Regional Variation and Use of Exception Letters for Cadaveric Liver Allocation in Children with Chronic Liver Disease
Abstract
The Pediatric End-Stage Liver Disease (PELD) score was designed to reduce subjectivity in liver allocation and to advantage patients with a higher probability of waiting list mortality. The aims of this study were to determine the impact of PELD implementation for children with chronic liver disease and to assess whether PELD met its goal of standardization of liver allocation for children. This study used data reported to the United Network for Organ Sharing (UNOS) registry for children with chronic liver disease receiving primary cadaveric liver transplant between January 2000 and December 2001 (pre-PELD) and March 2002 and July 2003 (PELD). PELD reduced the percentage of children transplanted while in an intensive care unit and as status 1. A calculated PELD score was used for allocation in only 52% of recipients. Thirty percent were status 1 at transplant and PELD scores granted by exception were used for allocation in 18% of patients. There was regional variation in PELD score at allocation and use of exception scores with a significant relationship between PELD score and percentage of exception cases. Regional variation suggests that PELD has not resulted in standardization of listing practices in pediatric liver transplantation.
Introduction
In 1998, the Institute of Medicine issued recommendations for allocation of cadaveric livers for transplantation. This report included directives that led to fundamental changes in liver allocation in the United States. Specifically, it required that medical urgency of the potential candidates replace waiting time as the primary determinant of allocation and that a process be developed to minimize futile transplants (1). Subsequently, the Department of Health and Human Services adopted the ‘Final Rule’, establishing that the nationwide allocation system of cadaveric livers for transplantation must benefit patients based primarily on their medical status (2).
In order to design and assess potential models that would estimate survival of patients on the liver waiting list and establish priority for allocation of cadaveric liver grafts, The Organ Procurement and Transplantation Network (OPTN) Liver and Intestine Committee formed a subcommittee, the Liver Disease Severity Score Committee (3). This subcommittee identified an existing model, the Model for End-Stage Liver Disease (MELD) as a score that could be adopted to prioritize adult patients for transplantation according to estimated short-term mortality (4,5). For the pediatric and adolescent population, the subcommittee developed a new model, which was designed to function in parallel with the adult MELD score. Model development utilized data that had been previously collected through the Studies of Pediatric Liver Transplant (SPLIT) database. The variables included in the model were limited to the existing SPLIT data set, which represented approximately 50% of children on the United Network for Organ Sharing (UNOS) waiting list. This model, the Pediatric End-Stage Liver Disease (PELD) score, was designed to predict death and/or transfer to an intensive care unit while on the waiting list. The statistical validation of the model using the area under the receiver operating characteristic curve (AUC ROC) showed a very high predictive ability of >0.80 (6).
In February 2002, UNOS adopted the MELD and PELD scores to allocate cadaveric livers, respectively, for the adult and the pediatric population in the United States. The new system was designed to prioritize patients by acuity of illness rather than waiting time. To improve standardization of organ allocation the MELD and PELD scores include only objective, verifiable measures of illness. Rigorous testing of both models has included validation in additional patient cohorts. The MELD score has been tested in different populations prior to and after its implementation (4,5,7–16) and the PELD score was successfully verified as a strong predictor of death on the waiting list in a second pediatric liver transplant database at the University of Pittsburgh (6).
Early reports analyzing the effect of the PELD score implementation on the cadaveric liver allocation system for the pediatric population have suggested that waiting list mortality, especially for younger infants may be reduced (17). However, the specific effect of the new system on children waiting for and receiving transplantation for the indication of chronic liver disease has not been examined. The aims of this study were to determine the effect of the PELD implementation on this specific population and to assess whether PELD met its intended goals of objectivity and standardization.
Methods
This is a retrospective cohort study of data from the OPTN database, supplied by UNOS. The OPTN collects waiting list additions, modifications and removals directly from the transplant centers through an internet-based data collection system. Patient inclusion criteria were: age at transplant <18 years, primary liver or liver/kidney transplant recipient and recipient of deceased donor graft. Patients having a diagnosis of acute hepatic necrosis of any etiology were excluded to focus the analysis on patients with chronic liver disease.
The analysis was based on data reported to UNOS as of 14 October 2003 and included data from pediatric patients transplanted during two historical eras: pre-PELD era (1 January 2000–31 December 2001): New Registrants n = 1292, Transplants n = 641 and PELD era (1 March 2002–31 July 2003): New Registrants n = 705, Transplants n = 463.
