( ) GGT 1.05 ( ) AST 1.70 ( ) ALT 1.28 ( ) Albumin 0.78 ( ) Prothrombin index 0.95 ( ) Platelet count 0.99 ( ) Serum HA 1.01 ( ) 16.4 < APRI 3.53 ( ) FIB ( ) AST/ALT ratio 0.72 ( ) Mayo score 6.11 ( ) 21.4 < LSM 1.10 ( ) 22.4 < DLSM/qt 1.26 ( ) 22.5 < TE evidence of F ( ) 21.6 < Paris I criteria* 5.42 ( ) 21.3 < Paris II criteria* 2.89 ( ) PBC-AIH 0.86 ( ) Combined therapy 0.85 ( ) Multivariate analysis DLSM/qt 1.30 ( ) 16.6 < Total bilirubin 1.08 ( ) Paris I criteria* 2.93 ( ) *Biochemical response as assessed at the time of first TE. Abbreviation: GGT, gamma-glutamyl transferase.

9 206 CORPECHOT ET AL. HEPATOLOGY, July 2012 Fig. 4. Comparisons of survival rates without adverse outcome according to baseline LSM (A) or DLSM/qt (B). Circles symbolize censored patients. Serum HA level was shown to have good PPV for the diagnosis of severe fibrosis. 19 We previously proposed a serum index combining HA and total bilirubin, but this index lacked sensitivity. 27 The European liver fibrosis (ELF) test, a paying composite score, including serum HA, tissue inhibitor of metalloproteinase 1, and N-terminal propeptide of type III procollagen, also showed acceptable diagnostic accuracy for significant fibrosis or cirrhosis. 28 Nonproprietary scores, including APRI, Forns score, PBC score, AST/ALT ratio, and FIB-4, were also assessed Herein, we show that TE is superior to APRI, FIB-4, HA, AST/ALT ratio, and also the Mayo risk score in predicting cirrhosis, severe fibrosis, and significant fibrosis in PBC. We found no evidence that the combination of TE with these scores significantly improves diagnostic accuracy. Whether the ELF test or PBC score might compete with TE remains undetermined, but their reported AUROCs suggest poorer performance. A challenge for clinicians in the care of patients with PBC is how to evaluate treatment efficacy with confidence. Biochemical response is clearly a major predictor of long-term outcome in UDCA-treated patients. 9,29-31 However, it is unknown whether it could be applied in other therapeutic settings. As a simple, reliable surrogate marker of liver fibrosis, TE appears to be well adapted to this aim. In the present study, we assessed retrospectively the time course of liver stiffness in 150 patients with UDCA-treated PBC followed-up for an average of 3 years (range, 1-5), a usual duration for clinical trials. Our results show that noncirrhotic patients have steady levels of liver stiffness over time, supporting a very low rate or even an absence of histological progression in this subset of patients, contrasting with a significant increase in liver stiffness in cirrhotic patients, a result suggesting a worsening of liver damage when cirrhosis is established. These results are at variance with our previous histological observations showing that the probabilities of developing cirrhosis from histological stages 1, 2, and 3 after an equivalent period of follow-up (i.e., 3 years) were 1%, 6%, and 44%, respectively. 32 One major difference between the two studies is that the monitoring of patients in the latter study was initiated while most of them were already treated with UDCA for a long time ( years), whereas in the former study, all the patients were followed-up from UDCA initiation. 32 In addition, our results may have been probably influenced by the use of adjuvant drugs, because a significant proportion of the patients were given corticosteroids and/or mycophenolate mofetil for PBC-AIH overlap syndrome (14%) or incomplete biochemical response to UDCA with significant histological activity grade (23%). This high proportion of patients under combined therapies is related to the specific recruitment of our tertiary referral center. In our experience, noncirrhotic patients in these disease settings seem to benefit greatly from combined therapies. 33 The significant progression of liver stiffness in cirrhotic patients likely explains the nonlinear relationship observed here in PBC, as in other chronic liver diseases such as chronic hepatitis C, between the METAVIR score and LSM. As underlined recently, 34 cirrhosis should not be seen as a static diagnosis reflecting the end stage of the wound-healing process, but rather as a dynamic and plastic state, independently of its etiology. Architectural changes, vascular shunts, liver cell regeneration, matrix modification with cross-linking, and elastin-rich scars are major events in the process of liver fibrosis development in cirrhosis. Parenchymal extinction, an epiphenomenon

