Group-based trajectory modeling for declines in hepatitis B surface antigen over time accurately identifies chronic...

Identifying chronic hepatitis B (CHB) patients likely to achieve hepatitis B surface antigen (HBsAg) sero-clearance upon receiving pegylated interferon α (PEG-IFNα) could aid in optimizing treatment strategies for "functional cure".

To develop a group-based trajectory model (GBTM) for identifying PEG-IFNα-treated CHB patients likely to achieve HBsAg sero-clearance.

Baseline characteristics were obtained from 342 PEG-IFNα-treated CHB patients, divided into training (205), validation (88), and testing (49) datasets via time series partitioning and random partitioning methods. GBTM was applied to the training dataset to identify the optimal trajectory groups for HBsAg levels over the follow-up period. Their accuracies for HBsAg sero-clearence were evaluated using receiver operating characteristic (ROC) and calibration curves with 1000 bootstrap samples. Clinical utility was examined using decision curve analysis (DCA), and the predictive GBTM-based 3-group trajectory model was incorporated into a website.

131 patients (38.30%) were functionally cured, and GBTM identified 3 trajectory groups, associated with rapid (traj_1), moderate (traj_2), and slow (traj_3) HBsAg declines. Traj_3 had the highest viral burden, plus lowered liver functions. The GBTM-based 3-group trajectory model was highly accurate under ROC, with areas under the curve >0.920, while calibration curves had high predicted-actual outcome correspondence, for all 3 datasets. Clinical utility was also high under DCA, and the website correctly found the trajectory group of a CHB patient, based on HBsAg from ≥2 different time points.

The GBTM-based 3-group trajectory model correctly identified PEG-IFNα-treated CHB patients likely to be "functionally cured", thereby providing a tool for devising personalized therapeutic strategies.