Model for end-stage liver disease (MELD) for liver allocation: A 5-year score card

Liver allocation has always been controversial, mostly because the donor supply is inadequate to provide treatment for all candidates who can potentially derive benefit from the liver transplantation (LT) procedure. In the late 1990s, most of the controversy centered around extreme differences in waiting time among LT candidates in different geographic regions and the use of waiting time and other subjective variables that did not necessarily reflect patient disease to assign priority. In response to these concerns, the Institute of Medicine report recommended,1 and the Department of Health and Human Services Final Rule for the Organ Procurement and Transplantation Network (OPTN) stipulated,2 that liver allocation should be based on objective medical criteria with a de-emphasis on waiting time. On the basis of the Institute of Medicine report1 and other studies indicating that mortality risk was a much more appropriate measure for ranking waiting LT candidates,3 policymakers selected the Model for End-Stage Liver Disease (MELD) score as the measure of disease severity because it utilized objective, reproducible, patient-based variables only, and because it was validated as an accurate predictor of short-term mortality for most adult patients with chronic liver disease.4 The Pediatric End-Stage Liver Disease (PELD) score was developed as a measure of mortality for children with end-stage liver disease along similar lines.5 The OPTN formulated the new liver allocation policy to incorporate the MELD/PELD scores as a disease severity scale and virtually eliminated waiting time as a prioritization variable in the system.6 During the MELD policy genesis, the developers realized that not all candidates for LT suffer from diseases that carry an immediate mortality risk and that these patients would not be well served by a priority system based solely on a mortality risk endpoint. Patients with hepatocellular cancer (HCC) are the largest such group because many of these LT candidates have a risk of their malignancy progressing beyond a curable stage that is much greater than their immediate risk of dying of intrinsic liver failure.7 For this reason, a peer review system was developed to allow centers to request extra priority for patients with these so-called exceptional diagnoses. The OPTN implemented this MELD/PELD-based system for liver allocation on February 27, 2002. In comparison with the year prior to MELD allocation, during the first year under the MELD system, there was a reduction in new LT waiting list registrations with the largest reductions seen in new registrants with low MELD/PELD scores, a reduction in the waiting list death rate, and an increase in deceased donor transplants. The reduction in waiting list mortality and increase in transplantation rates were evenly distributed across demographic and liver disease etiology strata, with some variation across geographic variables. Early patient and graft survival after deceased donor LT was not changed.8 Over the ensuing 5 years, much attention has been focused on this new system, with more than 400 titles in the literature mentioning MELD and liver allocation. Many other countries have adopted, or are considering adopting, MELD-based liver allocation policies using many of the same principles developed by the OPTN.9 At this point in time, it is appropriate to assess the status of MELD-based liver allocation in the United States and identify areas for future consideration. One of the most significant changes brought on by MELD is that patients and the system can be much more easily measured. By the removal of subjective variables, patients' natural history, physician behavior, and the effect that these variables have on the overall results of the system can be more readily compared and quantified. This is essential for any critical assessment of healthcare delivery systems and allows for the definition of metrics by which such a system is to be judged. Although many domains of the liver allocation system might be examined critically, in this review, I will focus on three main areas in which a scorecard for the MELD policy can be tabulated: individual justice, population utility, and overall equity. Justice is defined as “the maintenance or administration of what is just, especially by the impartial adjustment of conflicting claims or the assignment of merited rewards or punishments.”10 In the context of liver allocation, measuring individual justice requires measuring the degree to which the system impartially and fairly allocates organs to those with “competing claims” (need for LT) and fairly assigns “punishment” (not getting the organ) to other waiting candidates. To do this, one can examine which patients are allocated the deceased donor livers (whose “claims” are rewarded) and which patients are removed from the list because of death or because they are too sick (“punishments”) and compare these probabilities among various demographic or diagnostic groups. The degree to which one or another group is over- or underrepresented in these probabilities provides a “score” for individual justice within the MELD system. Over the years of MELD-based liver allocation, distributions of liver transplant recipient probabilities have remained relatively stable (Fig. 1). The proportion of recipients in the 50–70–year age groups has increased over time, likely because of the general aging of the population more than any effect of the allocation system. The MELD era data displayed in Fig. 1 suggest that African Americans generally have higher probabilities of transplant than other groups, perhaps because they are initially listed at higher MELD scores than other ethnic groups. Although Hispanics tend to have lower transplant probabilities than other groups, all groups have experienced increasing probability of transplant at 1 year during the MELD system's time frame. There are no comprehensive published reports of ethnic or racial disparities in transplant or removal rates addressing the MELD era. Transplantation probabilities stratified by diagnostic groups during the MELD era suggest that pediatric candidates and those with diagnoses