Liver allocation policy in the U.S. was recently changed to a continuous disease severity scale with minimal weight given to time waiting in an effort to better prioritize deceased donor liver transplant candidates. We compared rates of waiting list registrations, removals, transplants, and deaths during the year prior to implementation of the new liver allocation policy (2/27/01–2/26/02, Era 1) with the first year's experience (2/27/02–2/26/03, Era 2) under this new policy. Rates were adjusted for 1,000 patient years on the waiting list and compared using z-tests. A 1-sided test was used to compare death rates; 2-sided tests were used to compare transplant rates. Overall and subgroup analyses were performed for demographic, geographic, and medical strata. In Era 2, we observed a 12% reduction in new liver transplant waiting list registrations, with the largest reductions seen in new registrants with low MELD/PELD scores. In Era 2, there was a 3.5% reduction in waiting list death rate (P = .076) and a 10.2% increase in cadaveric transplants (P < .001). The reduction in waiting list mortality and increase in transplantation rates were evenly distributed across all demographic and medical strata, with some variation across geographic variables. Early patient and graft survival after deceased donor liver transplantation remains unchanged. In conclusion, by eliminating the categorical waiting list prioritization system that emphasized time waiting, the new system has been associated with reduced registrations and improved transplantation rates without increased mortality rates for individual groups of waiting candidates or changes in early transplant survival rates. (Liver Transpl 2004;10:7–15.)
Prior to 1997, patients waiting for deceased liver donors were prioritized by medical criteria based on the location in which these candidates received their care (home, hospital, intensive care unit) and by waiting time. Over time, these location criteria became more a measure of physician practice patterns and behavior and not necessarily indicative of the medical condition of patients in need of liver transplantation. In 1997, the federally funded Organ Procurement and Transplantation Network (OPTN), administered by the United Network for Organ Sharing (UNOS), developed new categories of medical urgency for waiting liver transplant candidates that were based on the well-established Child-Turcotte-Pugh score for liver disease.* Within these medical urgency categories, waiting patients were ordered according to time on the UNOS/OPTN list. As this waiting list grew in size, the numbers of patients in these categories increased and their severity of disease became much more heterogeneous. Consequently, waiting time became the major discriminator for patients in these large categories and reduced the role of medical condition in determining priority on the list. Subsequent studies found that waiting time has no correlation with liver transplant candidate mortality1 and a pilot allocation plan that deemphasized waiting time reduced the number of waiting list deaths.2 Furthermore, as care of patients with endstage liver disease has improved it became apparent that the CTP score does not adequately segregate patients with progressively abnormal laboratory tests.
The Model for Endstage Liver Disease (MELD) score3, 4 was chosen as a superior disease severity scale for adult liver allocation. Unlike the CTP score used in previous allocation policy, the MELD score had been rigorously tested and validated as a good predictor of mortality in diverse groups of patients with various types and degrees of chronic liver disease,5 including candidates on the waitlist for liver transplantation.6 A separate mortality risk score for pediatric patients, the Pediatric End Stage Liver Disease (PELD) score, was developed using the 3-month mortality risk endpoint.7 These risk scores use objective verifiable criteria to rank candidates in a continuous fashion and waiting time is used only to rank patients with identical MELD/PELD scores within blood group. However, these risk scores do not serve all potential liver transplant candidates well. For some patients with diagnoses not likely to progress to a life-threatening stage, but for whom liver transplantation is indicated, the MELD/PELD score does not assign a high priority. Patients with early-stage hepatocellular carcinoma (HCC) make up the largest group of such candidates. Under the new system, centers may propose such candidates who meet specific entry criteria for Stage I or Stage II8 HCC disease or other diagnoses for higher priority on the waiting list through a peer review process. These requests are forwarded to Regional Review Boards (RRBs) to assess their appropriateness.
The new liver allocation system was implemented on February 27, 2002. The previous Status 1 designation was maintained and the continuous scale based on the MELD models replaced the previous Status 2A, 2B, and 3 designations.6 Inherent in the institution of any new policy is the commitment to evaluate its effects and assess the results. This report compares the first year's experience with the new MELD/PELD-based liver allocation policy to the previous year's results.
