Primary biliary cirrhosis (PBC) is a rare chronic inflammatory CLINICAL LIVER, PANCREAS, AND BILIARY TRACT

Transcription

1 GASTROENTEROLOGY 2005;128: CLINICAL LIVER, PANCREAS, AND BILIARY TRACT The Effect of Ursodeoxycholic Acid Therapy on the Natural Course of Primary Biliary Cirrhosis CHRISTOPHE CORPECHOT,* FABRICE CARRAT, ABBAS BAHR,* YVES CHRÉTIEN, RENÉE EUGÉNIE POUPON, and RAOUL POUPON* *Hôpital Saint-Antoine, Service d Hépatologie, Paris; Faculté de Médecine Saint-Antoine, INSERM Unité 444, Paris; and Hôpital Necker-Enfants Malades, INSERM Unité 370, Paris, France See editorial on page 498. Background & Aims: We used a multistate modeling approach to assess the effect of ursodeoxycholic acid (UDCA) therapy on the natural course of primary biliary cirrhosis (PBC), which remains controversial. Methods: Our population included 262 patients with PBC who had received mg/kg UDCA daily for a mean of 8 years (range, 1 22 years). Data were analyzed using a multistate Markov model, with histologic stage progression, death, and orthotopic liver transplantation (OLT) as main end points. Survival without OLT was compared with that predicted by the updated Mayo model and with the expected survival in the control population. Results: Forty-five patients developed cirrhosis, 20 underwent OLT, and 16 died by the censor date. Ten deaths were due to liver disease. The overall survival rates were 92% at 10 years and 82% at 20 years. Survival rates without OLT were 84% and 66% at 10 and 20 years, respectively, which were slightly lower than the survival rate of an age- and sex-matched control population (relative risk [RR], 1.4; P.1) but better than the spontaneous survival rate as predicted by the updated Mayo model (RR,.5; P <.01). The survival rate of patients in stage 1 and 2 was similar to that in the control population (RR,.8; P.5), whereas the probability of death or OLT remained significantly increased in treated patients in late histologic stages (RR, 2.2; P <.05). Conclusions: Treatment with UDCA alone normalizes the survival rate of patients with PBC when given at early stages. However, there is a continued need for new therapeutic options in patients with advanced disease. Primary biliary cirrhosis (PBC) is a rare chronic inflammatory liver disease with a long natural course that can be divided into 3 distinct phases. The early phase is characterized by ongoing inflammation and destruction of interlobular bile ducts, accompanied in about one half of cases by lobular inflammation, characterized mainly by periportal lymphocytic piecemeal necrosis. The second phase is marked by irreversible loss of bile ducts and parenchyma remodeling, with nodular hyperplasia, extensive fibrosis, and eventually cirrhosis. The third phase, occurring when the serum bilirubin level reaches 6 mg/dl, is coined the terminal phase, because the mean survival of patients is 3 years unless orthotopic liver transplantation (OLT) is performed. Overall, the median survival of patients with PBC is years, which is much poorer than that of an ageand sex-matched control population. 1 4 Ursodeoxycholic acid (UDCA) is currently the only drug approved for the treatment of patients with PBC. 5 Several randomized trials, combined analyses, and longterm observational studies have shown that this agent not only improves biochemical liver indices 6 9 but also delays histologic progression and prolongs survival without OLT. 14,15 However, these findings have been recently challenged in 2 independent meta-analyses 16,17 and the actual impact of UDCA therapy on the longterm course of PBC remains controversial The low prevalence and long natural course of PBC explain, at least in part, the great difficulty in conducting controlled trials of sufficient size and duration to detect a beneficial effect on survival or histologic progression. 23,24 To overcome the lack of power of clinical trials in assessing long-term effect of treatments in chronic diseases, the multistate modeling approach (Markov model) can be used. 25,26 This approach is useful in analyzing incomplete disease history data, in which individuals are viewed over only a portion of their life history. We 12,27 and others 28 previously applied this method to PBC to assess the time course of its histologic progression. In this study, we Abbreviations used in this paper: OLT, orthotopic liver transplantation; PBC, primary biliary cirrhosis; RR, relative risk by the American Gastroenterological Association /05/$30.00 doi: /j.gastro

