Model for End-stage Liver Disease Can Predict Very Early Outcome After Liver Transplantation

Postoperative Model for End-stage Liver Disease (MELD) values have never been assessed to predict very early (<1 week) death after liver transplantation (OLT). We retrospectively reviewed 275 consecutive OLTs performed in 252 recipients reported in a prospective database. We calculated the MELD score (pre-MELD) and consecutive postoperative MELD (post-MELD) scores computed daily during the first postoperative week and on days 15 and 30 after OLT. Post-MELD scores from nonsurviving recipients displayed on a scatterplot of immediate probability of death were adjusted to the best goodness-of-fit curve, and, finally, depicted graphically as a receiver operating characteristic (ROC) curve. Nonsurviving recipients showed higher post-MELD scores: day 1: 23.5 versus 16.6 (P =.05); day 3: 25.1 versus 12.5 (P =.000); day 5: 25.7 versus 11.8 (P =.000); and day 7: 22.1 versus 10.2 (P =.000). Overall comparisons were performed using a time-dependent general linear regression model, revealing higher post-MELD scores for nonsurviving recipients, irrespective of postoperative time (P =.002). The best goodness-of-fit curve was displayed when adjusting to a theoretical exponential regression curve calculated as follows: Probability of dying within the first week (%) = 3.36 x e(0.079) x ((post-MELD)) (r =.89; P =.000). The area under the ROC curve was 0.783 (95% confidence interval, 0.630-0.935; P =.001). The model had a positive predictive value of 82.3%, a negative predictive value of 33.1%, and an accuracy of 79.2%. In conclusion, this study corroborated the suggestion that the MELD score may serve as a reliable tool to assess very early death after OLT.