Telephone: 613 9496 5468; FAX: 613 9458 1650
Liver transplant recipient selection: MELD vs. clinical judgment
Article first published online: 24 MAY 2005
Copyright © 2005 American Association for the Study of Liver Diseases
Volume 11, Issue 6, pages 621–626, June 2005
How to Cite
Fink, M. A., Angus, P. W., Gow, P. J., Berry, S. R., Wang, B.-Z., Muralidharan, V., Christophi, C. and Jones, R. M. (2005), Liver transplant recipient selection: MELD vs. clinical judgment. Liver Transpl, 11: 621–626. doi: 10.1002/lt.20428
- Issue published online: 24 MAY 2005
- Article first published online: 24 MAY 2005
Minimization of death while waiting for liver transplantation involves accurate prioritization according to clinical status and appropriate allocation of donor livers. Clinical judgment in the Liver Transplant Unit Victoria (LTUV) was compared with Model for End-Stage Liver Disease (MELD) in a retrospective analysis of the LTUV database over the 2-year period August 1, 2002, through July 31, 2004. A total of 1,118 prioritization decisions occurred. Decisions were concordant in 758 (68%), comparing priorities assigned by clinical judgment with those assigned by MELD, P < 0.01. A total of 263 allocation decisions occurred. Decisions were concordant in 190 (72%) and 203 (77%) of the cases, comparing donor liver allocation with prioritization by MELD and clinical judgment, respectively. Of the 52 patients allocated a liver, only 23 would have been allocated on the basis of MELD while 29 had been prioritized on the waiting list in the week prior to transplantation. A total of 10 patients died on the waiting list in the 2-year period (annual adult waiting list mortality is 9.3%). Patients who subsequently died waiting were 3 times as likely to be prioritized by MELD as clinical judgment (29% vs. 9%, respectively). One half (3 of 6) of the patients who could have received a donor liver but who died waiting would have been allocated the organ on the basis of MELD. In conclusion, an allocation process based on MELD rather than clinical judgment would significantly alter organ allocation in Australia and may reduce waiting list mortality. (Liver Transpl 2005;11:621–626.)
Death in patients waiting for liver transplantation occurs largely due to the imbalance between supply and demand. The annual liver transplant waiting list mortality reported by the United Network for Organ Sharing (UNOS) for 2002 was 10.6%.1 Management of liver transplant waiting lists is aimed at minimizing waiting list deaths by prioritization of those on the waiting list most likely to die, and ensuring allocation of available organs to these patients. These prioritization and allocation decisions require the repeated reassessment of patients as their clinical condition changes over time and comparison of their current state with other competing patients.
Each of the 6 liver transplant units in Australia and New Zealand maintains its own liver transplant waiting list and determines its own priorities for transplantation. In the absence of extremely urgent (category 1) patients, livers from donors within the procurement region of each unit (New South Wales and Australian Capital Territory for the Sydney Unit, Victoria and Tasmania for the Melbourne Unit, Queensland for the Brisbane Unit, New Zealand for the Auckland Unit, South Australia and Northern Territory for the Adelaide Unit, and Western Australia for the Perth Unit) are first offered to that unit and, if declined, to each of the other units on a strict rotation.2 The liver transplantation system in Australia thus resembles a series of single-center / single-organ procurement organization groupings.
