MELD score to predict outcome in adult patients with non-acetaminophen-induced acute liver failure

A model for end stage liver disease (MELD) score > 30 was proposed as an excellent predictor of mortality in patients with non-acetaminophen-induced acute liver failure (ALF). We analyzed the prognostic value of MELD score in our patients with ALF who were prospectively registered in our database since 1990. Overall, 106 patients met the criteria of ALF. Excluding seven patients with acetaminophen etiology, 99 patients (42 +/- 15 years, 40M/59F) were studied. Causes were cryptogenic (n=38), viral (n=29), drugs (n=20) and miscellaneous (n=12). Of these, 37% (n=37) survived with medical management alone (group I), 16% (n=16) died (group II) and 46% (n=46) underwent liver transplantation (group III). The strongest predictors of poor outcome were advanced encephalopathy, cryptogenic/drug-induced/hepatitis B etiology and a high MELD score. At the time of diagnosis, King's College Hospital criteria and MELD score > 30 had similar high negative predictive value (92% and 91%, respectively) and low positive predictive value (52% and 56%, respectively). The predictive values improved only slightly during follow-up. The best cut-off point for MELD score to discriminate between survivors and nonsurvivors was > 35, with a sensitivity and specificity of 86% and 75%, respectively. MELD score, which mostly takes into consideration the degree of liver impairment, has a similar prognostic value as King's College Hospital criteria to predict outcome in adult patients with nonacetaminophen-induced ALF. Overall, all current scores miss accuracy and therefore there is a clear need for factors that can better predict the regeneration of the liver in this setting.