Primary biliary cirrhosis (PBC) is an autoimmune

Early Biochemical Response to Ursodeoxycholic Acid and Long-Term Prognosis of Primary Biliary Cirrhosis: Results of a 14-Year Cohort Study Li-Na Zhang, 1,2 * Tian-Yan Shi, 1,2 * Xu-Hua Shi, 1,2 Li Wang, 1,2 Yun-Jiao Yang, 1,2 Bin Liu, 1,2 Li-Xia Gao, 1,2 Zong-Wen Shuai, 1,2 Fang Kong, 1,2 Hua Chen, 1,2 Wei Han, 3 Shao-Mei Han, 3 Yun-Yun Fei, 1,2 Quan-Cai Cui, 4 Qian Wang, 1,2 Min Shen, 1,2 Dong Xu, 1,2 Wen-Jie Zheng, 1,2 Yong-Zhe Li, 1,2 Wen Zhang, 1,2 Xuan Zhang, 1,2 and Feng-Chun Zhang 1,2 The biochemical response to ursodeoxycholic acid (UDCA) in primary biliary cirrhosis is a strong predictor of long-term outcome and thus facilitates the rapid identification of patients needing new therapeutic approaches. Numerous criteria for predicting outcome of treatment have been studied based on biochemical response to UDCA at 1 year. We sought to determine whether an earlier biochemical response at 3 or 6 months could as efficiently identify patients at risk of poor outcome, as defined by liver-related death, liver transplantation, and complications of cirrhosis. We analyzed the prospectively collected data of 187 patients with a median follow-up of 5.8 years (range, 1.3-14 years). The survival rates without adverse outcome at 5 years and 10 years were 86% and 63%. Under UDCA therapy, laboratory liver parameters experienced the most prominent improvement in the first 3 months (P < 0.0001) and then stayed relatively stable for the following months. The Paris, Barcelona, Toronto, and Ehime definitions, but not the Rotterdam definition, applied at 3, 6, and 12 months significantly discriminated the patients in terms of long-term outcome. Compared with biochemical responses evaluated after 1 year of UDCA therapy, biochemical responses at the third month demonstrated higher positive predictive value (PPV) but lower negative predictive value (NPV) and increased negative likelihood ratio (NLR) by all definitions; biochemical responses at the sixth month showed higher or the same PPV and NPV and lower NLR by all definitions. Conclusion: For the previously published criteria, biochemical responses at the sixth month can be used in place of those evaluated after 1 year of UDCA therapy. Our findings justify a more rapid identification of patients who need new therapeutic approaches. (HEPATOLOGY 2013;58:264-272) Primary biliary cirrhosis (PBC) is an autoimmune liver disease characterized by the presence of highly specific antimitochondrial antibodies and progressive destruction of intrahepatic bile ducts, resulting in chronic cholestasis, portal inflammation, and fibrosis, which can ultimately lead to cirrhosis and hepatic failure. 1,2 Ursodeoxycholic acid (UDCA) is currently the only approved medical treatment for PBC. Despite improved prognosis in many patients treated with UDCA, the transplant-free survival rate remains significantly lower in patients with a suboptimal biochemical response. 3-8 Thus, there is a continued need for new therapeutic options for treating PBC. Abbreviations: ALP, alkaline phosphatase; ALT, alanine aminotransferase; AST, aspartate aminotransferase; GGT, gamma-glutamyl transferase; IgM, immunoglobulin M; NLR, negative likelihood ratio; NPV, negative predictive value; PBC, primary biliary cirrhosis; PPV, positive predictive value; UDCA, ursodeoxycholic acid; ULN, upper limit of normal. From the 1 Department of Rheumatology and Clinical Immunology, the 3 Department of Epidemiology and Biostatistics, Institute of Basic Medical Sciences, and the 4 Department of Pathology, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College, Beijing, China; and the 2 Key Laboratory of Rheumatology & Clinical Immunology, Ministry of Education, Beijing, China. Received October 7, 2012; accepted February 5, 2013. This work was supported by grants from the National Science Technology Pillar Program in the 11th Five-Year Plan (2008BAI59B03), the National Major Scientific and Technological Special Project for Significant New Drugs Development (2012ZX09303006-002), and the Research Special Fund for Public Welfare Industry of Health (201202004). *These authors contributed equally to this work. 264