The outcomes variables compared between the two eras include: status 1 at transplant, admission to the intensive care unit (ICU) at transplant and 3-month post-transplant patient and graft survival. For the PELD era, the following variables were examined for regional variation: mean PELD score at allocation (exception and nonexception cases), number of patients that were allocated a graft based on an exception score, number of children transplanted while status 1 and admission to the ICU at transplant. ‘Exception cases’ refer to those patients who were transplanted with PELD scores assigned through the regional review process. Children transplanted as exception cases due to hepatocellular carcinoma (HCC) and hepatoblastoma (HBL) were excluded from all analyses evaluating the PELD score. PELD score at allocation is defined by UNOS as ‘patient allocation score at the time the match was run from which the patient received the offer that led to transplant’. For patients allocated an organ in a nonexception setting the PELD score at allocation represents the last updated, calculated PELD score based on laboratory data. For exception cases, the PELD score at allocation represents the score granted by exception used at allocation. The PELD scores calculated from laboratory values at transplant were not available in this data set for either group.
A separate cohort was created to examine median time-to-transplant (waiting time), waiting list mortality and reasons for removal from the waiting list between the two eras. This cohort consisted of patients added to the waiting list (‘new registrations’) in the pre-PELD and PELD eras. The mortality rates on the waiting list were calculated based on the number of patients and their cumulative time on the list and expressed as cases per 1000 patient years. In calculating the waiting list mortality, patients removed because of death or becoming too sick to transplant were treated as a combined endpoint because both represent a treatment failure of the allocation system.
Statistics
Results are expressed as mean ± standard deviation for continuous parametric data, median (interquartile range) for continuous nonparametric data and frequencies and percentages for categorical data. Time-to-transplant in each era was compared using a Cox proportional hazards regression model adjusting for age, gender, race and region. A separate Cox regression looked at regional differences in time-to-transplant in the PELD era adjusting for use of exceptions, age, gender and race. Comparisons between regional PELD scores at time of transplant were performed using analysis of variance (ANOVA) or Kruskal-Wallis, depending on data distribution. Waiting list mortality rate in the two eras was compared using the normal approximation to the binomial distribution. Actuarial survival of patients and allografts was determined by the Kaplan-Meier method and compared using the log-rank (Mantel-Cox) test. Statistical analyses were performed with SAS (SAS Institute, Cary, NC) statistical software. For all tests, p < 0.05 was considered significant.
Results
Demographics
Table 1 displays characteristics of patients receiving a primary liver or liver/kidney transplant in both eras. The median age at transplantation was 2 years (range: 0–10) in the pre-PELD era and 2 years (range: 0–11) in the PELD era. Of the 641 patients who were transplanted in the pre-PELD era, 30% were 12 months old or younger. Females represented 56% and Caucasian made up 51% of the transplanted population. Of the 463 patients who were transplanted in the PELD era, 30% were 12 months old or younger. Females represented 56% and Caucasian made up 53% of the transplanted population. Patients with biliary atresia, the most common diagnosis among pediatric liver transplant recipients, comprised 40% and 41% of the populations in pre-PELD and PELD eras, respectively. There were no statistically significant differences in these demographic variables by era.
| Variable | Pre-PELD era (n = 641) | PELD era (n = 463) | Total (n = 1104) | p-value |
|---|---|---|---|---|
| Median age (years) | 2.0 (0–10) | 2.0 (0–11) | 2.0 (0–11) | 0.93 |
| Age | ||||
| ≤1-year old | 30% (189) | 30% (138) | 30% (327) | 0.92 |
| >1-year old | 70% (450) | 70% (324) | 70% (774) | |
| Gender | ||||
| Male | 44% (280) | 44% (204) | 44% (484) | 0.90 |
| Female | 56% (361) | 56% (259) | 56% (620) | |
| Ethnicity | ||||
| Caucasian | 51% (325) | 53% (247) | 52% (572) | 0.39 |
| Others | 49% (316) | 47% (216) | 48% (532) | |
| Diagnosis | ||||
| Biliary atresia | 40% (259) | 41% (188) | 41% (447) | 0.93 |
| Others | 60% (380) | 59% (273) | 59% (653) | |
Comparison of waiting list dynamics between eras
The number of new registrations, median time-to-transplant and reasons for removal from the waiting list are shown in Table 2. Patients listed initially as status 1 (n = 304) and those with missing data (n = 15) were excluded. The number of new registrations to the waiting list were 1292 and 705 in the pre-PELD and PELD era, respectively. Liver transplant candidates were added to the waiting list at a rate of 1.76 per day in the pre-PELD era versus 1.36 per day in the PELD era. Cox regression analysis comparing time to transplantation included 776 transplants and 1221 censored observations. Observations were censored if the patient did not receive a transplant during the era or if the patient was removed from the list for any reason, including death. Time-to-transplant was significantly less in the PELD era, p < 0.0001. The adjusted medians for time-to-transplant were 365 days in the pre-PELD and 205 days in the PELD era. Table 3 shows median time-to-transplant in the PELD era by region and by exception and nonexception cases, after adjusting for age, gender and race. There is a significant difference between regions in time-to-transplant, p = 0.004. The unadjusted waiting list mortality rate for newly registered patients, including those listed as status 1, was not significantly different between the two eras, 199 versus 262 per 1000 patient years in the pre-PELD and PELD eras, respectively (p = 0.17).