likely to qualify for MELD exception (malignancies, metabolic diseases, and other) have a higher probability of transplant than the standard MELD cases encompassed by cholestatic and noncholestatic diseases. These exceptional cases represent approximately 25% of all transplants under the MELD system and will be addressed later. Probability of transplant 1 year after listing stratified by (A) age groups, (B) race, and (C) diagnosis since initiation of the Model for End-Stage Liver Disease/Pediatric End-Stage Liver Disease system. AA indicates African American; BA, biliary atresia. Data were taken from the Organ Procurement and Transplantation Network www.optn.org (August 3, 2007). Balancing competing “claims” and “punishments” for all diagnostic groups who can potentially benefit from LT is another critical aspect of preserving individual justice in the LT allocation system. Data in Fig. 2 indicate that there is only a slight and fluctuating variation in risk of removal from the list for death/being too sick among the demographic and diagnostic groups. The youngest children and older LT candidates have higher removal rates, the former mostly because of the difficulties in finding suitable size-matched organs for children < 1 year of age and the latter because of progression of comorbid conditions and individual center practice regarding risk-taking in the elderly. Balancing individual justice for patients whose “claim” to LT is based on a risk of dying without LT is relatively straightforward because it is difficult to justify other claims of quality of life “punishments” superceding mortality “punishments”. Precisely because one can justify using mortality risk as a waiting list endpoint for these LT candidates, the MELD liver allocation system appears to have functioned well. Probability of removal from the list for death/being too sick 1 year after listing stratified by (A) age groups, (B) race, and (C) diagnosis since initiation of the Model for End-Stage Liver Disease/Pediatric End-Stage Liver Disease system. AA indicates African American; BA, biliary atresia. Data were taken from the Organ Procurement and Transplantation Network www.optn.org (August 3, 2007). For other types of liver disease for which the MELD score is not an applicable measure of LT need, establishing claim justification (or equal punishment) requires determining appropriate waiting list endpoints and methods for predicting achievement of these endpoints. In the case of patients with HCC, system developers chose the risk of progressing beyond Milan criteria11 as an allocation endpoint because of the ample evidence at the time that this defined an acceptable “claim” on the donor pool. More recently, however, several studies have indicated that restricting priority to HCC candidates within Milan criteria only may impose an unreasonable “punishment” on some patients who have larger tumors and who have had outcomes that are comparable to most other indications for LT.12-14 The HCC example points up the difficulties of balancing LT claims based on mortality risk against claims based on other definitions of need. Policymakers have tried to adjudicate this conflict in competing claims by identifying more definitive endpoints for these so-called exceptional diagnoses where MELD does not apply15 (Table 1). This remains problematic, however, because the allocation endpoints for the exceptional diagnoses are not well defined and the models defining risks of achieving these endpoints are far from developed. Thus, to assess the liver allocation system's performance in achieving individual justice requires that scorekeepers weigh claims for LT based on mortality risk compared with claims based on risk of progressing to higher risk for poor outcome. For example, providing broader criteria for HCC inclusion because of potentially justifiable claims will increase the punishments of higher waiting list mortality rates for patients prioritized by MELD. Current results suggest that the probability of removal for death/being too sick is higher for standard MELD cases than for exceptional cases, especially in 2004 and 2005 (Fig. 2), but in the case of HCC at least, many candidates with potentially favorable tumors do not receive priority because they fall outside Milan criteria. AA, African American; BA, biliary atresia; BCS, Budd-Chiari syndrome; BMI, body mass index; DSA, donor service area; FEV1, forced expiratory volume at 1 second; HCC, hepatocellular cancer; ICP, intracranial pressure; LT, liver transplantation; MAP, mean arterial pressure; MELD, Model for End-Stage Liver Disease; MPAP, mean pulmonary artery pressure; OPO, organ procurement organization; OPTN, Organ Procurement and Transplantation Network; PELD, Pediatric End-Stage Liver Disease; PVR, pulmonary vascular resistance; RRB, regional review board; RV, right ventricle; TIPS, transjugular intrahepatic portosystemic shunt; UNOS, United Network for Organ Sharing. Merriam-Webster's On Line Dictionary defines utility as “fitness for some purpose or worth to some end.”16 Many observers assess the utility of the allocation system purely by the results achieved after LT. However, as with any allocation of scarce resource problem, the fitness of the LT allocation system as a means to achieve a maximal patient benefit “end” requires that scorekeepers account for all patients entering the system, whether they receive a transplant or not.17 Thus, the overall survival benefit of LT, that is, the survival of patients who receive the transplant plus the survival of candidates who are not allocated organs, should be used as the measure of utility in this situation.18 In other words, the balancing of the justice claims and punishments must be accounted for in the calculation of overall utility. Under the MELD-based system, allocating livers to candidates with higher mortality risks who still have reasonable survival after transplant results in significant survival benefit, whereas providing transplants for patients with low MELD scores may actually increase their risk of dying compared with no transplant.17, 19 Even in the case of HCC, where untreated median survival of early-stage HCC is less than 2 years but median survival with LT is more than 7 years, significant