Liver Retransplantation in Children: A SPLIT Database Analysis of Outcome and Predictive Factors for Survival

Authors

  • V. Ng,

    Corresponding author
    1. Hospital for Sick Children, Division of Gastroenterology, Hepatology, and Nutrition, 555 University Avenue Room 8262, Toronto, Ontario M5G1X8, Canada
    Search for more papers by this author
  • R. Anand,

    1. Hospital for Sick Children, Division of Gastroenterology, Hepatology, and Nutrition, 555 University Avenue Room 8262, Toronto, Ontario M5G1X8, Canada
    Search for more papers by this author
  • K. Martz,

    1. Hospital for Sick Children, Division of Gastroenterology, Hepatology, and Nutrition, 555 University Avenue Room 8262, Toronto, Ontario M5G1X8, Canada
    Search for more papers by this author
  • A. Fecteau

    1. Hospital for Sick Children, Division of Gastroenterology, Hepatology, and Nutrition, 555 University Avenue Room 8262, Toronto, Ontario M5G1X8, Canada
    Search for more papers by this author

Corresponding author: Vicky Ng, vicky.ng@sickkids.ca

Abstract

To examine outcomes and identify prognostic factors affecting survival after pediatric liver transplantation, data from 246 children who underwent a second liver transplantation (rLT) between 1996 and 2004 were analyzed from the SPLIT registry, a multi-center database currently comprised of 45 North American pediatric liver transplant programs. The main causes for loss of primary graft necessitating rLT were primary nonfunction, vascular complications, chronic rejection and biliary complications. Three-month, 1- and 2-year patient survival rates were inferior after rLT (74%, 67% and 65%) compared with primary LT (92%, 88% and 85%, respectively). Multivariate analysis of pretransplant variables revealed donor age less than 1 year, use of a technical variant allograft and INR at time of rLT as independent predictive factors for survival after rLT. Survival of patients who underwent early rLT (ErLT, <30 days after LT) was poorer than those who received rLT >30 days after LT (late rLT, LrLT): 3-month, 1- and 2-year patient survival rates 66%, 59%, and 56% versus 80%, 74% and 61%, respectively, log-rank p = 0.0141. Liver retransplantation in children is associated with decreased survival compared with primary LT, particularly, in the clinical settings of those patients requiring ErLT.

Introduction

Primary allograft failure after pediatric liver transplantation (LT) results from entities such as primary nonfunction, vascular complications, biliary complications with ensuing fibrosis/cirrhosis and chronic rejection (1). Except in the settings of extrahepatic sources of sepsis, disseminated malignancies and severe, irreversible neurologic injuries which are absolute contraindications to liver retransplantation (rLT), the approach of most pediatric LT programs is to offer the option of rLT to children with documented allograft failure. In an era when demand for livers far exceeds available supply, the judicious use of limited organs requires continual reevaluation of both indications for and outcomes after LT. To date, multiple single center experiences report 1-year patient survival rates of up to 77% after pediatric rLT (2–8), similar to results of rLT reported in adults (7,9–16), but inferior to the survival of children after primary LT. As single center experiences on the outcomes of pediatric rLT are subject to a number of limitations, we herein present the outcomes and identify prognostic factors affecting survival of children from the SPLIT database who have undergone rLT.

Patients and Methods

Prospectively collected patient data for all children who received a rLT since the 1995 inception of SPLIT were reviewed and analyzed. As of June 1, 2004, the time of data closure for this report, the SPLIT database was comprised of 39 pediatric LT centers in Canada and the United States who had registered 1611 children undergoing a first liver only transplant, with 181 (11%) of these proceeding on to receive rLT. An additional three patients received an rLT after a first combined liver-kidney transplant, and a further 62 patients were registered in the SPLIT registry for the first time at the time of rLT, for a total study population of 246 patients evaluated for outcomes after pediatric rLT. Children were excluded if they received any additional allografts at the time of rLT. Participation in the SPLIT registry was approved by each center's institutional review board or research ethics board and all parents or legal guardians provided written informed consent; subject assent was also obtained when appropriate.

