This study was presented in part at the American Transplant Congress, Toronto, Canada, June 2008.
This is publication 15 of the Adult-to-Adult Living Donor Liver Transplantation Cohort Study.
Supplemental data included in this article were supplied by the Arbor Research Collaborative for Health as the contractor for the Scientific Registry of Transplant Recipients. The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as an official policy of or interpretation by the Scientific Registry of Transplant Recipients or the US Government.
This study was supported in part by the National Institutes of Health through the National Institute of Diabetes and Digestive and Kidney Diseases (grants U01-DK62536, U01-DK62444, U01-DK62467, U01-DK62483, U01-DK62484, U01-DK62494, U01-DK62496, U01-DK62498, U01-DK62505, and U01-DK62531), by the American Society of Transplant Surgeons, and by the US Department of Health and Human Services through the Health Resources and Services Administration.
In response to the organ donor shortage, living donor liver transplantation (LDLT) was expanded to adult recipients, and there was rapid growth across the United States after the first reported case in 1998.1 Limited numbers of procedures were performed at most centers, and the approaches to the recipients and donors were so diverse across centers that it was difficult to provide reliable information about outcomes that could be generalized and used for patient education. Therefore, in 2002, the National Institutes of Health, with supplemental funding from the American Society of Transplant Surgeons and the Health Resources and Services Administration (US Department of Health and Human Services), organized a consortium of 9 liver transplant centers and a data coordinating center to accrue and follow sufficient numbers of patients being considered for and undergoing LDLT to provide results from adequately powered studies. The Adult-to-Adult Living Donor Liver Transplantation Cohort Study (A2ALL) was developed with the aim of providing accurate information on the risks and benefits of adult-to-adult LDLT for both donors and recipients. Retrospective and prospective studies were initiated with the primary goal of providing information on donor and recipient outcomes over the span of a decade (1998-2008). Follow-up data collection was completed in 2009, and a subsequent renewal of A2ALL with additional centers has now been funded by the National Institutes of Health.
The first report from the 9 clinical centers in the A2ALL consortium focused on the predictors of graft loss after LDLT. A learning curve was identified, with graft failure risk decreasing significantly after the first 20 adult-to-adult LDLT procedures in each center.2 In the initial report, additional risk factors for graft failure after LDLT were also identified; these included older recipient age and the duration of cold ischemia. Variables that were not found to be significant included the Model for End-Stage Liver Disease (MELD) score, the hepatitis C virus (HCV) status, and the donor age.
Because graft failure is only one contributor to posttransplant mortality, it is also important to determine whether the significant risk factors associated with graft failure that have been identified by A2ALL also apply to recipient mortality at A2ALL and non-A2ALL centers. As A2ALL begins additional studies in its second phase, it is important to analyze the first phase of A2ALL and compare the A2ALL experience to the experience of other US centers. If the findings correlate with outcomes at other centers, then the A2ALL reports can be viewed as representative of national results. Therefore, the aim of this study was to determine whether the previously reported findings concerning the importance of center experience and other predictors of outcomes in A2ALL are reflected in the national LDLT experience.
A2ALL, Adult-to-Adult Living Donor Liver Transplantation Cohort Study; CI, confidence interval; CIT, cold ischemia time; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HR, hazard ratio; ICU, intensive care unit; LDLT, living donor liver transplantation; MELD, Model for End-Stage Liver Disease; SD, standard deviation; SRTR, Scientific Registry of Transplant Recipients.
PATIENTS AND METHODS
Data for this study were obtained from the Scientific Registry of Transplant Recipients (SRTR) under a data use agreement. SRTR data are sourced from national transplant data voluntarily submitted by all transplant centers in the United States to the Organ Procurement and Transplantation Network, and they are supplemented by data from the Social Security Death Master File to identify deaths not reported by transplant centers. The multiple-source follow-up or censoring date was calculated as the transplant anniversary (e.g., 6 months, 1 year, or 2 years) immediately before the current SRTR database snapshot date (August 1, 2008); a 3-month lag was allowed to ensure the completion of forms. Graft failure was defined as the date of retransplantation or death (whichever was earlier).
