The new liver allocation system, based on the Model for End-Stage Liver Disease (MELD) scores, has been in effect for more than a year. The goal of the new system was—and continues to be—to use much more objective measures of liver transplant need to prioritize candidates with indications for liver transplantation. The system's architects chose to define liver transplant need as the risk of death from liver disease within 3 months because several studies showed that the MELD score was an excellent predictor of this endpoint and because waiting time did not correlate with the risk of death on the list. A major advantage of the MELD score is that it only requires 3 routine objective laboratory values to achieve this end. However, during the development phase, the policymakers recognized that there were some indications for liver transplantation that were not based on progressive intrinsic liver disease and that the MELD score was not completely accurate in predicting mortality risk for all cases. Rooted in this recognition is the fact that need for liver transplantation is not completely defined by mortality risk. For example, patients with early-stage hepatocellular carcinoma (HCC) are considered to be good candidates for liver transplantation because a large number can be cured if the transplant is performed before the HCC advances to a later stage. Thus, the need for liver transplantation in these cases is not based on the risk of death within 3 months but rather is based on the risk of progression of the HCC. There are other diagnoses, such as familial amyloidosis, other metabolic liver diseases, and hepatopulmonary syndrome, for which the MELD score may not be a good measure of transplant need.
To accommodate these so-called exceptional cases, and to allow an alternative pathway for clinicians to request increased priority for any liver transplant candidate judged to have an increased need, an alternative method for assigning priority was developed. This alternative pathway utilizes a regional peer review system consisting of regional review boards (RRB) that review all cases in which a request is made for increased priority by the treating liver transplant team. For all of these, prospective RRB approval is required before the requested increase in priority is assigned.
At the outset, the liver allocation policy provided guidelines for the review process to help requesting centers choose appropriate MELD/PELD scores. The original policy did not sanction specific diagnoses, symptoms, or conditions as being appropriate for exception priority requests, but it did specify a few conditions for which some criteria for the appropriate selection of transplant candidates with these exceptional diagnoses existed. In addition, because the MELD score had been validated against several patient cohorts with and without subjective clinical variables of liver disease (ascites, encephalopathy, variceal bleeding, spontaneous bacterial peritonitis), emphasis was placed on the fact that these variables were accounted for by the MELD score and that increased priority for candidates with these findings was not necessary.
One of the most important aspects surrounding the introduction of this new system was the commitment to assess effects and results and change policy based on these analyses. The report by Voigt et al. in this issue of Liver Transplantation is important because it is an attempt to address the effects of the alternate priority-setting pathway within the new system. Their approach is an interesting one: they have separated the candidates for whom a request for increased priority was made into those who were approved by the RRB and those who were denied. In addition, they have stratified these cases according to the diagnosis generating the request. Although the authors incorrectly use the term “sanctioned” to describe certain diagnoses (no diagnoses were sanctioned or nonsanctioned in the Organ Procurement and Transplant Network OPTN/United Network for Organ Sharing (UNOS) policy),3 they have identified cases in which the MELD score may not apply. Their exclusion of patients with portopulmonary hypertension, stage 3 and stage 4 HCC, cholangiocarcinoma, neuroendocrine tumors, adenomas, hemangiomas, and polycystic disease from the “sanctioned” category should not be justified solely because these diagnoses were not mentioned in the exceptional-case guidelines in the UNOS policy. In fact, the increased rate of RRB denials for the “non-sanctioned” cases as defined by Voigt et al.5 may be a reflection of the lack of guidelines in the UNOS policy rather than a more objective discernment of need by the RRB members. One could argue that all of the tumor and metabolic/structural cases, in addition to the “sanctioned” cases as defined by the authors, may deserve priority for liver transplantation based on variables other than the risk of 3-month mortality from intrinsic liver disease as defined by the MELD score. In most of these cases, patients face progression of malignant or benign disease that is unrelated to their hepatic synthetic function. To be fair, there were only 127 of these cases in the entire cohort; the vast majority of the “non-sanctioned” cases (79%) comprised diagnoses of cirrhosis or liver failure. Inclusion of these 127 in the “sanctioned” category may not have made a difference, but the distinction between mortality risk and risk of some other clinical event for priority endpoints is an important one.