Statistical Methods

Univariate statistical analyses included Pearson's chi-square test for comparisons between categorical variables (presented as the number with percentages), Wilcoxon test for comparisons of continuous variables (presented as the median with range) and log-rank test for comparison of time to event outcomes. Univariate analyses using the Cox proportional hazards model were performed to identify potential factors predictive of patient survival after rLT. Both donor and recipient factors were considered for inclusion in the model. Investigated factors concerning recipients were gender, age, race, weight Z-score, and blood type, primary diagnosis necessitating LT (biliary atresia vs. other diagnoses), patient status at time of rLT (not hospitalized vs. hospitalized or in intensive care unit) and etiology of allograft failure. Factors concerning donors were gender, age, donor/recipient (D/R) gender match and blood types. Factors concerning the transplant surgical procedures were cold ischemia time, duration of the anhepatic phase, total operation time, blood loss index and the type of graft (full size grafts vs. technical variant liver grafts). Other factors at time of rLT included presence of ascites, dialysis, aspartate transferase, international normalized ratio and total bilirubin.

For the development of the multivariate Cox regression model to predict post-rLT survival, factors achieving a significance level with p-value ≤0.20 in the univariate analyses were initially included. Model reduction was performed using the backward elimination method, and factors remaining significant at p-value ≤0.05 were maintained in the final model. All statistical analyses were performed using the SAS System for Windows, version 8.02 (SAS Institute, Cary, NC).

Actuarial survival was calculated using the Kaplan–Meier statistical method. Differences in survival were calculated using the log-rank test. Kaplan–Meier curves were generated to examine the impact of timing of rLT on graft and patient survival. Overall patient survival of the entire study population was assessed and rLT was categorized as early or late, depending on the time interval between primary transplantation and rLT. Early retransplantation (ErLT) was defined as rLT within the first 30 days after first LT, whereas late retransplantation (LrLT) was defined as rLT occurring more than 30 days after the initial LT. Kaplan–Meier curves were also generated to examine the impact of recipient age at rLT, (technical variant) donor organ type and donor age on patient survival.

Results

Patients

Patient characteristics of the study cohort are presented in Table 1. The median patient age at the time of rLT was 3.6 (range 0.06–18.1) years, of which 43% of patients were male. The median patient weight at the time of rLT was 15.3 (range 3.5–85) kg. The median time between initial LT and rLT was 80 (range 1–5685) days. The median time from listing for rLT to date of rLT was 10 days. The median PELD score on day of rLT was 16.9, with requirements of inpatient care, intensive care unit support and ventilatory support required by 77%, 61% and 41% of this study cohort respectively.

Table 1.  Characteristics of retransplantation patient cohort (N = 246)
  1. 1As 65 children joined the SPLIT registry at time of their rLT, first liver transplant data is available and reported on only 181 patients.

  2. 2Biliary complications including bile leakage, biliary strictures and ischemic type biliary lesions (nonanastomotic strictures in the intra- and/or extrahepatic bile ducts in the presence of a patent hepatic artery).

Patient ethnicity (%)
 White140 (56.9%)
 Black 46 (18.7%)
 Hispanic 36 (14.6%)
 Asian/Pacific Islander 7 (2.9%)
 American Indian 4 (1.6%)
 Other12 (4.9%)
 Missing 1 (0.4%)
Primary disease n (%)
 Biliary atresia110 (44.7%)
 Acute liver failure 36 (14.6%)
 Metabolic liver disease18 (7.3%)
 Tumor 7 (2.9%)
 Other cholestatic conditions 30 (12.2%)
 Crytogenic cirrhosis20 (8.1%)
 Other 25 (10.2%)
Allograft type used at primary LT n (%)
Deceased donor allograft
   • Whole 74 (30.1%)
   • Reduced 45 (18.3%)
   • Split 26 (10.6%)
   • Missing 3 (1.2%)
 Live donor allograft 33 (13.4%)
 Data unavailable1 65 (26.4%)
Indication for rLT n (%)
 Vascular complications 86 (34.9%)
 Primary graft dysfunction 47 (19.1%)
 Chronic rejection 36 (14.6%)
 Biliary complications217 (6.9%)
 Hyperacute rejection 2 (0.8%)
 Acute rejection 6 (2.4%)
 Recurrent liver disease 5 (2.0%)
 Other 33 (13.4%)
 Missing14 (5.7%)
Allograft type used at rLT n(%)
Deceased donor allograft
   • Whole120 (48.8%)
   • Reduced 78 (31.7%)
   • Split22 (8.9%)
   • Missing 8 (3.3%)
 Live donor allograft17 (6.9%)
 Data unavailable 1 (0.4%)
Time between 1st LT to rLT
 Median80.5 days
 Range1–5686 days
 
 Follow-up time (median, range) after rLT7.4 (0–85.3) months

Indications for retransplantation

The most common indication for pediatric rLT was hepatic artery thrombosis (n = 72, 29%). Primary graft dysfunction and chronic rejection were the second and third most common indications for rLT, seen in 47 (19%) and 36 (14.6%) of our study cohort.

