Journal of Hepatology
Volume 42, Issue 6 , Pages 790-792, June 2005

Developing concepts on MELD: delta and cutoffs

Unità di Gastroenterologia, Ospedale V Cervello,Via Trabucco 180, 90146 Palermo, Italy

published online 04 April 2005.

See Articles, pages 820--825 and 826--832

Article Outline

 

The Model for End stage Liver Disease [1] was originally developed to predict survival rates in patients with chronic liver disease undergoing transjugular intrahepatic porta-caval shunt [2]. The model, based on simple, readily available and reproducible variables (creatinine, bilirubin, international normalized ratio and etiology), allowed identifying patients with a median survival as short as 3 months. In the series of patients from which it was calculated [2], the model was superior to both the Child-Pugh classification and the Child-Pugh score [3] in predicting survival and its predictive accuracy was thereafter validated in several series of patients including 2278 hospitalized or ambulatory patients [1]. It was found to be highly accurate in predicting one-week, three-month and one-year mortality. Because of the objective nature of its components and because of its ability to identify increasing death risk, MELD was adopted by the UNOS in place of the Child-Pugh score for the allocation of livers from deceased donors. The introduction of the MELD system for the allocation of livers in the United States, resulted in a 12% decrease of the waiting list registrations, particularly for MELD values lower than 10 [4]. There was also a 3.5% reduction in waiting list mortality while early survival of liver transplant recipients remained unchanged despite the selection of sicker patients for transplantation [4]. Moreover, in a cohort of 12,996 adult patients listed for transplantation between 2001 and 2003, time-dependent Cox regression models allowed to estimate that overall mortality risk reduction for transplant recipients was 79% compared with candidates. Survival benefit increased with increasing MELD for scores higher than 18 while recipient mortality risk during the first post-transplant year was much higher than for candidates for scores lower than 14 [5], suggesting that patient selection for transplantation may be further improved by avoiding to list patients with MELD scores below a given cut-off value.

The introduction of MELD was followed by a large number of publications dealing with its clinical use. There are 105 citations in the MEDLINE (accessed march 15, 2005) searching the term MELD in the ‘title’ and 254 in the ‘title or abstract’. MELD has been found to be a significant predictor of death risk in alcoholic hepatitis [6], in patients with variceal bleeding [7], in cirrhotic patients undergoing cholecistectomy [8], in patients undergoing liver retransplantation [9], in fulminant hepatic failure without acetaminophen toxicity [10].

However, although it generated large interest, MELD has several limitations. First, the selection criteria for the candidate variables included in the multivariate analysis generating the final model were not reported [2]. It is therefore conceivable that other important as well as objective variables were not included in the analysis. Second, it is unclear what is the predictive accuracy for mortality of MELD in patients with or without ascites, the most consistent marker of decompensated cirrhosis. In fact, MELD validation studies have mostly included patients independently of the presence of ascites [1]. However, since patients with ascites have a median survival time of approximately 2 years compared to more than 10 years in patients without ascites at diagnosis [11], it may be expected that the predictive accuracy of MELD will be different according to the disease stage. In accordance with this hypothesis, the predictive accuracy for mortality of MELD in patients with refractory ascites has been found to be unsatisfactory [12]. Third, MELD calculation requires a computer, while the familiar Child-Pugh score is currently and easily calculated at bedside without computer and besides, several studies have shown that there is no significant difference in predictive accuracy for mortality between Child-Pugh and MELD score [13].

Some of these limitations are probably reflected in studies aimed at improving MELD performance. An example of this is that it has been consistently suggested that the addition of serum sodium level to the MELD score may improve its predictive accuracy [12], [14], [15]. Particularly, it has been shown that persistent ascites and low serum sodium identify cirrhotic patients with high mortality risk despite low MELD scores [12], suggesting that MELD may not be satisfactory in these end stage patients. Two recent studies addressed the issue of whether the variation of MELD, referred to as ΔMELD, is superior to actual MELD, with contrasting results [16], [17]. Both studies were retrospective and both compared MELD with ΔMELD in the previous 30 days. In the first [16], for any given MELD score, the magnitude and direction of change in MELD score during the previous 30 days was a significant independent mortality predictor. In the second study [17] ΔMELD was significant at univariate analysis, but less predictive when current MELD was included and not predictive when considered with both current MELD and the number of MELD scores determinations performed in the same patient. The results of this last study suggested that the predictive value of ΔMELD was in part explained by the current MELD because the higher the ΔMELD the higher the current MELD was. Moreover the loss of predictive value of ΔMELD when the number of MELD determinations was added in the model, clearly suggested a detection bias: more determinations in patients with clearly deteriorating disease [17].

