The aim of this study was to determine whether the changing patterns of quantitative hepatitis B e antigen (HBeAg) levels by serial monitoring could predict HBeAg seroconversion and viral breakthrough during lamivudine therapy. We retrospectively analysed 340 HBeAg positive naive chronic hepatitis B patients treated with lamivudine. The mean duration of lamivudine therapy was 18.7 (range 6-56) months. The changing patterns and reduction rates of pretreatment HBeAg levels by serial monitoring were categorized into three groups: Decrescendo group (n = 195), Decrescendo-crescendo group (n = 65) and no changing or fluctuating group (n = 80). Of 109 patients who had achieved HBeAg seroconversion, 105 (96.3%) were included in the decrescendo group. The decrescendo group, pretreatment quantitative HBeAg levels, alanine aminotransferase levels, and the duration of lamivudine therapy were independent predictive factors for HBeAg seroconversion. Of 82 patients who had viral breakthrough, 53 (64.6%) were in the decrescendo-crescendo group and 21 (25.6%) were in the no changing or fluctuating group. The only predictive factor for viral breakthrough was the changing patterns of quantitative HBeAg levels, especially, the decrescendo-crescendo group and the no changing or fluctuating group. The mean time of turning points in the decrescendo-crescendo group was 7.1 months earlier than the mean time of viral breakthrough (9.0 months vs 16.5 months). Therefore, the changing patterns of quantitative HBeAg levels by serial monitoring during lamivudine therapy may allow not only the prediction of treatment responses, but also an early recognition of a viral breakthrough.