Ryan P. Cauley, Khashayar Vakili, Jonathan A. Finkelstein, and Heung Bae Kim contributed to the study design, data collection, study analysis, and drafting of this article. Dionne A. Graham contributed to the study design, data collection, study analysis, and editing of this article. Nora Fullington and Kristina Potanos contributed to the study design, analysis, and drafting of this article.
This work was supported in part by the Health Resources and Services Administration (contract 231-00-0115), the Agency for Healthcare Research and Quality (grant 1T32HS019485-01), and the National Institute of Child Health and Human Development (grant 1K24HD060786). 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 US Government.
The authors of this article are not supported by any commercial associations and have no conflicts of interest to disclose.
Address reprint requests to Heung Bae Kim, M.D., Pediatric Transplant Center, Department of Sugery, Boston Children's Hospital, 300 Longwood Avenue, Fegan 3, Boston, MA 02115. Telephone: 617-355-8544; FAX: 617-730-0477; E-mail: email@example.com
With a severe shortage of size-matched whole organs, partial grafts from both living donors and deceased donors (DDs) have increasingly been seen as an alternative means of expanding the pediatric organ pool.[1-5] Although partial living donor grafts have been shown to have superior graft survival in comparison with both types of DD grafts,[1, 2] many pediatric patients do not have access to appropriate living donors. Although the use of living donors has increased since the introduction of the Pediatric End-Stage Liver Disease (PELD) scoring system in 2002, they still composed less than 11% of pediatric liver transplants in 2010.7 In contrast, since 2002, the use of partial DD grafts, including split grafts (in which the remainder of the organ is used) and reduced grafts (in which the remainder of the organ is discarded), has increased almost 8-fold to 25% to 32% of pediatric liver transplants.[4, 7, 8] Furthermore, split livers have the potential to become even more common; less than 10% of adult livers that met the criteria for splitting between 1996 and 2006 were actually made available for split liver transplantation.
Infants and young children awaiting liver transplantation have the highest wait-list mortality of all liver transplant candidates,[2, 3, 8, 9] and young children may be at the greatest risk of long-term morbidities and growth delays associated with an extended waiting time.[10-12] A significant increase in the number of partial DD livers could dramatically shorten the length of time that a child spends on the wait list and potentially decrease this high wait-list morbidity and mortality.
Although infants could potentially benefit the most from an expansion of the pediatric donor pool, they also may be at especially high risk when they accept partial DD grafts. In one study before the initiation of the PELD system, DD split liver recipients under the age of 2 years experienced a significantly greater risk of graft failure in comparison with recipients of whole grafts. Yet in older children, the risk of graft failure with split livers was comparable to the risk with whole livers. An analysis from a single academic center even suggested that all pediatric recipients may be subject to this increased risk of adjusted graft failure and mortality when they accept split grafts. Although other single-center studies have shown that the risk of partial graft failure in pediatric patients may be more comparable to the risk with whole grafts, there has been no recent national analysis of the specific risk of partial DD grafts in infants and young children. As partial grafts become more common for infants on the wait list, it will become increasingly important to understand the risk of accepting such grafts in this high-risk age group. The aims of this study were (1) to determine the current risks of graft failure and mortality in young recipients (<24 months) of partial and whole DD grafts, (2) to determine whether these risks have changed over time, and (3) to identify the effects of other transplant characteristics on the risks of both graft failure and mortality in this age group.
PATIENTS AND METHODS
All DD liver transplants reported in the United Network for Organ Sharing (UNOS) Standard Transplant Analysis and Research files were considered for analysis. Institutional review board approval was obtained from Boston Children's Hospital (IRB-P00002506).
All DD transplants from 1995 to 2010 in which the recipient was aged 0–24 months were analyzed. We excluded all retransplants and patients receiving organs from living donation and donation after death from circulatory arrest. All cases of multiorgan transplantation were also removed from the analysis. The follow-up period extended from January 1, 1995 to August 31, 2011. Patient survival and graft survival were truncated 1500 days after transplantation.