A comparison of A2ALL centers and non-A2ALL centers was conducted through an analysis of LDLT recipients at 9 A2ALL centers (n = 702) and 67 non-A2ALL centers (n = 1664) from January 1, 1998 to December 31, 2007. A case number was assigned to each LDLT procedure based on the number of adult LDLT procedures previously performed at that center up to that date. Cold ischemia time (CIT) was defined as the time from donor cross-clamping to the removal of the graft from ice. CIT data from the A2ALL retrospective cohort study were used to augment the data available in the SRTR database because the latter was missing a substantial amount of data.
Descriptive statistics included means, ranges, standard deviations (SDs), and proportions.
Potential predictors of the time from LDLT to death or graft failure were tested with multivariable Cox regression models that started at the time of transplantation; adjustments were made for center clustering with robust variances based on the sandwich estimator. Covariate effects are reported as hazard ratios (HRs) with 95% confidence intervals (CIs). The following covariates were evaluated: the center type (A2ALL versus non-A2ALL), the center-specific LDLT case number, the transplant year (on or before December 31, 2000 versus after December 31, 2000), the recipient age, the recipient weight, the donor age, the donor weight, diagnoses of hepatocellular carcinoma (HCC) and HCV, the presence of ascites, the pretransplant status [intensive care unit (ICU), hospital (non-ICU), or home], the CIT, the serum creatinine levels, and the biological relationship (biologically related versus unrelated). A2ALL data from SRTR were augmented with CIT data from A2ALL when they were not available from SRTR; this was not possible for the non-A2ALL data. Because of the level of missing CIT data in the SRTR database, separate CIT models were fitted with the subset of the cohort with complete CIT data. Statistical interactions between the center type and other predictors were tested. An assessment of the association of patient and graft survival with the case number was performed by the estimation of the HR by 5-case intervals (eg, cases 1-5 and cases 6-10); a case number >30 was used as the reference category. When individual center effects were added to the models with indicator variables, the variability in mortality among the centers was statistically significant (P < 0.001), even when it was limited to centers with more than 10 LDLT cases (P < 0.001); this motivated us to adjust the analysis for center clustering. P values ≤0.05 were considered to be statistically significant. All analyses were carried out with SAS 9.2 statistical software (SAS Institute, Inc., Cary, NC).
Protection of Human Subjects
The study was approved by the institutional review boards and privacy boards of the University of Michigan Data Coordinating Center and each of the 9 participating transplant centers.
The characteristics of patients from A2ALL and non-A2ALL centers are shown in Table 1. The mean donor and recipient ages were similar for the A2ALL and non-A2ALL centers. The A2ALL centers reported higher mean MELD scores and a higher percentage of recipients with HCV. A2ALL centers also had a higher percentage of left lobe recipients, but this percentage was very small in both types of centers (5% versus 4%). The overall mean creatinine levels were not different, but a significantly higher percentage of non-A2ALL patients had serum creatinine levels ≤0.7 mg/dL. The non-A2ALL group also had higher percentages of HCC patients and patients in the ICU or hospital. LDLT grafts at A2ALL centers had a significantly shorter mean CIT. Because most non-A2ALL centers never exceeded 15 LDLT cases in all, the proportion of recipient case numbers ≤15 was significantly higher in this group (33% versus 19%, P < 0.001).
Table 1. Characteristics of the Liver Transplant Recipients in the A2ALL and Non-A2ALL Centers
In all, 2366 LDLT procedures, including 702 (30%) at the 9 A2ALL centers and 1664 (70%) at 67 non-A2ALL centers, were performed during the study period. Approximately two-thirds (42/67) of the non-A2ALL centers performed 15 or fewer LDLT procedures over the 10-year study period. Five non-A2ALL centers and 3 A2ALL centers each performed more than 100 LDLT procedures. The number of years of activity (≥1 LDLT procedure per year) between 1998 and 2007 varied widely among the non-A2ALL centers (median = 4 years, range = 1-10 years), with only 1 of 67 programs (1.5%) active for all 10 years. In contrast, 3 of 9 A2ALL centers (33%) were active during each year (median = 9 years, range = 7-10 years).