The goals of waiting list prioritization and organ allocation in the Liver Transplant Unit Victoria (LTUV) are to minimize waiting list mortality and to avoid futile transplantation, while taking into account the likelihood of long-term survival in individual patients. Thus, patients are selected on the basis of medical urgency, while patients not expected to survive the transplantation process are removed from the waiting list. In addition, the age of the recipient and medical comorbidities are taken into account in contrast to a purely objective scoring system such as the Model for End-Stage Liver Disease (MELD). The system of waiting list prioritization in the LTUV involves a weekly meeting to assign clinical priorities. The patient list is presented, grouped by blood type, and records disease, age, activation date, weight, height, and girth. The MELD score is not displayed. At the meeting, each patient is assigned a clinical priority. This priority is represented by a number of stars, 0 for lowest priority to 3 for highest priority. The clinical priority is based on frequent clinical review of the patient by the 2 physicians (P.A. and P.G.) and represents the subjective assessment of the severity of illness and the risk of dying on the waiting list, while taking into account the risk of dying posttransplantation. It has been agreed within the unit that all pediatric patients will be assigned the maximal clinical priority of 3 stars. Clinical priorities determined at the meeting are printed on the waiting list and are available at the time of allocation of donor livers. Allocation is generally performed by the surgeons (R.J., R.B., and M.F.) in consultation with the physicians. Clinical priority is an important issue in the allocation decision, but a number of other factors are also taken into consideration, including donor / recipient size match, donor / recipient age match, donor age in relation to disease (for example hepatitis C), donor quality vs. recipient stability, and logistical issues.
The MELD score is a continuous points system, using objective criteria, originally developed to estimate mortality in cirrhotic patients undergoing transjugular intrahepatic portosystemic shunt procedures.3 It has subsequently been validated for its ability to predict mortality in a range of types of liver disease, including patients awaiting liver transplantation.4, 5 The system of organ allocation in the United States has recently been overhauled to deemphasize waiting time and take more account of clinical severity, using the MELD score to assign priority since February 2002.6 This has resulted in a 12% reduction in waiting list registrations, particularly patients with low MELD scores, and a 3.5% reduction in waiting list death, while patient and graft survival rates have been maintained.7
The aim of this study is to compare the current system of waiting list prioritization and organ allocation in the LTUV with the MELD scoring system.
Patients and Methods
The status of the clinical priority of each patient documented in the minutes of the weekly prioritization meetings for the 2-year period August 1, 2002, through July 31, 2004, was recorded. The MELD score was calculated from blood test results collected in the LTUV database. The MELD score was calculated using the most recent version of the formula documented on the UNOS website: 9.57 × loge(creatinine mg/dL) + 3.78 × loge(bilirubin mg/dL) + 11.2 × loge(international normalized ratio) + 6.43, rounded to the nearest integer.8 Laboratory values less than 1.0 were set to 1.0 and the maximum serum creatinine was set to 4.0 mg/dL. The serum creatinine was set to 4.0 mg/dL if the patient had received dialysis at least twice within a week prior to the serum creatinine test. The MELD score for patients with a single hepatocellular carcinoma (HCC) less than 2 cm diameter was set to 20 if the calculated score was less than 20, and for patients with a single HCC greater than or equal to 2 cm diameter and up to 5 cm diameter or 2 to 3 HCCs less than 3 cm diameter, it was set to 24 if the calculated score was less than 24. The maximum MELD score was set at 40. Records were created for each prioritization (waiting list meeting) and organ allocation decision. For each organ allocation, records were created for each waiting list patient of the same blood group as the organ recipient. For each allocation decision, the clinical priority recorded was that assigned at the most recent prioritization meeting. Within each blood group at each date, patients were ranked by both MELD score (rank by MELD) and clinical priority (rank by clinical judgment). Tied ranks were averaged. Patients with the highest rank by MELD in each group were considered to have been prioritized by MELD. Patients with the highest rank by clinical judgment in each group were considered to have been prioritized by clinical judgment. Patients were considered to have died on the waiting list if they were on the active transplant list at the time of death or if they had been removed from the waiting list and subsequently died.