2 298 CORPECHOT ET AL GASTROENTEROLOGY Vol. 128, No. 2 Figure 1. The PBC multistate Markov model. took a similar approach using a continuous-time multistate Markov model and end points that included occurrence of late histologic stages, OLT, and death to assess the long-term effect of UDCA therapy in patients with PBC. We show that the long-term survival of UDCAtreated patients with PBC in early histologic stages is nowadays similar to that of an age- and sex-matched control population. Patients and Methods A total of 262 patients with PBC treated with UDCA were followed up prospectively. The diagnosis of PBC was confirmed by positive antimitochondrial antibodies and compatible histology. Of the 262 patients, 116 were previously enrolled in a French multicenter, controlled trial of UDCA 29,30 and 146 were followed up under identical conditions at Saint- Antoine Hospital (Paris, France). Inclusion criteria included the following: (1) documented liver biopsy in the 3 months preceding the start of UDCA therapy, (2) absence of end-stage liver failure (as defined as any of the following criteria: serum bilirubin level 6 mg/dl, Mayo risk score 7, ascites, variceal bleeding, or hepatic encephalopathy) at the beginning of treatment, (3) a daily dose of UDCA of mg/kg, (4) absence of immunosuppressive therapy, (5) follow-up duration of at least 1 year, and (6) at least one documented liver biopsy during the follow-up period. Patients on UDCA therapy were enrolled from 1982 to 2001 and followed up until the date of their last known follow-up visit or death. Data were censored on July 1, Patients who developed end-stage liver failure were systematically considered for OLT in the absence of obvious contraindications. Patients having undergone OLT were also followed up until the time of their last known follow-up visit or death. Death was defined as liver related when it was due to a complication of liver disease or OLT. Clinical and biologic data used for the analysis were those recorded at the time biopsies were performed or at the time of OLT. The variables were coded as previously described. 27 Data analysis was based on the variations in histologic stage (as defined by Ludwig s classification 31 ) and clinical status (nongrafted, grafted, dead) during follow-up. The transition rates between histologic stages or clinical status and the predictive value of the clinical, biologic, or histologic variables were assessed using a continuous-time multistate Markov model with covariates. In such a model, all patients at a given stage at a given time are considered to have the same prognosis no matter how they got to this stage. This hypothesis, if verified, allows one to take into account all incomplete observations collected during the study period. In addition, the use of a multistate model allows one to consider the whole progression of the disease and to assess in the same analysis the respective effect of the prognostic parameters on the histologic stage progression, incidence of OLT, and mortality rate. The multistate model is illustrated in Figure 1, where arrows connect the various pathologic states and indicate possible progression or regression steps (state transitions) that disease may take over time. The first 4 disease states refer to the 4 histologic stages of Ludwig s classification, the fifth disease state refers to living patients having undergone OLT, and the sixth disease state refers to death. All histologic stages, including cirrhosis, were considered to be reversible. The model took into account the risk of death at any disease stage but did not allow for direct transition from an early stage (histologic stages 1 2) to OLT. Each transition was ruled by a transition rate depending on a baseline rate and on the regression coefficients of the prognostic variables. The regression equation linking the transition rates to the prognostic variables was expressed as i3j(v1...v n ) i3j(0) e ( 1V 1... n V n ), where i3j(0) was the baseline transition rate, V 1,...,V n the prognostic variables, and 1,..., n their regression coefficients. The estimates of transition rates and regression coefficients were simultaneously obtained by the method of maximum-likelihood estimation. The confidence intervals were assessed by a parametric bootstrap method using 1000 replicates. The disease durations at each pathologic stage were estimated. The results were expressed in terms of probability of being (1) at an early disease stage (stages 1 2), (2) at a late disease stage (stages 3 4), (3) dead, or (4) dead or grafted. The Table 1. Demographic and Baseline Characteristics of the Patients Characteristics Median (range) or proportion Age (y) 52 (21 76) % Women 91 % Pruritus (present/absent) 47/53 Total bilirubin (mg/dl).8 (.2 5.8) Serum albumin (g/l) 40 (29 56) Alkaline phosphatase ( ULN) 3.5 ( ) Alkaline aminotransferase ( ULN) 2.0 ( ) Prothrombin time (s) 11.5 ( ) Platelet count (10 3 / L) 250 (64 492) % Histologic stage (1 2/3 4) 62/38 % Interlobular bileduct paucity 64/36 (present/absent) % Lymphocytic piecemeal necrosis 51/36/14 (absent-mild/moderate/severe) Mayo risk score 4.7 ( ) ULN, upper limit of normal.

