ORIGINAL ARTICLE. Received April 30, 2007; accepted June

Transcription

1 LIVER TRANSPLANTATION 13: , 2007 ORIGINAL ARTICLE Human Leukocyte Antigen and Adult Living- Donor Liver Transplantation Outcomes: An Analysis of the Organ Procurement and Transplantation Network Database S. Simona Jakab, 1 Victor J. Navarro, 1 Beth W. Colombe, 2 Constantine Daskalakis, 3 Steven K. Herrine, 1 and Simona Rossi 1 1 Department of Medicine, Division of Gastroenterology and Hepatology, Jefferson Medical College, Thomas Jefferson University, Philadelphia, PA; 2 Department of Tissue Typing, Thomas Jefferson University Hospital, Thomas Jefferson University, Philadelphia, PA; 3 Division of Biostatistics, Thomas Jefferson University, Philadelphia, PA Human leukocyte antigen (HLA) compatibility has no clinically significant impact in cadaveric liver transplantation. Less is known regarding living-donor liver transplantation (LDLT). Our prior analysis of the Organ Procurement and Transplantation Network (OPTN) database suggested a higher graft failure rate in patients who underwent LDLT from donors with close HLA match. We further investigated the effect of HLA-A, -B, and -DR matching on 5-yr graft survival in adult LDLT by analyzing OPTN data regarding adult LDLT performed between 1998 and We evaluated associations between 5-yr graft survival and total, locus-specific, and haplotype match levels. Separate analyses were conducted for recipients with autoimmune (fulminant autoimmune hepatitis, cirrhosis secondary to autoimmune hepatitis, primary biliary cirrhosis, primary sclerosing cholangitis) or nonautoimmune liver disease. Multivariable Cox proportional hazard models were used to evaluate interactions and adjust for potential confounders. Among 631 patients with available donor/recipient HLA data, the degree of HLA match had no significant effect on 5-yr graft survival, even when analyzed separately in recipients with autoimmune vs. nonautoimmune liver disease. To be able to include all 1,838 adult LDLTs, we considered a first-degree related donor as substitute for a close HLA match. We found no difference in graft survival in related vs. unrelated pairs. In conclusion, our results show no detrimental impact of close HLA matching on graft survival in adult LDLT, including in recipients with underlying autoimmune liver disease. Liver Transpl 13: , AASLD. Received April 30, 2007; accepted June Adult live-donor liver transplantation (LDLT) offers several advantages over cadaveric liver transplantation: more expedient transplantation, scheduling of procedures on an elective basis, and a positive effect on the cadaveric donor organ shortage. Also, LDLT may have the theoretical advantage of a more favorable human leukocyte antigen (HLA) match profile among closely related donor and recipient pairs. However, whether close HLA match actually confers benefit or not in liver transplantation remains in question. HLA compatibility does not seem to have a significant impact on patient or graft survival in cadaveric liver transplantation. 1 Nevertheless, the role of HLA in liver transplantation has been a subject of controversy over the last 2 decades. Earlier studies observed that a higher degree of HLA matching could be associated with Abbreviations: HLA, human leukocyte antigen; LDLT, living-donor liver transplantation; OPTN, Organ Procurement and Transplantation Network. Supported by the Health Resources and Services Administration, contract number The content is the responsibility of the authors alone and does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government. Address reprint requests to Victor J. Navarro, MD, 132 South 10th Street, Suite 480 Main, Philadelphia, PA Telephone: ; FAX: ; victor.navarro@jefferson.edu DOI /lt Published online in Wiley InterScience ( American Association for the Study of Liver Diseases.

