Assessment of Severity of Alcoholic Hepatitis Episode: Comparison of Discriminant Function, MELD, and Glasgow Alcoholic Hepatitis Scores

Abstract

Introduction: Alcoholic hepatitis (AH) is a distinct syndrome among patients with chronic and active alcohol abuse and has a potential for 30-40% mortality at 1 month among patients with severe disease. Various scores have been developed to assess the disease severity, however, none is an ideal score. We performed this study with an aim to compare discriminant function (DF), model for end-stage disease (MELD), and Glasgow alcoholic hepatitis score (GAHS) on their accuracy in assessing the severity of AH as determined by in-hospital and 30-day mortality. Methods: Medical charts were reviewed for patients admitted to 1 tertiary care hospital (2004-2013) with a discharge diagnosis of AH (ICD-9 code 571.1). Patients meeting criteria for AH diagnosis and with total bilirubin >5 mg/dL formed the study population. Chi-squareD and t-tests were used for comparing categorical and continuous variables, respectively. Multivariate logistic regression models were built to compare DF, MELD, and GAHS in predicting in-hospital (IHM) and 30-day mortality. Results: A total of 105 AH patients (mean age 48±10 years, 66% males, 81% white, 56% underlying cirrhosis) were reviewed. Average (mean±SD) DF, MELD, and GAHS were 54±32, 25±10, and 8±2, respectively. About 79% patients had severe AH with DF ≥32 and/or hepatic encephalopathy (HE). A total of 52 (50%) received treatment with steroids (n=29), pentoxifylline (13), or both (n=10). A total of 17 (16%) had infections with sepsis in 11 and SBP in 7 (1 with sepsis). A total of 51 (49%) died within 30 days of initial hospitalization with 26 (25%) dying during the hospitalization with an average hospital stay of about 15 days. Non-survivors compared to 79 surviving the hospitalization differed for age (51±10 vs. 47±11 years, P=0.07), HE (100% vs. 32%, P<0.0001), white cell count (16±10 vs. 12±8, P=0.02), DF (68±27 vs. 50±33, P=0.01), MELD (30±8 vs. 24±10, P=0.006), and GAHS (9±1 vs. 8±2, P=0.003). On logistic regression analysis adjusting for age, sex, underlying cirrhosis, treatment, accuracy of DF, MELD, and GAHS were similar in predicting IHM with c statistic of 0.728, 0.726, and 0.729, respectively. The scores were also similar in predicting 30-day mortality. For independently predicting the IHM and 30-day mortality, GAHS was superior to DF and MELD with OR (95% CI) of 0.68 (0.47-0.98) vs. 0.99 (0.97-1.001) vs. 0.95 (0.9-0.99) and 0.69 (0.5-0.92) vs.0.99 (0.97-0.999) vs. 0.94 (0.9-0.99), respectively. Conclusion: Discriminant function, MELD score, and GAHS are similar in predicting in-hospital and 30-day mortality in alcoholic hepatitis patients. Studies are needed to refine and improve the existing scores in predicting severity of alcoholic hepatitis.