Indications for rLT differed between children undergoing ErLT and those requiring LrLT (Table 2). The ErLT and LrLT patients had similar acute rejection rates at 6 months (49.2% and 50.5%, respectively) and at 12 months (51.8% and 50.5%, respectively, log-rank p = 0.8868, data not shown). Of the three patients whose first SPLIT registered transplant was a multi-organ (combined liver-kidney) transplant, the need for liver only retransplantation occurred early in one child (for primary graft dysfunction) and late for two (for vascular complication and for chronic rejection). Recurrent liver disease as the primary indication for rLT was extremely rare, occurring in only five (2%) of our study cohort, as caused by primary sclerosing cholangitis (n = 3), autoimmune hepatitis (n = 1) and cirrhosis of unknown etiology (n = 1).

Table 2.  Indications for rLT differ between children who require early rLT and late rLT
Causes of graft failure n (%)Early rLT (<30 days)Late rLT (>30 days)
  1. 1Biliary complications including bile leakage, biliary strictures and ischemic type biliary lesions (nonanastomotic strictures in the intra- and/or extrahepatic bile ducts in the presence of a patent hepatic artery).

Vascular complications50 (46.7%)  36 (25.9%)
Primary graft dysfunction47 (43.9%)0 (0%)
Chronic rejection0 (0%)    36 (25.9%)
Biliary complications11 (0.9%)  16 (11.5%)
Hyperacute rejection2 (1.9%)0 (0%)
Acute rejection2 (1.9%)  4 (2.9%)
Recurrent liver disease0 (0%)    5 (3.6%)
Other2 (1.9%)  31 (22.3%)
Missing3 (2.8%) 11 (7.9%)
TOTAL107 (100.0%)  139 (100.0%)

Graft and patient survival

The 1- and 4-year patient survival rates after rLT in our study cohort were 67.4% and 58.8%, inferior to rates of 87.8% and 83.1% for 1611 children after first LT. Graft loss was defined as death or rLT. The 1- and 4-year allograft survival rates after rLT were similarly poorer at 58.6% and 49.5%, respectively, compared with 81.7% and 73.9% reported after primary LT for children in the SPLIT registry (Figure 1). Examination of the survival curves revealed that nearly all of the difference in survival outcomes between initial LT and rLT occurred in the first 6 months after transplant surgery. After the first 6 months, survival curves for rLT patients were parallel to those patients who have only ever received one LT (Figure 1). Patient and allograft survival were then evaluated by the timing of rLT. Early rLT occurred in 107 (43%) children at a median time of 6 (range, 1–29) days after first LT. Late rLT occurred in 139 children at median 618 (range, 31–5686) days after initial LT. The 1- and 4-year patient survival rates for ErLT were 59% and 56%, significantly lower than the 74% and 61% reported after LrLT (Figure 2A, log-rank p = 0.0141). The poorer outcomes for the ErLT group were accompanied by lower 1- and 4-year allograft survival rates of 52% and 46% after ErLT compared with 64% and 52% after LrLT (Figure 2B, log-rank p = 0.0444). There was no statistically significant difference in patient survival rates for rLT performed between 1995 and 2000 (n = 142) compared with rLT performed between 2001 and 2004 (n = 104). Further comparisons of era effect within ErLT (1 year Kaplan–Meier survival rates of 56.6% for patients with rLT performed between 1995 and 2000 vs. 65.1% between 2001 and 2004) and LrLT (1 year Kaplan–Meier survival rates of 69.7% for rLT performed between 1995 and 2000 vs. 78.2% between 2001 and 2004) groups were also not statistically different. Patient survival was decreased with both younger patient age at time of rLT (Figure 3A), and younger donor age at time of rLT (Figure 3B), particularly, in the age group less than 12 months. Graft type used at rLT is provided in Table 1. Comparing patient survival after rLT, reduced-size allografts were associated with decreased patient and allograft survivals after rLT compared with all other technical variants (Figure 3C).

Figure 1.

Patient and Graft Survival after rLT is lower than after primary liver transplantation.

Figure 2.

Figure 2.

(A) Patient survival after ErLT is poorer than after later retransplantation. (B) Graft survival after ErLT is poorer than after LrLT in children.

