Association of Genetic Variants Related to Serum Calcium Levels with Reduced Bone Mineral Density

Abstract

Context

The role of serum calcium in bone metabolism is unknown, even though calcium/vitamin D supplementations have been widely used and are expected to improve bone health. We aim to determine the independent role of serum calcium in bone mineral density (BMD).

Design and setting

Two epidemiological analyses with 5478 and 5556 participants from the National Health and Nutrition Examination Survey (NHANES) 2003 to 2006 and the Hong Kong Osteoporosis Study (HKOS) to evaluate the cross-sectional association of serum calcium with BMD. Two-sample Mendelian randomization (MR) studies using genetic variations as instrumental variables to infer causality. Summary statistics of genome-wide association study of serum calcium (N = 39 400) and lifelong whole-body BMD (N = 66 628) were used.

Main outcome measure

BMD measured by dual-energy X-ray absorptiometry

Results

In NHANES 2003–6 and HKOS, each standard deviation (SD) increase in serum calcium was significantly associated with 0.036–0.092 SD decrease in BMD at various sites (all P < .05). In multivariable inverse-variance weighted MR analysis, genetic predisposition to higher serum calcium level was inversely associated with whole-body BMD after adjustment for serum parathyroid hormone, vitamin D, and phosphate (–0.431 SD per SD increase in serum calcium; 95% CI: –0.773 to –0.089, P = .014). Similar estimates were obtained in sensitivity analyses.

Conclusions

Our study reveals that genetic predisposition to higher serum calcium level per se may have a negative impact on bone metabolism. Whether increased serum calcium caused by calcium/vitamin D supplementations would have the same negative effect on bone remains unknown, which warrants further investigation. In addition to other adverse clinical outcomes, careful use of high-dose supplementations is required.

Calcium plays a crucial role in bone metabolism. Calcium supplementation has long been promoted to improve bone health in older adults. A recent meta-analysis of randomized controlled trials (RCTs) reported that calcium supplementation has minimal effect on bone mineral density (BMD) regardless of dose and duration (improved BMD by 0.7–1.8% only), suggesting calcium supplementation might not lead to a significant reduction in fracture risk (1). Since the relationship between calcium supplementation and BMD is complex and many older adults use calcium supplementation as a means of fracture prevention, it is important to examine the effect of calcium on bone.

Calcium homeostasis is controlled by calcium absorption, excretion, and storage, mainly through hormonal systems including parathyroid hormone (PTH) and 1,25-dihydroxyvitamin, whereas dietary intake (such as calcium and vitamin D supplementation) can also affect the serum calcium level. A recent interventional study in humans showed that taking 1000 mg of calcium supplementation daily leads to both acute and sustained increase in serum calcium (2). A similar increase in serum calcium was also observed after vitamin D supplementation in an RCT (3), but it is uncertain if such an increase in serum calcium plays a role in bone metabolism. In addition, previous observational studies showed inverse (4) and null (5) association of serum calcium with BMD. It remains largely unclear if serum calcium is associated with BMD. Even if there is a consistent significant association, such association is confounded by multiple bone and mineral-related factors, such as vitamin D, phosphate, and PTH. Thus, it is difficult to evaluate the independent effect of serum calcium on bone metabolism amid other bone and mineral factors, especially in a human interventional study.

Mendelian randomization (MR) is an analytical approach that allows unbiased causal effects to be estimated, which is considered a complementary approach to RCT (6). This approach can avoid the effect of confounders, and is particularly suitable for evaluating the independent role of serum calcium in bone metabolism. In this study, we first examined the independent association of serum calcium with BMD by adjustment for serum phosphate, vitamin, and PTH in two epidemiological studies with distinct genetic compositions: the continuous National Health and Nutrition Examination Survey (NHANES) conducted in 2003 to 2006 and the Hong Kong Osteoporosis Study (HKOS). Next, we inferred causality of the lifelong effect of serum calcium on whole-body BMD using an MR approach.

Materials and Methods

Participants in observational studies

Two independent epidemiological studies of different genetic ancestries were employed using data from NHANES 2003–2006 (7) and HKOS.

NHANES included a stratified multistage probability sample of individuals representative of the civilian noninstitutionalized US population. Selection was based on counties, blocks, households, and individuals within households. It also included an oversample of non-Hispanic blacks and Mexican Americans to ensure adequate sample size for estimation by race/ethnicity. Participants were required to sign a consent form before their participation, and ethics approval was obtained from the Human Subjects Committee of the US Department of Health and Human Services. In this cross-sectional study, we included participants aged ≥20, who had both BMD and biomarkers of mineral metabolism measured in 2003 to 2006 (N = 5478 men and women).

