The GOLDEN Model Effectively Predicts Hepatitis B Surface Antigen Seroclearance in Untreated Patients in Real World Clinical Practice

ABSTRACT The GOLDEN model efficacy in predicting the seroclearance of hepatitis B surface antigen (HBsAg) had been validated in chronic hepatitis B patients on antiviral treatment. This study aims to investigate whether this model can also predict HBsAg seroclearance in untreated chronic hepatitis B virus (HBV) infection patients in the real‐world practice. A total of 3315 patients, followed for 72 months (median), were divided into GOLDEN‐favorable (GOLDEN score > 0.95) and GOLDEN‐unfavorable group. The incidence of HBsAg seroclearance between the two groups was compared with the Kaplan–Meier method. The predictive efficacy of the GOLDEN model, HepBLOSS model and baseline HBsAg levels was compared by receiver operating characteristic (ROC) and area under the ROC curve (AUC). At baseline, the median age was 41 (32, 49), 66.40% were male, and 11.04% had cirrhosis. 18.34% were HBeAg positive. 40.60% had undetectable HBV DNA levels. The cumulative HBsAg seroclearance in the GOLDEN‐favorable group was 13.27%, while 0% in the unfavorable group. The AUCs of the GOLDEN model for predicting HBsAg seroclearance were 0.96 (95%CI: 0.95–0.97), which were significantly higher than those of the HepBLOSS models ( p < 0.05). The GOLDEN‐favorable patients had significantly higher cumulative HBsAg seroclearance rates than those of GOLDEN‐unfavorable patients stratified by baseline HBsAg, HBeAg, HBV DNA, and cirrhosis status. The AUCs for predicting HBsAg seroclearance were significantly more robust than baseline HBsAg levels. The GOLDEN model is an effective tool, superior to that of the HepBLOSS model and baseline HBsAg levels, in predicting HBsAg seroclearance in untreated chronic HBV patients.