Comparison of the prognostic models for mortality in idiosyncratic drug-induced liver injury
Journal Articles
Overview
Research
Identity
Additional Document Info
View All
Overview
abstract
BACKGROUND: Several models have been proposed to predict acute liver failure/death in patients with drug-induced liver injury (DILI), but the predictive performances of them have not been systematically compared. We aim to compare the current models for their predictive potency of mortality at DILI onset. METHODS: DILI patients hospitalized at both Beijing Friendship Hospital and the Fifth Medical Center of PLA General Hospital were categorized into death/liver transplantation (LT) or survival without LT group. Predictive potency of 28-day, 90-day, 6-month and 12-month death/LT outcomes of Hy's Law, nHy's Law, Robles-Diaz Model, drug-induced liver toxicity (DrILTox ALF) Score, Model for End-stage Liver Disease (MELD) Score, and Ghabril Model was compared by Delong method. RESULTS: A total of 6.3% (83/1314) patients died or received LT within 12 months after DILI onset. The area under receiver operating characteristic of Hy's Law, nHy's Law, and Robles-Diaz Model was all lower than 0.750 for the prediction of within 12 months' mortality. DrILTox ALF Score, MELD Score and Ghabril Model showed better predictive potency of 28-day [0.896 (0.878-0.912), 0.934 (0.919-0.947), 0.935 (0.921-0.948), respectively], 90-day [0.883 (0.864-0.899), 0.951 (0.938-0.962), 0.952 (0.939-0.963), respectively], 6-month [0.820 (0.799-0.841), 0.905 (0.888-0.921) and 0.908 (0.891-0.923), respectively] and 12-month [0.801 (0.779-0.823), 0.882 (0.863-0.899) and 0.885 (0.866-0.902), respectively] mortality. CONCLUSION: Despite the difference of clinical characteristics and implicated-drug categories between China and industrialized countries, we demonstrate that MELD Score and Ghabril Model have the best predictive performance in the prediction of mortality within 12 months after DILI onset.
