Value of a new predictive model in evaluating short-term prognosis of patients with hepatitis B virus related acute-on-chronic liver failure

Abstract

To establish a predictive model and investigate its value in evaluating short-term prognosis of patients with hepatitis B virus related acute-on-chronic liver failure (HBV-ACLF). Patients with HBV-ACLF admitted to Tianjin Second People's Hospital and Beijing Youan Hospital, Capital Medical University from May 2015 to October 2018 were enrolled. The data of gender, age, laboratory markers at admission, model for end-stage liver disease (MELD) score and clinical complications were collected for analysis. According to the prognosis on 12-week, patients were divided into survival group and death group. Univariate analysis and binary Logistic regression analysis were used to test the risk factors for short-term prognosis of the patients with HBV-ACLF, and a prediction model was established. The accuracy of each index and the established model were verified by the receiver operating characteristic (ROC) curve. A total of 148 patients with HBV-ACLF were enrolled in the study, 91 cases survived while 57 cases died during the 12-week period. The age, total bilirubin (TBIL), neutrophil percentage (NEUT%), hepatitis B surface antigen (HBsAg), MELD score of death group were higher than those of survival group [age (years old): 50.00 (44.50, 55.00) vs. 43.00 (34.00, 53.00), TBIL (μmol/L): 310.30 (240.70, 405.70) vs. 266.40 (184.20, 360.20), NEUT%: (74.52±13.05)% vs. (66.64±12.35)%, lgHBsAg (kU/L): 3.72 (3.29, 3.92) vs. 2.97 (2.49, 3.78), MELD score: 24.27 (19.71, 27.40) vs. 21.88 (18.83, 24.38), all P < 0.05], while albumin (ALB), total cholesterol (CHO), prothrombin activity (PTA) and alpha-fetoprotein (AFP) were lower than those of survival group [ALB (g/L): 29.80 (27.05, 31.05) vs. 30.80 (28.00, 33.90), CHO (mmol/L): 1.98 (1.50, 2.38) vs. 2.49 (2.05, 3.01), PTA: (30.37±7.09)% vs. (32.94±6.03)%, AFP (μg/L): 21.54 (9.28, 51.54) vs. 66.16 (24.50, 152.80), all P < 0.05]. Logistic regression analysis showed that NEUT%, HBsAg and AFP were independent risk factors for short-term prognosis of patients with HBV-ACLF [odds ratio (OR) was 77.843, 1.439, 0.995, respectively, all P < 0.05]. According to the results of regression analysis, the NHA-ACLF model (NEUT%+HBsAg+AFP) was established. The formula was logit (NHA-ACLF) = -5.441+5.688×NEUT%+0.430×lgHBsAg-0.005×AFP. The area under the ROC curve (AUC) of the NHA-ACLF model for pred HBV-ACLF patients was 0.790, which was better than NEUT% (AUC = 0.696), lgHBsAg (AUC = 0.670), AFP (AUC = 0.703) and MELD score (AUC = 0.640). When the cut-off value of NHA-ACLF model score was 0.459, the sensitivity was 73.7%, and the specificity was 79.1%. NEUT%, HBsAg and AFP are independent predictive indicator for short-term prognosis in patients with HBV-ACLF. Compared with MELD score, the risk assessment model NHA-ACLF has a greater value in predicting the short-term prognosis of patients with HBV-ACLF.