10 HEPATOLOGY, Vol. 56, No. 1, 2012 CORPECHOT ET AL. 207 caused by innocent bystander injury and thrombosis of the local vessels, is assumed to progress long after cirrhosis is established. All these events may explain the spontaneous aggravation of liver stiffness and the loss of UDCA effectiveness in cirrhotic patients. Finally and importantly, the present results strongly suggest that the increase in LSM is an independent prognostic factor in PBC. The variation in LSM over time (DLSM/qt) was indeed related to the risk of deaths, LT, or LD, independently of the baseline serum bilirubin level and biochemical response to treatment. A DLSM/qt value over 2.1 kpa/year was associated with a 58% incidence rate of adverse events over a 3-year follow-up. All liver-related events occurred in patients with cirrhosis or severe fibrosis at baseline and a DLSM/qt value >2.1 kpa/year. Although confirmation in larger studies is needed (the small number of events being one of the limitations of our study), the present findings strongly suggest that the monitoring of TE in patients with PBC, especially at the stage of cirrhosis, provides significant prognostic information, in comparison to classical serum prognostic markers, and may be useful to predict outcome and select high-risk patients for future clinical trials. Because serum HA level was strongly associated with both the risk of liver stiffness increase and that of adverse outcome, it seems also reasonable to recommend its monitoring in PBC patients. In conclusion, this study confirms the diagnostic performance and accuracy of TE in the noninvasive quantification of liver fibrosis in PBC. It clearly indicates that TE is of high performance for the diagnosis of severe fibrosis or cirrhosis and is one, if not the best, current surrogate marker of liver fibrosis in PBC. This study also shows that, over an average follow-up of 3 years, on-treatment liver stiffness appears mostly stable in noncirrhotic patients with PBC, whereas it significantly increases in patients with compensated cirrhosis. Furthermore, it indicates that the absolute value of LSM and, even more significantly, its increase over time are major predictive factors of poor outcome in PBC. References 1. Lindor KD, Gershwin ME, Poupon R, Kaplan M, Bergasa NV, Heathcote EJ. Primary biliary cirrhosis. HEPATOLOGY 2009;50: EASL EAftSotL. EASL Clinical Practice Guidelines: management of cholestatic liver diseases. J Hepatol 2009;51: Vizzutti F, Arena U, Marra F, Pinzani M. Elastography for the noninvasive assessment of liver disease: limitations and future developments. Gut 2009;58: Friedrich-Rust M, Ong MF, Martens S, Sarrazin C, Bojunga J, Zeuzem S, Herrmann JE. Performance of transient elastography for the staging of liver fibrosis: a meta-analysis. Gastroenterology 2008;134: Corpechot C, El Naggar A, Poujol-Robert A, Ziol M, Wendum D, Chazouillères O, et al. Assessment of biliary fibrosis by transient elastography in patients with PBC and PSC. HEPATOLOGY 2006;43: Gomez-Dominguez E, Mendoza J, García-Buey L, Trapero M, Gisbert JP, Jones EA, Moreno-Otero R. Transient elastography to assess hepatic fibrosis in primary biliary cirrhosis. Aliment Pharmacol Ther 2008;27: Obara N, Ueno Y, Fukushima K, Nakagome Y, Kakazu E, Kimura O, et al. Transient elastography for measurement of liver stiffness measurement can detect early significant hepatic fibrosis in Japanese patients with viral and nonviral liver diseases. J Gastroenterol 2008;43: Chazouillères O, Wendum D, Serfaty L, Montembault S, Rosmorduc O, Poupon R. Primary biliary cirrhosis-autoimmune hepatitis overlap syndrome: clinical features and response to therapy. HEPATOLOGY 1998; 28: Corpechot C, Abenavoli L, Rabahi N, Chretien Y, Andreani T, Johanet C, et al. Biochemical response to ursodeoxycholic acid and long-term prognosis in primary biliary cirrhosis. HEPATOLOGY 2008;48: Corpechot C, Chazouillères O, Poupon R. Early primary biliary cirrhosis: biochemical response to treatment and prediction of long-term outcome. J Hepatol 2011;55: Ziol M, Handra-Luca A, Kettaneh A, Christidis C, Mal F, Kazemi F, et al. Noninvasive assessment of liver fibrosis by measurement of stiffness in patients with chronic hepatitis C. HEPATOLOGY 2005;41: Wai CT, Greenson JK, Fontana RJ, Kalbfleisch JD, Marrero JA, Conjeevaram HS, Lok AS. 