3 February 2005 EFFECT OF UDCA ON NATURAL COURSE OF PBC 299 Table 2. Causes of Death 10 liver-related deaths 5 end-stage hepatic failures 3 multifocal hepatocellular carcinomas 2 deaths post-olt 6 non liver-related deaths 2 myocardial infarctions 2 strokes 1 severe pneumonia 1 metastatic adenocarcinoma of unknown origin influence of the prognostic variables on the transition rates was tested using the Wald test. For each variable tested, 3 types of effect were assessed: (1) the effect on histologic stage progression, (2) the effect on the probability of OLT, and (3) the effect on mortality. To reduce the number of parameters in the model, it was considered that the influence of the variables on histologic stage progression was the same for all transitions but with a sign change between progression and regression ( ). All of the variables significantly associated with one or several transitions according to a single covariate model were subsequently introduced together in a multicovariate model. A backward stepwise selection procedure was then used to identify the independent prognostic variables. The multivariate models were compared using a penalized likelihood criterion for nonnested models. 32 Patient survival was compared with 2 different expected survival estimates: (1) the survival expected in the French population, matched for age, sex, and follow-up period, and (2) the survival predicted by the updated Mayo model. 33,34 The survival curve of the French population was calculated based on age, sex, and follow-up period conditional probabilities of death from official, published census tables. 35 For each patient, a yearly predicted cumulative survival rate was calculated from a subject of the same age and sex for a similar period of follow-up, beginning at the date of treatment introduction. All individual survival data were averaged to obtain a survival curve for a standard age-, sex-, and period-matched control cohort. Similarly, a predicted spontaneous survival curve for the first 7-year period was obtained for each patient by applying the updated Mayo model to the patient s risk score at the outset of treatment. All individual curves were averaged to obtain an overall predicted survival curve. The statistical comparisons between actual and predicted survivals and the calculation of the relative risk (RR) of death was performed using the log-rank test. Results The demographic and baseline characteristics of the 262 patients in the study are summarized in Table 1. The mean follow-up period was 8 years (range, 1 22 years). A total of 656 liver biopsy specimens were collected during this period, from which 394 pairs of consecutive biopsy procedures were analyzed. The mean interval between the biopsies for each patient was 3 years. Forty-five patients (19% of the noncirrhotic patients at entry) developed cirrhosis. Sixteen deaths and 20 OLTs occurred in 34 patients (13% of all patients). The causes of deaths are shown in Table 2. The model accuracy was assessed by comparing the predicted and observed transitions. The absence of divergence was attested by the Mantel Haenszel 2 test (P.5). The predicted percentages of patients in each disease stage were calculated for a 20-year period (Figure 2). Figure 2A shows the predicted course of patients followed up from histologic stages 1 2 (early stages). A total of 59% and 40% of these patients are predicted to remain at an early stage after 10 and 20 years of follow-up, respectively, while only 6% and 22% will progress to OLT or death during the same period. The predicted course of patients followed up from histologic stages 3 4 (late stages) is shown in Figure 2B. A total of 32% and 52% of these patients are predicted to progress to Figure 2. Predicted percentages of UDCA-treated patients with PBC in each disease state for a subsequent 20-year period. (A) Patients treated from stages 1 2 and (B) patients treated from stages 3 4. The probability of each state is represented graphically as an area.