2 1406 JAKAB ET AL. increased graft failure, 2,3 despite lower rates of rejection, but subsequent reports did not confirm this dualistic effect of HLA matching. 4 Our group recently conducted an analysis of 29,675 transplants using data from the Organ Procurement and Transplantation Network (OPTN) database, 5 the overall degree of HLA-A, -B, and -DR match had no effect on graft survival, both overall and for specific etiologies of liver disease. However, the subgroup with 0 HLA mismatches had significantly higher 5-yr graft failure (19.3% vs. 12.9%, P 0.02). Although this group was small (114 patients, 0.4% of all recipients), there was a high proportion of live donors. This analysis thus raised the question of whether close HLA match in the context of LDLT had a negative effect on graft survival. Using data that exists on adult LDLT donor/recipient pairs through the United Network for Organ Sharing, we performed an analysis of the effect of the degree of HLA match on adult LDLT outcomes. Our aim was to examine the impact of HLA matching on 5-yr graft survival. PATIENTS AND METHODS Patient Population We used data from the OPTN database as of December 27, 2005, to retrospectively analyze a cohort of patients who underwent adult LDLT in the United States. The OPTN database contains information on all liver transplants performed in the United States and reported to the OPTN since October 1, We included adult patients (age 18 yr or older) who received a liver graft from a live donor between 1998 and For patients who underwent retransplantation, only data regarding the first transplantation was analyzed. Donor and recipient pairs with missing data for HLA A, B, DR alleles were excluded. Variables To identify the effect of specific variables on the primary outcome, data for several characteristics were extracted. These included donor and recipient HLA A, B, and DR phenotype, age, gender, ethnicity, ABO blood group matching, and the relationship of the donor to the recipient. In addition, we collected the following information on recipients: etiology of liver disease, medical condition at transplantation (hospitalized in the intensive care unit, hospitalized but not in intensive care unit, nonhospitalized), year of transplantation, and time on the waiting list. Patient status (living, dead, retransplanted), and graft status (functioning, failed) were recorded at 5 yr after transplantation, as well as the date and the cause of death/graft failure, where applicable. We classified the recipient etiology of liver disease as autoimmune (which included cirrhosis secondary to autoimmune hepatitis, fulminant autoimmune hepatitis, primary biliary cirrhosis, and primary sclerosing cholangitis) vs. nonautoimmune. Donor/recipient HLA compatibility for HLA class I (A and B), and HLA class II (DR) was measured as the number of matches, locus-specific (0 to 2 matches) and overall for the A, B, and DR loci (0 to 6 matches). The primary outcome of our analyses was graft failure at 5 yr after transplantation. We felt this period of post- LDLT observation would be sufficiently long to permit any consequences of HLA effect to become detectable, with less effect from technical complications on graft survival which would more likely occur early following LDLT. Graft failure was defined as retransplantation or death from a transplant-related cause (according to the United Network for Organ Sharing categories of graft failure: biliary tract complications, primary graft failure, recurrent disease in the graft, de novo hepatitis, other infections, vascular thrombosis, graft-vs.-host disease, and other graft failure). We limited the analyses to the first 5 yr after liver transplantation, since it is unlikely that HLA compatibility would have any significant influence beyond that time point. Patients without graft failure at 5 yr after transplantation were censored at that time point. Patients lost to follow-up during the first 5 yr posttransplantation were censored at the last recorded visit. Patients who died from a nontransplant related cause were censored at the date of death. Statistical Analyses Statistical analyses were based on survival methods, specifically the Kaplan-Meier approach, and Cox proportional hazards regression. We first analyzed survival by the degree of the combined HLA matching at 3 loci: A, B, and DR (match levels 0 to 6). Univariate comparisons were made using the log-rank test. Separate survival curves were generated for recipients with autoimmune liver diseases and nonautoimmune liver diseases. We also investigated the effect of HLA matching at each locus (A, B, DR) individually (match levels 0 to 2), for the entire study population and separately for autoimmune and nonautoimmune liver diseases. Multivariable Cox proportional hazard models were developed to adjust for the effect of potential confounding variables, such as age, gender and ethnicity of the donor, age, gender and ethnicity of the recipient, donor/ recipient relationship, donor/recipient ABO blood group matching, diagnosis of autoimmune liver disease (cirrhosis secondary to autoimmune hepatitis, fulminant autoimmune hepatitis, primary biliary cirrhosis, primary sclerosing cholangitis), recipient medical condition, year of transplantation, and time on the waiting list. We also calculated the independent effect of each variable. Statistical significance was determined using the likelihood ratio test. P 0.05 was considered statistically significant. To investigate whether the effect of HLA compatibility depends on the etiology of liver disease (specifically autoimmune liver diseases) or on the donor/recipient relationship, we also tested the corresponding interactions between these variables. All analyses were conducted using SAS 8.2 (SAS Institute, Cary, NC).