Pediatric patients were excluded, because of automatic assignment to the maximum clinical priority. Decisions based on a single patient in a blood group were excluded. Patients suspended from the waiting list were also excluded. Numerical data are expressed as mean ± standard error. Data were compared using nonparametric statistics. The association between prioritization by MELD and clinical judgment, between organ allocation and prioritization by MELD, and between organ allocation and prioritization by clinical judgment was assessed by the log odds ratio. The correlation between MELD score and clinical priority was assessed by Spearman rank correlation. Comparison of ranking by MELD and clinical judgment of patients who subsequently died waiting was by the paired sign test. P values less than 0.05 were considered to be significant. Receiver operating characteristic curves, plotting sensitivity vs. (1 – specificity), were generated, and the area under each curve (also known as the c-statistic) was calculated to assess the ability of MELD score and clinical priority to predict waiting list death within 3 months.
A total of 77 patients were considered in at least 1 prioritization or allocation decision during the study period. The MELD score, corrected and uncorrected for HCC, and Child-Turcotte-Pugh score at the time of placement of these patients on the waiting list were 19.3 ± 0.7, 17.6 ± 0.7, and 9.9 ± 0.3, respectively. A total of 3% of patients had a MELD score (corrected for HCC) of less than 11; 45% had a score of 11-18; 35% had a score of 19-24; and 17% had a score >24. A total of 30 patients (39%) had a MELD score (corrected for HCC) of at least 21. A total of 9% of patients had a MELD score (uncorrected for HCC) of less than 11; 51% had a score of 11-18; 23% had a score of 19-24; and 17% had a score >24. A total of 20 patients (26%) had a MELD score (uncorrected for HCC) of at least 21. A total of 48 patients (62%) had a Child-Turcotte-Pugh score of at least 10. The Child-Turcotte-Pugh score did not differ between patients who died on the waiting list and those who did not die on the waiting list (10.6 ± 0.7 vs. 9.8 ± 0.3, respectively; P = 0.37), or between those in whom the decisions by MELD and clinical judgment were predominantly concordant and those in whom the decisions were predominantly discordant (9.9 ± 0.3 vs. 9.7 ± 0.5, respectively; P = 0.80). The MELD score at the time of transplantation was 21.6 ± 0.8 (corrected for HCC) and 20.6 ± 0.9 (uncorrected for HCC). The proportion of patients who underwent transplantation with a MELD score (corrected for HCC) of less than 11 was 2%; with a score of 11-18, 30%; with a score of 19-24, 41%; and with a score >24, 28%. The proportion of patients transplanted with a MELD score (uncorrected for HCC) of less than 11 was 7%; with a score of 11-18, 31%; with a score of 19-24, 33%; and with a score >24, 28%. During the study period, the 1-year actuarial patient (graft) survival was 90.7% (90.7%) for adults, 92.9% (85.7%) for children, and 91.1% (89.7%) overall.
A total of 91 prioritization meetings occurred over the 2-year period. A total of 329 decisions were made about groups of patients of the same blood type. The groups consisted of a single patient in 80 cases, leaving 249 groups for analysis. A total of 1,463 decisions were made about individual patients. After excluding the 80 decisions regarding a single patient in the group and 265 decisions regarding pediatric patients, there were 1,118 decisions for analysis. There was a mean of 4.5 patients per prioritization group.
Of the 1,118 prioritization decisions made, 314 were prioritized by MELD, and 334 were prioritized by clinical judgment (Table 1). The same patient was prioritized by both MELD and clinical judgment in 144 cases (12.9%). The same patient was not prioritized by both MELD and clinical judgment in 614 cases (54.9%). The decisions were concordant in 67.8% cases. The patient was prioritized by MELD but not by clinical judgment in 170 cases (15.2%). A total of 92 (54.1%) of the 170 decisions in which the patient was prioritized by MELD but not by clinical judgment involved patients with HCC. The patient was prioritized by clinical judgment but not by MELD in 190 cases (17.0%). A total of 77 (40.5%) of the 190 decisions in which the patient was prioritized by clinical judgment but not by MELD involved patients with conditions not well accounted for by MELD (hepatopulmonary syndrome, polycystic liver disease, Wilson's disease with severe neuropathy, familial hyperlipidemia, human immunodeficiency virus, and Osler-Weber-Rendu syndrome). Despite the added weight given to HCC in the MELD calculations, 31 of the 190 (16.3%) decisions in which the patient was prioritized by clinical judgment but not by MELD involved patients with HCC. Of the patients who in at least 1 prioritization decision were prioritized by MELD but not by clinical judgment, 43.8% were prioritized by clinical judgment but not by MELD in at least 1 other prioritization decision. Similarly, of the patients who in at least 1 prioritization decision were prioritized by clinical judgment but not by MELD, 48.3% were prioritized by MELD but not by clinical judgment in at least 1 other prioritization decision. There was a significant association between prioritization by the 2 systems (P < 0.01). There was also a significant correlation between the MELD score and clinical priority (P < 0.0001) and between the patient rank by MELD and by clinical judgment (P < 0.0001).