4 300 CORPECHOT ET AL GASTROENTEROLOGY Vol. 128, No. 2 Figure 3. Predicted survival rates with and without OLT of UDCAtreated patients with PBC compared with those of a control population matched for age, sex, and follow-up period and those predicted by the updated Mayo model. OLT or death after 10 and 20 years of follow-up, respectively. The predicted survival rates with and without OLT are shown in Figure 3. The survival rate of a control population matched for age, sex, and follow-up period is shown on the same graph, as well as the spontaneous survival without OLT as predicted by the updated Mayo model. In our study, the predicted survival rate of patients was 92% at 10 years and 82% at 20 years. The predicted survival rates without OLT were 84% and 66% for the same periods, respectively. These survival rates were not statistically different from that of the control population, even though the predicted survival rate without OLT was somewhat shorter (RR,1.4; P.1). On the other hand, the survival predicted by the updated Mayo model was far poorer than the survival rates predicted on UDCA therapy with (RR, 4.2; P.0001) and without OLT (RR, 2.1; P.01). The predicted survival rates according to early and late histologic stages are shown in Figure 4. In patients in early stages (stages 1 2), the survival rate without OLT was not different from that of the control population (RR,.8; P.5) and was significantly longer than the spontaneous survival rate without OLT as predicted by the updated Mayo model (RR,.2; P.001). In patients in late stages (stages 3 4), the survival without OLT was shorter than that of the control population (RR, 2.2; P.05) and was not statistically different from the survival predicted by the updated Mayo model (P.5). In patients with cirrhosis, the predicted survival rate without OLT was 57% at 10 years and 39% at 20 years (median survival without OLT, 13 years). The risk of death in these patients was multiplied by 4 as compared with the control population (P.05). The univariate analysis of prognostic factors affecting histologic stage progression, risk of OLT, or survival is shown in Table 3. Histologic stage progression was accelerated in patients with increased bilirubin and/or alkaline phosphatase levels, reduced levels of albumin or platelet count, elevated Mayo risk score, moderate or severe lymphocytic piecemeal necrosis, and/or interlobular bile duct paucity. Age, bilirubin level, alanine aminotransferase level, alkaline phosphatase level, albumin level, prothrombin time, and Mayo risk score influenced the incidence of OLT. Survival was impaired in patients with an increased serum bilirubin level and/or an elevated Mayo risk score. The multivariate analysis of prognostic factors was performed in 2 separate steps because it was not possible to introduce in the same model the Mayo risk score and some of the variables (age, bilirubin, albumin, prothrombin time) included in this index. The first part of the analysis consisted of introducing the separate prognostic Figure 4. Predicted survival rates with and without OLT of UDCA-treated patients with PBC according to (A) early and (B) late histologic stages.

5 February 2005 EFFECT OF UDCA ON NATURAL COURSE OF PBC 301 Table 3. Univariate Analysis of Factors Assessed Under UDCA Therapy Affecting Histologic Stage Progression, Incidence of OLT, or Mortality Histologic stage progression OLT Death Variable RR (95% CI) P RR (95% CI) P RR (95% CI) P Sex Male vs female 1.3 (.9 1.8) NS 1.6 (.4 5.7) NS 2.5 (.7 9.1) NS Age Age 10 vs age (y) 1.0 (.9 1.1) NS.4 (.3.7) ( ).001 Pruritus Present vs absent 1.3 (.9 1.8) NS 3.5 ( ) NS 1.1 (.4 2.7) NS Serum bilirubin level 1, 2 vs 1 (mg/dl) 2.0 ( ) ( ) ( ).05 2 vs 1 (mg/dl) 4.0 ( ) 16.4 ( ) 4.6 ( ) Serum albumin level 38 vs 38 (g/l) 1.4 ( ) ( ) ( ) NS Alkaline phosphatase activity 3 vs 3 (ULN) 1.4 (1 1.9) ( ).01.9 (.2 4.4) NS Alkanine aminotransferase activity 2 vs 2 (ULN) 1.3 (.9 1.8) NS 4.9 ( ).01.8 (.2 2.9) NS Prothrombin time 14 vs 14 (s) 1.7 (.9 3.1) NS 3.8 ( ) (.2 5.2) NS Platelet count 150 vs 150 (10 3 / L) 2.7 ( ) (.4 3.9) NS 2.1 ( ) NS Interlobular bile duct paucity Present vs absent 1.3 ( ) (.5 2.0) NS 1.3 (.5 3.5) NS Lymphocytic piecemeal necrosis grade Moderate vs absent/mild 1.4 ( ) (.8 3.2) NS 1.5 (.6 3.6) NS Severe vs absent/mild 1.8 ( ) 2.4 ( ) 2.3 ( ) Mayo risk score Score 1 vs score 1.6 ( ) ( ) ( ).001 CI, confidence interval; ULN, upper limit of normal. variables into the model without the Mayo risk score; in the second part of the analysis, the Mayo risk score was introduced in substitution of age, bilirubin level, albumin level, and prothrombin time. In the first model, the independent predictive factors were age, bilirubin level, albumin level, and lymphocytic piecemeal necrosis (Table 4); in the second model, the Mayo risk score and the grade of lymphocytic piecemeal necrosis were independently linked to disease progression (Table 5). When the 2 models were compared, the goodness of fit of the first model was found to be significantly higher than that obtained with the second model. Discussion Our primary objective was to assess the long-term effect of UDCA therapy on the survival of patients with PBC. Our data show that UDCA therapy alone normalizes patient survival when given at the early stages of the disease. This study provides new information about the Table 4. First Multivariate Model (Excluding the Mayo Risk Score) Variable Histologic stage progression OLT Death Age Age 10 vs age (y) NS.4 (.3.7) 6.6 ( ) Serum bilirubin level 1, 2 vs 1 (mg/dl) 2.1 ( ) 6.6 ( ) NS 2 vs 1 (mg/dl) 4.2 ( ) 43.0 ( ) Serum albumin level 38 vs 38 (g/l) NS 10.5 ( ) NS Lymphocytic piecemeal necrosis grade Moderate vs absent/mild 1.3 ( ) NS NS Severe vs absent/mild 1.7 ( ) NOTE. Results are expressed as risk ratio and 95% confidence interval.