| Pre-PELD era | PELD era | |
|---|---|---|
| Number of new registrations to the waiting list | 1292 | 705 |
| Reasons for removal from waiting list | ||
| Cadaveric transplant | 479 | 297 |
| Living donor transplant | 125 | 45 |
| Transferred to other center/multiple listing | 58 | 13 |
| Death/too sick | 101 | 51 |
| Improved condition | 15 | 9 |
| Refused transplant | 1 | 2 |
| Other | 14 | 2 |
| Remaining on list | 499 | 286 |
| Median time-to-transplant (days)** | 365 | 205 |
| 95% CI | 95% CI | |
| (300–470) | (175–265)* | |
- *p < 0.0001.
- **Median time-to-transplant is the time at which 50% of patients are expected to have received a transplant. Adjusted for age, gender, race and region.
| Region | PELD-era adjusted* median waiting time to transplant (days) | |
|---|---|---|
| Nonexception | Exception | |
| 1 | 228 | 153 |
| 2 | 443** | 307 |
| 3 | 104 | 79 |
| 4 | 211 | 144 |
| 5 | 267 | 182 |
| 6 | 103 | 78 |
| 7 | 122 | 92 |
| 8 | 179 | 123 |
| 9 | 299 | 208 |
| 10 | 197 | 136 |
| 11 | 273 | 191 |
- *Adjusted for age, gender and race.
- **53rd percentile—the 50th percentile (median) is not expected to be reached in this region for nonexception cases.
- There is a significant difference between regions in time-to-transplant, p = 0.004.
Patient status at transplant
Table 4 shows that both the percentage of children allocated an organ while in an ICU (24% in pre-PELD era vs. 18% in PELD era, p = 0.03) and as status 1 (36% in pre-PELD era vs. 30% in PELD era, p < 0.0001) has decreased in the PELD era. During the PELD era, 138 of 463 (30%) patients were status 1 at allocation, 83 of 463 (18%) met criteria for status 1 by being in an ICU and 51 of 463 (11%) were granted status 1 ranking by exception. Four patients (1%) had missing data for status 1 classification. Of the remaining children, 85 of 463 (18%) were allocated an organ based on a PELD score granted by exception and 240 of 463 (52%) were allocated based on a calculated PELD score (Table 5). Therefore, after the implementation of the new system of organ allocation, a calculated PELD score was used for cadaveric liver allocation in only 52% of children with chronic liver disease receiving transplants.
| Pre-PELD era (n = 641) | PELD era (n = 463) | p-value | |
|---|---|---|---|
| In ICU | 24% (151) | 18% (83) | 0.03 |
| Status 1 | 36% (230) | 30% (138) | <0.0001 |
| PELD exception case | — | 18% (85) | — |
| Region | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | Total |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Total transplants | 16 | 51 | 67 | 50 | 91 | 7 | 47 | 37 | 34 | 51 | 11 | 462* |
| Exception | 2 | 11 | 22 | 8 | 10 | 1 | 5 | 3 | 15 | 8 | 0 | 85 |
| (%) | (13) | (22) | (33) | (16) | (11) | (14) | (11) | (8) | (44) | (16) | (0) | (18) |
| Nonexception | 13 | 29 | 32 | 25 | 28 | 6 | 28 | 27 | 6 | 38 | 7 | 239 |
| (%) | (81) | (56) | (48) | (50) | (31) | (86) | (60) | (73) | (18) | (74) | (64) | (52) |
| Status 1 | 1 | 11 | 13 | 17** | 53*** | 0 | 14 | 7 | 13** | 5** | 4 | 138 |
| (%) | (6) | (22) | (19) | (34) | (58) | (0) | (29) | (19) | (38) | (10) | (36) | (30) |
| Exception | 0 | 5 | 0 | 4 | 27 | 0 | 5 | 2 | 5 | 2 | 1 | 51 |
| Nonexception | 1 | 6 | 13 | 12 | 23 | 0 | 9 | 5 | 7 | 2 | 3 | 81 |
| Admitted to ICU† | 1 | 3 | 14 | 14 | 25 | 0 | 11 | 4 | 5 | 3 | 3 | 83 |
| (% of transplants) | (6) | (6) | (21) | (28) | (28) | (0) | (23) | (11) | (17) | (6) | (30) | (18) |
- *Region is missing for one transplant.