Figure 2.

Figure 2.

(A) Patient survival after ErLT is poorer than after later retransplantation. (B) Graft survival after ErLT is poorer than after LrLT in children.

Figure 3.

Figure 3.

(A) Patient survival after rLT is decreased with younger recipient age. Patient age <12 months (n = 50), age 1–4 years (n = 86), 5–12 years (n = 63) and age >13 years (n = 47). (B) Effect of donor age on patient survival after rLT. (C) Use of technical variant graft is associated with decreased patient survival after rLT compared with use of whole organ allografts.

Figure 3.

Figure 3.

(A) Patient survival after rLT is decreased with younger recipient age. Patient age <12 months (n = 50), age 1–4 years (n = 86), 5–12 years (n = 63) and age >13 years (n = 47). (B) Effect of donor age on patient survival after rLT. (C) Use of technical variant graft is associated with decreased patient survival after rLT compared with use of whole organ allografts.

Figure 3.

Figure 3.

(A) Patient survival after rLT is decreased with younger recipient age. Patient age <12 months (n = 50), age 1–4 years (n = 86), 5–12 years (n = 63) and age >13 years (n = 47). (B) Effect of donor age on patient survival after rLT. (C) Use of technical variant graft is associated with decreased patient survival after rLT compared with use of whole organ allografts.

Recipient and donor variables affecting rLT outcomes

Table 3 provides details on recipient and donor characteristics at the time of rLT. Patient age at rLT was significantly younger at the time of ErLT compared with at LrLT (18.0 months vs. 74.3 months, p < 0.0001). ErLT recipients also had a higher median PELD score (21.8 vs. 12.9, p < 0.0001) and were more likely to be receiving care in the intensive care unit setting at time of organ availability (95% vs. 35%, p < 0.0001). More children undergoing ErLT (80%) required ventilatory support in an intensive care unit prior to rLT than LrLT children (42%, p < 0.0001). Technical variant liver grafts and older donor aged grafts were used more often in ErLT than in LrLT (Table 3). The median age at rLT was 9.9 years for children receiving a whole graft compared to age 1.9 years for children receiving a technical variant graft (p < 0.0001). However, the relationship between donor age and recipient age at rLT is not significantly strong (correlation coefficient of 0.13). Intraoperative variables including median cold ischemia time, median length of rLT operative time and median amount of blood transfusion support were decreased in ErLT compared with LrLT recipients (Table 3). A longer post-rLT hospitalization period was required by ErLT patients compared with LrLT recipients (18 days vs. 8.5 days, p = 0.0001).

Table 3.  Clinical, recipient and donor factors by time of rLT
 All rLT (N = 246)Early rLT (N = 107)Late rLT (N = 139)p-Value
Recipient variables
 Median patient age at rLT (months)43.418.074.3<0.0001
 Median PELD score at rLT16.921.812.9<0.0001
 Patients in PICU at time of rLT n (%)150 (61%)102 (95%)48 (35%)<0.0001
Clinical status
 Median weight Z-score0.9−1.1−0.80.3   
 Ascites present n (%)121 (50%)43 (41%)78 (57%)0.01  
 Requiring dialysis pre-rLT n (%)29 (19%)18 (18%)11 (23%)0.4   
 Median time from primary LT to rLT (days)806618<0.0001
Lab values
 Blood group A: n (%)93 (38%)47 (44%)46 (34%)0.08  
 Median AST (IU/L)2371149153<0.0001
 Median ALT (IU/L)20076795<0.0001
 Median INR1.52.11.4<0.0001
 Median total serum bilirubin (mg/dL)8.88.29.80.3   
Intraoperative variables
 Median length of rLT operative time (h)6.14.87.3<0.0001
 Median warm ischemia time (min)50.053.549.00.3   
 Median cold ischemia time (h)7.66.08.2<0.0001
 Median amount of PRBC transfused (cc)10007001400<0.0001
 Median days in PICU in first 30 days post-rLT11188.5   0.0001
Donor variables
 Whole organ graft n (%)120 (51%)28 (28%)92 (68%)   0.0001
 Median donor age (years)162210<0.0001

