Assessment of models for predicting response to corticoids treatment in alcohol-associated hepatitis: a global cohort study

Introduction and Objectives: Alcohol-associated hepatitis (AH) is a severe entity associated with high mortality. Corticosteroids might be used in cases with severe disease and several dynamic models can predict mortality and response to corticosteroids in AH patients. However, there is no consensus on the best of them. This study aimed to evaluate dynamic models to predict response to corticosteroid treatment based on short-term mortality in patients with severe AH based on a worldwide cohort. Materials and Methods: A retrospective cohort study of patients with severe AH (between 2009 – 2019). We included patients who received corticosteroid treatment and calculated the Lille model of day 4 (Lille-4), day 7 (Lille-7) (cut-off value ≥0.45), and the Trajectory of Serum Bilirubin (TSB)(cut-off value ≥0.8 of the ratio between bilirubin at admission and day 7) to predict mortality. We estimated up to 30-day survival using Kaplan-Meier curves, and we performed multivariable analyzes using Cox regression. Specifically, we constructed two models to compare Lille-4 vs. TSB and Lille-7 vs. TSB, adjusting by well-known clinical variables associated with higher mortality in AH (age, sex, and creatinine at admission). Results: 1,066 patients were included (30 centers, 10 countries), age 47.7 ± 10.9 years, 30% women. The MELD score on admission was 25 [21-30]. Responders were considered by Lille-4 49.1%, Lille-7 46.6%, and TSB 55.4%. In the first Cox regression, we observed that Lille-4 and TSB predicted 30-day mortality (HR 3.0, 95%CI: 1.7-5.1; p<0.0001, and HR 2.1, 95%CI: 1.3-3.5; p=0.005, respectively) (Table A). In the second Cox regression, Lille-7 also predicted 30-day mortality (HR 3.7, 95%CI: 2.1-6.7; p<0.0001) but not TSB (HR 1.5, 95% CI: 0.8-2.6; p=0.180) (Table B). Creatinine at admission was also statistically significant in both Cox-regressions. Conclusions: Different dynamic models can determine the response to corticosteroids in patients with severe AH. However, Lille-7 and Lille-4 have the best performance. New models are needed for better prognostication in AH.