A2ALL centers were among the first to implement LDLT programs. All 9 A2ALL centers performed their first LDLT procedures by the end of 2000, whereas approximately one-third of the 67 non-A2ALL centers performed their first LDLT procedures in 2001 or later (Fig. 1A). The distributions of LDLT procedures over time were relatively similar between the 2 groups (Fig. 1B). However, the non-A2ALL cases reflect many centers that started and stopped, whereas the A2ALL cases are from 9 centers that continued their activity throughout the study period. The national trend has been toward fewer centers performing LDLT. In 2008 and 2009, only 32 and 33 adult liver transplant programs, respectively, performed LDLT in the United States; 8 of these were A2ALL centers. A2ALL centers performed 30% and 37% of all US LDLT procedures in 2008 and 2009, respectively. The number of LDLT procedures performed in the United States peaked at 411 in 2001 and gradually declined to 198 in 2007 (Fig. 1C). Although the total number of adult LDLT procedures has continued to decline over the last 2 years in non-A2ALL centers (124 in 2008 and 106 in 2009), the A2ALL numbers slightly increased in 2008 (n = 54) and 2009 (n = 62).
The overall adjusted posttransplant mortality risk decreased after centers gained experience. When center experience was divided into 5-case intervals, separate models showed that case 15 was the threshold from higher mortality risk to lower mortality risk at A2ALL centers, and case 10 was the threshold at non-A2ALL centers (Fig. 2A). For subsequent analyses combining A2ALL and non-A2ALL transplants, we used case 15 as the threshold to differentiate earlier center experience (cases 1-15) from later center experience (case 16 and later cases).
In comparison with later center experience, earlier center experience was associated with significantly higher mortality risk at both A2ALL (HR = 2.24, P < 0.0001) and non-A2ALL centers (HR = 1.45, P = 0.005; Table 2 and Fig. 3A). In comparison with non-A2ALL centers that performed more than 15 LDLT procedures, mortality during the early experience was higher in both A2ALL centers (HR = 1.9, P < 0.001) and non-A2ALL centers that stopped by case 15 (HR = 1.44, P < 0.001). There was no statistical difference between A2ALL centers during the early experience and non-A2ALL centers that never performed more than 15 procedures (HR = 1.05, P = 0.80).
Table 2. Cox Regression Models for Patient Survival
NOTE: Adjustments were made for center clustering with robust variances.
Variables that were tested but were not significant included the recipient's weight, the use of dialysis, the presence of ascites, the donor's weight, the donor's biological relationship to the recipient, and the transplant year.
Interactions were tested one at a time in models including all the main effects listed as predictors in this table.
Results for CIT are based on separate models with a smaller sample size because of missing CIT values in 33% of cases.
Additional significant predictors of mortality (both groups combined) were identified (Table 2). Older donor age (HR = 1.13 per 10 years, P = 0.0002), older recipient age (HR = 1.20 per 10 years, P = 0.0003), a diagnosis of HCV (HR = 1.18, P = 0.026), a diagnosis of HCC (HR = 2.12, P < 0.0001), higher serum creatinine levels (HR = 1.52 per loge unit increase, P < 0.0001), the patient's medical condition (ICU versus home: HR = 2.52, P < 0.0001; hospitalization versus home: HR = 1.62, P < 0.0001), and early center experience (case number ≤ 15 versus case number > 15: HR = 1.61, P < 0.0001) were all significantly associated with a higher mortality risk. In a subset with available CIT data, a CIT > 4.5 hours was associated with a significantly higher mortality risk (HR = 1.79, P = 0.0006) but occurred in only 4% to 6% of the grafts. There was no significant difference in the patient mortality risk between the eras (on or before December 31, 2000 versus after December 31, 2000) after adjustments for center experience and other covariates.
In separate analyses of A2ALL centers and non-A2ALL centers, all associations were consistent in direction, although there were some differences in effect size and significance between A2ALL and non-A2ALL centers. The patient's medical condition was a significant predictor of patient mortality in non-A2ALL centers (ICU versus home: HR = 2.89, P < 0.0001; hospitalization versus home: HR = 1.67, P < 0.0001) but not in A2ALL centers. We tested for differences in the association of each factor by center type (ie, statistical interaction). The only significant interaction occurred with center experience; this effect was significantly stronger at A2ALL centers versus non-A2ALL centers, as reported previously (HR = 2.24 versus HR = 1.45, interaction P = 0.04).
