Association between BMD and coronary artery calcification: an observational and Mendelian randomization study

Abstract

Observational studies have reported inconsistent associations between bone mineral density (BMD) and coronary artery calcification (CAC). We examined the observational association of BMD with CAC in 2 large population-based studies and evaluated the evidence for a potential causal relation between BMD and CAC using polygenic risk scores (PRS), 1- and 2-sample Mendelian randomization (MR) approaches. Our study populations comprised 1414 individuals (mean age 69.9 yr, 52.0% women) from the Rotterdam Study and 2233 individuals (mean age 56.5 yr, 50.9% women) from the Framingham Heart Study with complete information on CAC and BMD measurements at the total body (TB-), lumbar spine (LS-), and femoral neck (FN-). We used linear regression models to evaluate the observational association between BMD and CAC. Subsequently, we compared the mean CAC across PRSBMD quintile groups at different skeletal sites. In addition, we used the 2-stage least squares regression and the inverse variance weighted (IVW) model as primary methods for 1- and 2-sample MR to test evidence for a potentially causal association. We did not observe robust associations between measured BMD levels and CAC. These results were consistent with a uniform random distribution of mean CAC across PRSBMD quintile groups (P-value > .05). Moreover, neither 1- nor 2-sample MR supported the possible causal association between BMD and CAC. Our results do not support the contention that lower BMD is (causally) associated with an increased CAC risk. These findings suggest that previously reported epidemiological associations of BMD with CAC are likely explained by unmeasured confounders or shared etiology, rather than by causal pathways underlying both osteoporosis and vascular calcification processes.

Keywords: BMD; Mendelian randomization; coronary artery calcification; osteoporosis; polygenic risk score.

Figure 1

The scatter plot of the causal association between BMD at different sites and CAC.^{a a}CAC: ln(CAC score + 1); each point represents a SNP, and SNP’s effect on CAC and BMD is shown in vertical and horizontal lines separately. The x-axis represents the SNPs effect on BMD and y-axis represents the SNPs’ effect on CAC. The slope of the regression line corresponds to the causal estimate between BMD at different sites and CAC using 3 different models: inverse variance weighted, MR-Egger, and weighted median. (A) Causal association between TB-BMD and CAC. (B) Causal association between LS-BMD and CAC. (C) Causal association between FN-BMD and CAC. (D) Causal association between e-BMD and CAC.