Is MELD-Plus Advantageous for Patients With Low MELD-Na Scores?
I was delighted to read that Mazumder et al1 emphasized the importance of using machine-learning techniques to potentially improve the prediction accuracy of model for end-stage liver disease (MELD)-Na. Specifically, I was glad to read under the “Potential Improvements to the MELD-Na Score” section that the authors referred to MELD-Plus2 as a possibility to achieve that; MELD-Plus is a highly accurate mortality prediction risk score in cirrhosis; the score’s components (creatinine, total bilirubin, international normalized ratio, sodium, albumin, white blood cell count, and age) are available for any patient and thus can be applied to any observational health repository.3 A critical component of the success of any risk prediction score is its ability to predict the outcome accurately for each patient regardless of the patient’s score value (ie, low versus high risk). As shown in Figure 1, which is extracted from the MELD-Plus study, the score performs uniformly across all risk quintiles. The strong correlation between the predicted 90-day mortality and the observed mortality rate throughout the range of risk across all quintiles in the entire cirrhotic population, as well as in the derivation and validation sets independently, indicates a possible significant advantage of MELD-Plus over MELD-Na. The strong correlation further indicates MELD-Plus’s potential to prioritize candidates for liver transplantation efficiently, regardless of whether a patient is at a very high or very low mortality risk (as opposed to individuals with low MELD-Na scores who, as extensively reported, are often associated with high mortality rates). As a result of MELD-Plus’s ability to predict uniformly and accurately 90-day mortality throughout the entire range of risk when applied to a large population of cirrhotic patients admitted to Massachusetts General Hospital or Brigham and Women’s Hospital, I thus ask the scientific community to assess these results in additional repositories of patient populations, especially for populations with low MELD-Na scores. Optimistically, MELD-Plus may be found advantageous for such patients and less likely to be passed over by the current system of organ allocation, as proposed by Mazumder et al.1
1. Mazumder NR, Atiemo K, Kappus M, et al. A comprehensive review of outcome predictors in low MELD patients. Transplantation. 2020; 104:242–250
2. Kartoun U, Corey KE, Simon TG, et al. The MELD-plus: a generalizable prediction risk score in cirrhosis. PLoS One. 2017; 12:e0186301
Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.
3. Kartoun U. MELD-plus: a call for an additional external validation. Eur J Gastroenterol Hepatol. 2019; 31:1603