The Pediatric End-Stage Liver Disease (PELD) scoring system is a new nationally utilized formula developed to provide a continuous numerical assessment of the risk of death in order to allocate livers to children for transplantation. A retrospective review of the clinical course of children undergoing liver transplantation at the Mount Sinai Medical Center was performed in order to assess the effectiveness of this scoring system in the first 24 months of its utilization. Forty-eight patients underwent liver transplantation with overall patient and graft survival rates of 98% and 96%, respectively. In 23 cases the PELD scoring system determined waiting time for transplantation. Of these 23 patients, 7 moved to the intensive care unit (ICU). Only 2 of 23 patients underwent transplantation with their actual PELD score. The rest required petition for exception (17) or status 1 listing (4). Significant morbidity occurred while awaiting transplantation: failure to thrive (78%), ascites (73%), hemorrhage (49%), infectious complications (39%), encephalopathy (30%), peritonitis (17%), pathologic bone fractures (13%), and hepatopulmonary syndrome (9%). In patients with PELD scores granted by exception the average score that did not yield a liver offer was 38 with an average waiting time of 55 days. At the time of transplantation actual PELD score averaged 22, while the petitioned score was 40. Based upon our center's initial experience, the current PELD scoring system is not adequate. Actual PELD scores did not lead to timely allocation of livers to children. It appears that this scoring system underestimates the near-term risk of death. Urgent reassessment is required to prevent potential morbidity and mortality in children. In conclusion the United Network for Organ Sharing policy that permits granting of exceptions has circumvented these problems with the PELD scoring system. (Liver Transpl 2005;11:788–795.)
Organ transplantation is a life-saving therapy for many individuals with acute and chronic diseases. The relative success of this therapeutic approach and expanding indications for transplantation coupled with persistent lack of long-term efficacy of other treatments for end-stage organ failure has led to continued relative shortages of organs for transplantation. In light of these shortages, a number of systems have been developed to ration available organs to individuals in need according to ethical, medical, and social criteria. Each system that has been developed strives to optimize organ allocation, with maximal benefit and minimal morbidity and mortality to those who await available organs. Several iterations of allocations systems have evolved. At the outset, each system is designed to be objective and attempts to avoid subject bias. Individual medical centers that participate in these allocation programs are faced with competing demands. One demand from society as a whole is to maintain a “fair” system for all individuals in need of a scarce organ. A competing demand, from individual transplant candidates, is to optimize the chance that the particular patient will receive a life-saving transplant. In light of these competing interests it is not surprising that each new system may be “gamed” and ultimately can fail to achieve its purpose of objective and fair allocation of organs.
In the early part of 2002 the liver transplant allocation system was changed in an attempt to establish, once again, a more objective listing paradigm. The prior system had a number of broad list status categories, and waiting time in a particular category was a potential tie-breaker in the allocation process. As with other allocation schemes, there was an evolution to “game” this system and many potential candidates for liver transplantation were listed at the earliest time point possible. As such waiting time became irrelevant and the system became relatively dysfunctional. A new system was developed that was predicated on eliminating waiting time and using objective laboratory data that were not easily manipulated or “gamed.” The system that was developed for adults was the Model for End-Stage Liver Disease (MELD) score.1 By necessity, the Pediatric End-Stage Liver Disease (PELD) scoring system was created; it was to be analogous to the MELD system.2 The PELD system was developed by analysis of a prospective registry (Study of Pediatric Liver Transplantation) of children listed for liver transplantation.3 A variety of parameters were assessed for their ability to predict either death while awaiting transplant or the need for transfer to the intensive care unit (ICU). The second parameter was equated with death. On the basis of the analysis of 74 children/events the PELD scoring system was developed and implemented. The essential elements of the PELD score include total bilirubin, International Normalized Ratio, albumin, age less than 1 year, and evidence of failure to thrive. The current report describes the first 24 months of utilization of the PELD scoring system for children who underwent pediatric liver transplantation at the Mount Sinai School of Medicine. This initial experience suggests that there may be major problems with this new scoring system as a means of prioritizing children for liver allocation. Actual PELD scores rarely led to liver allocation to children.
