Predictors of poor outcomes in 488 patients with herbinduced liver injury.

Background/Aims: Herb-induced liver injury (HILI) can lead to chronic liver injury, liver transplantation, or even death. This study aimed to identify the predictors of poor HILI outcomes, especially chronic HILI. Materials and Methods: Clinical data of 488 patients with HILI were retrospectively analyzed from a Chinese center between January 2010 and January 2014. Logistic regression and C-statistic were used to identify risk factors and prognostic models for HILI outcomes. Results: In all patients, 69 (14.1%) developed chronic HILI, and 20 (4.1%) died due to liver injury or underwent liver transplantation. To predict the fatal HILI prognosis, the model for end-stage liver disease (MELD) with a C-statistic of 0.981 (95% CI 0.968-0.995) was better than Hy's law (C-statistic 0.569; 95% CI 0.449-0.689). The latency, course of peak alanine aminotransferase decreasing >= 50% after discontinuation of herb application, peak triglyceride value, and platelet count at liver injury onset were identified as independent risk factors for chronicity with the adjusted odds ratios of 1.268 (95% confidence interval [CI] 1.034-1.554), 2.303 (95% CI 1.588-3.340), 0.580 (95% CI 0.343-0.978), and 0.183 (95% CI 0.091-0.368), respectively. A prognostic model for chronic HILI based on these four factors yielded the best prediction with a C-statistic of 0.812 (95% CI 0.755-0.868), compared with MELD (C-statistic 0.506; 95% CI 0.431-0.581) and Hy's law (C-statistic 0.418; 95% CI 0.343-0.492). Conclusion: Model for end-stage liver disease can be used to predict the fatal prognosis of HILI. A long latency, slow recovery, and low triglyceride value and platelet counts are important determinants for chronic HILI.