|Prioritized by MELD||Total|
|Prioritized by clinical judgment|
A total of 69 allocations, and thus transplants, were performed over the 2-year period. A total of 15 livers were allocated to pediatric patients and 2 to patients with no other potential recipients in the same blood group, leaving 52 allocations for analysis. A total of 263 allocation decisions were made regarding patients in these groups. There was a mean of 5.1 patients per allocation group.
Of the 52 livers allocated, 29 (55.8%) were transplanted into patients prioritized by clinical judgment (Table 2). Organ allocation and prioritization by clinical judgment were concordant in 77.1% of cases. There was a significant association between organ allocation and prioritization by clinical judgment (P < 0.025).
|Prioritized by clinical judgment||Total|
Of the 52 livers allocated, 23 (44.2%) were transplanted into patients prioritized by MELD (Table 3). Organ allocation and prioritization were concordant in 72.2% of cases. There was no significant association between organ allocation and prioritization by MELD (P < 0.1).
|Prioritized by MELD||Total|
Patients allocated organs had median ranks of 1.75 and 2.00 by clinical judgment and MELD, respectively, and those not allocated organs had median ranks of 4.5 and 4.0 by clinical judgment and MELD, respectively, out of a median of 4.5 patients per group.
Waiting List Deaths
A total of 10 patients died waiting for a liver over the 2-year period. The annual adult, pediatric, and overall waiting list mortality over the study period was 9.3%, 0%, and 8.4%, respectively. Of the 10 patients who died, 6 were female, 1 had fulminant hepatic failure, and 7 were blood group O. The median time from activation on the waiting list to death was 140 days. A total of 2 of the 10 patients who died waiting for a liver during the period studied were “on hold” throughout the study period and 1 died without an organ offer, within 1 day of activation, which occurred between prioritization meetings. These 3 patients, therefore, do not appear in the analysis that follows.
A total of 92 prioritization decisions were made in patients who went on to die on the waiting list. These patients were prioritized by MELD on 27 occasions (29.3%) and by clinical judgment on 8 occasions (8.7%). There was a median of 9 (range: 4-11) patients in the groups to which these patients belonged. Patients who subsequently died waiting were ranked significantly higher by MELD (median: 3, range: 1-10) than by clinical judgment (median: 6, range: 1-8) (P = 0.0016).
A total of 27 allocation decisions on 14 occasions were made in 6 of the patients who went on to die on the waiting list. A total of 3 of these patients were prioritized by MELD in 8 decisions (29.6%) (1 patient 5 times, 1 patient twice, and 1 patient on a single occasion). A total of 3 patients were never prioritized by MELD. None of the patients who went on to die on the waiting list were prioritized by clinical judgment at the time of organ allocation. Of the 3 patients prioritized by MELD who subsequently died waiting, 1 had progressively severe pulmonary hypertension that precluded transplantation. There was a median of 9.5 (range: 5-11) patients in the groups to which these patients belonged. Patients who subsequently died waiting were ranked higher by MELD (median: 4.5, range: 1-10) than by clinical judgment (median: 7, range: 2-8), but the difference was not significant (P = 0.12).