3 HLA AND ADULT LDLT 1407 TABLE 1. Descriptive Summary of Donor and Recipient Characteristics n % Donor age (years), mean SD Donor gender Female Male Donor race/ethnicity White (non-hispanic) African American, Hispanic, Other Recipient age (years), mean SD Recipient gender Female Male Recipient race/ethnicity* White, non-hispanic African American, Hispanic, Other Recipient primary diagnosis Non-autoimmune Autoimmune Recipient medical condition at transplant Non-hospitalized Hospitalized (non-icu, ICU) Year of transplant Donor/recipient relationship Unrelated Related (first-degree relative) Donor/recipient HLA matching (out of 6) 0 match match matches matches matches matches matches Abbreviations: SD, standard deviation; ICU, intensive care unit. *Counts do not add up to 631 because of missing values for some cases. RESULTS Study Cohort Characteristics The OPTN database contained information regarding 1,838 LDLTs in recipients of 18 yr of age and older, performed from 1998 to Only 7 LDLT were reported before Donor/recipient HLA data was available for 631 (34%) patients, which constituted the study cohort. Table 1 is a detailed summary of the donor and recipient characteristics. Of a total of 108 (17%) patients who developed graft failure in the first 5 yr after transplantation, 30 died and 78 underwent retransplantation. The time to graft failure among these patients ranged from 1 day to 58 months, with a median of 19 days. A total of 90 of the 108 graft failures occurred in the first year after transplantation. HLA Match Analyses The average number of HLA matches (A, B, and DR) between donor and recipient was 2.6 (out of 6). The majority of patients had 3 matches 233 (37%). Because of the small number of patients in the 5- and 6-match groups 42 (6.6%) and 13 (2.1%), respectively these patients were combined into a single category. As expected, matching among first-degree relatives was higher than for the rest of the patients (3.3 vs. 1.5). Of the 140 donor/recipient pairs with 4 or more matches, 125 (89%) were first-degree relatives; in contrast, of the 177 pairs with 0 or 1 match, 142 (80%) were not firstdegree relatives. Five-Yr Graft Failure by HLA Matching Kaplan-Meier survival curves (Fig. 1) showed no significant association between the time to graft failure and the degree of donor/recipient HLA match combined for A, B, and DR loci, considering 6 levels of match (0, 1, 2, 3, 4, and 5 or 6) log-rank P In multivariable Cox proportional hazard regression analysis (Table 2), donor/recipient HLA compatibility was not found to be a significant predictor (global P 0.274). The only variables associated with higher risk of graft failure were hospitalization at the time of transplantation (hazard ratio 2.32, P 0.004), increased donor age (P 0.027), and transplantation before the year of 2000 (P 0.011). Five-Yr Graft Failure by HLA Matching and Diagnosis We also investigated the hypothesis that the effect of donor/recipient HLA compatibility on graft failure may depend on the recipient s primary diagnosis. Figure 2a and b are Kaplan-Meier plots of time to graft failure, by degree of HLA compatibility, separately for nonautoimmune liver disease patients (87 graft failures among 457 transplants) and for autoimmune liver disease patients (21 graft failures among 174 transplants). The association between HLA matching and time to graft failure was not significant in either subgroup (P and 0.977, respectively). We also formally tested the interaction between HLA compatibility and recipient primary diagnosis in a multivariable Cox proportional hazards model. The estimated HLA effects among nonautoimmune and autoimmune cases are summarized in Table 3. Among the nonautoimmune cases, zero matches seemed to be at somewhat higher risk, although the differences were not significant (global P 0.198). Among the autoimmune cases, there was no clear pattern of difference by degree of HLA matching (global P 0.884). The interaction between HLA match-

