Integrative Epidemiology of Prognosis in Patients With Acute Alcoholic Hepatitis at AP-HP

Acute Alcoholic Hepatitis

Acute Alcoholic Hepatitis, Machine Learning, Accelerated failure time, RWE epidemiology, Prognosis prediction

Acute alcoholic hepatitis (AAH) is a severe liver disease that often occurs in individuals with prolonged excessive alcohol consumption. Patients face a high risk of liver failure, complications, and death, despite available treatments. Current prognostic scores based on blood tests provide limited accuracy and do not capture the full complexity of the disease. The purpose of this study is to improve the prediction of patient outcomes after a diagnosis of acute alcoholic hepatitis. By integrating clinical, biological, and histological information collected from the AP-HP data warehouse, the investigators aim to identify more reliable prognostic markers. This could help physicians better tailor treatments and improve survival of patients affected by this condition.

Acute alcoholic hepatitis (AAH) is characterized by acute liver inflammation caused by prolonged excessive alcohol intake. It is associated with jaundice, liver failure, and in severe cases, multi-organ failure and death. Current prognostic scores rely primarily on serum biomarkers of liver function but fail to account for systemic complications or histological features. This retrospective, multicenter observational study conducted at AP-HP hospitals will analyze a large cohort of patients diagnosed with AAH since 2017. Data extracted from the AP-HP Clinical Data Warehouse (EDS) will include demographics, comorbidities, biological markers (liver and renal function tests, coagulation parameters, lactate), histological findings (necrosis, fibrosis, neutrophil infiltration, bilirubinostasis), microbiology results, prescribed medications, and outcomes such as infections, need for liver transplantation, and mortality. The primary outcome is overall survival after diagnosis. Secondary outcomes include the impact of intermediate factors such as bacterial or fungal infections, alcohol withdrawal, organ failure, and transplantation on prognosis. Statistical analyses will combine classical survival models (Cox regression, Accelerated Failure Time models) and modern machine learning approaches (random forest, gradient boosting). The goal is to develop integrative prognostic models that provide a more accurate and personalized assessment of prognosis in AAH, ultimately guiding clinical decision-making and improving patient management.

No related papers found yet.

None listed

France