In this issue of the Journal two articles suggest further expansion of the use of MELD. Huo and colleagues [18] show that ΔMELD may be a more accurate predictor of death risk than initial MELD and Child-Pugh score. The idea underlying the study is clearly appealing: the worsening of death predictors should parallel the increase of death risk. However, the relationship between worsening predictor and increasing risk may be missed if death occurs before the second predictor measurement to assess the size of the variation (in this case ΔMELD). In fact the most recent MELD score available is used to prioritize donor organ allocation for patients with end stage liver disease awaiting liver transplantation [19]. Huo and colleagues found that a ΔMELD of 2.5 in a time period of 1–4 months was the only significant predictor of death risk in a multivariate analysis including initial MELD and Child-Pugh score. The study included 58 prospective and 293 retrospective patients surviving long enough to have a second MELD determination. Clearly, the design of the study was biased towards a more important prognostic role for ΔMELD because of the exclusion of patients dying after a single MELD, thus probably missing the most important prognostic information of initial MELD determination. Moreover, the inclusion of only 293 out of 1168 retrospective patients strongly calls for the inherent risk of selection bias. Yet, the exclusion rate from the prospective series is not reported. Another potential source of bias in this study might be the number of MELD determinations, which however was not accounted for. Finally, it is unclear why ΔMELD was compared with initial instead of last MELD determination in the 1–4 months considered period. Despite these limitations, however, the study by Huo and colleagues confirms the results of the two previous studies [16], [17] that ΔMELD is a significant predictor of death risk in advanced cirrhosis. From this study and the previous two [16], [17], it may be concluded that whether ΔMELD is superior to actual MELD should be assessed in well designed prospective studies, while for the present time, if the choice is made to use the MELD score as a death risk predictor, then the actual MELD should be used.

The second study, by Amitrano and colleagues [20], shows that a MELD score ≥15 or ≥18 together with advanced hepatocellular carcinoma are significant predictors of 6-week and, respectively, 3-month mortality after variceal bleeding. Using ROC curves, in a cohort of 172 cirrhotic patients with first variceal bleeding the Authors found that MELD and Child-Pugh scores were equally predictive of mortality and identified a MELD score of 15 as the best cutoff to predict 6 week mortality and of 18 for 3 month mortality. Although the predictive values for the positive tests for 6-week and 3-month mortality were not satisfactory (43 and 71%, respectively), the higher predictive values for the negative tests (92 and 83%, respectively) indicated that most of surviving patients may be correctly identified by these MELD cutoff values. However, 38 patients surviving more than 6 weeks out of 67 (57%) with MELD score >15 and 12 surviving more than 3 months out of 42 (29%) with MELD score >18 were incorrectly classified as being at the highest risk of death. Moreover, 8 and 17% of patients with terminal disease were erroneously classified as being at the lowest risk of death. Therefore, the performance of the proposed test seems to be far from satisfactory for clinical practice as also confirmed by the likelihood ratios (LR): LR+ 2.8 and LR− 0.3 for 6-week mortality and 5.8 and 0.46, respectively, at 3 months. Stratification of patients according to the MELD cutoff and the presence and size of HCC allowed to identify four groups of patients with significantly different mortality. However, how much HCC adds to the MELD in terms of predictive accuracy was not reported. Also, it is surprising that the MELD modification for HCC was not used in this study [19]. In fact, since HCC was such a robust predictor of mortality, particularly at 3 months, it is conceivable that the operative characteristics for the MELD modified for HCC would be significantly superior to that of MELD as assessed in the study.

Despite the limitations mentioned, the two studies on MELD in the present issue of the Journal, provide further insight in the evaluation of MELD. The first [18], validated increasing MELD score as a predictor of short term mortality, although the question on whether it should be preferred to the last known MELD score still remains unanswered: this requires fully prospective studies. The second study provides solid data on the predictive accuracy of MELD in cirrhotic patients with acute variceal bleeding, on the equivalence of Child-Pugh and MELD also in this setting and confirms the prognostic role of HCC.

These studies add to the ongoing debate on the clinical role of MELD, which seems to be still far from a conclusion. The best available scientific evidence supports the use of MELD for prioritizing liver allocation to patients on the waiting list for a liver transplant. However, no definite advantage from the use of MELD has been proven so far in clinical settings other than liver transplantation. It is therefore conceivable that physicians will continue to use the more familiar and friendly Child-Pugh score which is continuing to resist the MELD challenge.