- **Missing a status 1 classification of either standard or special for one case.
- ***Missing a status 1 classification of either standard or special for three cases.
- †Missing 12 patients ICU status.
Regional variation
The following analyses regarding PELD scores were computed from data in the PELD era and excluded patients allocated an organ as a status 1. The overall mean PELD score at allocation was 15.90 ± 14.43. The mean PELD score at allocation was statistically different between those utilizing a PELD score granted by exception (mean = 31.30 ± 8.88) and those utilizing a calculated PELD score (mean = 9.96 ± 11.51) (p < 0.0001). There was a statistical difference observed in the mean PELD scores between UNOS regions (p < 0.001). This regional variation was present when exception cases were considered separately (p = 0.004) and absent when only calculated cases were considered (p = 0.19) (Figure 1, Table 5).
Mean PELD scores at allocation by region among children with chronic liver disease after the implementation of the new organ allocation system (status 1 transplants excluded). Variability across regions for mean PELD score: Nonexception p = NS, Exception p = 0.004, All p < 0.001.
We further assessed the variability of patient status at allocation across UNOS regions during the PELD era by examining the relationship between the percentage of exception cases and the mean PELD score at allocation. Regions with a higher percentage of exception cases were more likely to require a higher mean PELD score at allocation (p = 0.001) (Figure 2, Table 5). Region 11 was not included in the above analysis because it did not have any exception cases reported in this data set.
Regional variation in exception cases and nonexception cases including mean PELD score (status 1 transplants excluded).
Figure 3 compares the percentage of patients admitted to the ICU at transplant to the percentage of patients that were status 1 at transplant between regions (Table 5). There was a significant difference between regions for both the frequency of admission to the ICU at time of transplant (X2= 22.93, p = 0.006) and status 1 at time of transplant (X2= 57.54, p < 0.0001). Region 6 was not included in the above analysis because it did not have a patient admitted to the ICU at transplant or a patient transplanted as a status 1 in this data set.
Patients admitted to intensive care unit and transplanted as status 1 in the PELD era. p = 0.006 (ICU). p < 0.001 (status 1).
Post-transplant outcomes
Kaplan-Meier survival rates were calculated for patient and graft survival at 90 days post-transplant. There were no significant differences between eras in overall patient and graft survival curves. The pre-PELD era and the PELD era 90-day patient survival rates were 94.8% and 95.4%, respectively and the 90-day graft survival rates were 89.0% and 89.4%, respectively. Patient and graft survival rates at 90 days were not significantly different between era when status 1 and nonstatus 1 patients were considered separately. For the status 1 patients, the pre-PELD and PELD era 90-day patient survival rates were 94.1% and 90.3%, respectively and 90-day graft survival rates were 86.4% and 83.4%, respectively. For nonstatus 1 patients the pre-PELD and the PELD 90-day patient survival rates were 95.2% and 97.5%, respectively and 90-day graft survival rates were 90.4% and 91.8%, respectively.
Discussion
This study focuses on children with end-stage chronic liver disease and examines how the MELD/PELD system has impacted their access to liver transplantation. The analysis describes how transplant centers use the allocation system to acquire organs for their patients and specifically examines regional variations in patient status and listing practices. The most important finding in this study is that during this early experience with the MELD/PELD system only 52% of children with chronic liver disease, who received a deceased donor graft, utilized a calculated PELD score for liver allocation. The remainder of the children accessed organs bypassing the PELD system either as status 1 candidates or by PELD exception.
