Background: Acute-on-chronic hepatitis B liver failure (ACHBLF) is associated with poor short-term prognosis. The aim of the present study was to construct and validate a model for end-stage liver disease (MELD)-based nomogram for the 3-month mortality estimation for patients with ACHBLF.
Methods: A total of 551 patients with ACHBLF were prospectively enrolled from 2 independent medical centers and divided into 2 cohorts of training and validation, respectively. The 3-month mortality was recorded as the outcome. The MELD-based nomogram was constructed to predict the 3-month mortality for ACHBLF using the training group of 335 patients and validated using an independent cohort of 216 patients. The predictive capability of MELD-based nomogram was compared with the MELD score system by calibration analysis, receiver operating characteristics (ROC) and decision curve analysis in both training cohort and validation cohort.
Results: Multivariate analysis suggested that age, serum sodium, and MELD score were independent prognostic indicators associated with the 3-month mortality for ACHBLF, and therefore used for developing the nomogram. In terms of calibration, the predicted survival by the MELD-based nomogram was found to be extremely in line with the observed 3-month mortality both in training cohort and validation cohort. Additionally, both ROC and decision curve analyses showed that the MELD-based nomogram was better than MELD, MELD-Na, MELDNa, and iMELD for ACHBLF prognosis prediction. The results were confirmed in the external cohort of validation.
Conclusions: The MELD-based nomogram provided a user-friendly, accurate and reproducible tool for predicting 3-month mortality of patients with ACHBLF.
Keywords: Acute-on-chronic hepatitis B liver failure; Calibration; Decision curve; Nomogram; Prediction; Receiver operating characteristics.
Copyright © 2017 Elsevier B.V. All rights reserved.