Details of the HKOS have been described elsewhere (8). The HKOS was initiated in 1995 and the cohort participants were community-dwelling southern Chinese men and women of Han descent recruited in Hong Kong from 1995 to 2010, with a total of 9449 participants recruited. In this cross-sectional study, we included participants aged ≥20, who had whole-body BMD, BMD at lumbar spine and hip, and biomarkers of mineral metabolism measured in the years 2003 to 2006 (N = 5556 men and women). Ethics approval was obtained from the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster (reference number: UW 15–236).

BMD measurements

For NHANES, BMD was measured using Hologic QDR-4500A dual-energy X-ray absorptiometry (DXA) (9). For HKOS, BMD was measured using Hologic QDR-2000plus and QDR-4500 plus systems, with in-house validation and in vivo precision tested for the two machines (8).

Serum biomarkers of mineral metabolism and other covariates

Details of measurements of serum biomarkers of mineral metabolism in NHANES are provided on the NHANES website (http://www.cdc.gov/nchs/nhanes.htm) (7). Details of measurements of serum biomarkers of mineral metabolism in HKOS have been described previously (10, 11).

For smoking and drinking status in NHANES, we used the same definition for never, current, and former smokers/drinkers as that used in the National Health Interview Surveys (NHIS) conducted by the Center for Disease Control (http://www.cdc.gov/nchs/nhis.htm). In HKOS, smoking status (never, former, and current smokers) and alcohol intake were self-reported as described elsewhere (12).

Epidemiological analyses of serum calcium and BMD

For the analyses in NHANES, serum calcium and BMD were standardized with a mean of 0 and standard deviation (SD) of 1 in a sex- (for BMD only) and ethnic/race-specific manner. For the analyses in the HKOS, serum calcium was standardized with a mean of 0 and SD of 1, while BMD T-score (a standardized BMD score that is used for diagnosis of osteoporosis) was used in the analysis. The relationship between serum calcium and BMD was evaluated using multivariable linear regression with adjustment for age, sex, ethnicity/race (NHANES only), height, and weight in the simple model, and further adjusted for smoking status, alcohol intake, serum phosphate, PTH, and 25-hydroxyvitamin D [25(OH)D] in the full model. For the NHANES analysis, sample weights that account for the unequal probabilities of selection, oversampling, and non-response were applied for all analyses using complex sampling module in SPSS version 22.0 software (SPSS Inc, Chicago, IL). All values presented were weighted to represent the US civilian population.

Study design and data sources for MR

We utilized publicly available summary statistics from large-scale genome-wide association study (GWAS) and meta-analysis (https://osf.io/m35bv/ (13)). Seven independent single nucleotide polymorphisms (SNPs) associated with serum calcium (with genome-wide significance) in the largest GWAS meta-analysis of 17 cohorts involving 39 400 individuals (14) were selected. The estimates of these 7 SNPs in other related biomarkers of mineral metabolism (serum PTH (15), phosphate (16), and/or 25(OH)D (17)) were also retrieved. Summary statistics for the association of the 7 SNPs with BMD were extracted from the largest GWAS of DXA-measured BMD (whole-body BMD) conducted by GEnetic Factors for OSteoporosis Consortium. The dataset comprised 66 628 participants from 30 epidemiological cohorts, of which 86% originated from European ancestry (18). These studies were approved by the relevant ethics committees, as mentioned in the respective GWAS publications (14–18).

MR analyses

MR analyses, including the primary analysis of inverse variance weighted (IVW) (19) and multivariable IVW methods (20), sensitivity analyses of weighted median (21), and MR–Egger (22), were conducted using the “MendelianRandomization” package in R (23). We conducted MR analyses to infer causality of serum calcium on DXA-measured whole-body BMD. Summary-level data were utilized to test for causal association between an exposure (serum calcium) and an outcome (whole-body BMD) by using genome-wide significant SNPs as instrumental variables. All the SNPs were oriented such that the effect alleles were positively associated with serum calcium. The effect alleles were matched between the summary data of serum calcium and BMD datasets. Primary analysis was done using the conventional IVW method (19). A multivariable IVW analysis was conducted by adjusting for beta estimates from serum PTH, phosphate, and 25(OH)D to account for their potential pleiotropy effects (20). Although IVW is the conventional method, the major drawback is that it assumes all instrumental variables are valid. Therefore, two sensitivity analyses were conducted: weighted median (21) and MR–Egger (22). The weighted median method provides consistent estimates even when up to 50% of the information comes from invalid instrumental variables (21). One of the major assumptions in MR analysis is the absence of horizontal pleiotropy, meaning that the genetic variants are not associated with the outcome except via the exposure. We additionally performed MR–Egger regression to detect bias arising from unbalanced pleiotropy in MR studies. The intercept represents the average pleiotropic effects across all SNPs, under the assumption that the magnitude of the pleiotropic effects are independent of the SNP–risk factor associations across all variants, also known as the INstrument Strength Independent of Direct Effect assumption (22). We conducted an additional sensitivity analysis by excluding 3 genetic instruments which were also significantly associated with other confounders affecting mineral metabolism. No bidirectional causation was tested for the effect of BMD on serum calcium as this relationship is well documented. All statistical analyses were conducted using R or SPSS.

