Establishing a metabolomic model for the prognosis of hepatitis B virus-induced acute-on-chronic liver failure treated with different liver support systems

Liver failure induced by hepatitis B virus (HBV) is a severe disease with a high mortality rate. Liver support treatment is a powerful method for treating liver failure and is a bridge to liver transplantation. Patients with similar liver function indices, however, have different outcomes following treatment, and no satisfying prediction parameters exist. In this study, we used ultraperformance liquid chromatography-mass spectrometry (UPLC-MS) to investigate plasma metabolites in groups of acute-on-chronic liver failure patients with different prognosis. An orthogonal partial least squares (OPLS) model, with satisfactory explanatory and predictive ability (R ² Y?=?0.943, Q ²?=?0.913) was established using SIMCA-P?+?12.0. The model was based on samples collected just before the first artificial treatment for comparable model efficacy. The concordance statistics of our model was 0.968 (95% CI 0.951, 0.985]) which is superior to that of the MELD (the Model for End-stage Liver Diseases) score (0.737, 95% CI 0.578, 0.896]). Three groups of markers were identified: lysophosphatidylcholine, primary fatty acid amides and conjugated bile acids. Lysophosphatidylcholine and conjugated bile acids were protection factors for survival and primary fatty acid amides were risk factors. The cut-off point for the predictive value of our model was greater than or equal to 0.196, at which the model showed the best discrimination between the recovery and non-recovery groups, with a sensitivity of 95% and a specificity of 87%. This metabolomic model, based on plasma UPLC-MS profiles, provides not only excellent discrimination and prognostic ability for HBV-induced acute-on-chronic liver failure but also early and precise warning of possible liver transplantation.