Abstract
OBJECTIVE: While the phenotypic associations of menstrual traits, such as age at menarche (AAM) and age at natural menopause (ANM), with bone mineral density (BMD) have been well observed, the understanding of their shared genetic mechanisms is lacking. We aimed to systematically explore the underlying genetic basis connecting AAM and ANM with BMD. METHODS: We performed a large-scale genome-wide cross-trait analysis by leveraging summary statistics from the hitherto largest genome-wide association studies conducted among the European population for AAM (n = 556,124), ANM (n = 201,323), and heel estimated BMD (eBMD, n = 426,824), a robust validated predictor of osteoporosis risk. RESULTS: We identified significant genetic correlations for eBMD with both AAM (r g = -0.082, P = 1.83 × 10-8) and ANM (r g = 0.044, P = 0.007). Cross-trait meta-analysis yielded 203 AAM-eBMD shared loci, of which 3 were novel and 77 ANM-eBMD shared loci, of which two were novel. Gene-based analysis revealed 409 AAM-eBMD shared genes and 179 ANM-eBMD shared genes. Mendelian randomization demonstrated that genetically predicted later AAM (β = -0.054, 95% CI = -0.069 to -0.040, P = 6.08 × 10-14) and genetically predicted earlier ANM (β = 0.010, 95% CI = 0.004-0.017, P = 0.003) were significantly associated with decreased levels of eBMD. No evidence of reverse causality was found. CONCLUSION: Our work provides evidence in support of a substantial shared genetic basis and causal relationships between menstrual traits and BMD. The findings could be instrumental in developing risk stratification strategies and formulating novel pharmaceutical interventions for osteoporosis. STRENGTHS AND LIMITATIONS OF THE STUDY: Genome-wide cross-trait design provided insights into the shared genetic basis and biological mechanisms underlying the phenotypic associations between AAM, ANM, and eBMD. The adoption of the hitherto largest GWAS summary statistics ensured statistical power. The genetic data were exclusively from European ancestry, limiting the generalizability of our results. Current findings are primarily derived from bioinformatic analyses and interpreted based on existing, incomplete biological knowledge of the SNPs and genes.