MELD, Model for End-stage Liver Disease; PELD, Pediatric End-stage Liver Disease; OPTN, Organ Procurement and Transplantation Network; UNOS, United Network for Organ Sharing; CTP, Child-Turcotte-Pugh; HCC, hepatocellular carcinoma; RRBs, Regional Review Boards; DSA, Donor Service Area.
Patients and Methods
Data from the OPTN database for liver transplantation candidates were collected from 2 time periods; February 27, 2001, through February 26, 2002 (Era 1), corresponding to the year prior to implementation of the new MELD/PELD allocation plan, and February 27, 2002, through February 26, 2003 (Era 2), representing the first year's experience with the new plan. The OPTN captures waiting list additions, modifications, and removals directly from the member transplant centers in real time through UNetsm, an internet-based data collection system. Demographic and clinical data on candidates and recipients are entered using online forms. 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. Pretransplant death data was augmented by linking waiting list removals other than for death or transplant to the Social Security Death Master File (SSDMF). Linking records were counted as waiting list deaths if the SSDMF death date occurred prior to the OPTN removal date or if a later SSDMF death date occurred 90 or fewer days from OPTN removal date. MELD/PELD scores were obtained from the UNetsm waiting list system where MELD/PELD data are recorded at registration and removal and linked to the online registration forms by way of unique patient identifiers. Mandatory entry of the laboratory data required to calculate the MELD/PELD score at candidate registration was begun in September of 2001. For this reason, there was insufficient MELD/PELD data from Era 1 to make valid comparisons of MELD/PELD scores at the time of waiting list removal. Except where noted, all MELD/PELD scores are calculated based on the laboratory values at the time point of interest and termed “calculated scores.” MELD/PELD scores assigned through the regional peer review process are termed “exceptional scores.” Rates for removals for death / too sick and transplantation were calculated based on the number of patients and their length of time on the list and expressed as cases per 1,000 patient-years. Because registrations do not have any time accumulated on the list, rates of registration were not calculated or compared. Comparisons between rates for demographic, clinical, and geographic strata for the 2 eras were performed using the normal approximation to the binomial distribution. A 1-sided Z-test was used for the comparison of death rates and a 2-sided Z-test was used for the comparison of transplant rates. The current UNOS/OPTN regions defined geographic strata.9 Posttransplant survival data were collected on the UNOS/OPTN liver transplant follow-up forms. At this early date complete follow-up data (94%) is available only for the 90-day follow-up period. Posttransplant survival rates were compared using Kaplan-Meier methods.
Significantly fewer liver transplant candidates were added to the waiting list after the implementation of the new MELD/PELD-based allocation system (Table 1). There were fewer registrations in the emergent Status 1 category and fewer registrations for both genders, all ethnic, and all diagnostic groups. All regions except Region 6 had fewer registrations in Era 2 under the new system (Table 1). There were fewer recipients relisted and fewer retransplants in Era 2 that may explain the reduced number of registrations in the Status 1 category (data not shown). Over the course of implementation of the new system, marked changes in the distribution of MELD score at listing were observed (Fig. 1). Once MELD submission became mandatory in September 2001, the number of newly listed candidates with MELD less then 10 dropped sharply and has remained low. Conversely, listing of patients with MELD greater than 20 has increased.
Table 1. Comparison of Registrations to the UNOS/OPTN Liver Transplant Waiting List, Era 1 vs. Era 2
Because the rate of additions to the list has no baseline rate to which comparisons can be made, no statistical comparisons were performed for these variables comparing Era 1 and Era 2.
Candidates removed from the waiting list because of death or becoming too sick are summarized in Table 2. There were fewer removals for this combined endpoint in Era 2 compared with Era 1. However, the reduction in the rate of removals per 1,000 patient years did not reach statistical significance (P = .076). Reduced rates of death/too sick were seen across age group and gender strata. Whites, Hispanics and Asians had reduced rates of removal for death / too sick under the MELD/PELD system with the reductions for the Hispanic and Asian groups reaching statistical significance. Black candidates had slightly higher rates of removal for the composite endpoint in Era 2. Regions 5, 7, and 9 had significant reductions in this endpoint under the new system.