Table 4 provides estimates of hazard ratio (HR) and 95% confidence intervals (CI) for each individual risk factor for death post-rLT identified by univariate analysis. Patients receiving an allograft from a donor under the age of 12 months had the highest post-rLT mortality (HR = 5.49; 95% CI: 2.51–11.98; p < 0.0001). Patients were also at higher risk for post-rLT mortality if ICU care (HR = 2.29; 95% CI: 1.16–4.49; p = 0.017) or hospitalization care (HR = 2.92; 95% CI: 1.33–6.37; p = 0.007) was required while awaiting organ availability for rLT. With increasing INR, an approximate risk in mortality is increased (HR = 2.04, 95% CI: 1.31–3.15; p = 0.0015). To control for other confounders and to better understand the impact of rLT on patient survival, we performed a multivariate Cox proportional hazards regression. The factors that were initially included in the multivariate model were those with p < 0.2 identified in univariate analyses. The three most important predictors for post-rLT mortality were donor age less than 1 year, use of a technical variant allograft and increased INR level at time of rLT. Table 5 provides the hazard ratio estimates for each of these factors in the final multivariate model. Patients receiving a graft from a donor under the age of 12 months are at 21 times higher risk of death after rLT compared with patients who received a second allograft from a donor greater than 1 year of age (HR = 21.06; 95% CI:7.39–60.01; p < 0.0001). Although there were only 10 children in this category, 7 had died before 2 months.

Table 4.  Univariate analysis of factors included in the multi-variate model to predict patient survival after rLT
FactorComparison groupReference groupHazard ratio95% CIp-Value
  1. 1Chol = Cholestatic liver disease.

Donor organ typeWhole organTechnical variant0.460.29–0.740.001
Donor age<1 year 5.49 2.51–11.98<0.0001
(overall p = 0.0002)(18, 50] years(1, 18) years1.300.81–2.080.3
 50+ years 0.680.21–2.220.5
Recipient age(1–5) years 0.500.29–0.850.01
(overall p < 0.003)(5–13) years<1 year0.450.25–0.820.008
 13+ years 0.270.12–0.590.001
Time to rLTLateEarly0.580.37–0.900.01
Patient statusICUNot hospitalized2.291.16–4.490.01
(overall p < 0.03)Hospital/no ICU 2.921.33–6.370.007
Cause of graft failurePrimary graft dysfunctionOther cause1.761.06–2.920.03
Primary diseaseChol1/metabolicBiliary atresia1.050.59–1.850.8
(overall p < 0.05)Fulminant 1.360.75–2.470.3
 Cirrhosis 0.250.06–1.030.05
 Other 0.430.18–1.010.05
Blood transfusions<1000 cc≥1000 cc1.480.93–2.370.1
Era in which rLT performed2001–20041995–20000.730.45–1.180.2
Natural logarithm of INRContinuous2.041.31–3.150.001
Warm ischemia timeContinuous1.011.00–1.020.003
ASTContinuous1.001.00–1.000.1
Weight Z-scoreContinuous0.920.82–1.030.1
Table 5.  Final multivariate patient survival model for children undergoing retransplantation
FactorComparison groupReference groupHazard ratio95% CIp-Value
Donor age (overall p < 0.0001)<1 year 21.06 7.39–60.01<0.0001
 (18, 50) years(1, 18) years1.400.67–2.960.3733
 50+ years 0.250.03–1.960.1875
Donor organ typeWhole organTechnical variant0.410.20–0.850.0155
Natural logarithm of INRContinuous1.831.08–3.120.0248

Complications following rLT

One hundred forty-eight (60%) children returned to the operating room in the first 30 days after rLT, for indications including intra-abdominal bleeding (26%), bowel perforation (11%), vascular complications (13%) and biliary tract complications (10%). Forty (16%) patients developed biliary complications within the first 30 days post-rLT, with biliary leaks developing in 31 (13%) children. Hepatic artery thrombosis and portal vein thrombosis developed in 21 (9%) patients and 13 (5%) children after rLT respectively. Ten children developing early hepatic artery thrombosis subsequently underwent third LT. Thirty-eight (15%) children required kidney dialysis within the first 30 days of rLT. Early postoperative complications after rLT including cerebral edema (ErLT, n = 6) and dialysis requirements (ErLT, n = 25) were more commonly seen in ErLT group than in LrLT group (cerebral edema, n = 0; dialysis requirements, n = 13). Table 6 provides details on the causes of death after rLT and after a single LT during this study period. Multi-organ failure was more common in children after rLT (24.6%) than in those children with only a single LT (9.3%), p = 0.007. To determine the influence of rLT on death, a Cox regression model with rLT as a time-varying covariate was utilized. This model uses the 1611 patients with a first LT registered within the SPLIT registry, of which 181 progressed to rLT. The results of an analysis confirmed that the risk of death was significantly increased after rLT compared to patients without an rLT (hazard ratio 7.88, p-value <0.0001, 95% CI = 5.83–10).