PELD, Pediatric End-Stage Liver Disease; MELD, Model for End-Stage Liver Disease; ICU, intensive care unit.
Patients and Methods
This retrospective analysis was performed from a chart review of all children who underwent liver transplantation or died awaiting liver transplantation between February 27, 2002, and February 26, 2004. This time period includes the first 24 months of utilization of the PELD scoring system. The PELD system applies to children less than 18 years of age; thus, this analysis was limited to the same age range. Inpatient and outpatient records at the Mount Sinai School of Medicine were comprehensively evaluated to assess waiting times and clinical events or complications that occurred during that waiting period. Clinical and laboratory data were reviewed. Prothrombin time prolongation was only recorded if there was documentation that it would not respond to parenteral vitamin K administration. Hypoalbuminemia was noted if it was persistent and not temporally related to acute gastrointestinal hemorrhage or sepsis. Hyponatremia was identified in patients with chronic liver disease and no other identifiable cause for that hyponatremia (e.g., acute gastroenteritis, pseudohyponatremia from diabetes, etc.). Hyperbilirubinemia was always associated with a significant proportion of direct reacting bilirubin (typically greater than 50%, always more than 20%).
A number of specific complications of chronic liver disease were sought and are characterized below. Gastrointestinal hemorrhage was defined as blood loss that could only be attributed to the gastrointestinal tract, with special identification of episodes requiring red blood cell transfusion. Ascites was defined as obvious on abdominal sonography or clinical examination and requiring diuretic therapy (typically both spironolactone and furosemide). Large volume paracentesis as a therapy for intractable ascites was defined as the percutaneous drainage of greater than 50 cc/kg of peritoneal fluid at a single paracentesis. Our clinical practice always includes the use of diuretics so large volume paracentesis is always for symptomatic and intractable ascites. Spontaneous bacterial peritonitis was defined by the growth of an appropriate bacteria from a percutaneously obtained peritoneal fluid specimen, the finding of more than 500 white blood cells/mL or more than 250 neutrophils/mL in a patient who required intravenous antibiotic therapy. Cholangitis was diagnosed in the setting of unexplained fever associated with a cholestatic change in biochemical liver profile and/or positive blood culture consistent with biliary infection. Failure to thrive is difficult to retrospectively assess in children with chronic liver disease associated with organomegaly and/or ascites. We defined failure to thrive by two different relatively objective criteria: (1) z score < −2 for either weight or height, or (2) the need for nasogastric tube feedings to treat clinically relevant failure to thrive. Hepatopulmonary syndrome was screened for by upright transcutaneous oxygen saturation measurements in all patients with cirrhosis. Saturation values of less than 97% were subsequently assessed by bubble echocardiography and macroaggregated albumin scanning to confirm the presence or absence of hepatopulmonary syndrome. Severe pruritus was defined as scratching that led to cutaneous mutilation despite maximal standard medical therapy. Pathologic bone fractures were those observed without obvious significant trauma in the setting of markedly decreased bone density as measured by dual energy x-ray absorption scanning.
All liver organ offers to the Mount Sinai Medical Center during the time period of this study were reviewed. The primary rationale for refusal of an offer was recorded and included the following possibilities; (1) donor factors—age, serologies (e.g., hepatitis B or C), organ size (if large for size, organ characteristics did not permit splitting or reducing of the organ), liver function tests, liver biopsy findings, miscellaneous other factors (prolonged hypotension, multiple pressors, etc. in the donor), and (2) recipient factors—acute illness precluding transplant or improvement in the recipient.
This retrospective analysis was approved by the Institutional Review Board of the Mount Sinai School of Medicine.