HEPATOLOGY, Vol. 58, No. 1, 2013 ZHANG, SHI, ET AL. 265 The biochemical response to UDCA, especially changes in the serum activities of alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT), alanine aminotransferase (ALT), and aspartate aminotransferase (AST), may serve as a strong predictor of long-term outcome in patients with PBC 6-10 and thus could have a role in clinical practice and therapeutic trials by identifying patients with a poor prognosis. Previously published criteria for predicting outcome of treatment were mainly based on biochemical response after 1 or 2 years of UDCA therapy. 6-9 However, it is helpful to identify as soon as possible patients who will get optimal benefit from alternative therapy. It has been recommended that therapeutic trials should target patients with incomplete biochemical response after 3 to 6 months of UDCA treatment. 11 However, a biochemical response as early as 3 to 6 months was evaluated in only a few large independent cohorts of patients, including two studies using the Mayo criteria and Ehime criteria. 12,13 Today, more and more patients are diagnosed at an early stage of PBC. Given the slow disease progression and limited availability of study participants, traditional hard endpoints, such as the occurrence of death or liver transplantation, are considered unfeasible in clinical trials. 11 Accordingly, more extended endpoints in homogeneous cohorts of patients are required to define clinically relevant criteria of biochemical response in patients with PBC. 14 In the current study, our aim was: (1) to determine whether an earlier biochemical response at 3 or 6 months (using previously published biochemical criteria) could as efficiently identify patients at risk of poor outcome, as defined by liver-related death, liver transplantation, and complications of cirrhosis; and (2) to evaluate the prognostic impact of the multiple criteria in an independent cohort of Chinese patients with PBC. Patients and Methods Patient Population. All patients included in the study were diagnosed and followed up at Peking Union Medical College Hospital between 1995 and 2012. Diagnosis of PBC was based on liver function tests, presence of serum antimitochondrial antibodies, and histopathological findings. Patients were treated with UDCA at a daily dose of 13 to 15 mg/kg and had documented biochemical analyses (serum bilirubin concentration; activities of ALP, GGT, AST, and ALT; serum albumin concentration) before and after 3, 6, and 12 months of treatment with UDCA. Ineligibility criteria were as follows: a diagnosis of autoimmune hepatitis overlap syndrome, a positive serological test for hepatitis B or C virus, or the use of corticosteroids or immunosuppressive drugs for a duration of more than 2 months. Patients with complications of cirrhosis and those who underwent or were awaiting liver transplantation were also excluded. The patients were followed regularly every 3 months during the first year of UDCA treatment, and biochemical analyses during each visit were documented. The patients were then shifted to a half-yearly followup program including physical examination, liver function tests, abdominal ultrasonography, and gastroscopy in case of suspicion of portal hypertension. Liver biopsy at entry was optional. The histological stage of PBC was defined according to the Ludwig classification. 15 Definitions of Biochemical Response and Endpoint. The biochemical response to UDCA was evaluated after 3, 6, and 12 months of treatment according to five previously published definitions: (1) the Barcelona definition, a decrease in ALP level >40% of the baseline level or a normal level 6 ; (2) the Paris definition, ALP level 3 upper limit of normal (ULN), together with AST level 2 ULN and a normal bilirubin level 7 ; (3) the Rotterdam definition, normal bilirubin and albumin concentrations when one or both parameters are abnormal before treatment, or normal bilirubin or albumin concentrations after treatment when both are abnormal at entry 8 ; (4) the Toronto definition, an ALP level <1.76 ULN 9 ; and (5) the Ehime definition, a decrease in GGT level >70% of the baseline level or a normal level. 10 Of note, the first three biochemical definitions have been investigated after 1 year of UDCA treatment using death and liver transplantation as endpoints; the Toronto definition Address reprint requests to: Feng-Chun Zhang, M.D., Department of Rheumatology and Clinical Immunology, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College, Key Laboratory of Rheumatology & Clinical Immunology, Ministry of Education, No. 41 Da Mu Cang, Western District, Beijing 100032, China. E-mail: zhangsamfch@hotmail.com; fax: þ86-10-6915-8799. Copyright VC 2013 by the American Association for the Study of Liver Diseases. View this article online at wileyonlinelibrary.com. DOI 10.1002/hep.26322 Potential conflict of interest: Nothing to report. Additional Supporting Information may be found in the online version of this article.