Back to Article Outline

References 

  1. Kamath PS, Wiesner RH, Malinchoc M, Kremers W, Therneau TM, Kosberg CL, et al. A model to predict survival in patients with end-stage liver disease. Hepatology. 2001;33:464–470
  2. Malinchoc M, Kamath P, Gordon F, Peine C, Rank J, Ter borg P. A model to predict poor survival in patients undergoing transjugular intrahepatic portosystemic shunts. Hepatology. 2000;31:864–871
  3. Pugh RN, Murray-Lyon IM, Dawson JL, Pietrni MC, Williams R. Transection of the esophagus for bleeding oesophageal varices. Br J Surg. 1973;60:646–649
  4. Freeman RB, Wiesner RH, Edwards E, Harper A, Merion R, Wolfe R. United Network for organ sharing organ procurement and transplantation network liver and transplantation committee. Results of the first year of the new liver allocation plan. Liver Transpl. 2004;10(1):7–15
  5. Merion RM, Schaubel DE, Dykstra DM, Freeman RB, Port FK, Wolfe RA. The survival benefit of liver transplantation. Am J Transplant. 2005;5:307–313
  6. Dunn W, Jamil LH, Brown LS, Wiesner RH, Kim WR, Menon KV, et al. MELD accurately predicts mortality in patients with alcoholic hepatitis. Hepatology. 2005;41:353–358
  7. Chalasani N, Kahi C, Francois F, Pinto A, Marathe A, Bini E, et al. Model for end-stage liver disease (MELD) for predicting mortality in patients with acute variceal bleeding. Hepatology. 2002;35:1282–1284
  8. Perkins L, Jeffries M, Patel T. Utility of preoperative scores for predicting morbidity after cholecystectomy in patients with cirrhosis. Clin Gastroenterol Hepatol. 2004;2:1123–1128
  9. Rosen HR, Prieto M, Casanovas-Taltavull T, Cuervas-Mons V, Guckelberger O, Muiesan P, et al. Validation and refinement of survival models for liver retransplantation. Hepatology. 2003;38:460–469
  10. Kremers WK, van IJperen M, Kim WR, Freeman RB, Harper AM, Kamath PS, et al. MELD score as a predictor of pretransplant and posttransplant survival in OPTN/UNOS status 1 patients. Hepatology. 2004;39:764–769
  11. D'Amico G, Morabito A, Pagliaro L, Marubini E. Survival and prognostic indicators in compensated and decompensated cirhhosis. Dig Dis Sci. 1986;31:468–475
  12. Heuman DM, Abou-Assi SG, Habib A, Williams LM, Stravitz RT, Sanyal AJ, et al. Persistent ascites and low serum sodium identify patients with cirrhosis and low MELD scores who are at high risk for early death. Hepatology. 2004;40:802–810
  13. Durand F, Valla D. Assessment of the prognosis of cirrhosis: Child-Pugh versus MELD. J Hepatol. 2005;42:S100–S107
  14. Biggins SW, Rodriguez HJ, Bacchetti P, Bass NM, Roberts JP, Terrault NA. Serum sodium predicts mortality in patients listed for liver transplantation. Hepatology. 2005;41:32–39
  15. Ruf AE, Kremers WK, Chavez LL, Descalzi VI, Podesta LG, Villamil FG. Addition of serum sodium into the MELD score predicts waiting list mortality better than MELD alone. Liver Transpl. 2005;11:336–343
  16. Merion RM, Wolfe RA, Dykstra DM, Leichtman AB, Gillespie B, Held PJ. Longitudinal assessment of mortality risk among candidates for liver transplantation. Liver Transpl. 2003;9:12–18
  17. Bambha K, Kim WR, Kremers WK, Therneau TM, Kamath PS, Wiesner R, et al. Predicting survival among patients listed for liver transplantation: an assessment of serial MELD measurements. Am J Transpl. 2004;4:1798–1804
  18. Huo T-I, Wu J-C, Lin H-C, Lee F-Y, Hou M-C, Lee P-C, et al. Evaluation of the increase in model for end-stage liver disease (ΔMELD) score over time as a prognostic predictor in patients with advanced liver cirrhosis: risk factor analysis and comparison with initial MELD and Child-Turcotte-Pugh score. J Hepatol 2005;42:826–832.
  19. Policy 3.6 organ distribution. Allocation of livers. http://www.unos.org/PoliciesandBylaws/policies/pdfs/policy_8.pdf; 2003.
  20. Amitrano L, Guardiascione MA, Bennato R, Manguso F, Balzano A. MELD score and hepatocellular carcinoma identify patients at different risk of short-term mortality among cirrhotics bleeding from esophageal varices. J Hepatol. 2005;42:820–825

PII: S0168-8278(05)00198-4

doi:10.1016/j.jhep.2005.03.009

Journal of Hepatology
Volume 42, Issue 6 , Pages 790-792, June 2005

Article Tools