Between February 27, 2002, and February 26, 2004, there were a total of 48 liver transplants performed in the Pediatric Liver/Liver Transplant Program at the Mount Sinai School of Medicine (Table 1). The overall patient and graft survivals for this cohort were 98% and 96%, respectively, with follow-up ranging from 6 to 30 months. The underlying diagnoses leading to liver transplantation in this cohort were; biliary atresia (n = 15), acute liver failure (n = 12), chronic cholestasis (n = 6, [Jeune's syndrome 2, Alagille syndrome 1, ductal plate malformation 1, sclerosing cholangitis 1, and progressive intrahepatic cholestasis 1]), autoimmune hepatitis (n = 4), metabolic disease (n = 4 [tyrosinemia 2, citrullinemia 1, Factor H deficiency 1]), retransplant (n = 3 [acute 2, chronic 1]), cryptogenic cirrhosis (n = 2), autosomal recessive polycystic kidney disease (n = 1), and liver tumor (n = 1). The following types of liver transplants were performed in this cohort; living donor 7, split liver 24, reduced liver 3, whole liver 15.
Table 1. Listing Status
Total number of transplants
Excluded from analysis of waiting time/petitioning and PELD
Age greater than 17
Acute liver failure: status 1
Split not allocated to recipient
Elective living donor
Petition status 1
Moved to ICU status 1
The analysis of the impact of PELD concentrated on patients where the PELD score determined waiting time (Table 1). Four patients (8%) who were older than 17 years of age were excluded from further analysis since MELD was used in their cases. Ten patients (21%) were listed as status 1 for fulminant hepatic failure, as such PELD changes had no impact on their status and they were excluded from further analysis. Five additional patients were excluded from analysis of the effects of PELD for the following reasons; 2 combined liver and kidney transplant, 2 early retransplant for hepatic artery thrombosis and primary nonfunction (status 1), 1 split liver transplant not allocated to the recipient. In three cases (6%) the primary approach to transplantation was living donor and PELD was not relevant. In three cases (6%), status 1 was granted by petition for the concern of hepatocellular carcinoma (subsequently identified in all three explanted livers). Thus there were 23 transplants where it was appropriate to assess the impact of the PELD scoring system. In 4 cases (8%) the patient was made status 1 after moving to the ICU for a life-threatening complication of their liver disease. Nineteen patients (40% of the total in the cohort) underwent liver transplantation with their PELD score determining allocation of the liver. Only 2 (10%) of these 19 patients were allocated livers on the basis of their actual PELD score, while the rest required granting of a wait-list exception. The blood types in the two patients whose actual PELD score was used were A and B.
Petitioning and Wait Time Analysis
A total of 38 exception petitions were submitted for the 26 patients where PELD score potentially determined waiting times. Eleven patients had a single petition, 8 had two petitions, while 4 had three petitions. Thirty-five of the petitions were accepted by the regional review board (Region 9). Three of the accepted petitions were for status 1 because of a high index of suspicion of hepatocellular carcinoma. In two cases the child had an underlying diagnosis of tyrosinemia type 1, while the third child had a massive tumor that developed in the setting of growth hormone therapy for short stature associated with Turner's syndrome. In all three cases hepatocellular carcinoma was confirmed at pathologic evaluation of the explanted liver. In no case was extrahepatic spread documented and all three patients are currently alive and well. Thus the Status 1 exception in these three cases appeared to have been justified. The remaining 35 petitions were submitted on behalf of 23 patients. The rationales for the petitions can be seen in Table 2. The waiting time at each petitioned status can be seen in Figure 1. The average waiting time since original listing for liver transplantation and prior to petitioning was 311 days (standard deviation [SD], 619 days; median, 132 days; range, 6-3041 days). The average petitioned PELD score, which did not lead to an offer for liver transplantation, was 38 (SD 5), with an average waiting time at the inadequate score of 55 days (SD, 31 days). The average petitioned PELD score that led to liver allocation was 40 (SD, 7), with an average waiting time of 25 days (SD, 20 days). Actual PELD scores at the time of liver transplantation averaged 22 (SD, 10).
Table 2. Rationale for Petition for Exception
NOTE. More than one rationale may be used for a single patient.