Results

Cross-sectional analyses

We evaluated the relationship between serum calcium and BMD in NHANES and HKOS. Demographic characteristics of participants are shown in Tables 1 (NHANES) and 2 (HKOS). In the NHANES, each SD higher serum calcium was associated with lower whole-body BMD and BMD at the lumbar spine, with an estimate of –0.065 SD (95% CI: –0.099 to –0.031; P = .001) and –0.058 SD (95% CI: –0.087 to –0.029; P = .0003), respectively, in the simple model adjusted for age, sex, height, weight, and ethnicity/race (Table 3). No association was observed for BMD at the pelvis. In the full model after further adjustment for confounding factors such as serum phosphate, PTH, and vitamin D levels, the associations with whole-body BMD and BMD at the lumbar spine became stronger, whereas the association with BMD at the pelvis became statistically significant (estimate: –0.036 SD, 95% CI: –0.065 to –0.007, P = .016; Table 3).

Table 1.

Characteristics of NHANES 2003–6, (N = 5478)

Variables	Mean	SD
Age (years)	44.15	14.96
Female	2643	50.1%
Ethnicity/Race
Mexican American	1182	8.1%
Other Hispanic	186	3.7%
Non-Hispanic White	2779	73.1%
Non-Hispanic Black	1107	10.0%
Other Race	224	5.1%
Smoking
Never smoker	2773	49.9%
Ex-smoker	1341	23.9%
Current smoker	1364	26.2%
Current drinker	3892	74.6%
Height (m)	1.69	0.10
Weight (kg)	78.86	16.85
Serum calcium (mg/dL)	9.54	0.35
Serum phosphate (mg/dL)	3.81	0.55
Serum parathyroid hormone (pg/mL)	42.07	25.30
Serum 25(OH)D (ng/mL)	24.06	9.18
Bone mineral density (g/cm2)
Whole body	1.162	0.117
Lumbar spine L1-L4	1.040	0.151
Pelvis	1.300	0.184

Weighted mean and standard deviation (SD) are reported. For categorical variables, unweighted N and weighted % are reported.

Table 2.

Characteristics of the HKOS (N = 5556).

Variables	Mean	SD
Age (years)	54.57	16.34
Female	3828	68.9%
Height (m)	1.58	0.08
Weight (kg)	57.46	10.78
Current smoker	306	5.5%
Current drinker	478	8.6%
Serum calcium (mg/dL)	9.58	0.36
Serum phosphate (mg/dL)	3.48	0.48
Serum parathyroid hormone (pg/mL)	36.73	16.31
Serum 25(OH)D (ng/mL)	21.66	6.59
Bone mineral density (g/cm2)
Lumbar spine L1-L4	0.899	0.173
Femoral neck	0.688	0.135
Trochanter	0.606	0.126
Total hip	0.797	0.149

For categorical variables, N and % are reported.

Table 3.

Association of serum calcium with bone mineral density in HKOS (N = 5556) and NHANES 2003–2006 (N = 5478).

		Simple modela				Full modelb
			95% CI				95% CI
Cohort	BMD site	Beta	Lower	Upper	P-value	Beta	Lower	Upper	P-value
HKOS	Lumbar spine L1–L4	–0.095	–0.125	–0.065	<.0001	–0.091	–0.122	–0.06	<.0001
	Femoral neck	–0.026	–0.05	–0.002	.032	–0.039	–0.064	–0.015	.002
	Trochanter	–0.035	–0.061	–0.009	.009	–0.045	–0.072	–0.018	.001
	Total hip	–0.027	–0.054	–0.001	.043	–0.041	–0.068	–0.014	.003
NHANES 2003–6	Whole body	–0.065	–0.099	–0.031	.001	–0.092	–0.125	–0.059	<.0001
	Lumbar spine L1–L4	–0.058	–0.087	–0.029	.0003	–0.072	–0.102	–0.042	<.0001
	Pelvis	–0.019	–0.047	0.008	.166	–0.036	–0.065	–0.007	.016

^aAdjusted for age, sex, height, weight, and race/ethnicity (NHANES 2003–6 only).