6 302 CORPECHOT ET AL GASTROENTEROLOGY Vol. 128, No. 2 Table 5. Second Multivariate Model (Excluding Age, Bilirubin Level, Albumin Level, and Prothrombin Time) Variable Histologic stage progression OLT Death Mayo risk score Score 1 vs score 1.6 ( ) 1.7 ( ) 2.6 ( ) Lymphocytic piecemeal necrosis grade Moderate vs absent/mild 1.4 ( ) NS NS Severe vs absent/mild 1.9 ( ) current time course of PBC in the era of UDCA, thus constituting an important resource for design and feasibility of future clinical trials aimed at identifying effective combinations of drugs in the management of PBC. The multistate Markov model was designed and validated in 262 patients with PBC, including (1) 116 patients previously enrolled in a French multicenter, controlled trial of UDCA 29,30 and (2) 146 patients followed at Saint-Antoine Hospital (Paris, France) according to the same protocol. Patients with a serum bilirubin level 6 mg/dl, edema, or digestive bleeding were excluded. Compliance with UDCA therapy was not taken into account. The model accurately predicted the probability of developing the chosen end points, because observed and predicted data were virtually the same. This allowed for computing of the predicted percentages of patients at each disease stage in a time-dependent manner. To quantify the impact of UDCA therapy, the predicted survival rates with and without OLT were compared with the survival rate in the French population after matching for age, sex, and period of follow-up and with the survival rate predicted by the updated Mayo model. 34 The relevance of this model has been recently confirmed as a good predictor of patient survival in 2 large series of patients from the United States and the United Kingdom. 3,4 The overall survival rate of patients in the present study was not different from that expected in the French control population and was far better than that predicted by the updated Mayo model. It is also markedly higher than the survival rate recently reported in patients with PBC from northeast England. 4 A possible explanation for such a large difference between the data from French and English patients may be due to the fact that the majority of patients from northeast England were not given UDCA therapy or received a low daily dose. Our study shows unequivocally that a large part of the benefit from UDCA therapy lies with its impact on the course of PBC in patients in early histologic stages of the disease. In fact, 59% and 40% of these patients, respectively, are predicted to remain at early histologic stages after 10 and 20 years of follow-up, and only 6% and 22%, respectively, are expected to progress to OLT or death during the same periods. The decrease in efficacy of UDCA therapy in advanced histologic stages is not surprising given the assumed mode of action of UDCA. 36 This partial loss of therapeutic benefit probably explains why trials of short duration, in which a majority of the enrolled patients still were in the remodeling phase of the liver parenchyma, thus with marked bile duct destruction, could not detect the efficacy of UDCA. Our study confirms that serum bilirubin level and parenchymal inflammation with periportal or periseptal lymphocytic piecemeal necrosis are independent predictors of histologic progression in UDCA-treated patients. 27,37 This finding, together with the lack of a reliable surrogate marker of lymphocytic piecemeal necrosis, 38 suggests that liver biopsy should be recommended for the appropriate management and counseling of the patients. As shown previously, bile duct paucity is an important determinant of the progression of liver fibrosis in PBC. 27 However, probably because of high sampling variability, the severity of bile duct paucity does not appear to be a reliable predictive factor of histologic progression. In our study, age, serum bilirubin level, and serum albumin level were prognostic indicators of OLT or death, confirming many previous studies. 