All variables in the standard UNOS database considered to be possible risk factors for graft failure in previous analyses were considered.[2, 16, 17] Transplants were categorized by the time of their occurrence: 1995-2000, 2001-2005, or 2006-2010. The time periods before and after the initiation of the PELD system in March 2002 were also examined. No variable included in the analysis had more than 20% missing data. Missing data were categorized as missing in order to maximize the power in the multivariate analysis. The transplant type was defined according to the UNOS database as partial (in which the remainder of the organ was discarded), split (in which the remainder of the organ was transplanted), or whole. An adjustment for the regional volume of first-time transplant recipients who were 0 to 24 months old was made in the final multivariate analysis. Primary diagnoses that accounted for less than 3% of all diagnoses were combined into an “other diagnosis” category and included parenteral nutrition–related liver disease and cholestasis. The primary outcome variable of interest was the time to graft loss as defined by retransplantation or death. Patients were followed until the date of graft failure or until they were lost to follow-up. Mortality was analyzed as a secondary outcome measure.
The association of the transplant type with all other variables was determined with a bivariate analysis. Fisher's exact test was used to compare categorical variables. Normally distributed data were compared with an analysis of variance, whereas skewed continuous variables [eg, the wait-list time, Model for End-Stage Liver Disease (MELD) score, and cold ischemia time (CIT)] were compared with rank-sum tests.
A Kaplan-Meier survival analysis (log-rank test) was used to examine the unadjusted association of risk factors with graft failure. Because reduced and split grafts were noted to have similar crude outcomes, all partial DD grafts were combined to increase the power of the multivariate analysis. Variables with P values < 0.1 in the bivariate analysis were considered for the multivariate analysis. Backwards selection was used to build the final multivariate models (P < 0.1). Covariates that were investigated but excluded under backwards selection included the following: laboratory and final PELD scores, history of exceptions, donor diabetes, coronary artery disease, previous malignancy, hepatitis status, and number of vasopressors. Height and body mass index were excluded because they were collinear with weight, and recipient creatinine was collinear with the use of dialysis. Donor cause of death, ABO matching, final status 1 designation, and regional transplant volume were forced into the final model, although they were not independently associated with the outcome. Cox proportional hazards modeling was used to estimate the adjusted association of all variables with graft failure as a primary outcome. A second Cox proportional hazards model was used to estimate the adjusted associations of all variables with patient mortality. Interaction tests were performed to determine the effects of all significant independent risk factors on the adjusted risk of partial liver grafts. Using subset analyses, we then examined the possibility of effect modification of significant risk factors with the transplant type. Ten-fold cross-validation was used to evaluate the stability of all estimates in the multivariate models.[18, 19] A P value ≤ 0.05 was considered to be significant. Analyses were performed with JMP Pro 10.0.0 (SAS Institute, Inc., Cary, NC) and SAS 9.3 (SAS Institute).
DDs of partial grafts were significantly older than DDs of whole livers (P < 0.001; Table 1). As expected, the donor weight was also significantly lower for whole grafts versus partial grafts (P < 0.001). Whole organs were also more likely to be shared nationally (P < 0.001). In comparison with whole grafts, a larger proportion of partial grafts were given to recipients less than 12 months old, and the weight of these partial graft recipients was also lower. Recipients of partial grafts were sicker than recipients of whole grafts according to the following: (1) a greater proportion of partial graft recipients required a ventilator before transplantation, (2) a greater proportion were designated as status 1, and (3) they had higher laboratory PELD scores. The wait-list time in the PELD era also varied significantly with the graft type: young children who received a whole graft tended to wait much longer than those given partial grafts, and this trend was consistent when all recipients (P = .004) and just the subgroup designated as status 1 (P < 0.001) were examined. Only among recipients with a PELD score greater than 30 was the graft type not significantly associated with differences in the wait-list time (P = 0.29). There were 698 graft failures and 345 deaths across the study cohort; 92% of these occurred within 1500 days of transplantation.
Table 1. Donor and Recipient Characteristics by the Types of Cadaveric Liver Grafts in Infant Recipients (n = 2683)
Partial DD Grafts (n = 1118)
Whole DD Grafts (n = 1565)
NOTE: Infant recipients were less than 24 months old.
The data are presented as medians and interquartile ranges.