266 ZHANG, SHI, ET AL. HEPATOLOGY, July 2013 Table 1. Biochemical Characteristics of the Patients (n 5 187) at Baseline and After 1 Year of UDCA Treatment Characteristics At Baseline After 1 Year P* Female sex 94% Age (years) 51 6 9 ALT (ULN) 2.52 6 1.79 1.26 6 1.00 <0.0001 AST (ULN) 2.60 6 1.55 1.34 6 0.89 <0.0001 ALP (ULN) 3.36 6 2.14 1.79 6 1.37 <0.0001 GGT (ULN) 6.70 6 5.34 3.04 6 3.96 <0.0001 Albumin (g/l) 40.48 6 4.69 42.14 6 4.11 <0.0001 Bilirubin (ULN) 1.35 6 1.47 0.93 6 0.92 <0.0001 IgM (ULN) 1.31 6 0.96 0.92 6 0.70 <0.0001 Platelet 202 6 67 198 6 69 NS Prothrombin index (%) 96 6 9 Histological stage (%) Early (1-2) 61 (33%) Late (3-4) 11 (6%) Not available 115 (61%) Quantitative data are expressed as the mean 6 SD. *Wilcoxon signed-rank test for paired data. Available in 124 (66%) patients. Abbreviation: NS, not significant. has been investigated after 2 years of treatment using an increase in two histological stages as a key endpoint; and the Ehime definition was based on biochemical response as early as 6 months after UDCA treatment. For the purpose of the present study, all five definitions were applied at 3, 6, and 12 months after UDCA therapy and evaluated using the same endpoint, which was the occurrence of adverse outcome as defined by at least one of the following events: liver-related death, liver transplantation, or complications of cirrhosis (ascites, variceal bleeding, hepatic encephalopathy, or hepatocellular carcinoma). Data were censored at the time of death or liver transplantation for the patient who died or underwent transplantation, and at the time of presenting with a cirrhosis-related complication or of last follow-up for the living nontransplanted patients. If a living nontransplanted patient developed more than one cirrhosis-related complication during follow-up, data were censored at the time of the first presentation of cirrhosis-related complications. Data Analysis. Comparisons between biochemical variables before and after 3, 6, or 12 months of UDCA treatment were performed using the Wilcoxon signed-rank test for paired data. Survival rates without adverse outcome were estimated using the Kaplan- Meier method. The effects of baseline factors and of biochemical responses to UDCA on survival rate were estimated using the Cox proportional hazards regression model. The average hazard ratio and 95% confidence interval were used to quantify the strength of the statistical links between the tested variables and survival. The sensitivity (Se), specificity (Sp), positive (PPV) and negative (NPV) predictive values, and positive and negative (NLR) likelihood ratios were calculated for all definitions to assess their performance for prediction of long-term outcome. Quantitative data are expressed as the mean 6 SD and qualitative data as a percentage. All analyses were two-sided, and P < 0.05 was considered statistically significant. The statistical package SPSS 16.0 (SPSS Inc, Chicago, IL) was used to perform the analysis. Results Descriptive Data. A total of 187 patients (94% females; age, 51 6 9 years) met the inclusion criteria. Biochemical data were available in 128 patients for the third month after UDCA treatment, 145 patients for the sixth month, and 157 patients for 1 year. To take full advantage of the available data, we used all of them for analyses. Table 1 shows the characteristics of the patients at baseline and after 1 year of UDCA therapy. Figure 1 shows the evolution of ALT, AST, ALP, GGT, serum bilirubin, albumin, and immunoglobulin M (IgM) levels within the first year of UDCA treatment. A prominent decline in the serum activities of ALP, GGT, AST, and ALT was noted in the first three months (P < 0.0001), which was accompanied by a significant decrease of bilirubin (P < 0.0001) and IgM (P < 0.0001) and elevation of albumin (P < 0.0001). Serum bilirubin, ALP, GGT, AST, ALT, and IgM levels gradually decreased subsequently, with a maximum decrease observed at the sixth month for serum bilirubin, ALP, and IgM levels, and at 1 year for GGT, AST, and ALT levels. After 1 year of UDCA treatment, the serum albumin level was slightly but significantly elevated (P < 0.0001), the serum activities of ALP, GGT, AST, and ALT were decreased by about 50% (P < 0.0001), and the serum concentrations of total bilirubin and IgM were decreased by 30% (P < 0.0001) compared with baseline values. The patients were followed up under UDCA therapy for a mean period of 5.9 6 2.6 years (median, 5.8 years; range, 1.3-14 years). An adverse outcome was recorded in 37 patients, including eight liverrelated deaths, four liver transplantations, and 25 complications of cirrhosis (six ascites, nine variceal bleeding, five with both ascites and variceal bleeding, four with hepatic encephalopathy and ascites, and one hepatocellular carcinoma). The survival rates without adverse outcome at 5 years and 10 years were 86% and 63%, respectively (Fig. 2). In univariate analysis, the baseline factors associated with an adverse outcome

