Early prediction of short-term development of hepatic encephalopathy in patients with acute liver disease unrelated to paracetamol. A prospective study in Japan

Abstract

The aim of our study was to provide a predictive model for early recognition of the risk of short-term development of hepatic encephalopathy (HE) in patients with symptomatic acute liver disease (ALD). From a retrospective analysis of 220 patients with ALD, prothrombin time (PT) equal to, or lower than, 80% of normal, was set as the registration criteria in the subsequent patient cohorts of the study. Then, a HE-prediction model was derived by a logistic regression analysis of data in 259 new patients, and prospectively validated in 124 other patients, both groups of patients were affected by ALD unrelated to paracetamol (non-P ALD). The following HE-prediction model was established: lambda = [0.692 x ln(1 + total bilirubin(mg/dL))] - 0.065 x PT(%) + [1.388 x Age(years)] + [0.868 x Etiology] - 1.156, where Age is 1 in patients older than 50 years and Etiology is 1 when the cause of non-P ALD is flare-up of type B hepatitis, auto-immune hepatitis or unknown, and 0 otherwise. In the validation group, according to the model-based risk of subsequent development of HE, sensitivity and specificity of the model were 100% and 69.0%, respectively, in patients with an evaluated risk lower than 20%, and 62.5% and 93.1%, respectively, in those with an evaluated risk equal to, or greater than, 50%. In Japanese patients with symptomatic non-P ALD, our model, which includes four of the five items used in the King's College Hospital criteria, represents an acceptable, effective model to allow early detection of the risk of short-term development of HE. Using this model in other populations requires further validation specific to each of them.