This analysis reveals that a calculated PELD score is being used for allocation in only half of the pediatric population with chronic end-stage liver disease. With the current system, exceptions granted by regional review boards are not unusual. The exception process is used to obtain grafts for 30% of children, 18% via a PELD score and 11% as status 1 exceptions (data missing for 1%). These children are still utilizing relatively subjective parameters for organ allocation. Regional variation observed in the mean PELD scores at allocation and the percentage of exception cases show the system is not utilized uniformly across the country (18,19). Although some of the variability observed in mean PELD scores at allocation may be due to regional differences in severity of illness, these data suggest that the frequent use of exceptions plays a role as well. It is presumed that regional review boards in areas that are highly competitive for adult liver allocation would grant higher PELD scores for children with compelling cases. It is unclear whether granting higher scores or having a higher percentage of exception cases in a given region contributes to increasing the mean MELD/PELD scores required for allocation in that region. The purported strength of the MELD/PELD system is the use of objective parameters that are intended to avoid interobserver, interinstitutional, local, regional and national variability while benefiting the sickest patients. This early analysis suggests the system has not been used to its full potential.
The new system has substantially reduced the percentage of children with decompensated liver disease who require ICU support and who are status 1 at the time of transplant. The decline in the number of pediatric patients that require ICU care at the time of transplant appears to be the most positive early effect of the PELD system. This outcome may be the result of increased access to transplantation for the deteriorating patient. Since PELD was modeled to reduce both waiting list mortality and transfer to the ICU, in this respect the PELD system has performed as expected. However, the high rate of status 1 exception cases and the regional variation in status 1 exceptions could threaten this positive outcome by allowing some candidates to use nonstandardized criteria to move ahead of patients with high calculated PELD scores. This analysis also revealed significant regional variation in the percentage of children that are in an ICU at the time of transplant suggesting regional variation in acuity of illness. Future studies are needed to confirm these findings and address the determinants of these observed differences.
This examination of waiting list dynamics demonstrates a decrease in the number of new registrations after the implementation of the PELD system. Similar results have been reported for adults and children by UNOS and by Freeman et al. in an early analysis of the new allocation system (20,21). Since the new system minimizes the impact of waiting time on allocation, there is little advantage to ‘early’ listing of patients with low levels of medical acuity. This change in listing practices is probably the most important determinant in the reduction of median time to transplantation that was observed in the PELD era. Waiting list mortality for children appears to be unchanged by the new system. However, this analysis did not adjust for severity of illness at the time of transplantation and therefore comparison of waiting list mortality between the two eras has limited value.
A second objective of the ‘Final Rule’ was to improve patient and graft survival. Children with chronic liver disease have not experienced a survival advantage by implementation of the new system. This is not surprising since the PELD score was modeled to predict pre-transplant mortality and thus may not be a valid predictor of post-transplant outcomes. Analysis to establish the variables that predict post-transplant survival and incorporate estimates of post-transplant outcomes into the allocation scheme are important future objectives.
As the new system matures, future studies should reevaluate the current factors used to rank pediatric patients and investigate new parameters that could better be used to rank children on the liver waiting list. These studies should test the PELD score in disease specific populations and use post-transplant outcomes in an AUC ROC model. A previous study from our group demonstrated that the PELD score did not strongly predict post-transplant mortality, number of ventilated days or post-transplant length of ICU stay (22). Larger, prospective studies to examine PELD as a predictor of post-transplant outcomes are ongoing at this time. It is well accepted that young age at transplant and malnutrition impact outcomes in pediatric liver transplant recipients, especially those with biliary atresia (23,24). These parameters are included in the current PELD score. Other variables, not currently included, may be important predictors in other disease states. Analysis of general physiologic variables and measurements of pulmonary, cardiac and renal function may prove to be important indicators in children with more advanced liver disease. Thus, it becomes important to identify and include other variables in the allocation system if our ultimate goal is to improve post-transplant outcomes.
In summary, we have assessed whether the PELD system has improved liver allocation for children as measured by changes in recipient status and regional variation in listing practices. Overall waiting list mortality has not changed and there appears to have been no impact on post-transplant survival. The use of exceptions is common and may have an adverse effect on the intended functions of PELD. Regional variation suggests that PELD has not resulted in standardization of liver allocation in pediatric liver transplantation. Adjustments in the PELD score and elaboration of standard criteria for granting points by exception may improve the outcomes for children awaiting liver transplantation.
Acknowledgments
We would like to thank Roopa Seshadri, Ph.D. and Christine Sullivan, MBA, MS for their valuable assistance in statistical analyses and preparation of this manuscript. We also wish to thank Sarah Taranto and Maureen McBride, Ph.D. from UNOS for the important suggestions and for helping in the acquisition of the data. The data reported here have been supplied by UNOS as the contractor for the OPTN. The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as an official policy of or interpretation by the OPTN or the U.S. government.
This study was supported by the Siragusa Transplantation Center of Children's Memorial Hospital.
