Background: We investigated the patterns and predictors for virological relapse (VR), clinical relapse (CR), and sustained clinical response (SCR) and the outcomes of retreatment after nucleos(t)ide analogue (NUC) therapy discontinuation. Methods: Patients with chronic hepatitis B who were discontinuing NUC therapy were prospectively enrolled. Viral and host predictors of relapse were evaluated, including hepatitis B virus (HBV) surface antigen (HBsAg) level, anti-HBV core antibody level, and presence of single-nucleotide polymorphisms in the genes encoding the receptors NTCP (rs2296651) and CTLA4 (rs231775) and in the 3' untranslated regions of the genes encoding HLA-DPA1 (rs3077) and HLA-DPB1 (rs9277535); posttherapy predictors of relapse were also investigated. Information about NUC retreatment and outcomes were recorded. Results: Overall, 100 patients discontinuing 3-year entecavir (ETV) or tenofovir (TDF) therapy were enrolled. Patients discontinuing TDF exhibited significantly higher rates of VR (52.9% vs 6.1%; P < .001) and CR (15.2% vs. 1.5%, P = .007) at 3 months than those discontinuing ETV, but relapse rates at 12 months were comparable. The end-of-therapy HBsAg levels predicted VR (hazard ratio [HR], 1.62; 95% confidence interval [CI], 1.19-2.21), CR (HR, 1.78; 95% CI, 1.13-2.81), and SCR (OR, 0.57; 95% CI, .35-.94). The CTLA4 (rs231775) non-GG genotype predicted VR (HR, 1.74; 95% CI, 1.01-3.00) and CR (HR, 2.06; 95% CI, 1.04-4.11), while the HLA-DPA1 (rs3077) AA genotype predicted SCR (OR, 10.84; 95% CI, 1.12-105). The HBV DNA 1 month after NUC treatment cessation was an early predictor of subsequent relapse. Conclusions: Discontinuation of tenofovir disoproxil fumarate treatment rather than entecavir treatment is associated with earlier relapse, and NUC-specific posttherapy monitoring is necessary.