Background: We aimed at investigating causal associations between matrix metalloproteinases (MMPs) and bone mineral density (BMD) by the Mendelian randomization (MR) analysis.Methods: From genome-wide association studies of European ancestry, we selected instrumental variables for MMP-1, MMP-3, MMP-7, MMP-8, MMP-10, and MMP-12. Accordingly, we retrieved summary statistics of three site-specific BMD, namely, forearm, femoral neck, and lumbar spine. We conducted an inverse variance weighted MR as the primary method to compute overall effects from multiple instruments, while additional MR approaches and sensitivity analyses were implemented. Bonferroni-adjusted significance threshold was set at p < 0.05/18 = 0.003.Results: Totally, there was no evidence for causal effects of genetically-predicted levels of MMPs on BMD measurement at three common sites. MR results indicated that there were no causal associations of circulating MMPs with forearm BMD (all p ≥ 0.023) by the inverse variance weighted method. Similarly, there were no causal effects of MMPs on femoral neck BMD (all p ≥ 0.120) and MR results did not support causal relationships between MMPs and lumbar spine BMD (all p ≥ 0.017). Multiple sensitivity analyses suggested the robustness of MR results, which were less likely to be biased by unbalanced pleiotropy or evident heterogeneity.Conclusion: We found no evidence for the causal relationship between MMPs and BMD in the European population.