4 1408 JAKAB ET AL. Figure 1. Time to graft failure by donor/recipient HLA compatibility. There is no significant association between the time to graft failure and the degree of donor/recipient HLA match combined for A, B, and DR loci, considering 6 levels of match (0, 1, 2, 3, 4, and 5 or 6). ing and recipient primary diagnosis was not significant (P 0.616). Five-Yr Graft Failure by HLA Haplotype Matching We investigated HLA haplotype matching among the 383 related donor/recipient pairs (first-degree relatives). We assumed that, among such relatives, a single antigen match on each of the 3 loci indicates a 1-haplotype match, while a 6 antigen (full) match on all 3 loci indicates a 2-haplotype match. There were 12 graft failures among the 63 cases with a 0-haplotype match (19%), 48 among the 284 cases with a 1-haplotype match (17%), and 10 among the 36 cases with a 2-haplotype match (28%). Kaplan-Meier analysis did not demonstrate any correlation of time to graft failure with HLA haplotype matching (Fig. 3, log-rank P 0.425). Estimated hazard ratios from a multivariable Cox proportional hazards regression did not show any relationship between HLA haplotype matching and graft failure (0.73 for 1 vs. 0 haplotype matches, and 1.37 for 2 vs. 0 matches, global P 0.270), not even when separating patients with autoimmune vs. nonautoimmune etiologies of liver disease. HLA Locus-Specific Analyses We carried out analyses to evaluate the impact of degree of matching separately for each of the 3 HLA loci (A, B, and DR). Matching at any single HLA locus was not significant, as analyzed by Kaplan-Meier plots of time to graft failure by HLA matching on HLA A, B, and DR loci (P 0.489, 0.930, and 0.234, respectively). Multivariable Cox proportional hazards models showed possible protective effects of HLA matching for each of the 3 loci, but not statistically significant (Table 2). Donor/Recipient Relationship as a Surrogate for HLA Matching During the period 1998 to 2005, there were 1,838 firsttime LDLTs with recipients 18 yr of age or older. However, we had to restrict our main analyses to the 631 cases (34%) for which both donor and recipient HLA data were available. The excluded cases had a higher fraction of male donors (55% vs. 51%, P 0.091), minority donors (22% vs. 15%, P 0.001), minority recipients (24% vs. 16%, P 0.001), and recipients who were hospitalized at the time of the transplantation (19% vs. 12.5%, P 0.001). However, the graft failure rates were almost identical among cases excluded because of missing HLA data and in the main analyses sample (17.7% vs. 17.1%). As a proxy for the association between donor/recipient HLA compatibility and graft failure in the full set of 1,838 adult LDLT, we examined the association between donor/recipient relationship and graft failure.

5 HLA AND ADULT LDLT 1409 TABLE 2. Multivariate Analysis: Cox Proportional Hazards Model (Selected Variables) Graft failure n (%) HR 95% CI P value Donor age (years) (19) 1.00 Ref (13) , (17) , (24) , 3.01 Donor sex Female 53 (17) 1.00 Ref Male 55 (17) , 1.49 Donor race/ethnicity White (non-hispanic) 85 (16) 1.00 Ref Non-white 23 (24) , 2.23 Recipient age (years) (14) 1.00 Ref (15) , (22) , (15) , 3.85 Recipient sex Female 40 (14) 1.00 Ref Male 68 (19) , 2.03 Recipient race/ethnicity White (non-hispanic) 81 (16) 1.00 Ref Non-white 25 (25) , 5.22 Recipient primary diagnosis Non-autoimmune 87 (19) 1.00 Ref Autoimmune 21 (12) , 1.33 Recipient medical condition at transplant Non-hospitalized 85 (15) 1.00 Ref Hospitalized 23 (29) , 3.64 Year of transplant (28) 1.00 Ref (13) , 0.88 Donor/recipient relationship Related (first-degree) 38 (15) 1.00 Ref Unrelated 70 (18) , 1.42 Donor/recipient HLA matching (out of 6) match 18 (22) 1.00 Ref 1 match 15 (16) , matches 11 (14) , matches 42 (18) , matches 10 (12) , or 6 matches 12 (22) , 1.63 Donor/recipient HLA-A matching match 29 (19) match 56 (15) , matches 23 (20) , 1.45 Donor/recipient HLA-B matching match 36 (17) match 61 (17) , matches 11 (18) , 1.54 Donor/recipient HLA-DR matching match 36 (21) match 56 (15) , matches 16 (21) , 1.65 Abbreviations: HR, hazard ratio; CI, confidence interval, Ref, reference group. The results of the multivariable Cox proportional hazards model in these analyses which did not directly account for HLA matching were qualitatively similar to the ones obtained in our HLA analyses presented above (summarized in Table 2). For example, in the analyses of all 1,838 cases, the estimated hazard ratio for unrelated vs. related donor/recipient pairs was 1.21 (0.80 in Table 2), for autoimmune vs. nonautoimmune disease 0.96 (0.79 in Table 2), for historically later vs. earlier transplants 0.65 (0.57 in Table 2), and for hospitaliza-