HEPATOLOGY, Vol. 58, No. 1, 2013 ZHANG, SHI, ET AL. 267 Fig. 1. ALT, AST (A), ALP (B), GGT (C), serum bilirubin (D), albumin (E), and IgM (F) levels over time for patients with PBC. Data are expressed as mean 6 SEM. *P < 0.0001 versus baseline. were a serum activity of ALP >3 ULN, GGT >5 ULN, AST >2 ULN, an abnormal serum concentration of total bilirubin, and a decreased serum level of albumin (Table 2). In multivariate analysis, a serum activity of ALP >3 ULN, elevated bilirubin level, and decreased albumin level were independent risk factors significantly associated with an adverse outcome (Supporting Table 1). Barcelona, Paris, Rotterdam, Toronto, and Ehime Criteria. The Barcelona, Paris, Rotterdam, Toronto, and Ehime definitions of biochemical responses to Table 2. Univariate Analysis of Baseline Characteristics Associated With Adverse Outcome Variable HR (95% CI) P* Fig. 2. Survival rates without adverse outcome (an adverse outcome being defined as liver-related death, liver transplantation, or complication of cirrhosis) in patients with PBC. Age >55 years 1.87 (0.97-3.63) NS Female sex 2.36 (0.32-17.29) NS ALT >2 ULN 1.62 (0.83-3.17) NS AST >2 ULN 2.39 (1.12-5.08) <0.05 ALP >3 ULN 4.97 (2.27-10.87) <0.0001 GGT >5 ULN 2.74 (1.35-5.57) <0.01 Albumin <35 g/l 4.79 (2.34-9.80) <0.0001 Bilirubin >1 ULN 4.08 (2.01-8.28) <0.0001 IgM >1 ULN 0.89 (0.44-1.80) NS *Cox regression analysis. Abbreviations: CI, confidence interval; HR, hazard ratio; NS, not significant.

268 ZHANG, SHI, ET AL. HEPATOLOGY, July 2013 Table 3. Rates of Biochemical Response and Adverse Outcome According to the Barcelona, Paris, Rotterdam, Toronto, and Ehime Definitions Response Definition Rate of Biochemical Response Adverse Outcome in Responders Adverse Outcome in Nonresponders HR (95% CI) P* Paris Third month 86/128 (67.2%) 6/86 16/42 7.46 (2.90-19.17) <0.0001 Sixth month 103/145 (71.0%) 10/103 19/42 6.19 (2.86-13.39) <0.0001 1 year 108/157 (68.8%) 10/108 23/49 6.71 (3.19-14.13) <0.0001 Barcelona Third month 86/128 (67.2%) 11/86 11/42 2.58 (1.11-5.97) <0.05 Sixth month 108/145 (74.5%) 15/108 14/37 2.67 (1.29-5.53) <0.01 1 year 114/157 (72.6%) 18/114 15/43 2.63 (1.31-5.28) <0.01 Rotterdam Third month 27/59 (45.8%) 8/27 13/32 1.27 (0.53-3.07) NS Sixth month 30/63 (47.6%) 10/30 14/33 1.48 (0.65-3.34) NS 1 year 33/68 (48.5%) 11/33 16/35 1.98 (0.91-4.30) NS Toronto Third month 75/128 (58.6%) 4/75 18/53 9.53 (3.19-28.42) <0.0001 Sixth month 100/145 (69.0%) 7/100 22/45 8.84 (3.76-20.81) <0.0001 1 year 100/157 (63.7%) 7/100 26/57 9.24 (4.00-21.35) <0.0001 Ehime Third month 51/128 (39.8%) 5/51 17/77 3.45 (1.23-9.69) <0.05 Sixth month 79/145 (54.5%) 6/79 23/66 4.70 (1.91-11.61) <0.001 1 year 91/157 (58.0%) 10/91 23/66 3.26 (1.55-6.86) <0.001 Biochemical data were available in 128 patients for the third month after UDCA treatment, 145 patients for the sixth month, and 157 patients for 1 year. For the Rotterdam criteria, data from only 59, 63, and 68 patients were applicable for analysis at 3 months, 6 months, and 1 year following UDCA therapy. *Cox regression analysis. Abbreviations: CI, confidence interval; HR, hazard ratio; NS, not significant. UDCA were evaluated for their ability to discriminate patients according to the long-term outcome (Table 3). For each definition, the rates of biochemical response after 3, 6, or 12 months of UDCA therapy were comparable. The highest rate of biochemical response was observed at the sixth month according to the Paris (71.0%), Barcelona (74.5%), and Toronto (69.0%) definitions, whereas the highest level occurred after 1 year of UDCA therapy according to the Rotterdam (48.5%) and Ehime (58.0%) definitions. The Paris, Barcelona, Toronto, and Ehime definitions significantly discriminated the patients in terms of long-term outcome, whereas no significant association was found with the Rotterdam definition (Table 3 and Fig. 3). Responders differed significantly from nonresponders and had lower