The Decreasing Predictive Power of MELD in an Era of Changing Etiology of Liver Disease

Elizabeth L. Godfrey; Tahir H. Malik; Jennifer C. Lai; Ayse L. Mindikoglu; N. Thao N. Galván; Ronald T. Cotton; Christine A. O'Mahony; John A. Goss; Abbas Rana


American Journal of Transplantation. 2019;19(12):3299-3307. 

In This Article

Materials and Methods

Study Population

The United Network for Organ Sharing (UNOS) provided de-identified patient-level data, sourced from the Organ Procurement and Transplantation Network (OPTN) liver registry. Data were used from 120 393 patients over the age of 18 listed for liver transplantation between March 1, 2002 and December 2, 2016, with additional listing data from patients through January 1, 2018 for a total of 131 644. Multi-visceral or combined transplants including the heart, intestines, or lungs were excluded, as were living donor transplants. Liver retransplants or combined transplants including the pancreas or kidneys were included if the previous transplant date was recorded. Any patients approved for waitlist exceptions were excluded, except where specifically stated otherwise.

Statistical Analysis

Data were analyzed with the standard statistical software package, Stata 15.1 (Stata Corp, College Station, TX). Continuous variables were reported as a mean and standard deviation. The Student t test and contingency and chi squared tables were used to compare continuous and categorical variables, respectively. Results were considered significant at a two-tailed P-value of <.05.

Area under the receiver operating curve analyses were performed against listing year and diagnosis to determine the correspondence between 90-day waitlist mortality and score. The primary outcome measure was the c-statistic of the MELD score plotted against 90-day waitlist mortality. Mortality was defined using a multilayered definition of death. The primary definition, referred to here as 90-day mortality, included only death on the waitlist. The secondary definition added waitlist death according to the Social Security death index in available years (2002–2012), and the tertiary definition added presumptive death after removal from the waitlist for being too sick to transplant. The same analysis was conducted using the current MELD-Na.

A competing risk analysis using the Fine and Gray model integrated into Stata was also performed to assess the relationship between the MELD and MELD-Na scores and death within 90 days, with transplant during that time period as the competing risk. An opposing analysis with transplant as the primary outcome and death on the waitlist within 90 days as the competing risk was performed for comparison. Removal from the waitlist was not investigated here because it didn't change the multi-layered death analysis. Brier scores were also calculated to evaluate the concordance of MELD against an odds-of-mortality prediction, based on the mortality probabilities established by the original Mayo Clinic and UNOS score committee paper by Wiesner et al.[17]

The population was also characterized in terms of hepatitis C virus (HCV)–positive status, age, score at listing and removal, and percentage of individuals receiving transplants. An analysis of indication for listing by year was also conducted using diagnosis code and year of listing.