33,39,40 In summary, the present study provides for the first time an accurate description of the whole course of PBC in the era of UDCA therapy and shows that this agent, when given at the early histologic stages, normalizes patient survival rates. In the future, those data may help to better define the subset of patients who could benefit from combination treatment at the optimal time in the course of PBC. References 1. Balasubramaniam K, Grambsch PM, Wiesner RH, Lindor KD, Dickson ER. Diminished survival in asymptomatic primary biliary cirrhosis. A prospective study. Gastroenterology 1990;98: Mahl TC, Shockcor W, Boyer JL. Primary biliary cirrhosis: survival of a large cohort of symptomatic and asymptomatic patients followed for 24 years. J Hepatol 1994;20: Kim WR, Lindor KD, Locke GR III, Therneau TM, Homburger HA, Batts KP, Yawn BP, Petz JL, Melton LJ III, Dickson ER. Epidemi-

7 February 2005 EFFECT OF UDCA ON NATURAL COURSE OF PBC 303 ology and natural history of primary biliary cirrhosis in a US community. Gastroenterology 2000;119: Prince MI, Chetwynd A, Newman W, Metcalf JV, James OF. Survival and symptom progression in a geographically based cohort of paients with primary biliary cirrhosis: follow-up for up to 28 years. Gastroenterology 2002;123: Heathcote EJ. Management of primary biliary cirrhosis. Hepatology 2000;31: Poupon RE, Balkau B, Eschwege E, Poupon R, UDCA-PBC Study Group. A multicenter, controlled trial of ursodiol for the treatment of primary biliary cirrhosis. N Engl J Med 1991;324: Lindor KD, Dickson ER, Baldus WP, Jorgensen RA, Ludwig J, Murtaugh PA, Harrison JM, Wiesner RH, Anderson ML, Lange SM, et al. Ursodeoxycholic acid in the treatment of primary biliary cirrhosis. Gastroenterology 1994;106: Heathcote EJ, Cauch-Dudek K, Walker V, Bailey RJ, Blendis LM, Ghent CN, Michieletti P, Minuk GY, Pappas SC, Scully LJ, et al. The Canadian multicenter double-blind randomized controlled trial of ursodeoxycholic acid in primary biliary cirrhosis. Hepatology 1994; 19: Combes B, Carithers RL Jr, Maddrey WC, Lin D, McDonald MF, Wheeler DE, Eigenbrodt EH, Munoz SJ, Rubin R, Garcia-Tsao G, et al. A randomized, double-blind, placebo-controlled trial of ursodeoxycholic acid in primary biliary cirrhosis. Hepatology 1995;22: Angulo P, Batts KP, Therneau TM, Jorgensen RA, Dickson ER, Lindor KD. Long-term ursodeoxycholic acid delays histological progression in primary biliary cirrhosis. Hepatology 1999;29: Poupon RE, Lindor KD, Pares A, Chazouilleres O, Poupon R, Heathcote EJ. Combined analysis of the effect of treatment with ursodeoxycholic acid on histologic progression in primary biliary cirrhosis. J Hepatol 2003;39: Corpechot C, Carrat F, Bonnand AM, Poupon RE, Poupon R. The effect of ursodeoxycholic acid therapy on liver fibrosis progression in primary biliary cirrhosis. Hepatology 2000;32: Pares A, Caballeria L, Rodes J, Bruguera M, Rodrigo L, Garcia-Plaza A, Berenguer J, Rodriguez-Martinez D, Mercader J, Velicia R. Longterm effects of ursodeoxycholic acid in primary biliary cirrhosis: results of a double-blind controlled multicentric trial. J Hepatol 2000;32: Poupon RE, Lindor KD, Cauch-Dudek K, Dickson ER, Poupon R, Heathcote EJ. Combined analysis of randomized controlled trials of ursodeoxycholic acid in primary biliary cirrhosis. Gastroenterology 1997;113: Lindor KD, Poupon R, Heathcote EJ, Therneau T. Ursodeoxycholic acid for primary biliary cirrhosis. Lancet 2000;355: Goulis J, Leandro G, Burroughs AK. Randomised