Table 6.  Cause of death after primary LT and after rLT
Cause of deathSingle LT 1430 patients (150 deaths)Total rLT 181 patients (65 deaths)ErLT 104 patients (41 deaths)LrLT 77 patients (24 deaths)
Infection 41 (27.3%)18 (27.7%) 9 (22.0%) 9 (37.5%)
Organ failure14 (9.3%)16 (24.6%)11 (26.8%) 5 (20.8%)
Malignancy/tumor recurrence12 (8%) 2 (3.1%)2 (4.9%)0 (0.0%)
Cardiopulmonary complications   20 (13.3%) 9 (13.8%) 7 (17.1%)2 (8.3%)
Vascular complications6 (4%)2 (3.1%)0 (0.0%)2 (8.3%)
Central nervous system complications   20 (13.3%)6 (9.2%) 5 (12.2%)1 (4.2%)
Gastrointestinal complications   4 (2.7%)4 (6.2%)2 (4.9%)2 (8.3%)
Allograft dysfunction   20 (13.3%)5 (7.7%)3 (7.3%)2 (8.3%)
Unknown/unspecified 13 (8.7%)3 (4.6%)2 (4.9%)1 (4.2%)

Discussion

The present study was an analysis of the outcome and potential risk factors affecting the survival of 246 children undergoing rLT at one of the 39 centers participating in the SPLIT Research Group during the study period of 1995–2004. During this 9-year study period, the rate of rLT of our multicenter study cohort was 11.2%, similar to rLT rates of 5–22% reported in adult series (9,11,13–15,17–19) and pediatric single center experiences in the United States and Europe (3,6,8).

Patient survival rates after rLT are inferior (1-and 4-year rates of 67% and 59%) compared to children registered with a primary LT in the SPLIT registry (88% and 83%). Our univariate results demonstrated worse overall survival after rLT for children under age 1 year at time of rLT, requiring intensive care or hospitalization at the time of rLT, having primary graft dysfunction as the indication for rLT, receiving a graft from a donor younger than 12 months of age, and receiving a technical variant graft. However, on a multivariate analysis of patient survival, only the variables of young donor age under 12 months, use of a technical variant graft and INR level at the time of rLT, were pretransplant factors predictive for survival after rLT. Survival after rLT is likely to be dependent upon the interactions of a number of factors, such as the timing of the rLT. Children who required rLT within the first 30 days after first LT (ErLT) had even poorer survival (1- and 4-year patient survival rates of 59% and 56%, respectively) compared with those children who underwent LrLT (1- and 4-year rates of 74% and 61%), as previously reported in both pediatric and adult single center reports (2,3,6,8,17). Conversely, improved outcomes in children after LrLT versus ErLT are likely explained by their being less debilitated to withstand the trials of a second LrLT, even within the context of the increased technical difficulty involved with fibrosis surrounding the graft and blood vessels complicating the LrLT surgery (17). In our study cohort, the operative time and amount of blood transfusion for LrLT was significantly longer and increased compared with the needs for ErLT.

In our study cohort, a whole organ allograft was used more frequently in children undergoing LrLT (68%) than in ErLT (28%, p < 0.0001). These rates, in addition to the inferior patient and graft survival of the ErLT cohort, support findings previously suggestive that outcome after rLT is less successful for indications such as primary graft dysfunction (20,21). Impaired outcomes after rLT, particularly ErLT, are due to both liver-related, as well as extra-hepatic pre-LT morbidities and complications. These include cerebral edema and dialysis requirements, which were observed more frequently in patients undergoing ErLT than in those patients requiring LrLT. In our multivariate analysis, children receiving a whole organ allograft were two and a half times more likely to survive after rLT than those receiving a technical variant graft, although recipient age, significant in the univariate analysis, did not remain significant in our multivariate analysis. In a recent SPLIT analysis examining the impact of graft type on outcome after primary pediatric LT, a poorer outcome for children receiving technical variant (both split and reduced) allografts compared with those in receipt of whole organs was similarly noted, although clearly the overall results for each of the transplant types were still excellent (22). In our current study, the observed worse outcome associated with rLT using reduced-size allografts is similarly again a consequence of sicker patients receiving such technical-variant allografts. Noting that children less than 3 years of age undergoing rLT with reduced-size grafts had the worse outcome, Newell had previously proposed a strategy of awaiting whole organ graft availability for their young children requiring rLT, in the belief that the benefit of improved survival justified the theoretical risk of clinical deterioration while awaiting a full-size allograft (6). Arguably, such strategy would also allow livers that would have been used as reduced-size grafts in such high-risk patients to be used with a far better outcome in adults, children undergoing primary LT, or older children undergoing rLT. Ultimately, the poorer outcomes of technical variant techniques must be put into perspective on a case-by-case basis for high-acuity smaller sized recipients who, in the absence of receiving technical variant grafts for either primary LT or for rLT, have no other therapeutic options.