Abbreviation: HCC, hepatocellular carcinoma.
Failure to thrive
Portal hypertensive hemorrhage
Risk of HCC
Urea cycle defect
Clinical Course and Complications Awaiting Transplantation
Complications of the underlying liver disease were carefully assessed by chart review in the 23 patients where PELD scoring determined time awaiting liver transplantation. This includes the 19 patients who were transplanted with a PELD score and the 4 patients who ultimately moved to the ICU and were listed as status 1 because of a complication of their liver disease. A summary of the diagnoses, clinical course, and complications in these 23 patients is seen in Table 3. Overall, significant complications were seen in all patients, with a range of between 1 and 8 complications per patient (mean, 4 complications per patient). The most common complications were significant failure to thrive and ascites, which were observed in 78% and 73% of patients, respectively. Other potentially life-threatening complications occurred in a significant number of patients and included, severe bacterial infections (not peritonitis) requiring intravenous antibiotics (39%), spontaneous bacterial peritonitis (17%), hepatopulmonary syndrome (9%), and hemorrhage (49%), which required transfusion in all but two of the patients. Many of the complications although not life-threatening had a large impact on quality of life, including intractable pruritus (9%), pathologic bone fractures (13%), and chronic encephalopathy (30%).
NOTE. In the Complications column, note for specific and distinct complications the number of events is listed in parentheses after the complication; e.g., Al (4), indicates 4 large volume paracenteses performed for intractable ascites.
Abbreviations: BA, biliary atresia; Crypt, cryptogenic; urea, citrullinemia; PSC, primary sclerosing cholangitis; DPM, ductal plate malformation; AIH, autoimmune hepatitis; S1, status 1 on the basis of complications requiring intensive care treatment; PIHC, progressive intrahepatic cholestasis; S, split liver transplant; W, whole liver transplant; L, living donor transplant; ht, height; and wt, weight. In the Complications column: E, encephalopathy; Fz, failure to thrive (z indicates z-score ≤ 2 [for either height or weight]), Fn, ng tube feeding; I, infection/sepsis/cholangitis; Pr, pruritus; Pa, pancreatitis; Pe, peritonitis; Ad, ascites diuretics; Al, ascites large volume paracentesis; H, gastrointestinal hemorrhage; T, gastrointestinal hemorrhage requiring packed red blood cell transfusion; Hp, hepatopulmonary syndrome; and Fx, pathologic fracture.
Hospitalization at referring hospital not quantifiable.
Not applicable because of exchange transfusion and continuous blood product administration.
Hospitalization primarily for social reasons, excluded from statistical analysis.
A total of 7 children (30%) in this cohort were transferred to the ICU while awaiting liver transplantation. Four of the seven children who were transferred to the ICU were made status 1 based upon the severity of their illness. One of these children was intubated for pulmonary hemorrhage, one was in renal failure, and two required large volume paracenteses for intractable ascites associated with significant respiratory distress. All four had profound failure to thrive, synthetic liver dysfunction, and hyponatremia. The average PELD score of 42.7 (SD, 3.5; n = 3) at the time of ICU transfer in these status 1 patients was similar to the petitioned PELD scores required for liver transplantation (40 ± 7, P > 0.05) and significantly greater than the actual PELD scores in the 19 patients who were transplanted on the basis of their PELD score (22 ± 10, P < 0.05).
Liver Transplantation Based on PELD Scores
Nineteen children underwent liver transplantation in the setting of a liver allocated on the basis of their PELD score. Seventeen of these PELD scores (90%) were the result of a petition for exception. Six of the 19 transplantations were performed with whole organs, while the remainder were split livers. The average age of the donor was 19 (SD 10), with an age range for the donors of between 3 and 41 years. Only 5 (26%) of the 19 livers were from pediatric donors.