When the center experience was divided into 5-case intervals, the threshold from higher graft failure risk to lower graft failure risk was identified as case 15 at A2ALL centers and as case 10 at non-A2ALL centers (Fig. 2B). As we did with the mortality analyses, we used case 15 as a threshold to differentiate earlier center experience from later center experience in models that included A2ALL and non-A2ALL centers.
A2ALL and non-A2ALL centers had similarly higher risks of graft failure during the early experience period (combined HR = 1.61, P < 0.0001) versus the later experience period (ie, after the completion of at least 15 cases; Table 3 and Fig. 3B). During their early experience, non-A2ALL centers that never performed more than 15 procedures had significantly higher graft failure rates than the non-A2ALL centers that performed more than 15 procedures (HR = 1.30, P = 0.002). This significant difference in the early experience was not seen when non-A2ALL centers that performed ≤15 LDLT procedures were compared with A2ALL centers during their early experience (HR = 1.23, P = 0.25).
NOTE: Adjustments were made for center clustering with robust variances.
Variables that were tested but were not significant included the use of dialysis, the presence of ascites, the donor's weight, the donor's biological relationship to the recipient, and the transplant year.
Interactions were tested one at a time in models including all main effects listed as predictors in this table.
Results for CIT are based on separate models with a smaller sample size because of missing CIT values in 33% of cases.
Significant predictors of graft failure (both groups combined) were identified (Table 3). Older donor age (HR = 1.13 per 10 years, P = 0.0002), a diagnosis of HCC (HR = 1.87, P < 0.0001), higher serum creatinine levels (HR = 1.26 per loge unit increase, P = 0.05), the patient's medical condition (ICU versus home: HR = 2.67, P < 0.0003; hospitalization versus home: HR = 1.49, P < 0.0001), and early center experience (case number ≤ 15 versus case number > 15: HR = 1.61, P < 0.0001) were associated with an increased risk of graft failure. Older recipient age (P = 0.26), heavier recipient weight (P = 0.06), and a diagnosis of HCV (P = 0.23) were also associated with increased graft failure and were retained in the model for face validity, even though they were not statistically significant. In the subset with CIT data, a CIT > 4.5 hours (HR = 1.49, P = 0.05) was associated with a higher graft failure risk.
In separate analyses of A2ALL and non-A2ALL centers that considered factors associated with graft failure, all associations except recipient age, recipient weight, and hospitalization versus home were consistent in direction. Older recipient age (HR = 1.26 per 10 years, P = 0.01), serum creatinine levels (HR = 1.43, P = 0.02), and a diagnosis of HCV (HR = 1.41, P = 0.02) were more significant predictors of graft failure at A2ALL centers versus non-A2ALL centers. The effect of CIT on the graft failure risk was significantly stronger at A2ALL centers (HR = 3.25, P < 0.0001) versus non-A2ALL centers (HR = 1.29, P = 0.34). The recipient's weight (HR = 1.09 per 10 kg, P = 0.011) and medical condition (ICU versus home: HR = 3.12, P < 0.0001; hospitalization versus home: HR = 1.58, P = 0.0002) were significant predictors of graft failure in non-A2ALL centers but not in A2ALL centers. We tested for differences by the center type (ie, statistical interactions) and found that the center type had a significant statistical interaction only with recipient age (P = 0.03) and CIT (P = 0.03).
As the transplant community continues to make efforts to define the most appropriate role for LDLT, it is important to identify the significant clinical risk factors associated with graft failure and recipient mortality. Within the A2ALL consortium, one of the first observations about adult-to-adult LDLT was the significant learning curve: improved graft survival was found after the first 20 cases at each center.2 We have also recently described a decrease in the incidence of recipient and donor complications after a period of experience.3, 4 Similar findings have been reported in large single-center reports: patient and graft survival has improved significantly after the initial center experience.5–7 Significant clinical characteristics associated with graft loss, including older recipient age and CIT, were also identified. With the first phase of A2ALL completed and the second phase beginning, it was important to determine whether the findings of the A2ALL consortium are representative of centers throughout the United States with respect to experience and posttransplant outcomes (specifically graft failure and patient mortality). Equally important were the goals of (1) identifying any significant similarities and differences between A2ALL and non-A2ALL centers with respect to other factors affecting outcomes that might alter the applicability of A2ALL findings to the general pool of patients undergoing LDLT and (2) providing evidence regarding the most appropriate recipients of LDLT.
