The number of SNPs selected as instrumental variables that were significantly correlated with exposure ranged from 4 to 1079. Their explained variances varied from 2.5 to 31.7%. The F statistics for each SNP and the general F statistics were all greater than 10 (Table 1).
Bidirectional MR analysis
In the MR analysis from breast cancer to BMD, there was an inverse causal relationship between breast cancer, its ER+ subtype and HE-BMD. Breast cancer might reduce HE-BMD and was recognized as a risk factor for osteoporosis (OR 0.980, 95% CI 0.970–0.990, FDR = 1.51 × 10−4) as was the ER+ subtype (OR 0.979, 95% CI 0.969–0.989, FDR = 1.51 × 10−4) (Fig. 2A, Supplemental Table S1). The ER− subtype had no causal relationship with HE-BMD (OR 1.034, 95% CI 0.998–1.072, FDR = 7.69 × 10−2). No causal relationship was found between breast cancer, its subtype and BMD in other sites—LS, FN, FA. Heterogeneity was found in Mendelian randomization of breast cancer, ER+ subtype to HE-BMD, LS-BMD and FN-BMD. When heterogeneity existed in sensitivity analysis, the statistics of weighted median method were in the same direction as that of IVW models, and the weighted median method was selected as the main statistical effect. Horizontal pleiotropy was found for the ER− subtype to HE-BMD. The MR-Egger result was adopted as the main effect size. The original results of IVW, weighted-median, and MR-Egger between breast cancer and BMD can be found in Supplemental Table S2, together with the heterogeneity and pleiotropy tests.
The forest plot of bidirectional Mendelian randomization results. (A) is the Mendelian randomization results of the association of breast cancer and its subtypes with bone mineral density at different sites. (B) is the Mendelian randomization results of the association of bone mineral density at different sites with breast cancer and its subtypes.
In the MR analysis from BMD to breast cancer, no significant association was found (FDR > 0.05, Fig. 2B). There was heterogeneity in the MR statistics analysis of HE-BMD in breast cancer and its ER+ and ER− subtypes. No horizontal pleiotropy was found for BMD in any region to breast cancer in MR analysis. Supplemental Table S3 shows the results of IVW, weighted-median, and MR-Egger between BMD and breast cancer, together with the heterogeneity and pleiotropy tests.
Multivariate mediated MR analysis
In the MR analysis of hormone levels to BMD, there was a positive causal relationship between FT, TT and HE-BMD. Higher FT could increase HE-BMD and was considered a protective factor for osteoporosis (OR 1.116, 95% CI 1.086–1.147, FDR = 1.04 × 10−14); similar results could be seen with TT (OR 1.039, 95% CI 1.013–1.066, FDR = 2.60 × 10−4) (Fig. 3, Supplemental Table S1). FT and TT had no causal relationship with LS-BMD, FN-BMD, or FA-BMD. Heterogeneity was found in the MR analysis of FT, TT to HE-BMD, LS-BMD, FN-BMD, and FA-BMD. The weighted median was adopted as the main method. No horizontal pleiotropy was found for any of the hormone levels to BMD in MR analysis. The original results of IVW, weighted-median, and MR-Egger between hormone levels and BMD can be found in Supplemental Table S4, together with the heterogeneity and pleiotropy tests. The multivariate MR analysis suggested that elevated FT may be an independent protective factor for HE-BMD (adjusted OR = 1.076, P = 0.033), while TT was not significant in multivariate MR model (adjusted OR = 1.023, P = 0.520). The results of mediating MR analysis suggested that FT mediated 71.5% of the causal relationship between TT and HE-BMD. There were causal relationships between breast cancer, its ER+ subtype, FT and HE-BMD. The multivariate MR analysis showed that breast cancer corrected for FT mediating factor had no significant adverse effect on HE-BMD (adjusted OR = 0.982, P = 0.077), while both its ER+ type and FT were independent factors of HE-BMD (ER+: adjusted OR = 0.977, P = 0.021; FT: adjusted OR = 1.111, P = 6.88 × 10−6). The mediating MR analysis showed that FT only mediated the causal relationship between breast cancer and HE-BMD by 2.9%. The effect of breast cancer and its ER+ subtypes adjusted FT on HE-BMD was shown in Fig. 4.
The forest plot of Mendelian randomization results between testosterone levels and bone mineral density at different sites.
The forest plot and mediating ratios of breast cancer and its ER+ subtypes before and after FT adjustment on the bone density of the heel.
In the MR analysis of hormone levels in breast cancer, there was a positive causal relationship between FT and TT and breast cancer and its ER+ subtype. Higher FT could increase the risk of breast cancer (OR 1.137, 95% CI 1.054–1.226, FDR = 1.21 × 10−3) as could TT (OR 1.130, 95% CI 1.040–1.227, FDR = 5.81 × 10−3) (Fig. 5, Supplemental Table S1). Similarly, higher FT could increase the risk of ER+ subtype (OR 1.235, 95% CI 1.136–1.344, FDR = 2.51 × 10−6) as could TT (OR 1.213, 95% CI 1.096–1.343, FDR = 5.55 × 10−4) (Fig. 5, Supplemental Table S1). FT and TT had no causal relationship with the ER− subtype of breast cancer (FDR > 0.05, Fig. 5). Heterogeneity was found in the MR analysis of FT and TT in breast cancer and its ER+/ER− subtype, so the weighted-median results were adopted as the main effect. No horizontal pleiotropy was found for any of the hormone levels in breast cancer in MR analysis. The original results of IVW, weighted-median, and MR-Egger between hormone levels and breast cancer can be found in Supplemental Table S5, together with the heterogeneity and pleiotropy tests.
The forest plot of Mendelian randomization results between testosterone levels and breast cancer.
The statistical power for exposure to FT and TT to outcomes HE-BMD, breast cancer, and ER+ were all 100%; however, the statistical power of exposure to breast cancer and ER+ in HE-BMD outcomes was 74% and 69%, respectively.
