This article is more than 5 years old. Information may no longer be current.

Joint model better predicts survival in patients with alcoholic hepatitis

Researchers combined results from static and dynamic scoring systems to create a joint-effect model that accurately predicted survival of patients with alcoholic hepatitis, according to study data.

“At present, to adapt the treatment strategy in patients with severe [alcoholic hepatitis], prognosis is determined by the use of either static or dynamic models,” the researchers wrote. “A new notion is to combine the information from static and dynamic models to predict the risk of death on a continuum. This approach might make it possible to stratify the risk of death and provide a more precise estimation of outcome based on severity when management begins and after a short period of treatment.”

Researchers collected and analyzed data from several international databases of patients with severe alcoholic hepatitis (AH) treated with corticosteroids in France and the United Kingdom to create a model to predict patient survival. This group of patients was known as the derivation cohort (n = 538).

The researchers then compared the efficacy and outcomes of 3 joint-effect models: the Maddrey discriminant function plus the Lille model, MELD score plus Lille model and age, bilirubin, international normalized ratio, creatinine score (ABIC) plus Lille mille, to determine which combination had “the best prognostic value,” based on patient outcomes, according to the research. These results were then validated using data from clinical trials testing the efficacy of corticosteroids among patients in the United States, France, Korea and Belgium (n = 604).

The joint-effect model was used to predict patient survival after 2 and 6 months and researchers found that this model predicted patient outcomes better than either static or dynamic models alone in both cohorts (P < .01 for all comparisons). Also, the joint model accurately predicted patient survival despite the patient risk level.

The MELD score and Lille combination was better than the other combinations for predicting patient survival. In this model, the predicted 6-month mortality of complete responders with MELD scores between 15 and 45, with a Lille score of 0.16, was between 8.5% and 49.7%, compared with 16.4% and 75.2% for non-responders.

In the joint-effect model, two patients had the same baseline MELD score of 21, but the patient with a Lille score of 0.45 had a 1.9-fold higher risk of mortality compared with the patient with a Lille score of 0.16 (23.7% vs. 12.5%), according to the research.

The researchers concluded: “We propose a new approach to predict outcome in patients with severe AH in terms of a continuum of mortality risk. This can help improve management of these patients as well as to develop clinical trials for future molecules and/or therapeutic strategies.” – by Melinda Stevens

Disclosures: The researchers report no relevant financial disclosures.