controlled trials of ursodeoxycholic-acid therapy for primary biliary cirrhosis: a meta-analysis. Lancet 1999;354: Gluud C, Christensen E. Ursodeoxycholic acid (UDCA) in primary biliary cirrhosis (PBC) a Cochrane hepato-biliary systemic review. J Hepatol 1999;30(Suppl 1): Kaplan MM. Primary biliary cirrhosis: past, present, and future. Gastroenterology 2002;123: Chan CW, Papatheodoridis GV, Goulis J, Burroughs AK. Ursodeoxycholic acid and histological progression in primary biliary cirrhosis. J Hepatol 2003;39: Burroughs AK, Leandro G, Goulis J. Ursodeooxycholic acid for primary biliary cirrhosis. J Hepatol 2001;34: Gluud C, Christensen E. Ursodeoxycholic acid for primary biliary cirrhosis: lesson for the future? J Hepatol 2001;34: Combes B, Luketic VA, Peters MG, Zetterman RK, Garcia-Tsao G, Munoz SJ, Lin D, Flye N, Carithers RL Jr. Prolonged follow-up of patients in the U.S. multicenter trial of ursodeoxycholic acid for primary biliary cirrhosis. Am J Gastroenterol 2004;99: Poupon RE. Ursodeoxycholic acid for primary biliary cirrhosis: lessons from the past issues for the future. J Hepatol 2000; 32: Carithers RL Jr. Primary biliary cirrhosis: specific treatment. Clin Liver Dis 2003;7: Beck JR, Pauker SG. The Markov process in medical prognosis. Med Decis Making 1983;3: Gentleman RC, Lawless JF, Lindsey JC, Yan P. Multi-state Markov models for analysing incomplete disease history data with illustrations for HIV disease. Stat Med 1994;13: Corpechot C, Carrat F, Poupon R, Poupon R. Primary biliary cirrhosis: incidence and predictive factors of cirrhosis development in ursodiol-treated patients. Gastroenterology 2002;122: Locke GR III, Therneau TM, Ludwig J, Dickson ER, Lindor KD. Time course of histological progression in primary biliary cirrhosis. Hepatology 1996;23: Poupon RE, Poupon R, Balkau B, UDCA-PBC Study Group. Ursodiol for the long-term treatment of primary biliary cirrhosis. N Engl J Med 1994;330: Poupon RE, Huet PM, Poupon R, Bonnand AM, Nhieu JT, Zafrani ES, UDCA-PBC Study Group. A randomized trial comparing colchicine and ursodeoxycholic acid combination to ursodeoxycholic acid in primary biliary cirrhosis. Hepatology 1996;24: 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: Schwartz G. Estimating the dimension of a model. Ann Stat 1978;6: Dickson ER, Grambsch PM, Fleming TR, Fisher LD, Langworthy A. Prognosis in primary biliary cirrhosis: model for decision making. Hepatology 1989;10: Lindor K, Therneau T. Survival in ursodeoxycholic acid-treated patients with primary biliary cirrhosis. Gastroenterology 1996; 111: Vallin J, Meslé F. Tables de mortalité françaises pour les XIXe et XXe siècles et projections pour le XXIe siècle. Institut National d Etudes Démographiques, Paumgartner G, Beuers U. Ursodeoxycholic acid in cholestatic liver disease: mechanisms of action and therapeutic use revisited. Hepatology 2002;36: Degott C, Zafrani ES, Callard P, Balkau B, Poupon RE, Poupon R. Histopathological study of primary biliary cirrhosis and the effect of ursodeoxycholic acid treatment on histology progression. Hepatology 1999;29: 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: Roll J, Boyer JL, Barry D, Klatskin G. The prognostic importance of clinical and histologic features in asymptomatic and symptomatic primary biliary cirrhosis. N Engl J Med 1983;308: Christensen E, Altman DG, Neuberger J, De Stavola BL, Tygstrup N, Williams R. Updating prognosis in primary biliary cirrhosis using a time-dependent Cox regression model. Gastroenterology 1993;105: Received May 5, Accepted October 28, Address requests for reprints to: Christophe Corpechot, MD, Hôpital Saint-Antoine, Service d Hépatologie, Paris, France. christophe.corpechot@sat.ap-hop-paris.fr; fax: (33)