In this report, we have again shown an association between center experience and both patient and graft outcomes after LDLT in the United States. Non-A2ALL centers that never performed more than 15 procedures had significantly higher rates of graft failure than centers that performed more procedures. After the first 15 cases, both A2ALL and non-A2ALL centers demonstrated significant decreases in posttransplant mortality. Although the A2ALL centers had significantly higher mortality rates for their first 15 cases, the results for non-A2ALL centers improved after only 10 LDLT cases, perhaps because many non-A2ALL centers started in later years, with experienced teams moving from centers that had already performed LDLT. In comparison, the compositions of surgery and hepatology teams remained stable in most A2ALL centers over the years.
Learning curves are often described after the introduction of a new procedure, but no new complex procedures have been introduced into the field of liver transplantation in the last 10 years except for right lobe LDLT. Relationships between experience and outcomes for kidney, liver, and heart transplantation have been reported,8, 9 with outcomes being better at high-volume centers, and a strong relationship has been reported between higher surgeon volume and decreased morbidity and mortality with other complex surgical procedures.10, 11 Therefore, the learning curve noted here was not unexpected because the introduction of adult-to-adult LDLT was a major technical development from deceased donor liver transplantation.
It was equally important to determine what other clinical factors contribute to mortality after LDLT and whether these factors are comparable between A2ALL and non-A2ALL centers. These findings can help centers to select the most appropriate donors and recipients for LDLT to achieve the best outcomes. For all centers, we found that older recipient age, donor age, a diagnosis of HCC, higher creatinine levels (a major component of the MELD score), and hospitalization or ICU admission contributed to recipient mortality. When we explored the clinical risk factors for mortality in A2ALL and non-A2ALL centers that were available, we found that these risk factor effects were not significantly different between A2ALL and non-A2ALL centers. The separate A2ALL and non-A2ALL models gave results of consistent direction and fairly similar magnitudes.
With respect to graft failure, we also found similar risk factor estimates for the A2ALL and non-A2ALL groups. When the cohorts were combined, older donor age, a diagnosis of HCC, higher serum creatinine levels, and ICU admission were all associated with a higher risk of graft loss. There were some differences in the effects of individual predictors between the 2 groups; recipient age and CIT were the only 2 variables for which statistically significant differences in their effects were demonstrated between the A2ALL and non-A2ALL cohorts. The CIT discrepancy may be explained by the fact that we were able to supplement missing SRTR data with A2ALL data for this field, and although the missing CIT data in the SRTR database may have affected the coefficient estimates, we have previously reported that CIT is significantly associated with various outcomes in A2ALL.2, 3, 12, 13 It is more difficult to explain the difference in the significance of recipient age in the A2ALL group versus the non-A2ALL group. Nonetheless, each of these factors may be clinically important, either alone or in combination, and should be considered when decisions regarding donor and recipient selection for LDLT are being made.
The A2ALL multicenter consortium was formed to address outcomes for both donors and recipients with detailed prospective data collection; its goal is to generalize these results to the national experience because detailed data collection and reporting are not feasible on a national scale. This report demonstrates that the A2ALL study results are generally representative of national outcomes. Although incomplete data collection and a lack of granularity in the SRTR data may have contributed to some of the differences noted in this study, we have shown comparability between the groups with respect to the magnitude and direction of risk factor effects.
Limitations of this study may include missing data and misclassification in the SRTR database for the covariates and outcomes investigated14 and potential secular trends not captured in the statistical modeling, although era effects (on or before December 31, 2000 versus after December 31, 2000) were tested. Also, because of the limitation of variables available in the SRTR database, we were not able to demonstrate the comparability of A2ALL and non-A2ALL centers for other published A2ALL results, such as donor and recipient complications and graft size.
From the results of all US centers performing LDLT, we have shown that gaining initial experience is important for improving survival after LDLT, regardless of when a center starts an adult LDLT program. The data also demonstrate a continuing decline in the number of adult LDLT procedures performed outside the A2ALL consortium, and this may demonstrate the natural tendency of these procedures to gravitate to experienced centers.