baseline bilirubin, ALT, AST, ALP, and GGT levels and higher albumin levels (Supporting Table 2). The responders were more likely to have early disease (by histological stage, P < 0.05; by the Dutch prognostic class, 5 P < 0.001). The biochemical responses evaluated at 3, 6, and 12 months of UDCA treatment were highly comparable. We also examined the influence of the initial severity of disease on the prognostic performance of biochemical response at 3, 6, and 12 months. Both histological and nonhistological (the Dutch prognostic class 5 ) criteria were used to grade the initial severity of the disease. Compared with nonresponders, patients with optimal biochemical responses were less prone to adverse outcomes and tended to have a higher survival rate without adverse outcome, irrespective of the initial severity of the disease (Supporting Figs. 1A-C and 2A-C). However, after stratifying the data by histological stages, the impact of biochemical response on survival was not statistically significant. The prognostic impact of biochemical response on survival remained significant after stratifying the data by Dutch prognostic class (biochemical response at the third month, P < 0.01; at the sixth month, P < 0.05; at 1 year, P < 0.01). Predictive Performance of Biochemical Response for Long-Term Outcome. The performance of biochemical response after 3, 6, and 12 months of UDCA therapy for prediction of long-term outcome was assessed using the Paris, Barcelona, Toronto, and Ehime definitions (Table 4). For that purpose, we used Corpechot et al. s calculation method and considered biochemical response as a positive test and the absence of adverse outcome as an event. 14 Compared with biochemical responses evaluated at 1 year, biochemical responses at the third month demonstrated higher PPV (Paris criteria, 0.93 versus 0.91; Barcelona criteria, 0.87 versus 0.84; Toronto criteria, 0.95 versus 0.93; Ehime criteria, 0.90 versus 0.89) but lower NPV (Paris criteria, 0.38 versus 0.47; Barcelona criteria 0.26

HEPATOLOGY, Vol. 58, No. 1, 2013 ZHANG, SHI, ET AL. 269 Fig. 3. Survival without adverse outcome for UDCA-treated patients with PBC according to biochemical response at 3, 6, and 12 months as defined by the Barcelona, Paris, Rotterdam, Toronto, and Ehime criteria. a, responders; b, nonresponders.

270 ZHANG, SHI, ET AL. HEPATOLOGY, July 2013 Table 4. Performance of Biochemical Response for Prediction of Long-Term Outcome Response Definition Sensitivity Specificity PPV NPV PLR NLR Paris Third month 0.75 0.73 0.93 0.38 2.77 0.34 Sixth month 0.80 0.66 0.90 0.45 2.32 0.30 1 year 0.79 0.70 0.91 0.47 2.61 0.30 Barcelona Third month 0.71 0.50 0.87 0.26 1.42 0.58 Sixth month 0.80 0.48 0.86 0.38 1.55 0.41 1 year 0.77 0.45 0.84 0.35 1.42 0.50 Toronto Third month 0.67 0.82 0.95 0.34 3.68 0.40 Sixth month 0.80 0.76 0.93 0.49 3.32 0.26 1 year 0.75 0.79 0.93 0.46 3.54 0.32 Ehime Third month 0.43 0.77 0.90 0.22 1.91 0.73 Sixth month 0.63 0.79 0.92 0.35 3.04 0.47 1 year 0.65 0.70 0.89 0.35 2.16 0.50 Biochemical response was considered as a positive test and the absence of adverse outcome as an event. Abbreviation: PLR, positive likelihood ratio. versus 0.35; Toronto criteria, 0.34 versus 0.46; Ehime criteria 0.22 versus 0.35), and increased NLR (Paris criteria, 0.34 versus 0.30; Barcelona criteria, 0.58 versus 0.50; Toronto criteria, 0.40 versus 0.32; Ehime criteria, 0.73 versus 0.50), suggesting that biochemical responses at the third month were superior in selecting patients with good prognosis yet inferior in selecting high-risk patients. In contrast, biochemical responses at the sixth month showed higher or the same PPV (Paris criteria, 0.90 versus 0.91; Barcelona criteria, 0.86 versus 0.84; Toronto criteria, 0.93 versus 0.93; Ehime criteria, 0.92 versus 0.89), higher or the same NPV (Paris criteria, 0.45 versus 0.47; Barcelona criteria, 0.38 versus 0.35; Toronto criteria, 0.49 versus 0.46; Ehime criteria, 0.35 versus 0.35), and lower NLR (Paris criteria, 0.30 versus 0.30; Barcelona criteria, 0.41 versus 0.50; Toronto criteria, 0.26 versus 0.32; Ehime criteria, 0.47 versus 0.50) compared with biochemical responses evaluated after 1 year of UDCA therapy. This result suggests that biochemical responses at the sixth month may more accurately identify patients with good or poor prognosis compared with evaluation at 1 year of UDCA treatment. Discussion The identification of PBC patients with poor longterm outcome among those treated with an adequate dose of UDCA is an important issue in clinical practice as well as in the design of therapeutic trials. The biochemical response to UDCA serves as a strong predictor of long-term outcome 6-10 and was recommended as one of the study endpoints in clinical trials where traditional endpoints were deemed unfeasible. 