Although there was not a significant increase in the proportion of partial DD grafts over time (P = 0.15; Table 2), an increasing proportion of partial liver transplants were split liver grafts (P < 0.001). The proportion of very small recipients (<6 kg), the median CIT, the organ sharing status, the recipient medical condition, and the proportion of status 1 designees all varied over time. Although the overall wait-list time decreased in more recent time periods (P < 0.001), there was no significant improvement in the median wait-list time for status 1 designees (P = 0.24).
Table 2. Donor and Recipient Characteristics for Liver Transplants in Recipients Between the Ages of 0 and 24 Months by the Time Period (n = 2683)
1995-2000 (n = 936)
2001-2005 (n = 827)
2006-2010 (n = 920)
The data are presented as medians and interquartile ranges.
In the bivariate analysis, there was a significant improvement in graft survival for both whole and partial liver transplantation in the PELD era versus the pre-PELD period (P = 0.01 and P < 0.001, respectively; Fig. 1A). Additionally, although there were significant differences in graft survival between partial and whole grafts in the pre-PELD era (P = 0.002), there was no significant difference in graft survival by the graft type in the PELD era (P = 0.68). Reduced and split grafts had comparable survival in both the pre-PELD and PELD eras (P = 0.76 and P = 0.39, respectively). Although there were significant differences in crude graft survival by the graft type in 1995-2000 (P < 0.001; Fig. 1B), graft survival rates with partial and whole grafts were comparable in 2001-2005 (P = 0.43) and 2006-2010 (P = 0.36). Only livers from donors more than 40 years old had significantly worse graft survival (P < 0.001). In the PELD era, donors weighing less than 10 kg led to significantly inferior graft survival in comparison with donors in higher weight categories (P < 0.001), and recipients less than 6 kg also had diminished survival (P = 0.01). Patients who were in the intensive care unit (ICU) on a ventilator had a significantly higher risk of failure (P < 0.001). Graft survival also differed significantly with final status score, with status 1 recipients having a significantly increased risk of graft failure (P = 0.003).
The multivariate adjusted associations of select donor and recipient characteristics are shown in Table 3. Partial liver transplants were associated with a case-mix adjusted hazard ratio (HR) for graft failure of 1.14 [confidence interval (CI) = 0.89-1.44, P = 0.30] in comparison with whole livers for the entire study cohort. Donor age was an independent predictor of graft failure (P = .007). Similarly to the bivariate analysis, a low donor weight (≤10 kg), a low recipient weight (≤6 kg), and a longer CIT (>9 hours) also were associated with an increased risk of failure. The earliest time period in the study (1995-2000) appeared to be associated with a 63% increased risk of graft failure in comparison with the most recent time period (2006-2010), even after we controlled for other factors (P < 0.001). The age of recipients was not independently associated with outcomes. The pretransplant medical status, as measured by both ventilator status in the ICU and dialysis, also independently affected graft outcomes.
Table 3. Adjusted Associations of Select Donor and Recipient Variables With Graft and Patient Survival Between 1995 and 2010 (n = 2545)
Graft Failure: HR (95% CI)/P Value
Mortality: HR (95% CI)/P Value
NOTE: The models were also adjusted for the donor cause of death, ABO matching, status 1 designation for recipients, and regional transplant volume; however, these were not independently associated with outcomes.
Includes Parenteral Nutrition Related Liver Disease.
Overall, there was a trend toward an increased risk of patient mortality for recipients of partial DD grafts versus recipients of whole DD grafts (HR = 1.35, P = 0.08; Table 3). As with graft failure, variables such as older age, low weight (≤10 kg), and longer CIT (>9 hours) were significantly associated with greater adjusted risks of mortality. The earliest time period of 1995-2000 was associated with a 79% increased case-mix adjusted hazard of mortality in comparison with the most recent time period of 2006-2010 (P < 0.001). Recipient factors such as low weight (≤6 kg), primary diagnosis of other, ventilator status, need for dialysis, and previous transplantation were all independently associated with increased risks of mortality. Although local sharing was associated with a slightly increased risk of patient mortality overall (P = 0.04), local shares had an adjusted risk of mortality comparable to that of national shares in the most recent time period (HR = 1.33, 95% CI = 0.65-2.7). Furthermore, the risks of graft failure were comparable across all share types in all time periods. The final status score was not independently associated with either graft failure or mortality. We did not find an independent association between the regional volume of transplants (for young recipients who were 0-24 months old) and either type of outcome. Except for the time period, we did not identify any significant effect modification between the use of partial grafts and any other risk factors considered in the model, including the final status score, age of recipient, medical condition, diagnosis, and CIT.