Deaths Awaiting Liver Transplantation
During the time period of this review four children died who had been listed for liver transplantation. One child died from brain stem herniation associated with increased intracranial pressure in the setting of acute liver failure. This child had been listed at United Network for Organ Sharing Status 1. One child died of overwhelming sepsis. She had biliary atresia and a history of recurrent cholangitis. At the time of her death her PELD score was 17 and plans were in progress to petition for a higher priority score. Two children were removed from the transplant list and subsequently died. In one case the child had systemic manifestations of mitochondrial disease and in the second case severe porto-pulmonary hypertension was diagnosed. In both cases the children were deemed to be noncandidates for liver transplantation.
Liver Organ Offers Refused
During the time period of this review there were a total of 162 offers of livers to the 23 children in this cohort where the PELD scoring system impacted upon waiting time for liver transplantation (Fig. 2). The majority (91.4%) of the reasons for refusal were due to donor characteristics. The primary reason for refusal was the advanced age of the donor. All but one of the recipient-related refusals (8.6%) were due to acute illness that precluded liver transplantation. In many circumstances there were more than one reason for refusal of a potential liver offer, although only the primary reason is reported.
This single center experience demonstrates significant potential problems with PELD, the current scoring system for liver allocation in pediatrics. The problems are most notable in the fact that the scoring system in its current format could only be used in 10% of eligible candidates. Seven out of 23 children moved to the ICU while on the transplant list. Movement to the ICU was equated with death in the development of PELD, and as such this represents a 30% wait-list mortality rate in this limited experience. Reasons for transfer to the ICU are quite subjective, and thus this may not be an accurate assessment of the degree of illness of these children. Our policy for transfer to the ICU is relatively strict and involves a clinical need for intensive care (e.g., not simply severe failure to thrive) and does not reflect our prediction of near term mortality. In fact the number of children at The Mount Sinai Medical Center who underwent liver transplantation at Status 1 without acute liver failure was 20% compared with 32% nationally.4 Thus, it is unlikely that our criteria for ICU transfer were too lenient. The children in this cohort suffered a wide range of complications, with morbidity seen in all patients. Quality of life was seriously compromised by problems including failure to thrive, chronic encephalopathy, intractable pruritus, various infectious complications, and metabolic bone disease. This center's problems with the PELD scoring system did not translate into a significant number of actual deaths while awaiting transplantation. This is due to the fact that the United Network for Organ Sharing allocation system permits submission of petitions for exception that can increase a recipients PELD score.
A number of factors may be implicated as potential problems with PELD, but the most prominent is the fact that it seems to underestimate near-term risk of mortality. This is especially important since it is predicted 3-month risk of mortality that is used to compare MELD and PELD scores, thus mediating competition between adults and children for livers from adult donors. This seems to be a more problematic at the higher end of the scoring system. Organ shortage in Region 9 appears to be more of an issue than other regions of the country and thus our center's experience may have uncovered issues with relatively high PELD scores. The PELD system was derived to be a continuous scoring system that was purported to predict the likelihood of death in the ensuing 3 months. PELD scores of 27, 35, 40, 43, and 46 are supposed to predict 10%, 20%, 30%, 40%, and 50% risks of death, respectively, in the ensuing three months. Although PELD appears to stratify the degree of illness of children with similar types of liver disease, the absolute prediction of the risk of death appears to be erroneous. In our limited series, children with PELD scores of approximately 40 were typically transferred to the ICU in a clinical state that justified listing at status 1. The 30% risk of death in the next three months in these patients appears to be markedly understated. In Figure 3, a graphical depiction of the clinical elements that are required to generate a PELD score of 40 are shown. Careful analysis of the parameters reveals clinical situations that are not typically seen in children with chronic liver disease. Biliary atresia is the leading indication for liver transplantation in children. It is very unusual for children with biliary atresia to develop uncorrectable and severe coagulopathy until they are extremely ill. This is certainly the experience in our center where there is routine supplementation with adequate doses of vitamin K. Similarly, total bilirubin levels are rarely greater than 20 to 25 mg/dL with the exception of associated renal failure or sepsis. Thus, it is unlikely that a child will have a PELD score of 40 and be clinically stable. In the Yokohama experience, the mean PELD score less than one month prior to death in children with biliary atresia awaiting liver transplantation was 23, with a range of 2 to 41.5 Of note there were almost no children who had a PELD score in excess of 40.5 Interestingly, the patients in our cohort had a mean PELD score of 20 (predicted three-month mortality of 10%) at the time of petition for transplantation. Based upon the Yokohama experience, the timing of the petitions was certainly warranted.