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, Sola R, Correa MC, Montaner J, et al. Development of a simple noninvasive index to predict significant fibrosis in patients with HIV/HCV coinfection. HEPATOLOGY 2006;43: Dickson ER, Grambsch PM, Fleming TR, Fisher LD, Langworthy A. Prognosis in primary biliary cirrhosis: model for decision making. HEPATOLOGY 1989;10: Ludwig J, Dickson ER, McDonald GS. Staging of chronic nonsuppurative destructive cholangitis (syndrome of primary biliary cirrhosis). Virchows Arch A Pathol Anat Histol 1978;379: Bedossa P, Poynard T. An algorithm for the grading of activity in chronic hepatitis C. The METAVIR Cooperative Study Group. HEPA- TOLOGY 1996;24: Efron B, Tibshirani R. An introduction to the bootstrap. New York, NY: Chapman & Hall/CRC; Delong E, Delong D, Clarke-Pearson D. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44: Poupon RE, Balkau B, Guechot J, Heintzmann F. Predictive factors in ursodeoxycholic acid-treated patients with primary biliary cirrhosis: role of serum markers of connective tissue. HEPATOLOGY 1994;19: Nyblom H, Bj ornsson E, Simren M, Aldenbourg F, Almer S, Olsson R. The AST/ALT ratio as an indicator of cirrhosis in patients with PBC. Liver Int 2006;26: F arkkil a M, Rautiainen H, K arkk ainen P, Karvonen AL, Nurmi H, Niemel a O. Serological markers for monitoring disease progression in noncirrhotic primary biliary cirrhosis on ursodeoxycholic acid therapy. Liver Int 2008;28: Alempijevic T, Krstic M, Jesic R, Jovanovic I, Sokic Milutinovic A, Kovacevic N, et al. Biochemical markers for non-invasive assessment of disease stage in patients with primary biliary cirrhosis. World J Gastroenterol 2009;15: Meaney C, Kumagi T, Al-Harthy N, Acarsu U, Coltescu C, Hirschfield GM. The Toronto Western Hospital PBC Study: the AST to platelet ratio index (APRI) is an effective predictor of long-term outcome in patients with primary biliary cirrhosis. HEPATOLOGY 2010;52:491A. 24. Castera L, Foucher J, Bernard PH, Carvalho F, Allaix D, Merrouche W, et al. Pitfalls of liver stiffness measurement: a 5-year prospective study of 13,369 examinations. HEPATOLOGY 2010;51:

11 208 CORPECHOT ET AL. HEPATOLOGY, July Poynard T, de Ledinghen V, Zarski JP, Stanciu C, Munteanu M, Vergniol J, et al. FibroTest((R)) and Fibroscan((R)) performances revisited in patients with chronic hepatitis C. Impact of the spectrum effect and the applicability rate. Clin Res Hepatol Gastroenterol 2011;35: Vaubourdolle M, Gufflet V, Chazouillères O, Giboudeau J, Poupon R. Indocyanine green-sulfobromophthalein pharmacokinetics for diagnosing primary biliary cirrhosis and assessing histological severity. Clin Chem 1991;37: Corpechot C, Poujol-Robert A, Wendum D, Galotte M, Chretien Y, Poupon RE, Poupon R. Biochemical markers of liver fibrosis and lymphocytic piecemeal necrosis in UDCA-treated patients with primary biliary cirrhosis. Liver Int 2004;24: Mayo MJ, Parkes J, Adams-Huet B, Combes B, Mills AS, Markin RS, et al. Prediction of clinical outcomes in primary biliary cirrhosis by serum enhanced liver fibrosis assay. HEPATOLOGY 2008;48: Pares A, Caballeria L, Rodes J. Excellent long-term survival in patients with primary biliary cirrhosis and biochemical response to ursodeoxycholic acid. Gastroenterology 2006;130: Kuiper EM, Hansen BE, de Vries RA, den Ouden-Muller JW, van Ditzhuijsen TJ, Haagsma EB, et al. Improved prognosis of patients with primary biliary cirrhosis that have a biochemical response to ursodeoxycholic acid. Gastroenterology 2009;136: Kumagi T, Guindi M, Fischer SE, Arenovich T, Abdalian R, Coltescu C, et al. Baseline ductopenia and treatment response predict long-term histological progression in primary biliary cirrhosis. Am J Gastroenterol 2010;105: Corpechot C, Carrat F, Poupon R, Poupon RE. Primary biliary cirrhosis: incidence and predictive factors of cirrhosis development in ursodiol-treated patients. Gastroenterology 2002;122: Rabahi N, Chretien Y, Gaouar F, Wendum D, Serfaty L, Chazouillères O, et al. Triple therapy with ursodeoxycholic acid, budesonide and mycophenolate mofetil in patients with features of severe primary biliary cirrhosis not responding to ursodeoxycholic acid alone. Gastroenterol Clin Biol 2010;34: Garcia-Tsao G, Friedman S, Iredale J, Pinzani M. Now there are many (stages) where before there was one: in search of a pathophysiological classification of cirrhosis. HEPATOLOGY 2010;51: