Allocation of donor livers — is MELD enough?

The success of liver transplantation as an effective and cost-effective means of increasing both the length and quality of life of people with end-stage liver disease has brought its own problems. The lack of effective liver support and the shortage of cadaveric donor livers have meant that the process of liver allocation is, for the patient, quite literally a matter of life and death. Any system has to be, fair, ethical, and based on objective evidence before it can be widely accepted. In 2002, after extensive public debate, UNOS (the United Network of Organ Sharing) adopted a new approach to organ allocation based on giving priority to the sickest patient. Assessment of disease severity was based on the MELD (model for end-stage liver disease) score. The MELD score was developed initially to predict survival in patients undergoing transjugular intrahepatic shunts. Based on just three laboratory variables (international normalized ratio [INR], bilirubin, and creatinine) the model has proved remarkably robust in predicting short-term survival, and confounding those skeptics and critics (including myself) who did not believe that one model fits all types of liver disease.1 The model has many strengths, not least that these variables are all easily measured and, unlike for example the Child-Turcotte-Pugh score, these is no subjectivity in the components of the score. Of course, the three components may be affected by nonhepatic factors (for example, the serum bilirubin will be inappropriately increased by the hemolysis and the INR will be affected by vitamin K deficiency) and there remain significant variations in how laboratories measure both the INR and serum creatinine; nonetheless population studies have shown that these considerations do not affect the model's power to predict survival in the population. The model does not appear to be as reliable in those with malignancy in the context of cirrhosis, nor in those with metabolic diseases, nor does the model take into account, quality of life.2 As always, with any model, extrapolation to the individual will always need to be tempered by common sense and the system of allocation can be (and has been) modified to incorporate these limitations; for example, those with hepatocellular cancer will be allocated additional points. There have been numerous reports, many in this journal, evaluating the impact of liver allocation on the MELD era3 and the conclusions are consistent: introduction of the MELD (and its corresponding PELD for children) has greatly reduced the mortality of patients on the waiting list without significantly altering the outcome post-transplant.4, 5 Prognostic models have both strengths and weaknesses and they should be used with caution. Good prognostic models will accurately identify those variables that are associated with prognosis and give an estimate of the strength of the association. While most of these variables will be apparent intuitively, sometimes surprises are uncovered: these may be statistical quirks, surrogate markers, or genuine factors that give rise to new insights. Prognostic models will give an objective prediction of outcome and can be applied widely. By giving an estimate of survival probability, assessment of interventions can be made and comparisons of outcomes between centers can also be provided; where differences are uncovered, explanations sought. Limitations are present: while such models will give information for the population, application to an individual may be misleading or inappropriate. Models can only incorporate those variables that are assessed and some important prognostic variables omitted, because the data were not collected, cannot readily be measured, or just not considered. Furthermore, models based on retrospective data can take no account of developments in diagnosis and treatments and inappropriate application may lead to inappropriate conclusions. Thus, while prognostic models are an invaluable aid in the management of patients but will of necessity be an aid and not the sole basis on which to make clinical decisions. UNOS, United Network for Organ Sharing; MELD, model for end-stage liver disease; PELD, pediatric end-stage liver disease. There are several important dimensions required in selecting the optimal approach to organ allocation. Ethical issues: the ethical issues surrounding the process of organ allocation are complex and beyond the scope of this editorial. The conflicting and competing priorities of justice, equity, fairness, and utility need resolution as far as possible. Utility: because liver transplantation is expensive and, because of the shortage of donor livers, a rationed procedure, practical issues need to be balanced with ethical ones. Thus, both the resource utilization and the outcome of patients following liver transplant should also be considered in the allocation process. Resource utilization and outcome must consider both patients needing liver transplantation (and not necessarily on the waiting list) and those who are fortunate enough to receive a liver graft. It is uncertain to what extent the gain in survival and the absolute survival need to be balanced — sickest patients are likely to have the lowest post-transplant survival but also the greatest gain — should 1, 5, or 10 year survival be considered? Patient and graft outcome, both short and long-term, is dependent on many factors — donor, recipient, and technical.6 Assessing the interaction between these three factors may be difficult. While most surgeons can usually effectively identify a marginal liver, it has been difficult to define what constitutes a marginal (or extended indication) liver. Clearly, there is a continuum from a high quality graft to a marginal liver, and to assume a simple dichotomy is probably unjustified. When selecting the appropriate recipient (given the size and blood group issues are resolved), surgeons will usually use a marginal liver only in a relatively healthy recipient. Thus, those studies that analyze outcomes will inevitably be limited since the matching of donors and recipients is not random. It is intuitive that the condition of the recipient and donor graft will have an impact on the short-term outcome after transplantation, whereas other factors, such as age, co-existing disease, disease recurrence, and the unwanted consequences of immunosuppression will impact on longer term outcome. However, as is clear with recurrent hepatitis C viral infection, the graft (as donor age) is a major factor in determining graft survival in the medium term. MELD score pretransplant is loosely associated with 1-year outcomes in cadaveric donor transplantation, living donor transplantation, and retransplantation.7–10 The current evidence suggests that only when the MELD score exceeds 25 is survival post-transplant reduced7; conversely, when the MELD score is below 11, the available data suggest that liver transplantation is associated with a reduction in short-term survival; this increased risk of death may be a reasonable risk for effective treatment of symptoms of liver disease such as intractable encephalopathy.11 The three publications12–14 included in this volume of Liver Transplantation approach, from different aspects, the correlations between pretransplant MELD score and outcomes. It is worth stressing that all three papers are dependent on the free availability of data from large national registries (in North America, UK, and Canada). While it is acknowledged that there are potential flaws in the use of large data sets from national registries, the potential for studying large numbers of patients allows for valuable conclusions to be drawn with confidence. Roberts and colleagues12 used the UNOS database to analyze disease-related outcomes in over 17,000 adult liver allograft recipients, grafted between 1990 and 1996. They conclude that survival is a function of both disease and severity and severity of disease. This study, as other similar analyses before, confirmed that outcome is dependent on indication, with 5-year survival being greatest for patients grafted for primary biliary cirrhosis and for primary sclerosing cholangitis; the MELD score at transplantation was associated with outcome but not by much, varying by just 7% between those with a MELD score less than 10 and greater than 24. When disease specific models were considered, significant differences in prognostic variables and their importance was identified. These models add further to some of the models already published.15 Similar conclusions were reached by Jacob and colleagues13 who found that the ability of the MELD score to predict short-term outcome (at 90 days) was poor (the c-statistic was just 0.58). The burden of disease, the approach to organ donation, and the system of health care delivery are different between the US and the UK, so how far can lessons learned in the US be extrapolated elsewhere? The UK has adopted a different approach to organ allocation, using an approach based mainly on utility rather than need16: for those patients who are in need of a liver graft, transplantation is offered when there is a greater then 50% probability that the recipient will be alive and well 5 years after transplantation, with a quality of life acceptable to the person. For other than those with fulminant hepatic failure (who have priority over all), organs are offered initially to the center that will select the appropriate recipient. The center will also decide whether to use the graft as a whole, reduced, or split graft. If there is no suitable recipient, or if there are logistical issues (for example, the team is already transplanting) then the organ will be offered elsewhere. Stell and colleagues14 examined the US, UK, and Canadian databases assessing patients with chronic liver disease transplanted in the year 2000; they reported that, while the US and Canada were similar with respect to the severity of disease at transplantation (assessed by the MELD score, 17 and 16 respectively), the median MELD score of patients grafted in the UK was significantly lower at 11. There were fewer UK patients listed for transplantation but the real need for transplantation is not known so simple comparisons may not be helpful. In the UK, fewer patients were transplanted from the ventilator and fewer patients died or were withdrawn from the waiting list and the waiting time was shorter. This difference in disease severity did not translate into better survival for the UK patients. In contrast, the median MELD score of the cohort of UK patients grafted between March 1994 and September 2002 is similar to that in North America at 16. Thus, these findings are in keeping with the previous observations that the severity of disease is a relatively poor predictor of outcome in those with chronic liver disease, with graft failure and acute liver failure.2, 17, 18 The transplant community can, with every justification, congratulate itself on the successful outcomes after transplantation. We need to move away from assessing short-term outcomes and consider long-term outcomes, quality of life and, inevitably, costs. Introduction of the MELD score as a basis of allocation has resulted not only in a welcome reduction in the mortality of patients on the waiting list but an objective and validated system for allocation of donor livers. Now the system is in place, it may be appropriate to consider including a measure of outcome as well as urgency for listing in the allocation of organs.