The cessation criteria for lamivudine treatment vary in published articles and their results are contradictory, especially factors predicting relapse. To clarify these contradictions, this long-term follow-up study of 125 Chinese hepatitis B e antigen (HBeAg)-positive chronic hepatitis B patients was designed with stringent cessation criterion. All patients received lamivudine and achieved HBeAg seroconversion (group A, n = 82) or loss (group B, n = 43) with undetectable hepatitis B virus (HBV) DNA by PCR assay during the treatment. Lamivudine was withdrawn ≥6 months after HBeAg seroconversion/loss occurred. The median treatment durations were 24 (12-54) months and 36 (18-89) months in group A and group B, respectively. Patients were followed up for median 24 (2-84) months. The cumulative relapse (defined as serum HBV DNA ≥10 4 copies/mL) rates in the two groups at months 12, 24, 36 and 48 were 23.4%vs 35.0%, 25.0%vs 37.7%, 25.0%vs 41.1% and 29.4%vs 41.1%, respectively (log-rank test, P = 0.119). For patients whose total treatment duration ≥18 months in group A, the cumulative relapse rates at months 12, 24, 36 and 48 were 18.3%, 20.1%, 20.1% and 25.1%, which was significantly lower than those with a shorter duration (log-rank test, P = 0.002). The mean age and median total duration were statistically different between relapsers and nonrelapsers in group A (33.9 ± 13.6 vs 23.1 ± 11.0 years, P < 0.001 and 24 vs 26 months, P = 0.003). Cox regression revealed that age was the only predictive factor for relapse (RR, 1.069; 95% CI, 1.032-1.106, P < 0.001). Patients aged <30 years relapsed less frequently in 5 years (12.3%vs 53.5%, P = 0.001). In conclusion, for patients who maintained HBeAg seroconversion for ≥6 months and total duration for ≥18 months, lamivudine withdrawal is a reasonable option. Prolonged treatment may be required for patients aged greater than 30 years to reduce relapse. © 2009 Blackwell Publishing Ltd.