Voigt et al.5 have found some important results of the RRB process. A significant number (46.6%) of the patients denied increased priority by the RRBs still went on to receive transplants, and waitlist mortality rates were not different for the approved or denied cases. Thus, denial by the RRB does not necessarily restrict access to transplant treatment but merely results in a lower probability of transplantation relative to those approved by the RRB. Also interesting is the finding that, for the “nonsanctioned” group, cases that were approved by the RRB had a higher waitlist mortality rate than those that were denied. As the authors point out, this suggests that the RRBs were able to identify patients who were much more likely to die while waiting. This would further suggest that there may have been factors in addition to the MELD score that informed the RRB reviewers' decision to agree to a higher mortality risk for the approved cases than that determined by MELD score. On the other hand, for those “nonsanctioned” cases denied by the RRB, the calculated MELD score was a better predictor of mortality risk than the risk determined by the requested MELD score.
However, the fact that there was little correlation between the requested MELD score–or “delta MELD”–and mortality risk is not surprising. As mentioned above, the purpose of the RRB process overall is to allow patients who have a low mortality risk from synthetic liver disease, but for whom there is some other compelling reason to accelerate the probability of transplantation, to have that opportunity. Therefore, for cases reviewed by the RRBs, one would not expect that the requested MELD score would have any relationship to mortality risk. For example, a MELD score of 29, initially arbitrarily assigned to candidates meeting stage 2 HCC criteria, does not indicate that these patients have a 30% mortality risk. It was assigned as an estimate (in retrospect, an overestimate) of the risk of progression beyond stage 2 disease, and one would not expect a close correlation in these patients between the assigned, exceptional MELD score and their mortality risk. It is noteworthy, however, that for the liver failure cases for whom upgrade requests were made (usually because of symptoms shown to not improve the predictive accuracy of the MELD model4), one would expect the calculated MELD score to correlate closely with mortality risk, as has been documented in other studies.4
The current liver allocation plan actually uses a mixture of endpoints to define transplant priority. As mentioned, HCC prioritization is based on an estimate of the risk of progression beyond stage 2 disease. This endpoint is not well defined because there are no precise predictive models for defining this risk, and the diagnostic accuracy of the studies used to determine eligibility for increased priority is poor. For other exceptional diagnoses, other endpoints can be envisioned. For example, patients with familial amyloidosis have progressive neurologic and myocardial deterioration that is independent of synthetic liver dysfunction. The risk of death in these patients, and their need for liver transplantation, is not defined by their MELD score and is more accurately measured by their risk of deterioration to an unacceptable operative risk. Risk of progression to an unacceptable operative risk can also be used to define need for patients with hepatopulmonary syndrome. More subjective variables might be included in such models, such as risk of progression to malignancy for hepatic adenomas or risk of progression to narcotic dependence for bulky neuroendocrine or polycystic liver disease. Along the continuum, these endpoints become more subjective, are less well defined, and begin to become measures of the risk of quality-of-life deterioration more than measures of mortality risk. In fact, one could imagine using the risk of deterioration to an unacceptable quality of life as the endpoint for all liver candidates if such an endpoint could be precisely defined for all of the various diseases thought to be treatable with liver transplantation and if a mathematical model of sufficient predictive power could be constructed. This leads to the question of fairness as suggested by the article's title.
Ultimately, the allocation of scarce resources requires some measure of fairness. However, this is an ill-defined term. One can equate fairness with individual justice. In this case, the allocation of scarce resources is based on what is just for the individual without much weight given to the overall benefit to the population at risk. In a system weighted toward individual justice, one would need to define which individual attributes are most important for determining who receives the scarce resource. This then raises some important questions: Is it justified to give priority to candidates with a high risk of deterioration in quality of life and a low mortality risk over those with a higher mortality risk? Are there ever circumstances in which the risk of quality of life deterioration should supercede the risk of death? Should risk of progression to an unacceptable operative risk supercede the risk of progression of malignancy when the operative risk remains low and constant for the candidate with cancer? Ultimately, we have to ask: Is it fair to prioritize candidates to receive the extremely constrained donor liver resource based on nonmortality endpoints, especially when they are not well established, as we are asking our RRBs to do now?
Of course, individual justice is not the only measure of fairness for the liver allocation system. Utility, or the net result for the entire population, must also be considered. In this sense, the liver allocation system must provide for responsible stewardship of the scarce resource. Refining the system to be more objective and evidence-based is one essential ingredient to this end. For those areas where there is a lack of evidence, peer review, consensus of opinion, and review of results are important methods for enhancing utility.
Is the RRB process fair? It depends on how the viewer perceives that the measures of individual justice are balanced with systemic utility in the liver allocation system. At least analyses like the one by Voigt et al. provide methods for and insight into how these questions might be answered in the future.