The most common causes of graft failure requiring rLT were vascular complications (hepatic artery thrombosis and portal vein thrombosis) in over a third of cases, followed by primary graft dysfunction (19%) and chronic rejection (14.6%). Contrary to postulations that quickly replacing a graft which became defective from donor factors (such as is the case where primary nonfunction is the indication for rLT) should result in the similar better outcomes reported in the comparable adult scenario (12), children who undergo ErLT very likely have much less physical reserve to withstand the physical stress of a second major surgery, amidst the background of severe malnutrition and other complications of chronic liver failure in a small body with median age of under 2 years at the time of rLT. The recurrence of primary disease remains rare and is almost never the indication of rLT in children (23). Chronic rejection accounted for almost 15% of rLT indications within our cohort, all occurring as LrLT despite an increasing array of immune suppression agents and its diverse and frequently complementary mechanisms of action. Advances in the knowledge of the postulated immunological roles of HLA antigen matching and/or mismatching, positive lymphocytotoxic cross-matching (24,25) and older donors with preexisting arteriopathy in the allograft and the biliary epithelium rendering increased susceptibility to immunologic insult (26) will further add to our ability to decrease this entity as a common indication for rLT in children. The role of self-discontinuation of immune suppression on the longevity of allografts in noncompliant teenagers is increasingly being put forth as a focus of concern for pediatric programs (27,28).

In this day and age when the demand for livers is far greater than the supply, the ethical question logically arising each time a patient is put forth for reevaluation for rLT includes assessing the fairness of allocating precious organs to a rLT group with known worse outcome, particularly because of the implications that it is at the expense of other patients awaiting a chance for primary LT (29). Nonetheless, health care providers and families of patients with a failing first liver graft are faced with the reality that rLT is the only alternative to death.

This data analysis from a multi-center registry has a number of limitations and biases, both known and unknown. First of all, it is not possible to perform a complete verification of accuracy and completeness of data submitted to the SPLIT data-coordinating center. Site visits are conducted every 2–3 years. Secondly, participation in SPLIT is voluntary and therefore may be subject to bias. Third, with 39 participating centers contributing data to SPLIT at the time of data closure for this analysis, discrepancies between the outcome of a large series such as this and smaller single center series may be related to center-specific effects. Given the low actual numbers of rLT in children, it is currently impossible to demonstrate any center-specific effect. Nonetheless, despite the limitation afforded by registry data, our results are important both as they provide insight into the cumulative experience in outcomes of children after rLT, but also in contributing to the generation of hypotheses for future studies.

In conclusion, the highest risk group for death after rLT was those patients receiving an allograft from a donor less than 12 months of age, children receiving a reduced allograft, and recipients with increased INR at the time of rLT. The rLT in children is associated with decreased patient survival, particularly for rLT undertaken within 1 month of primary LT. Factors related to decreased survival after rLT include younger recipient age, poorer overall clinical condition at the time of rLT surgery, donor age and use of a reduced graft. Impaired outcomes after rLT, particularly ErLT, are due to both liver-related, as well as extrahepatic pre-LT morbidities and complications. Continued improvements in both immunologic and technical aspects of LT should help reduce the need for rLT in children, and thereby allow the imbalance in supply and demand of deceased-donor organs to be utilized more optimally. As the SPLIT registry continues to accumulate data, we shall continue to actively track the outcome and advances in managing these challenging and important patients.

Acknowledgments

The following study was presented in part at the 2005 American Transplant Congress in Washington, DC. The Studies of Pediatric Liver Transplantation (SPLIT) research group wishes to acknowledge the contributions of the following people at the data coordinating center (EMMES Corporation: A Lindblad PhD, D Brown, G Fraser, N Hornbeak, J Mitchell, L Covington, N Patel). SPLIT is funded by a grant from National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health (U01DK061693-01A1). Additional support is provided by unrestricted grants from Astellas Pharma US, Inc and Roche Laboratories. Dr. Ng also wishes to acknowledge Professor Philip Sherman for his critical and insightful review of this manuscript. This study would not be possible without the ongoing dedicated efforts of the study coordinators at the participating centers.