Evaluation of receiver operating characteristic curves of the prediction of waiting list death within 3 months by MELD score and clinical priority revealed similar predictive value for each score (Fig. 1). The area under the curve was 0.69 for both MELD score and clinical priority.
The major aim of liver transplant waiting list management is the minimization of waiting list mortality, achieved by waiting list prioritization and organ allocation systems, selecting patients most at risk of dying, while ensuring good long-term posttransplantation outcome. These systems may be centralized or unit-based, as in Australia and New Zealand. Because the number of waiting list patients in each blood group in our unit is small at any 1 time, a system of prioritization based on frequent clinical review by a small number of physicians is likely to accurately reflect disease severity and potential mortality. Countries with a large number of competing potential recipients and multiple transplant units in any 1 procurement region require more complex and systematized algorithms to achieve optimal allocation. In the United States, the MELD score has been used in this capacity.6 We have compared the clinical prioritization and organ allocation system currently used in our unit with prioritization using the MELD score.
The distribution of MELD scores among the LTUV waiting list patients indicates that the LTUV patient population is sicker than the UNOS patients. A total of 39% of our patients had a MELD score, corrected for HCC, of at least 21, compared with 24% in the UNOS annual report.1 A total of 23% of our patients had a MELD score, uncorrected for HCC, of 19-24, compared with 7% of UNOS patients, and 17% of our patients had a MELD score >24, compared with 2% of UNOS patients.9 Although the size of the unit corresponds to a small organ procurement organization, the distribution of MELD scores at transplantation is intermediate between those of small and large organ procurement organizations.9
The 67.8% concordance of MELD and clinical judgment, the significant association between prioritization by the 2 systems, the correlation between MELD score and clinical priority, and the similar ranking by MELD and by clinical judgment of patients allocated organs suggest that a prioritization system based on frequent clinical review of liver transplant waiting list patients assesses patients similar to a prioritization system based on a mathematical formula incorporating objective criteria that have been shown to independently predict survival in cohorts of patients with liver disease. We have only established this, however, in the setting of prioritization from within small groups of patients. Some of the differences between patients prioritized by 1 system and not the other were as a result of patients who were prioritized by 1 system, and ranked highly, but not first, by the other system. This is reflected in the fact that the same patient was frequently prioritized by MELD on 1 occasion and by clinical judgment on another occasion.
It has been recognized that the MELD score does not accurately predict mortality in potential liver transplant recipients in whom cirrhosis is not the major indication for transplantation.4 This includes patients with HCC. This issue has been accounted for in the Organ Procurement and Transplantation Network /UNOS organ allocation policy, which currently assigns an initial MELD score to patients with a single HCC greater than or equal to 2 cm diameter and up to 5 cm diameter or 2 to 3 HCCs less than 3 cm diameter equivalent to a 15% probability of death within 3 months.8 An increased MELD score was also applied to HCC patients in the current study (20 points for single tumors less than 2 cm diameter and 24 points for single tumors greater than or equal to 2 cm diameter and up to 5 cm diameter or 2 to 3 tumors less than 3 cm diameter). As a result, more than one half of the prioritization decisions in which patients were prioritized by MELD, but not by clinical judgment, occurred in patients with HCC. However, almost one sixth of the prioritization decisions in which patients were prioritized by clinical judgment, but not MELD, also occurred in patients with HCC.