It is not clear why the absolute prediction of risk of death by a specific PELD scores appears to be significantly lower than reality. Rigorous testing of this system was not possible given the short time period that existed between the development of the PELD scoring system and its implementation. In fact, PELD was developed with a very limited set of data elements. Clinical data in SPLIT is only collected every six months prior to liver transplantation. The outcome variable that was tested was death or movement to the ICU. The most definitive end-point was death prior to transplant, although it was not specified that the death was directly related to liver disease. Fewer than 50 children died awaiting liver transplantation. In light of the limited number of deaths awaiting liver transplantation, an additional a more subjective end-point was utilized that being movement to the ICU. A recent national conference reviewing the initial experience with MELD/PELD concluded, “pediatric data are not robust enough to draw specific conclusions as to how PELD should be modified to enhance its ability to stratify children awaiting transplantation.6” In our opinion, it is likely that without the potential for exceptions death rates in children awaiting liver transplant will rise as a result of the discrepancy between predicted and actual risk of death. It is worrisome that the percentage of children nationally who undergo liver transplantation at Status 1 by exception has increased from 23.3% to 29.7% (P = 0.006 derived from data in Figure 7 in McDiarmid et al.4) after the institution of PELD.
Whether local practices impacted on the utility of PELD can be questioned. The current allocation system gives preference to children for livers from pediatric donors. There are not enough pediatric organs to meet the current need for liver transplantation in children. The significant number of children at Status 1 at our center and in our region are likely to have utilized many of the potential pediatric organs that were available. In this experience, 13 of 19 transplantations in children affected by PELD were performed using livers from adult donors. It might be argued that the criteria for accepting liver offers at the Mount Sinai Medical Center were too stringent and therefore the required PELD scores were inflated. Review of the data in Figure 2 negates this argument. Organs from significantly older donors and/or seropositive for hepatitis C and hepatitis B (including core antibody positive) were offered and generally refused by this center. A large number of the refused organs had already been refused by other centers nationwide. Poor long-term outcome has been well documented with the use of organs from very old donors.7, 8 In light of the potential life expectancy in children, long-term outcome is a critical issue and we feel that refusal of these organs was justified. It is difficult to compare these data to other center's experience as this information is rarely published. One might argue that Region 9 was too liberal in granting exceptions. Review of the waiting times with the exceptions, makes this argument untenable. It is very unlikely that candidates in this cohort would have received livers in a timely fashion utilizing their native PELD scores. The current data cannot definitively prove that the actual PELD score would not have led to organ allocation. In our opinion, it would not have been in our patient's best interest to continue listing these patients with their actual PELD scores.
In summary, the experience described here illustrates significant potential problems with the PELD scoring system at a single institution in the first 24 months of its use. Relative organ scarcity in Region 9 and the current allocation approaches in that region appear to have uncovered specific issues with the prediction of risk of death at the higher end of the PELD scoring system. The scoring system as a marker of liver disease severity may have merit, but the quantitative translation of the score into risk of death appears to be invalid. As such, direct comparison and simultaneous use with the analogous MELD system is problematic. There is an urgent need to reassess the PELD scoring system and organ allocation to children in order to avoid unnecessary morbidity and mortality in children. Analyses similar to the current study should be considered at other centers and through national databases. End-points other than death or ICU transfer may need to be analyzed as surrogate markers of advancing and critical illness in order to generate sufficiently large data sets for a robust scoring system. Currently, the United Network for Organ Sharing policy that permits granting of exceptions has circumvented these issues in the PELD scoring system.