Epidemiological studies suggested that the frequency of tooth brushing might be associated with the risk of diabetes mellitus (DM), but the results were inconsistent, and no systematic review was conducted to focus on this topic. In this meta-analysis, we synthesized available observational epidemiological evidences to identify the association between tooth brushing and DM risk and investigate the potential dose-response relationship of them. We searched PubMed and Embase from their inception through December 2017 to identify observational studies examining the association between tooth brushing and the risk of DM. Reference lists from retrieved articles were also reviewed. We quantitatively combined results of the included studies using a random-effects model. Dose-response meta-analysis was conducted to further examine the effect of tooth brushing frequency on DM risk. We identified 20 relevant studies (one cohort study, 14 case-control studies, and 5 cross-sectional studies) involving 161 189 participants and 10 884 patients with DM. Compared with the highest tooth brushing frequency, the lowest level was significantly associated with an increased risk of DM (OR 1.32; 95% CI, 1.19-1.47), and there was no significant heterogeneity across the included studies (p=0.119, I2 =28.1%). Exclusion of any single study did not materially alter the combined risk estimate. The dose-response analysis indicated that the summary odds of DM for an increment of one time of tooth brushing per day was 1.20 (95% CI, 1.16-1.24). Integrated epidemiological evidence supports the hypothesis that low frequency of tooth brushing may be a risk factor of DM, and lower frequencies of tooth brushing were significantly associated with higher risk of DM.