OBJECTIVE Oral nucleoside/nucleotide analogues (NAs) have been advocated for chronic hepatitis B (CHB) treatment with good efficacy. However, less attention has been put on their adverse events. Therefore, a Bayesian network meta-analysis (NMA) was performed to evaluate the relative safety of five NAs (lamivudine, adefovir dipivoxil, entecavir, telbivudine, and tenofovir disoproxil fumarate) in CHB treatment among adults. METHODS Eligible randomized clinical trials (RCTs) and prospective cohort studies were systematically and thoroughly searched until May 1, 2019. Poisson-prior-based Bayesian NMA was performed to synthesize both direct and indirect evidence with reporting hazard ratios (HRs) and 95% credible intervals (CrIs) for serious adverse events (SAEs) and hepatic/renal impairments. RESULTS Thirty-three RCTs and 11 prospective cohort studies were identified. As to SAEs, no statistically significant difference was found of any comparison among five NAs. In terms of hepatotoxicity, lamivudine was safer than telbivudine (HR 0.45; 95% CrI 0.21, 0.85), and entecavir increased the risk by 102% (entecavir vs lamivudine: HR 2.02; 95% CrI 1.19, 3.27). CONCLUSIONS The findings from this large NMA could influence clinical practice, and the methodological framework of this study could provide evidence-based support to analyze sparse safety data in the field.