BackgroundHBeAg loss is an important endpoint for antiviral therapy in chronic hepatitis B (CHB), however there are no reliable biomarkers to identify patients who will respond to the addition of pegylated interferon to nucleos(t)ide analogue (NA) therapy. AimTo evaluate the use of serum biomarkers to predict HBeAg loss. MethodsHBeAg positive CHB participants on NAs who switched-to or added-on 48 weeks pegylated interferon alpha2b (clinicaltrial.gov NCT01928511) were evaluated at week 72 for HBeAg loss. The predictive ability of qHBeAg, qHBsAg, HBV RNA and clinical variables for HBeAg loss were investigated. ResultsHBeAg loss occurred in 15/55 (27.3%) participants who completed 48 weeks of pegylated interferon. There was a lower baseline qHBeAg (1.18 IU/mL [2.27] versus 10.04 IU/mL [24.87], P = 0.007) among participants who lost HBeAg. Baseline qHBeAg (OR = 0.15, 95% CI 0.03–0.66, P = 0.01) and detectable HBV DNA at baseline (OR = 25.00, 95% CI 1.67–374.70, P = 0.02) were independent predictors of HBeAg loss. In addition, on-treatment qHBeAg was also a strong predictor of HBeAg loss (OR = 0.39, 95% CI 0.18–0.81, P = 0.012). The models combining detectable baseline HBV DNA with baseline (C-statistic 0.82) and on-treatment (C-statistic 0.83) had good accuracy for predicting HBeAg loss. A rise in qHBeAg ≥ 10 IU/ml was a predictor of flare (ALT ≥ 120 U/ml) on univariable analysis but not after adjustment for treatment arm. ConclusionsBaseline and on-treatment qHBeAg is a useful biomarker that can identify participants on NA therapy who may benefit from adding or switching to pegylated interferon.