Effect of the Time Period
Although the adjusted hazards of partial graft failure and mortality were 1.40 (95% CI = 1.05-1.89) and 1.41 (95% CI = 0.95-2.09), respectively, in 1995-2000, the adjusted risks of graft failure and mortality were comparable for partial and whole organs in 2006-2010 (HR for graft failure = 0.81, 95% CI = 0.56-1.18; HR for mortality = 1.02, 95% CI = 0.66-1.71; Table 4). An interaction analysis demonstrated that there was a significant change in the risk of partial graft failure over time, even after a case-mix adjustment (P = 0.009). We did detect a trend toward an increased risk of patient mortality for partial grafts in the 2001-2005 time period, but the CIs were wide (HR = 1.70, 95% CI = 0.99-2.92). Furthermore, this was not found to be significantly different from the current time period in interaction analyses.
Table 4. Effects of the Time Period on the Hazard of Partial Grafts: Interaction Analyses
Graft Failure: HR (95% CI)
Interaction P Value
Mortality: HR (95% CI)
Interaction P Value
Transplant era: 1995-2000
Transplant era: 2001-2005
Transplant era: 2006-2010
Ten-fold cross-validation of both multivariate models indicated that these estimates were highly stable across our study samples (Table 5).
Table 5. Ten-Fold Cross-Validation of the Adjusted HRs for Partial and Whole Liver Transplantation in Young Children: 2006-2010.
Whole DD livers—especially those that are size-matched for young recipients—continue to be a severely limited resource. Infants and young children have been shown to have the highest morbidity and mortality rates on the wait list for liver transplantation.[3, 8-12] Although partial DD liver grafts could be used to expand the pediatric donor pool, past analyses have suggested an increased risk of graft failure when these organs are transplanted into high-risk recipients, including adults with a high risk of wait-list mortality based on the MELD scoring system[2, 14, 20] and infants and young children.[1, 14]
In the present study, we noted that although there was an increased risk of graft failure with partial grafts in infants in 1995-2000 (as suggested by previous analyses), this risk became negligible in the most recent time period of 2006-2010. In both crude and adjusted analyses, we found that partial DD grafts became significantly safer over time. Both split and reduced DD grafts now have outcomes comparable to those of whole grafts in terms of both graft and patient survival. Although several other risk factors were independently predictive of graft failure and patient mortality, none of these variables were found to be associated with disproportionate risks in partial grafts versus whole grafts, and this signifies that all infants, regardless of their status score, medical condition, diagnosis, age, or weight, could accept a partial DD liver graft (instead of a whole DD liver) without increasing their relative risk of graft failure or death.
We found that although the median wait-list time appears to have decreased in the most recent time period, there have been no significant improvements in the wait-list time for the highest risk status 1 designees or those with elevated laboratory PELD scores. With the demand for whole organs presumably outpacing the supply for this young recipient cohort, we also noted that recipients of whole livers tended to wait significantly longer before transplantation than recipients of partial livers, even though more whole organs are shared nationally. It is possible that differences in wait-list times could suggest an underlying preference for whole grafts versus partial grafts for risk-equivalent patients. This preference would make sense because of the increased relative risk of partial DD grafts noted in analyses of the pre-PELD era. However, because we found that partial DD grafts had outcomes comparable to those of whole grafts in the most recent era (2006-2010), we believe that these 2 types of grafts should now be preferred equally for all young pediatric candidates. If the grafts are considered comparable, it is likely that there will be fewer differences in the wait-list times by the graft type in the future.