Table 2. Comparison of Removals for Death/Too Sick From the UNOS/OPTN Liver Transplant Waiting List, Era 1 vs. Era 2
We observed a significant increase in the overall rate of deceased donor liver transplantation under the new system in Era 2 (Table 3). The increases were significant for adults and for males. The improved transplantation rate in Era 2 occurred as a result of an increase in the total number and rate of livers recovered and transplanted in Era 2, along with decreased absolute numbers of waiting candidates. Prior to the MELD/PELD system, annual deceased donor utilization rates have fluctuated between a high of 92.2% in 1990 and a low of 87.2% in 1994. Deceased donor utilization rates for 1996 through 2001 were 89.1, 87.8, 89.8, 89.5, 90.1, and 90.0%, indicating that the increased utilization rate of 92.5% in Era 2 was not just the latest result in a consistently increasing trend. White, Black, and Asian ethnic groups had significant increases in deceased donor transplantation rates under the MELD/PELD system. As expected, because the new system gave increased priority to candidates with hepatocellular cancer, we observed a significant increase in transplantation rates for candidates with malignancies. In addition, recipients with primary sclerosing cholangitis had a significant increase in transplantation rate in Era 2. OPTN/UNOS Regions 2, 7, 9, and 10 had significant increases in transplantation rates under the new system. Region 8 had a significant reduction in transplantation rate. Not surprisingly, because the MELD system gives higher priority to candidates with renal dysfunction (due to the fact that serum creatinine is one of the components of MELD), we observed a significant increase in the rate and number of combined liver/kidney transplants in Era 2 (Era 1: 7.36; Era 2: 12.24; P < .0001). Regional variation exists in the distribution of MELD scores at transplant (Fig. 2).
Table 3. Comparison of Patients Removed From the UNOS/OPTN Liver Transplant Waiting List for Reason of Cadaveric Transplant, Era 1 vs. Era 2
Twenty-three percent of deceased donor liver transplants were performed for candidates who received RRB approval for extra priority due to hepatocellular cancer, with another 5% of transplants performed for other exceptional diagnoses (data not shown). Eighty-two percent of the transplants for hepatocellular cancer were performed for patients for whom a request for the Stage II priority was made.
Early patient and graft survival rates are comparable among Eras 1 and 2 (Figs. 3, 4). Follow-up data is complete to 90 days in 94% of the cases in Era 2. The relative risk of mortality within 30 days of liver transplant is increased for candidates with MELD/PELD scores >25 and sharply increases for those with MELD/PELD scores >35, with these candidates having 2 times higher risk of death than those with MELD/PELD scores of 25 (Fig. 5).
Overall, the new liver allocation system seems to be functioning well. Although there is no indication that the prevalence or incidence of endstage liver disease is falling, registrations to the liver transplant waiting list are fewer since adoption of the MELD/PELD system. This is likely due to the fact that waiting time is no longer a criterion for ranking candidates. Removal of waiting time as a major factor for allocation obviates the need to list patients early to gain waiting time and reduces the number of relatively well patients on the list. Our data showing that all demographic, diagnostic, and regional groups had fewer registrations under the new system support that this is not just due to changes in indications for liver transplant, or regional changes in practice patterns. In addition, MELD/PELD scores at registration are increased under the new system across all strata, again suggesting that the system change has resulted in practitioners registering patients later in the progression of their liver disease. Supporting this view, we observed that among new registrants the numbers of patients with MELD scores less than 10 has dropped substantially. Despite higher MELD scores at registration and transplantation there has been a slight decrease in the number of re-registrations overall. This may explain the reduction in registration for Status 1 listings but more detailed analysis will be required to accurately dissect the cause for fewer Status 1 registrations. Our data suggest that practitioners are using the MELD score to estimate the mortality risk for their evaluees and more frequently registering only those with mortality risks that exceed the mortality risk of the transplant itself. Implications for this trend are many, but foremost is the possibility of developing minimal listing criteria based not on how long a patient may be able to wait, but on a MELD/PELD score indicative of a higher risk of death without a transplant than with one.
