The model of end-stage liver disease (MELD) score is often used for liver graft allocation, and patients with hepatocellular carcinoma (HCC) receive exception points (22 in the US). A better model is desirable for patients with HCC as they tend to have a privileged access to transplantation, without taking HCC characteristics into account. A new simpler model designed from a training set of US patients (n = 49 026) was tested on two validation sets (US and UK patient cohorts with, respectively, n = 20 475 and n = 1781). The risk of dropout was between 3.2 and 7.8% at 3 months in patients with HCC, and was captured into a score, including HCC size, HCC number, AFP, and MELD (-37.8 +1.9*MELD+5.9 if HCC Nb ≥ 2 + 5.9 if AFP > 400 + 21.2 if HCC size > 1 cm). This new model could be validated on external US and UK liver candidate cohorts. It provides a dynamic and more accurate assessment of dropout than the use of exception MELD (C-indices of 66.2-73.7% vs. 52.7-56.6%). In addition, the model shows a similar distribution as MELD for patients with non-HCC, and both scores could be used in parallel for the management of waiting-list patients with and without HCC.
Keywords: MELD; allocation; deMELD; score; validation.
© 2014 Steunstichting ESOT.