Anti-resorptive therapy (AT) increases insulin resistance and decrease insulin secretion through reduced undercarboxylated osteocalcin in mice. However, there are inconsistent findings regarding the impact of AT use on the risk of diabetes mellitus in humans. We examined the association between AT and incident diabetes mellitus using classical and Bayesian meta-analysis. We searched Pubmed, Medline, Embase, Web of Science, Cochrane, and Google Scholar for studies listed from database inception to 25 February 2022. Randomized controlled trials (RCTs) and cohort studies reporting associations of estrogen therapy (ET) and non-estrogen anti-resorptive therapy (NEAT) with incident diabetes mellitus were included. Two reviewers independently extracted research data such as ET and NEAT, diabetes mellitus, risk ratios (RRs), and 95 % confidence intervals (CIs) for incident diabetes mellitus associated with ET and NEAT from individual studies. This meta-analysis included data from nineteen original studies, consisting of fourteen ET and five NEAT studies. ET was associated with reduced risk of diabetes mellitus in the classical meta-analysis (RR:​ 0.90;​ 95 % CI:​ 0.81–0.99). Slightly stronger results were found in the meta-analysis of RCTs (RR:​ 0.83;​ 95 % CI:​ 0.77–0.89). The probability that RR < 1.0 was 95 % in the overall analysis and 99 % in RCTs under weakly informative prior. Although NEAT was associated with reduced risk of diabetes mellitus overall (RR:​ 0.80;​ 95 % CI:​ 0.67–0.97), this was not found in the RCT meta-analysis (RR:​ 0.90;​ 95 % CI:​ 0.75–1.10). Under weakly informative prior, the probabilities that NEAT reduces diabetes mellitus by >0 % were 99 %, and 73 % in the overall and RCT meta-analysis, respectively. In conclusion, meta-analysis provided consistent evidence against the hypothesis that AT increases diabetes risk. ET may reduce the risk of diabetes mellitus. Whether NEAT reduces the risk of diabetes mellitus is uncertain and requires additional evidence from RCTs.