A number of indications for liver transplantation involve conditions in which the risk of death does not relate to the degree of liver dysfunction, or in which a prolonged wait for transplantation may result in deterioration in health, again not reflected in liver dysfunction. These conditions include metabolic and pathophysiologic syndromes affecting nonhepatic organ systems. The mortality risk and risk of deterioration in these conditions is not accurately predicted by the MELD score. Currently, in the Organ Procurement and Transplantation Network / UNOS allocation system, such patients are assigned a MELD score considered appropriate to their clinical condition after application to the regional review board.8 In our patient population, such patients included those in whom liver disease resulted in respiratory failure (hepatopulmonary syndrome), cardiac failure (Osler-Weber-Rendu syndrome), coronary artery disease (familial hyperlipidemia), neuropathy (Wilson's disease), and abdominal pain and inferior vena cava compression (polycystic liver disease). Not surprisingly, these patients tended to be prioritized by clinical judgment, but not by MELD.
It may be expected that allocation would occur exactly according to priorities assigned at the most recent prioritization meeting. However, only just over one half of our liver recipients had been prioritized by clinical judgment. This is because considerations other than clinical priority are taken into account when assessing potential recipients at the time of a donor organ offer. These include the need for comparable donor and recipient size. While it is possible to transplant a relatively small liver into a large recipient, transplantation of a large (whole) liver into a small recipient is not feasible. In addition, selection of a large recipient for a small donor liver may disadvantage the smaller patients on the waiting list (often women). There is also a tendency to match donor and recipient age, to a certain extent. Young recipients may live many decades with their donor liver, and the function of a liver from an elderly donor over this period is not yet known. In addition, the outcome of liver transplantation in hepatitis C has been shown to be inferior with increasing donor age.10–15 The quality of the donor may also affect the choice of recipient. It would be inadvisable, for example, to use a liver from a “marginal” donor in an unstable recipient, who is likely to have a high clinical priority and MELD score.16 Some donors, such as hepatitis B core antibody–positive or hepatitis C–positive donors, are only suitable for a select cohort of potential recipients. Finally, logistical issues, such as the number of team members available at any given time, may have a bearing on recipient selection. Our current system of organ allocation is sufficiently flexible to allow us to take such factors into account.
Since the aim of waiting list prioritization and organ allocation is minimization of waiting list mortality while maintaining optimal posttransplantation outcome, it is important to examine the prioritization of our patients who went on to die waiting for a suitable organ. These patients were 3 times as likely to be prioritized by MELD as by clinical judgment and were ranked significantly higher by MELD than by clinical judgment in the prioritization meetings. Of greater clinical importance are the organ allocation decisions. These represented the opportunities to offer lifesaving therapy to patients who subsequently died without such intervention. Of the 6 such patients, one half would have been allocated organs if a system of organ allocation based on MELD score were in use at the time, while none were prioritized by clinical judgment. However, clinical judgment takes into account other factors that may impact long-term survival.
In conclusion, prioritization of liver transplant waiting list patients by MELD score and by clinical judgment were significantly associated and the resulting rankings by the 2 systems were correlated in our patient population. However, MELD score was more likely than clinical judgment to prioritize for organ allocation patients who subsequently died waiting for a donor liver. An allocation process based on MELD rather than clinical judgment would significantly alter organ allocation in Australia and may reduce waiting list mortality.
- 12003 OPTN/SRTR annual report. Available at: http://www.optn.org/AR2003/903_li.pdf. Accessed: August 14, 2004.
- 2Transplantation Society of Australia and New Zealand. General organ donor criteria, liver. Available at: http://www.racp.edu.au/tsanz/oap3.htm. Accessed: August 14, 2004.
- 6The OPTN waiting list, 1988-2001. In: CeckaJM, TerasakiPI, eds. Clinical Transplants 2002. Los Angeles: UCLA Immunogenetics Center; 2002: 79–92., , .
- 8Organ distribution: allocation of livers. Available at: http://www.optn.org/PoliciesandBylaws/policies/pdfs/policy_8.pdf. Accessed: August 14, 2004.
- 13Liver transplantation for HCV-associated liver cirrhosis: predictors of outcomes in a population with significant genotype 3 and 4 distribution. Australian and New Zealand Liver Transplant Clinical Study Group. Liver Transpl 2003; 9: 339–347., , , , , , et al.