Early on-treatment dynamics predicting hepatitis B e antigen seroconversion in chronic hepatitis B

Area of Science:

• Hepatology
• Virology
• Immunology

Background:

• Hepatitis B e antigen (HBeAg) loss and anti-HBe appearance signify a key milestone in nucleos(t)ide analogues (NAs) therapy.
• Static baseline biomarkers poorly predict this durable immunological event.
• Early dynamic parameters during the initial 12 weeks of NA therapy may offer superior prognostic information.

Purpose of the Study:

• To develop a predictive model for week-60 HBeAg seroconversion (SR).
• To integrate week-12 changes in serum biochemical and virological markers for enhanced prediction.
• To evaluate the model's performance against baseline values.

Main Methods:

• Collected serum biochemical and virological data from HBeAg-positive patients on NA treatment.
• Derived difference, ratio, and absolute week-12 values for key parameters.
• Utilized least absolute shrinkage and selection operator (LASSO) logistic regression with ten-fold cross-validation to build a multi-variable model.

Main Results:

• 38.24% of patients achieved HBeAg SR.
• An eight-variable signature involving dynamic changes in HBeAg, HBsAg, GGT, and albumin was selected.
• The multi-variable model achieved high AUC (0.975 in training, 0.721 in testing) with excellent sensitivity and specificity.

Conclusions:

• Integrated early on-treatment dynamics of HBeAg, HBsAg, GGT, and albumin predict subsequent HBeAg SR.
• The developed eight-variable model can aid in patient monitoring.
• This model may inform alternative treatment strategies for patients undergoing NA therapy.