^bFurther adjusted for smoking status, drinking status, serum phosphate, parathyroid hormones, and 25(OH)D.

Next, we evaluated the association of serum calcium with BMD in HKOS. Using the simple model, higher serum calcium was significantly associated with lower BMD at the lumbar spine, femoral neck, trochanter, and total hip, with an estimate of –0.095 SD (95% CI: –0.125 to –0.065; P < .0001), –0.026 SD (95% CI: –0.05 to –0.002; P = .032), –0.035 SD (95% CI: –0.061 to –0.009; P = .009), and –0.027 SD (95% CI: –0.054 to –0.001; P = .043), respectively. After further adjustment for additional confounding factors, such as smoking and drinking status, serum phosphate, PTH, and vitamin D levels, the associations with BMD became stronger (Table 3), except for BMD at the lumbar spine where a similar association was observed.

MR analyses

To infer causality of serum calcium level on BMD, MR analyses were performed. The characteristics of the calcium-associated SNPs are provided in Table 4. Among the 7 SNPs included, 3 of them (rs1550532, rs1570669, and rs1801725) were also significantly associated with serum PTH, phosphate, and/or 25(OH)D; while 2 (rs7336933, rs780094) of them were significantly associated with whole-body BMD after Bonferroni correction, respectively (https://osf.io/m35bv/ (13)).

Table 4.

Characteristics of the SNPs associated with serum calcium levels and their association with serum parathyroid hormone (PTH), phosphate, and 25-hydroxyvitamin D (25[OH]D) (all units are in standard deviation [SD]).

			Serum calcium			Serum PTH			Serum phosphate			Serum 25(OH)D
SNP	EA/ NEA	Variance explained (%)	Beta	SE	P-value	Beta	SE	P-value	Beta	SE	P-value	Beta	SE	P-value
rs10491003	T/C	1.05	0.027	0.005	4.8×10–09	–0.018	0.041	0.397	–0.001	0.01	.900	–0.006	0.004	.109
rs1550532	C/G	0.12	0.018	0.003	8.2×10–11	0.016	0.008	0.001a	–0.013	0.006	.030	–0.002	0.002	.311
rs1570669	G/A	0.11	0.018	0.003	9.1×10–12	–0.025	0.006	3.1×10–11a	–0.004	0.006	.500	0.010	0.002	5.3× 10–06a
rs1801725	T/G	0.11	0.071	0.004	8.9×10–86	0.027	0.008	3.2×10–07a	–0.038	0.008	3.4×10–07a	–0.005	0.003	.128
rs7336933	G/A	0.10	0.022	0.004	9.1×10–10	0.009	0.005	0.003	–0.0115	0.008	.100	0.000	0.003	.907
rs7481584	G/A	0.10	0.018	0.003	1.2×10–10	–0.002	0.016	0.774	–0.011	0.006	.080	–0.002	0.002	.541
rs780094	T/C	0.12	0.017	0.003	1.3×10–10	–0.017	0.025	0.208	0.011	0.005	.030	0.000	0.002	.838

Variance explained r² was calculated using the equation: 2 × minor allele frequency × (1 − minor allele frequency) × (β/SD)², where 1 SD equals 0.35 mg/dL (according to NHANES).

EA/NEA: Effect allele/ non-effect allele.

^aExcluded from the MR analysis due to pleiotropic association with other markers of mineral metabolism (significant threshold after Bonferroni correction = 0.05/28 = 0.0018).

Results of MR analyses are provided in Table 5. In univariable IVW analysis, higher genetically predicted serum calcium levels (per 1 SD, 0.35 ng/dl) was associated with lower whole-body BMD (estimate: –0.139 SD, 95% CI: –0.290 to 0.012; P = .072), though the association was marginally significant. After adjustment for serum levels of PTH, 25(OH)D, and phosphates in multivariable IVW analysis, the association became stronger and statistically significant (estimate: –0.431 SD; 95% CI: –0.773 to –0.089, P = .014). As an additional sensitivity analysis, 3 SNPs (rs1550532, rs1570669, and rs1801725) that were associated with other potential pleiotropic mineral metabolisms were excluded. A significant association between lifelong genetic exposure to increased serum calcium and whole-body BMD was observed in univariable IVW (estimate: –0.289 SD; 95% CI: –0.559 to –0.02, P = .036), multivariable IVW (estimate: –0.348 SD; 95% CI: –0.613 to –0.082, P = .01), and weighted median method (estimate: –0.251 SD; 95% CI: –0.427 to –0.076, P = .005). In the MR–Egger analysis, null association was observed for whole-body BMD. There was no evidence of directional pleiotropy (MR–Egger intercept: –0.012; P = .5).