11 Biochemical response has been defined by numerous criteria, yet irrespective of the definition for response that was used, the prognosis of responders was significantly better than that of nonresponders. The Paris criteria were recognized as the best validated and easiest to use. 7,11 Using published criteria, we sought to determine whether a biochemical response as early as 3 to 6 months instead of 1 year would similarly identify patients with poor long-term outcome; if true, it could facilitate a more rapid selection of patients suitable for new therapeutic approaches. In the present study, we analyzed prospectively collected data of 187 patients with a mean follow-up period of 5.9 years. First, we found that serum bilirubin, ALP, GGT, AST, ALT, and IgM levels most prominently decreased within the first 3 months of UDCA therapy. These laboratory parameters continued to decrease gradually, with the maximum response seen at either 6 months or 1 year. Second, we found that the Paris, Barcelona, Toronto, and Ehime definition applied at 3, 6, and 12 months all significantly discriminated the patients in terms of long-term outcome, whereas no significant association was found with the Rotterdam definition (Table 3 and Fig. 3). Finally, we found that biochemical response at the sixth month can more accurately identify patients with good or poor prognosis compared with that at 1 year. The long-term evolution of laboratory liver parameters beyond 1 year UDCA therapy has been documented, suggesting that biochemical response to UDCA can be maintained for up to 15 years. 3,16 In contrast, laboratory parameters within the first year were seldom reported in a large cohort of patients. Our cohort consisted of 187 patients who were followed at 3-month intervals. Laboratory investigations were performed and data were collected prospectively. All of the laboratory parameters studied showed a prominent improvement in the first 3 months and then stayed relatively stable for the following months within the first year of UDCA treatment (Fig. 1). This led us to hypothesize that an early biochemical response as short as 3 to 6 months may be used in place of that after 1 year of UDCA therapy. We then evaluated the prognostic impact of multiple criteria in our patients. By all definitions except the Rotterdam criteria, biochemical response at 3, 6, and 12 months significantly discriminated our patients in terms of long-term outcome (Table 3 and Fig. 3). Our results tend to agree with those of the study

HEPATOLOGY, Vol. 58, No. 1, 2013 ZHANG, SHI, ET AL. 271 recently published by the Paris group. 14 The Paris group s study included 165 patients with early PBC, and no significant association was found between the long-term outcomes and the Rotterdam definition. Since the Rotterdam criteria have been demonstrated to be more potent prognostic indicators of long-term outcome in late rather than early stages of PBC, 8 they may not be applicable in a cohort of patients that contains high proportions of early PBC. Of note, for each definition, biochemical responses at 3 or 6 months were compatible with those at 1 year in discriminating patients with poor long-term outcome. Our patients showed significantly better prognosis compared with previously published studies of Chinese patients. 17,18 Wong et al. 17 reported that 14 (35.9%) out of 39 patients developed hepatic decompensation or hepatocellular carcinoma during a median followup of 4 years. In Zhao et al. s cohort, 18 65 (44.2%) out of 147 patients developed hepatic decompensation, and 36 (24.5%) patients died or underwent liver transplantation. We have followed up for a relatively longer period (median 5.8 years), during which 12 (6.4%) out of 187 patients died or underwent liver transplantation and 25 (13.4%) patients developed complications of cirrhosis or hepatocellular carcinoma. This probably reflects the variation in the severity of the disease in the patient populations. We excluded at study entry patients with autoimmune hepatitis overlap syndrome and/or cirrhosis-related complications. These exclusion criteria may explain why our cohort contained a higher