6 Figure 2.

7 HLA AND ADULT LDLT 1411 TABLE 3. Association of Overall Donor/Recipient HLA Compatibility and Time to Graft Failure, by Recipient Primary Diagnosis Non-autoimmune Autoimmune Donor/recipient HLA matching (out of 6) Graft failure n (%) HR Graft failure n (%) HR 0 match 16 (26) (10) match 12 (17) (13) matches 8 (13) (14) matches 35 (21) (10) matches 7 (11) (15) or 6 matches 9 (25) (16) 1.20 Abbreviation: HR, hazard ratio. Figure 3. Time to graft failure by donor/recipient HLA haplotype matching. The number of HLA haplotype matches (0 to 2 matches) in related donor/recipient pairs, does not significantly influence graft failure. tion at time of transplantation 1.58 (2.16 in Table 2). This, although not conclusive, is indirect evidence that any selection bias introduced by the missing HLA data is probably of modest magnitude. Although there was Figure 2. Time to graft failure by donor/recipient HLA compatibility for: (A) non-autoimmune disease and (B) autoimmune disease. Kaplan-Meier survival curves, plotted separately for recipients with nonautoimmune or autoimmune liver disease, show no significant association between graft failure and the degree of donor/recipient HLA match combined for A, B, and DR loci, considering 6 levels of match (0, 1, 2, 3, 4, and 5or6). no difference in graft survival in related vs. unrelated pairs, when separating recipients based on the underlying liver disease, we found that receiving a graft from a first-degree relative had a beneficial effect in recipients with nonautoimmune liver disease (odds ratio 1 vs. 1.4, P 0.017), and was potentially associated with worse graft survival (did not reach statistical significance) in recipients with autoimmune liver disease (odds ratio 1 vs. 0.74, P 0.256). DISCUSSION Our prior research on the effect of HLA compatibility in liver transplantation suggested that graft failure tends

8 1412 JAKAB ET AL. to occur more frequently in patients who undergo LDLT and in the setting of a close HLA match. 5 Therefore, we designed this study to determine the impact of HLA matching on 5-yr graft survival in adult LDLT. We also explored the hypothesis that the effect of HLA matching may be different in recipients with underlying autoimmune diseases, as other studies have suggested. 4 Finally, we explored the hypothesis that graft outcomes may be different in closely related donor/recipient pairs, as first-degree relationship may be considered a surrogate for a closer HLA matching. Our findings show that any degree of HLA matching is not likely to have a negative impact on 5-yr graft survival in LDLT recipients. Moreover, close HLA match in recipients with underlying autoimmune liver disease does not reduce graft survival, compared with other recipients. Finally, using the degree of relatedness between the recipient and donor as a surrogate for close HLA match, we could analyze a larger population (1,838 adult LDLT vs. 638 with available HLA data), but we did not find any clinically significant effect on 5-yr graft survival. Markus, in 1988, proposed a dualistic effect of HLA matching. He found increased nonrejection graft failure with higher HLA match, 1 an observation suggesting that too much of a good thing, i.e., HLA compatibility, might be harmful. Twenty years later, this concept is still controversial. Several investigators proposed that transplanting a graft with a closely matched HLA phenotype could be associated with graft injury by enhancing immune-mediated mechanisms involved in recurrence of hepatitis B, C, and autoimmune liver diseases, 4 or in predisposing for CMV hepatitis. 3,6,7 These observations may have been confounded by center-specific factors that would have had less of an effect on the large multiinstitutional OPTN database. Nonetheless, the dualistic effect of HLA is of particular interest in LDLT because there is a higher degree of HLA matching when the graft is from a blood-related donor. The studies regarding the role of HLA compatibility in LDLT have generated inconsistent results. Most of them analyzed pediatric data, were limited by sample size, and lacked specific subgroup analysis. 8,-12 In adult LDLT, HLA matching was associated with lower incidence of rejection and no effect on graft survival. 11,12 In contrast, a very recent study evaluated the outcomes of LDLT in 50 adult patients with primary biliary cirrhosis. A higher number of HLA A/B/DR matches was associated with increased recurrence of primary biliary cirrhosis, but also with increased 5-yr survival (92.3% 5-yr survival estimate for 0-2 HLA A/B/DR mismatches, vs. 40.4% for more than 3 mismatches). 