The analyses presented here support the application of findings from the A2ALL consortium to other centers in the United States performing LDLT. As we embark on further studies in the second phase of A2ALL, the goals of the A2ALL consortium are that (1) analyses of detailed data and lessons learned will contribute to future advances on a national scale, and (2) A2ALL findings will be used to provide guidance for center performance and clinical decision making in the field of adult-to-adult LDLT.
The authors gratefully acknowledge Abraham Shaked, M.D., Ph.D., for the concept and development of this article and Jay E. Everhart, M.D., for his intellectual contributions to this project. The authors also acknowledge the A2ALL principal investigators and their teams for their contributions to the overall study. The following individuals were instrumental in the planning, conduct, and/or care of patients enrolled in this study at each of the participating institutions (National Institute of Diabetes and Digestive and Kidney Diseases grant numbers are shown in parentheses):
Columbia University Health Sciences, New York, NY (DK62483): Jean C. Emond, M.D. (principal investigator); Robert S. Brown Jr., M.D., M.P.H. (coprincipal investigator); and Rudina Odeh-Ramadan, Pharm.D., and Scott Heese, B.A. (study coordinators).
Northwestern University, Chicago, IL (DK62467): Michael M. I. Abecassis, M.D., M.B.A. (principal investigator); Laura M. Kulik, M.D. (coprincipal investigator); and Patrice Al-Saden, R.N., C.C.R.C. (study coordinator).
University of Pennsylvania Health System, Philadelphia, PA (DK62494): Abraham Shaked, M.D., Ph.D. (principal investigator); Kim M. Olthoff, M.D. (coprincipal investigator); and Brian Conboy, P.A., M.B.A., and Mary Shaw, R.N., B.B.A. (study coordinators).
University of Colorado Health Sciences Center, Denver, CO (DK62536): Gregory T. Everson, M.D. (principal investigator); Igal Kam, M.D. (coprincipal investigator); and Carlos Garcia, B.S., and Anastasia Krajec, R.N. (study coordinators).
University of California Los Angeles, Los Angeles, CA (DK62496): Johnny C. Hong, M.D. (principal investigator); Ronald W. Busuttil, M.D., Ph.D. (coprincipal investigator); and Janet Mooney, R.N., B.S.N. (study coordinator).
University of California San Francisco, San Francisco, CA (DK62444): Chris E. Freise, M.D., F.A.C.S. (principal investigator); Norah A. Terrault, M.D. (coprincipal investigator); and Dulce MacLeod, R.N., and Vivian Tan, M.D. (study coordinators).
University of Michigan Medical Center, Ann Arbor, MI (DK62498): Robert M. Merion, M.D. (principal investigator), and Anna S. F. Lok, M.D., Akinlolu O. Ojo, M.D., Ph.D., Brenda W. Gillespie, Ph.D., Margaret Hill-Callahan, B.S., L.S.W., Terese Howell, B.S., Lan Tong, M.S., Tempie H. Shearon, M.S., Karen A. Wisniewski, M.P.H., Monique Lowe, B.S., and Abby Smith, B.A. (Data Coordinating Center staff).
University of North Carolina, Chapel Hill, NC (DK62505): Paul H. Hayashi, M.D., M.P.H. (principal investigator), and Tracy Russell, M.A. (study coordinator).
University of Virginia (DK62484): Carl L. Berg, M.D. (principal investigator); Timothy L. Pruett, M.D. (coprincipal investigator); and Jaye Davis, R.N. (study coordinator).
Medical College of Virginia Hospitals, Virginia Commonwealth University, Richmond, VA (DK62531): Robert A. Fisher, M.D., F.A.C.S. (principal investigator); Mitchell L. Shiffman, M.D. (coprincipal investigator); Andrea Lassiter, B.S. (transplant data analyst and study coordinator); and April Ashworth, R.N. (study coordinator).
Division of Digestive Diseases and Nutrition, National Institute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD: James E. Everhart, M.D., M.P.H., Leonard B. Seeff, M.D., Patricia R. Robuck, Ph.D., and Jay H. Hoofnagle, M.D.