proportion of patients with early PBC and thus demonstrated better prognosis compared with the other two Chinese cohorts. We also examined the influence of the initial severity of disease on the prognostic performance of biochemical response at 3, 6, and 12 months. Because both histological and nonhistological criteria (the Dutch prognostic class 5 ) have been used in recent studies to grade the severity of disease, 8,14 we applied both criteria in our analyses. The prognostic impact of biochemical response on survival remained significant after stratifying based on the Dutch prognostic class. However, after stratifying based on histological stages, the impact of biochemical response was not statistically significant. This discrepancy may be due to the fact that only 72 (39%) patients had available biopsy specimens. This reduced sample size may result in insufficient power to detect a certain effect. Nevertheless, by using the nonhistological criteria, we were able to show that biochemical response was an independent prognostic factor for survival without adverse outcome, irrespective of the initial severity of the disease. We also realized that liver biopsy is a very useful tool for assessing the stage of the disease at diagnosis. However, the number of patients with available biopsy specimens was relatively small in the present study. In the future, we will perform liver biopsies for histological assessment at diagnosis if the patient s condition allows for a liver biopsy. In this study, we used PPV, NPV, and NLR to compare the discriminatory capabilities of each test. PPV and NPV give the probabilities that an individual is truly positive given that they tested positive or truly negative given that they tested negative. NLR estimates the extent to which a negative test decreases the likelihood that a patient has that disease. These values help the clinician to interpret the accuracy of an individual test. We defined a positive test and an event as suggested by Corpechot et al. 14 : biochemical response was considered to be a positive test, and the absence of adverse outcome was considered to be an event. Thus, a higher PPV suggests more accuracy of a test in identifying patients with good prognosis; a higher NPV suggests more accuracy of a test in identifying patients with poor prognosis; and a lower NLR suggests more accuracy of a test in identifying patients with poor prognosis. The performance of biochemical response after 3, 6, and 12 months of UDCA therapy for prediction of long-term outcome was assessed (Table 4). We found that biochemical responses at the sixth month may identify patients with good or poor prognosis with no less accuracy than evaluation at 1 year. The former showed higher or the same PPV and NPV and lower NLR than the latter by all definitions tested. In contrast, biochemical responses at the third month demonstrated higher PPV, lower NPV, and increased NLR by all definitions compared with biochemical responses at 1 year. It can be inferred that biochemical responses at the third month were superior in identifying patients with good prognosis, yet were less potent in selecting high-risk patients for therapeutic trials. Our findings are in accordance with and provide more direct evidence to the statement that therapeutic trials should target patients with incomplete biochemical response after 3 to 6 months of UDCA treatment. 11 Our study was limited by small sample size and relatively short duration of follow-up. Another limitation is that a considerable proportion of patients did not strictly follow the regular 3-month examination interval, making biochemical data available for 128 (68.4%) patients for the third month after UDCA treatment, 145 (77.5%) patients for the sixth month, and 157 (84.0%) patients for 1 year. To make full use