13 Our results suggesting that HLA matching has no clinically significant impact in adult LDLT are supported by a large database and robust analysis, taking into account many potential confounding and interacting variables that could have affected the primary outcome of 5-yr graft survival. The importance of these findings are to reaffirm the current day practice of performing all types of liver transplantation, cadaveric and LDLT, without regard to HLA, the presence of an underlying autoimmune disease, or the degree of relatedness of the donor and recipient. The significance of our findings notwithstanding, we made several observations that must be interpreted in light of our study s potential limitations. First, with regards to haplotype matching, there was an apparent difference in the group with 2-haplotype match; however, there were only 36 cases in this category. Second, although the majority of graft losses occurred in the first year after transplantation, our analysis was intended to focus upon the 5-yr graft survival for reasons previously stated. Third, we encountered the problem, as with most large multiinstitutional databases, of missing data and centerspecific activities, not captured by the database. These included specific causes of graft failure, types of immunosuppression, differences in surgical technique, and missing HLA information (we had to exclude 1,207 patients without HLA data, which limited subgroup analysis and decreased the possibility to achieve statistical significance, given the low rate of graft failure). Finally, it is possible that other residual confounders that affect graft survival have not been captured, because of unmeasured factors (e.g., immunosuppression regimens, surgical technique, specific HLA alleles implicated in disease recurrence) or genetic loci other than HLA that may influence immune responsiveness to a graft. REFERENCES 1. Gordon RD, Fung JJ, Markus B, Fox I, Iwatsuki S, Esquivel CO, et al. The antibody crossmatch in liver transplantation. Surgery 1986;100: Markus BH, Duquesnoy RJ, Gordon RD, Fung JJ, Vanek M, Klintmalm G, et al. Histocompatibility and Liver Transplant Outcome: Does HLA exert a dualistic effect? Transplantation 1988;46: Donaldson P, Underhill J, Doherty D, Hayllar K, Calne R, Tan KC, et al. Influence of human leucocytes antigen matching on liver allograft survival and rejection: the dualistic effect. Hepatology 1993;17: Neumann UP, Guckelberger O, Langrehr JM, Lang M, Schmitz V, Theruvath T, et al. Impact of Human Leukocyte Antigen Matching in Liver Transplantation. Transplantation 2003;75: Navarro V, Herrine S, Katopes C, Colombe B, Spain CV. The effect of HLA class I (A and B) and class II (DR) compatibility on liver transplantation outcomes: an analysis of the OPTN database. Liver Transpl 2006;12: Manez R, White LT, Linden P, Kusne S, Martin M, Kramer D, et al. The influence of HLA matching on cytomegalovirus hepatitis and chronic rejection after liver transplantation. Transplantation 1993;55: Belli LS, Zavaglia C, Alberti AB, Poli F, Rondinara G, Silini E, et al. Influence of immunogenetic background on the outcome of recurrent hepatitis C after liver transplantation. Hepatology 2000;31:

9 HLA AND ADULT LDLT Hirata M, Harihara Y, KitaY, Hisatomi S, Miura Y, Yoshino H, et al. Impact of HLA matching in living-related liver transplantation. Transplant Proc 2000;32: Harihara Y, Makuuchi M, Kawasaki S, Hashikura Y, Kawarasaki H, Takayama T, et al. Influence of HLA compatibility on living-related liver transplantation. Transplant Proc 2000;32: Kasahara M, Kiuchi T, Uryuhara K, Uemoto S, Fujimoto Y, Ogura Y, et al. Role of HLA compatibility in pediatric living-related liver transplantation. Transplantation 2002; 74: Sugawara Y, Makuuchi M, Kaneko J, Saiura A, Imamura H, Kokudo N, et al. Risk factors for acute rejection in living donor liver transplantation. Clin Transplant 2003;17: Suehiro T, Shimada M, Kishikawa K, Shimura T, Soejima Y, Yoshizumi T, et al. Influence of HLA compatibility and lymphocyte cross-matching on acute cellular rejection following living donor adult liver transplantation. Liver Int 2005;25: Morioka D, Egawa H, Kasahara M, Jo T, Sakamoto S, Ogura Y, et al. Impact of human leukocyte antigen mismatching on outcomes of living donor liver transplantation for primary biliary cirrhosis. Liver Transpl 2007;13: