Evidence-based evolution of the MELD/PELD liver allocation policy

To improve is to change. To be perfect is to have changed a lot. Sir Winston Churchill The implementation of the Model for End-Stage Liver Disease / Pediatric End-Stage Liver Disease (MELD/PELD) allocation system in February 2002 launched a major paradigm shift in how deceased donor organs are allocated in the United States. Previous allocation schemes were based primarily on waiting time, subjective clinical parameters such as ascites and portosystemic encephalopathy, and the location of the patient (i.e., intensive care unit, continuous hospitalization, and home). These previous schemes all failed to appropriately prioritize allocation of donor organs to the sickest patients because of the subjective nature of the parameters and the ability to easily manipulate the variables in favor of an individual patient. In addition, the US government dictated in the Final Rule Mandate that the United States allocation policy should use objective medical criteria that reflects disease severity to prioritize patients for donor allocation.1 Thus, the United Network for Organ Sharing was charged with the development of more objective measures of liver disease severity, which was to be the basis of future liver donor allocation. MELD, model for end-stage liver disease; PELD, pediatric liver disease. The MELD model originally developed to predict survival rates in patients undergoing transjugular intrahepatic portal caval shunt procedure was based on three simple biochemical variables (creatinine, bilirubin, and international normalized ratio).2 These 3 simple laboratory tests are readily available, highly reproducible, and inexpensive. In addition, when used within a weighted formula, the MELD model has been shown to be highly predictive of short-term mortality in both retrospective and prospective studies.3, 4 The MELD model is consistent across diverse etiologies of chronic liver disease, and has been found to be fairly independent of the complications of portal hypertension such as severity of ascites, portosystemic encephalopathy, spontaneous bacterial peritonitis, and variceal bleeding.3 Of importance, the model is also free of using politically charged variables such as recipient age, gender, race, and transplant center, all of which have been shown to be predictive of both pre- and posttransplant outcomes, but which can not be readily used for our national donor allocation policy. Finally, the MELD score has been shown to correlate well with other liver function tests, and has been shown to be significantly better than the Child-Turcotte-Pugh score in predicting short-term survival.4, 5 The impact of using the MELD for allocation system in the United States was a 12% reduction in liver transplant waiting list registrations, particularly in those patients having MELD scores of less than 10.6 In addition, there was a 3.5% reduction in waitlist mortality, and an increase in transplantation rates, which were evenly distributed across all demographic and epidemiologic strata. Furthermore, early patient and graft survival after deceased donor liver transplantation remained unchanged despite the fact that we were transplanting sicker patients. In addition, the simplicity and objectivity of the MELD model allows for close scrutiny and continued assessment regarding its impact on liver allocation, waitlist death, transplantation rates, and outcomes of liver transplantation. It also allows for easy assessment of allocation disparity between regions, organ procurement organizations, and individual centers. While MELD has been a major advancement in liver allocation policy, like all mathematical models, MELD is not perfect. In its development and during its use for liver donor allocation, it has been shown that the MELD score will not accurately predict mortality in approximately 15% to 18% of patients with chronic liver disease.4 Patients with intrinsic renal disease, patients with hyperbilirubinemia secondary to Gilbert's syndrome, and patients on anticoagulant therapy with warfarin (brand name Coumadin) for thromboembolic disease, are inappropriately advantaged beyond the severity of their liver disease. In addition, recent studies have shown variability in measuring international normalized ratio in various laboratories performing this test.7 Furthermore, there has been a perception by a number of clinicians who have identified patients with refractory ascites in whom it is felt the MELD allocation system seems to underestimate mortality. Such a subgroup of patients was recently pointed out by Heuman et al.,8 who identified a group of patients with MELD scores less than 21, persistent ascites, and low serum sodium levels, who had a higher mortality than predicted by their MELD score alone. In this issue of Liver Transplantation, Ruf et al.9 demonstrate that the addition of serum sodium to the MELD score identifies a subgroup of patients with ascites, who have a significant increase in wait list mortality. These authors further demonstrate that the risk of death across all MELD scores is higher for patients with hyponatremia as compared to patients without hyponatremia.10 The study suggests that hyponatremia appears to be an earlier and more sensitive marker than creatinine to detect renal impairment and / or circularly dysfunction in patients with advanced cirrhosis. This study was based on 262 consecutive adult patients with decompensated cirrhosis listed for liver transplantation at a single center in Buenos Aries, Argentina between June 1995 and January 2003. The MELD score was calculated using retrospective data collected sometime during the pre-liver transplant evaluation but not always at the actual time of listing. In this patient cohort, 34 had hyponatremia (serum sodium < 130 mEq/L), and 68 underwent liver transplant during the initial 3 months of follow-up. One downside to the study was that there were only 19 deaths, making the study less than definitive. In addition, only 25% of the patients studied had hepatitis C as the underlying etiology, compared to over 45% of patients on the present day United Network for Organ Sharing waiting list. Thus, the study population in regard to etiology is somewhat different than that seen on the US waiting list. Of importance, the paper confirmed that MELD is an excellent predictor of 3-month mortality as indicated by a C-statistic at 0.894, which is even better than that achieved using the United Network for Organ Sharing data. Yet despite MELD's astounding ability to predict waitlist mortality, the incorporation of serum sodium into the MELD formula significantly further increased the C-statistic for 3-month mortality due to the ability of hyponatremia to identify a subgroup of high risk patients who died with lower MELD scores. Indeed, these findings fit well into the overall goal of the MELD / PELD allocation policy. This new policy was based on a considerable amount of evidence that the MELD risk score accurately predicts risk of death in patients waiting for liver transplantation. However, inherent in the implementation of the new policy, was the commitment to constantly evaluate the policy's results and update the scheme as necessary as new findings and developments occurred. Therefore, if prospective studies can confirm that hyponatremia will enhance the ability of the MELD model to predict survival in patients with ascites, serum sodium could be easily added to the present MELD allocation model. Indeed serum sodium—like serum bilirubin, creatinine, and international normalized ratio—is a laboratory value that is objective, readily available, reproducible, and has already been shown in independent studies to be a measure of liver disease severity and prognosis.10 Thus, additions and adjustments of MELD variables have been a long-term goal in our quest for evidence-based medicine to maximize the outcome of patients waiting for liver transplantation. At the present time, serum sodium levels are being prospectively collected by the United Network for Organ Sharing to further assess these preliminary findings. If confirmed, serum sodium would be added to the MELD model for prioritization of liver donor allocation. In summary, we believe that it is important that the liver allocation policy continue to evolve based on objective data using an evidence-based approach. Ultimately, the goal is to reduce waiting list mortality, maximize long-term outcomes, and optimize the use of our scarce donor organ resource.