The SPLIT Research Group

List of Centers, Investigators, and Coordinators (as of June 1, 2004).

Alfred I Dupont Hospital for Children, Wilmington, DE (S Dunn, D Guattery).

Boston Children's Hospital, Boston, MA (M Jonas, L Krawczuk).

Cardinal Glennon Children's Hospital, St. Louis, MO (R Kane, H Solomon, E Phillips).

Children's Healthcare of Atlanta, Atlanta, GA (T Heffron, J DePaolo, T Pillen, G Smallwood, L Davis).

Children's Hospital of Cincinnati, Cincinnati, OH (J Bucuvalas, W Balistreri, F Ryckman, K King, G Arya).

Children's Hospital of Denver, Denver, CO (F Karrer, M Narkewicz, R Sokol, K Orban-Eller).

Children's Hospital of Philadelphia, Philadelphia, PA (E Rand, C Goodsell).

Children's Hospital of Pittsburgh, Pittsburgh, PA (G Mazariegos, A Smith).

Children's Hospital Western Ontario, London, ON, Canada (P Atkinson).

Children's Hospital of Wisconsin, Wauwatosa, WI (G Telega, S Lerret).

Children's Medical Center of Dallas, Dallas, TX (J Roden, N Mittal, B Friedman, L Cutright).

Children's Memorial Medical Center, Chicago, IL (E Alonso, R Superina, S Kelly, J Lokar, K Neighbors).

Children's Mercy Hospital, Kansas City, MO (W Andrews, J Daniel, V Fioravanti).

Hospital for Sick Children, Toronto, ON, Canada (A Fecteau, C Koziokek, V Ng).

Johns Hopkins Hospital, Baltimore, MD (P Drusso, P Colombani, M Alford, C Baldino, L Wilson).

LeBonheur Children's Medical Center, Memphis, TN (H Shokough-Amiri, H Grewal, S Powell).

Mayo Medical School, Rochester, MN (D Freese, M Ishitani, J Weckworth).

Medical College of Virginia, Richmond, VA (R Fisher, M Akyeampong, M Shah).

Mount Sinai Medical Center, New York, NY (S Emre, L Arnott, R Gagliardi,).

New York Presbyterian Hospital, New York, NY (S Lobritto, P Harren, D Zimmerman).

Sainte-Justine Hospital, Montreal, QC, Canada (F Alvarez, S Martin, C Viau).

St. Louis Children's Hospital, St. Louis, MO (R Shepherd, M Nadler).

Stanford University Medical Center, Palo Alto, CA (W Berquest, K Cox, M Kreisl, S Alvarez,R Berquest).

Texas Children's Hospital, Houston, TX (R Quiros, J Goss, J Scott, L Bistrow).

UCLA Medical Center, Los Angeles, CA (S McDiarmid, S Fiest).

University of Alberta, Edmonton, Endmonton, AB, Canada (L Smith, L Greet).

University of California, San Diego, CA (J Lavine, A Khanna, R Clawson, D Valmassoi).

University of California, San Francisco, CA (P Rosenthal, S Stritzel).

University of Chicago, Chicago, IL (MJ Millis, P Boone).

University of Florida – Shands, Gainesville, FL (R González-Peralta, M Langham, E Richmond).

University of Miami/Jackson Memorial Hospital, Miami, FL (A Tzakis, T Kato, D Weppler, B Miller, M Gonzalez, K Marshall).

University of Michigan, Ann Arbor, MI (J Lopez, J Magee, V Sheick).

University of Minnesota, Minneapolis, MN (A Humar, B Durand).

University of Nebraska, Omaha, NE (S Horslen, B Shaw, D Andersen, B Fleckten).

University of North Carolina, Chapel Hill, NC (J Fair, S Lichtman, P Yee).

University of Rochester, Rochester, NY (M Orloff, T Rossi, C Hill-Sober, T Shisler).

University of Texas, HSC San Antonio, San Antonio, TX (A Mejia, J Silva).

University of Wisconsin, Madison, WI (M Kalayoglu, A D'Alessandro, S Knechtle, E Spaith).

Ancillary