Main clinical factors influencing early mortality in a cohort of patients with severe alcoholic hepatitis, and evaluation trough ROC curves of different prognostic scoring systems

Abstract

BackgroundSevere alcoholic hepatitis (SAH) presents high early mortality-rate (90 days) and is related to several complications. The aim of this study was to evaluate the impact on early mortality-rate of developed complications in patients with SAH, and to evaluate the accuracy of different prognostic scoring systems to predict early mortality. Subjects and methodsCohort study. Were included 110 patients with SAH. We collected data about development of complications: acute renal failure (ARF), hepatic encephalopathy (HE), variceal bleeding (VB), infections; alcohol intake (g/day) and presence of cirrhosis by ultrasonography (USG). Child-Pugh, Maddrey's modified discriminant function (DF), Model for End stage Liver Disease (MELD); Age, Bilirubin, INR, and Creatinine score (ABIC); Lille score, and Glasgow Alcoholic Hepatitis Score (GAHS) were calculated. Primary endpoint was 90-day mortality. To evaluate survival according to the development of complications we performed a Cox regression model. Accuracy of different prognostic scoring systems was evaluated trough ROC curves. Results90-day mortality-rate was 71 patients (64.5%). 79 patients (71.8%) had evidence of cirrhosis in the USG, 59 (53.6%) developed HE, 54 (49.1%) ARF, 41 (37.3%) VB, and 41 (37.3%) infection. In the Cox-regression model significant association was found between greater risk of mortality and the development of HE (HR 8.0; IC al 95% 3.0 a 21.4; P=0.0001) and presence of cirrhosis in the USG (HR 3.0; 95% CI 1.0 to 8.7; P=0.045). Regard to prognostic scoring systems we found that Lille score ≥ 0.45 was the best predictor of early mortality in patients with SAH (AUROC=0.83; 95% CI 0.75 to 0.91, P<0.0001). ConclusionsThe development of HE is the main factor associated to early mortality. Coexistence of cirrhosis is a factor that worsen the prognosis. Lille score is the most accurate for predict early mortality.