272 ZHANG, SHI, ET AL. HEPATOLOGY, July 2013 of the available data, we used all data for the statistical analyses. To avoid bias, we compared the baseline characteristics of patients with or without recorded biochemical data. They did not differ statistically in age, sex, baseline biochemical data, and long-term outcomes (data not shown). Thus, the influence of missing data would mostly likely not effect the results. The strict criteria for selecting patients at entry, the careful follow-up of patients, and the prospective nature of the collected data can partly compensate for these limitations. The evolution of laboratory liver parameters within the first year of UDCA therapy, the compatibility of biochemical response at 3, 6, and 12 months in discriminating patients with good or poor outcome, and similar prognostic impact of biochemical responses at 6 and 12 months together contributed to our conclusion that an early biochemical response at 6 months would as efficiently identify patients at risk of poor outcome as evaluation at 1 year. In conclusion, our cohort of PBC patients showed the evolution of laboratory liver parameters within the first year of UDCA therapy: serum bilirubin, albumin, ALP, GGT, AST, ALT, and IgM levels showed a prominent improvement in the first 3 months and then stayed relatively stable for the following months. The Paris, Barcelona, Toronto, and Ehime definition applied at 3, 6, and 12 months significantly discriminated the patients in terms of long-term outcome. A biochemical response as early as 6 months after UDCA therapy predicts long-term outcome of PBC. For the previously published criteria, biochemical responses at the sixth month can be used in place of those evaluated after 1 year of UDCA therapy. Our findings provide important information that will be helpful in clinical evaluation of PBC patients. It may also facilitate a more rapid identification of patients who need new therapeutic approaches. References 1. Kaplan MM, Gershwin ME. Primary biliary cirrhosis. N Engl J Med 2005;353:1261-1273. 2. Selmi C, Bowlus CL, Gershwin ME, Coppel RL. Primary biliary cirrhosis. Lancet 2011;377:1600-1609. 3. Jorgensen R, Angulo P, Dickson ER, Lindor KD. Results of long-term ursodiol treatment for patients with primary biliary cirrhosis. Am J Gastroenterol 2002;97:2647-2650. 4. Corpechot C, Carrat F, Bahr A, Chretien Y, Poupon RE, Poupon R. The effect of ursodeoxycholic acid therapy on the natural course of primary biliary cirrhosis. Gastroenterology 2005;128:297-303. 5. ter Borg PC, Schalm SW, Hansen BE, van Buuren HR. Prognosis of ursodeoxycholic acid-treated patients with primary biliary cirrhosis. Results of a 10-yr cohort study involving 297 patients. Am J Gastroenterol 2006;101:2044-2050. 6. Pares A, Caballeria L, Rodes J. Excellent long-term survival in patients with primary biliary cirrhosis and biochemical response to ursodeoxycholic acid. Gastroenterology 2006;130:715-720. 7. 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;48:871-877. 8. 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;136:1281-1287. 9. 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;105:2186-2194. 10. Azemoto N, Kumagi T, Abe M, Konishi I, Matsuura B, Hiasa Y, et al. Biochemical response to ursodeoxycholic acid predicts long-term outcome in Japanese patients with primary biliary cirrhosis. Hepatol Res 2011;41:310-317. 11. Silveira MG, Brunt EM, Heathcote J, Gores GJ, Lindor KD, Mayo MJ. American Association for the Study of Liver Diseases endpoints conference: design and endpoints for clinical trials in primary biliary cirrhosis. HEPATOLOGY 2010;52:349-359. 12. Angulo P, Lindor KD, Therneau TM, Jorgensen RA, Malinchoc M, Kamath PS, et al. Utilization of the Mayo risk score in patients with primary biliary cirrhosis receiving ursodeoxycholic acid. Liver 1999;19:115-121. 13. Azemoto N, Abe M, Murata Y, Hiasa Y, Hamada M, Matsuura B, et al. Early biochemical response to ursodeoxycholic acid predicts symptom development in patients with asymptomatic primary biliary cirrhosis. J Gastroenterol 2009;44:630-634. 14. Corpechot C, Chazouilleres O, Poupon R. Early primary biliary cirrhosis: biochemical response to treatment and prediction of long-term outcome. J Hepatol 2011;55:1361-1367. 15. Ludwig J, Dickson ER, McDonald GS. Staging of chronic nonsuppurative destructive cholangitis (syndrome of primary biliary cirrhosis). Virchows Arch A Pathol Anat Histol 1978;379:103-112. 16. Kuiper EM, Hansen BE, Lesterhuis W, Robijn RJ, Thijs JC, Engels LG, et al. The long-term effect of ursodeoxycholic acid on laboratory liver parameters in biochemically non-advanced primary biliary cirrhosis. Clin Res Hepatol Gastroenterol 2011;35:29-33. 17. Wong GL, Hui AY, Wong VW, Chan FK, Sung JJ, Chan HL. A retrospective study on clinical features and prognostic factors of biopsy-proven primary biliary cirrhosis in Chinese patients. Am J Gastroenterol 2005;100:2205-2211. 18. Zhao DT, Liao HY, Liu YM, Zhao Y, Feng X, Yan HP. Prognostic factors and survival analysis of antimitochondrial antibody-positive primary biliary cirrhosis in Chinese patients. Dig Dis Sci 2011;56:2750-2757.