Trend patterns of HBsAg kinetics in chronic hepatitis B patients during nucleos(t)ide analogue therapy based on ARMA models

Abstract

Background: Trend pattern analysis are lacking for hepatitis B surface antigen (HBsAg) kinetics in chronic hepatitis B (CHB) patients during nucleos(t)ide analogue (Nuc) therapy. We evaluated the trend patterns of HBsAg kinetics by time series analysis and forecasting times to HBsAg seroclearance accordingly. Methods: A total of 116 CHB patients with documented three-month HBsAg levels during the previous more than five years of Nuc therapy were included. The piecewise linear trends of the autoregressive-moving average (ARMA) model were used for time series analysis of HBsAg kinetics trends. Best fitted models were created for each patient using HBsAg datasets with backtracking capability. Predicted time to HBsAg seroclearance was calculated accordingly. Results: Four trend patterns of HBsAg kinetics were found: no trend (n = 22, 19.0%), single trend (n = 16, 13.8%), biphasic trend with rapid-slow decline (n = 56, 48.2%) and biphasic trend with rise-decline (n = 22, 19.0%). Except for no-trend patients, the trend became slow reduction as HBsAg declined. Only 6.1% of patients continued rapid decline when the initial HBsAg of the last trend reached <100 IU/mL. Last trend slopes < −10 and rise-decline patterns indicate greater chances of achieving HBsAg seroclearance within two years. Conclusion: Best fitted ARMA models of HBsAg kinetics can be created individually for patients during Nuc therapy. About 67.2% patients have biphasic trend patterns, suggesting the dynamic nature of HBsAg kinetics over time. Trend patterns and last trend slopes predict individual times to HBsAg seroclearance.

Keywords

• ARMA model
• HBsAg kinetics
• Nucleas(t)ide analogue
• Time series analysis