survival benefit accrues. The change to a MELD-based system has put much more emphasis on survival benefit as a utilitarian measure. However, MELD does not completely define survival benefit after LT, and the use of this as the final utilitarian measure will require a thorough discussion of the potential negative effects of using factors such as age, which has profound survival benefit implications, and race, which carries deep legal and societal implications, on individual justice and fairness.20 Furthermore, if the allocation system's utility is defined purely with survival benefit calculations, patients with exceptional diagnoses with relatively low waiting list mortality may not have large increases in survival benefit over the first few years after entry into the system. Thus, MELD is not completely fit for defining the LT allocation end of survival benefit, and survival benefit may not be the one single utilitarian end for organ allocation overall. Resolution of these issue requires further societal discussions aimed at the tradeoffs between qualitative and quantitative benefits from the scarce donor resource. The MELD system of prioritizing waiting candidates defines the means by which candidates are ranked; it does not select the group of candidates over which this ranking system is to apply. That is, MELD does not affect the distribution of organs; it affects only the prioritization of patients within distribution units. The U.S. distribution system as currently defined by the local/regional/national unit definitions is based on artificial administrative boundaries. The country is divided into regions that, in turn, are divided by the donor service areas (DSAs) served by the organ procurement organization (OPO) operating within each region. Deceased donor livers are offered first to candidates within the local DSA of the procuring OPO, then to centers within the region in which the procuring OPO DSA resides, and finally to the national list. Recent changes in LT distribution require donor offers to patients with MELD scores ≥ 15 within a region first before offers to candidates with MELD < 15. This revised distribution policy is based on the finding that patients with MELD scores < 15 have a greater probability of surviving 1 year without a transplant than with one. This so-called Share 15 Policy has directed deceased donor livers to patients who are more likely to achieve greater survival benefit.21 This policy change illustrates how utility and individual justice in the allocation system can be affected by the distribution system under which the allocation policy operates. Geographic differences in deceased donor distribution can have profound effects on individual justice and/or utility and have often been cited as the most significant inequity in the system.22, 23 However, in the United States, the MELD-based priority system is overlaid on the organ distribution system, and the MELD allocation system per se does little to change the distribution divisions already established years before. Consequently, MELD alone could not have changed the distribution inequities inherent in the system. Nonetheless, scoring of the MELD system should include some assessment of the geographic issues because, as evidenced by the Share 15 Policy, organ distribution and allocation policy can be intertwined. In assessing geographic variations, one must note that there is a profound mismatch of donors and waiting candidates among the US DSAs. Figure 3 illustrates the almost 6-fold difference in candidates per deceased donor among the 63 DSAs and clearly illustrates why transplant and removal rates vary. Other effects of the differential mismatch of donors and candidates include variations in MELD score at transplant (Fig. 4) among the DSAs and in waiting list death rates. The size of the waiting list also affects MELD score at transplant.23 Liver transplant candidates per deceased donor by donor service area (DSA). Some DSAs have no liver transplantation centers. Data were taken from the Organ Procurement and Transplantation Network (2005). Model for End-Stage Liver Disease (MELD) scores at transplant among Organ Procurement and Transplantation Network regions for standard MELD and exceptional cases (January 1, 2003 through December 31, 2005). The left bar indicates standard MELD cases, and the right bar indicates exceptional cases. There are other contributors to disparities in LT results that are not necessarily directly a result of the MELD system but are much more readily measured and appreciated because of the increased objectivity inherent in the MELD system. For example, the MELD score provides for better candidate risk adjustment, so observed outcomes versus expected outcomes can be calculated. These results are computed biannually by the Scientific Registry for Transplant Recipients for every LT center in the United States and are publicly available.24 Figure 5 indicates that there is a broad range in the difference between observed and expected 1-year survival rates after transplantation among the US LT centers which also contributes to inequities that are not necessarily attributable to the MELD system directly. Difference between observed and expected survival for the 134 liver transplantation (LT) programs in the United States. Data were taken from http://www.ustransplant.org (accessed July 4, 2007). These geographic and other differences in utilitarian outcomes contribute to the widely cited inequities of the system much more than the MELD-based allocation. In fact, the MELD-based allocation system makes it possible to understand these other inequities much more objectively because the MELD score provides a much more reliable, reproducible, and transparent measure. The completion of a scorecard implies that “the game is over.” However, organ allocation policy, like medicine in general, should be viewed as a constant, progressive, and improving endeavor that can be concluded only when there is no need to prioritize because of resource scarcity. Until then, the MELD-based system has provided a more credible basis for assessing the system and will provide a more solid foundation on which to build future policy revisions aimed at reducing the remaining inequities.