We did not observe a statistically significant reduction in overall waiting list mortality rates under this new system. These overall results may be perceived as disappointing, but they reflect the changing dynamic of the waiting list population. As the cohort of waiting patients shifts from one of relatively low disease severity with lengthy waiting time to one of relatively higher disease severity and less time on the list, increased mortality rates might be expected, especially if there is no increase in cadaveric donation rates. The fact that we saw a small reduction in mortality rates overall suggests that the system has provided some benefit. Patients who are referred later in the course of their liver disease or those receiving care where competition among centers drove listing of patients at earlier stages of their disease are now served more equitably by the elimination of waiting time. In the future, if trends toward reduction in less ill candidates on the list continue, and absent a dramatic increase in organ availability, it is likely that waiting list mortality rates will increase. It will be difficult to show reduced rates of mortality when this rate is based both on time and number of patients waiting, both of which are decreasing under this new system. Elimination of waiting time has improved organ access for the more ill patient compared with the old system, but over time this improvement is likely to be diminished because of the imbalance in organ supply and demand.
Increased transplantation rates under the new system likely reflect an increase in the number of cadaveric donors deemed suitable for transplantation and a reduction in the time waiting and number of candidates on the list. With a slight increase in the number of transplants performed and a decrease in the patient-years on the list, a significant rate increase was observed. Interestingly, the rate of liver utilization from deceased donors was significantly higher in Era 2. Many factors may explain this rate increase but a more accurate assessment of recipient need as defined by mortality risk under the MELD/PELD system may have influenced clinical decision making to improve matching high-risk organs with recipients likely to tolerate graft dysfunction. While geographic differences in the severity of disease (as measured by the MELD/PELD score) at the time of registration on the waiting list are very slight, the severity of disease among the regions at the time of transplant varies more widely. This is evident not only in the mean MELD/PELD scores among the regions at the time of transplant, but also in the distribution of MELD/PELD scores at transplant in these regions. When MELD/PELD scores at transplant are stratified by Organ Procurement Organization Donor Service Area (DSA), these differences are even greater in some areas of the country, although the number of transplants in many DSAs are small and limits definitive conclusions (data not shown). There are many important factors influencing these differences, as has been pointed out before,10 only some of which are affected by the allocation system. However, the implementation of this new system provides a much more objective measure of these differences because subjective assessments of medical condition are eliminated. This makes the system more transparent in that patients and centers in different geographic areas can be more easily compared using the MELD/PELD scores. Hopefully, these more objective measures of disease severity can be used to adopt policies to reduce the geographic variations noted in our data.
Hepatocellular cancer remains a difficult issue. With more than 16,000 Americans affected by this disease,11 this indication alone can potentially overwhelm the liver transplantation system. Relatively insensitive diagnostic imaging and a paucity of natural history data make it difficult to develop risk models for HCC progression that would assign priority for liver transplantation more scientifically. Emerging evidence suggests that smaller HCC tumors, which are more difficult to diagnose with current imaging techniques, have relatively low risk of progression to advanced stages, at least for the first year.12 These findings may provide sufficient evidence in the future for further modifications for prioritizing these candidates.
At this early stage of the new policy, posttransplant survival remains equivalent to experience under the previous allocation plan. The increase in the proportion of HCC recipients who have relatively less severe liver disease at the time of transplantation, and consequently better expected early posttransplant results, may be offsetting a decreased survival for the more severely ill recipients that did not always receive priority in the previous era. This is supported by the fact that candidates with the highest MELD/PELD scores have the highest risk of death (Fig. 5) and higher risks of graft failure (data not shown) relative to candidates with MELD/PELD scores of 25. These increased rates of failure need to be balanced with the extremely high risk of death for these candidates if they are not given transplant. In contrast, however, our data showing more overall transplants with fewer relists and fewer retransplants suggests that, even though there may be more severely ill candidates receiving transplants now, there has been no reduction in efficient use of the donor pool. Of course, much longer follow-up times will be required to more accurately assess the success of liver transplantation under this new allocation plan and future debate will need to adjust the appropriate balance between candidate mortality and recipient survival.
The implementation of the MELD/PELD-based continuous disease severity scale represents a dramatic paradigm shift in liver allocation policy. The MELD/PELD system provides a much more transparent and measurable framework on which to build future refinements. Nonetheless, the system is not perfect and the waiting list continues to evolve and change. The results reported herein represent only a starting point in this evolutionary process. We have outlined some future issues that must be addressed. Until there is no longer an organ shortage, there will be a constant need to adjust, improve, reevaluate, and measure the system using the evidence-based approach that drove this initial plan.