Recipients of partial grafts tended to have higher median laboratory PELD scores than those receiving whole grafts, and in comparison with whole liver recipients, a greater proportion of partial liver recipients were designated as status 1. Indeed, reduced and split liver recipients were also more likely to be in the ICU or on a ventilator before transplantation in comparison with whole liver recipients. The longer wait-list time for whole livers, even for status 1 recipients, implies that for sick pediatric patients, partial livers may be a more expedient means of obtaining an organ. Because size-matched whole livers are relatively scarce in comparison with the large number of adult livers, sicker candidates may not have the time to wait for a whole size-matched organ to become available. Fortunately, even though partial livers tended to be transplanted into sicker pediatric recipients, we found that partial grafts had unadjusted and adjusted graft survival comparable to that of whole grafts and a similar adjusted risk of mortality. Although previous studies have suggested that adults with higher final status scores (status 1) are associated with a higher risk of graft failure when they accept a partial or split graft,[2, 14, 20] we found no evidence that status 1 designation, ventilator status, or dialysis use was associated with a disproportionate risk of graft failure or mortality in young pediatric recipients of partial liver grafts versus whole liver grafts.
Although recipients who were designated as status 1 appeared to be associated with worse graft survival in the unadjusted analysis, the status score was not predictive of outcomes in the full multivariate models. Conversely, the medical status, including the need for a ventilator or dialysis, was highly associated with both graft failure and patient mortality in the adjusted models. Previous studies have also suggested that the need for life support may increase the risk of graft failure.[22, 23] Because status 1 designees are more likely to require life support, it may be that the status score is simply confounded by the medical status. Although recipient age was not shown to affect graft or patient survival, we did note that a low donor or recipient weight was associated with diminished outcomes in both the unadjusted and adjusted analyses. Previous studies have also suggested that a low donor or recipient weight may be associated with an increased risk of graft failure and mortality,[2, 23] possibly because of the increased rate of vascular thrombosis in the small-caliber vessels of these grafts. Because whole grafts are rare for this small cohort, it is notable that we did not find a difference in risk when these patients were given partial grafts instead of size-matched whole grafts.
We found that both partial and whole pediatric liver transplants became safer in the most recent PELD time period. Previous studies have also shown that the outcomes of pediatric liver transplantation have improved since the initiation of PELD, possibly because of improvements in recipient and donor selection, organ allocation, surgical techniques, postoperative care, and immunosuppression.[3, 4] However, we noted that the relative outcomes of partial liver transplants also greatly improved over time. Although there was a significant difference in graft survival between partial and whole liver transplants in the earliest time periods of our study, the outcomes now appear comparable. Interaction analyses suggest that the risk of graft failure in partial grafts has become significantly safer over time, even after adjustments for other significant donor and recipient factors. We found that outcomes appear to have gradually improved between 1995 and 2010. Although the initiation of the PELD system may be a part of the explanation for improved outcomes, the gradual nature of this progress across 3 time periods suggests that it is most likely due to numerous improvements in pediatric transplantation over time.
Because of the large sample size, the UNOS database can provide power for more extensive multivariate analyses in comparison with many single-center studies. Nevertheless, the database is retrospective and is limited to the risk factor and outcome variables collected for the allocation process. Although the findings of the present study suggest that partial and whole grafts have comparable primary outcomes such as graft and patient survival, we could not accurately compare graft types by the relative proportion of specific complications or the causes of death because of the limitations of missing data. Additionally, because there were fewer deaths in the study in comparison with graft failures, we had less power to detect differences in the risk of mortality. Further prospective evaluations will be required to validate our findings.
Infants continue to have the highest mortality rate on the liver wait list. Although this mortality rate has improved over time, it remains almost twice that of adult recipients. Because size-matched whole livers remain a scarce resource, maximizing the use of partial livers from both deceased and living donors will be the key to decreasing this high mortality rate. However, we note that there has been no increase in the proportion of partial DD grafts over time. Although we firmly believe that living donors should continue to be pursued, split grafts from adult donors should also be used to optimally expand the pediatric donor pool. Despite an increase in the number of adult livers that are split since the introduction of PELD/MELD, the current liver allocation system is not designed to optimize the use of this valuable resource; less than 10% of livers from donors that met the criteria for splitting between 1996 and 2006 were actually split. Studies have suggested that if even half of appropriate donor livers are split, the pediatric wait list could be eliminated. We believe that split liver transplants have outcomes comparable to those of transplants of size-matched whole grafts in all young pediatric recipients, and this makes them a safe alternative to size-matched whole organs. Through the expanded use of partial liver grafts in young recipients, it may be possible to reduce the unacceptably high morbidity and mortality rates of high-risk pediatric transplant candidates.
The authors thank Naomi R. Shatz for her support and editorial assistance.