A Model for Predicting Development of Overt Hepatic Encephalopathy in Patients With Cirrhosis

BACKGROUND & AIMS: Overt hepatic encephalopathy (HE) affects patients' quantity and quality of life and places a burden on families. There is evidence that overt HE might be prevented pharmacologically, but prophylaxis would be justified and cost effective only for patients at risk. We aimed to identify patients with cirrhosis at risk for overt HE. METHODS: We collected data from October 2009 through December 2012 for 216 consecutive patients with cirrhosis (based on liver biopsy, 96 patients with minimal HE), admitted to the Gastroenterology Unit at the University of Rome. Patients were followed up and evaluated for an average of 14.7 11.6 months; development of overt HE was recorded. We analyzed end-stage liver disease scores, shunt placement, previous overt or minimal HE, psychometric hepatic encephalopathy score (PHES), and levels of albumin, bilirubin, creatinine, and sodium to develop a prediction model. We validated the model in 112 patients with cirrhosis seen at the University of Padua and followed up for 12 +/- 9.5 months. RESULTS: During the follow-up period, 68 patients (32%) developed at least 1 episode of overt HE. Based on multivariate analysis, the development of overt HE was associated with previous HE, minimal HE (based on PHES), and level of albumin less than 3.5 g/dL (area under curve [AUC], 0.74). A model that excluded minimal HE but included albumin level and previous HE also identified patients who would develop overt HE (AUC, 0.71); this difference in AUC values was not statistically significant (P =.104). Both models were validated in the independent group of patients (3 variables: AUC, 0.74; 95% confidence interval, 0.66-0.83; and 2 variables: AUC, 0.71; 95% confidence interval, 0.63-0.78). CONCLUSIONS: We developed and validated a model to identify patients with cirrhosis at risk for overt HE based on previous HE, albumin levels, and PHES. If PHES was not available, previous HE and albumin levels still can identify patients at risk. Psychometric evaluation is essential for patients with no history of HE. These findings should aid in planning studies of pharmacologic prevention of overt HE.