Table 5.

Causal estimates for bone mineral densities (in standard deviation) per 1 SD (0.35 mg/dL) increase in genetically predicted serum calcium levels.

	Full model (7 SNPs)				Sensitivity analysis (4 SNPs)
		95% CI				95% CI
Analysis	Beta	Lower	Upper	P-value	Beta	Lower	Upper	P-value
Conventional IVW	–0.139	–0.290	0.012	.072	–0.289	–0.559	–0.02	.036
Multivariable IVWa	–0.431	–0.773	–0.089	.014	–0.348	–0.613	–0.082	.01
Weighted median	–0.061	–0.139	0.018	.128	–0.251	–0.427	–0.076	.005
MR–Egger	0.036	–0.207	0.279	.771	0.306	–1.448	2.059	.732
MR–Egger intercept	–0.006	–0.013	0.001	.092	–0.012	–0.047	0.023	.5

CI, confidence interval; IVW, inverse variance weighted; MR, Mendelian randomization.

^aAdjusted for serum parathyroid hormone (for both models), 25(OH)D (for both models), and phosphate (for full model only).

Discussion

Although bone is a reservoir and main regulator of serum calcium, the role of serum calcium in BMD and bone metabolism is not known in humans. To the best of our knowledge, this study is the first to demonstrate a potential negative causal effect of serum calcium on BMD, which is independent of vitamin D, phosphate, and PTH.

Although the relationship between calcium and bone metabolism is well established, inconsistent associations of serum calcium with BMD have been reported in the literature. In the current study, we have demonstrated a consistent, inverse, and independent association of serum calcium with BMD. The cross-sectional estimates of the association between serum calcium and BMD were further strengthened after adjustment for serum PTH, phosphate, and vitamin D in both NHANES and HKOS, suggesting that serum calcium plays an independent role in bone metabolism. This is in agreement with a recent study showing that among 127 patients with elevated serum calcium and normal PTH levels at baseline, a larger proportion of patients developed osteoporosis after a 10-year follow up than those who had normal serum calcium and PTH levels at baseline (24). However, these cross-sectional analyses are subjected to reverse causation. For example, hypercalcemia is known to be associated with low BMD due to increased bone resorption. Therefore, the cross-sectional associations observed could be due to increased bone resorption leading to increased serum calcium level.

Serum calcium level is tightly regulated by complicated homeostatic mechanisms, including influx from/to gut, kidney, and bone, which are controlled by PTH and vitamin D (25). They interact with each other to regulate calcium homeostasis. One example is at the time of hypocalcemia, there is increased secretion of PTH, which stimulates renal production of 1,25(OH)₂D₃ (1 of the major active metabolites of vitamin D). 1,25(OH)₂D₃ subsequently increases absorption of calcium in the intestine (25), increasing the serum calcium level. During hypercalcemia, PTH secretion is lowered, synthesis of 1,25(OH)₂D₃ is reduced, and absorption of dietary calcium is decreased (25). The serum calcium level will be brought back to normal. Owing to the complexity of homeostasis, we are not able to evaluate the effect of serum calcium per se on bone metabolism without taking into account other bone- and calcium-related factors in observational studies. In this scenario, the MR approach as used in the current study seems to be a justified approach to clarify the effect of serum calcium on BMD. Our MR analysis confirmed that serum calcium plays a causal and independent role on BMD. Using all 7 serum calcium-associated SNPs, no significant association was observed between lifelong genetic exposure to increased serum calcium and BMD variation initially in univariable MR analysis. However, the associations became statistically significant after adjustment for serum PTH, 25(OH)D, and phosphate levels, which are related to both calcium homeostasis and bone metabolism. Similar results were observed in most of the sensitivity analyses, except in the MR–Egger method. The MR–Egger method is well-known to be underpowered in rejecting causal null effects (22), but well-powered for detecting directional pleiotropy (22). Since the MR–Egger intercept test was insignificant, directional pleiotropy is unlikely to be present. All these findings highlighted that serum calcium per se plays an independent and causal role in reduced BMD.

Using the univariable MR approach with 7 calcium-associated SNPs, a recent study found that genetic predisposition to increased serum calcium level had null causal association with estimated BMD (eBMD), which was measured at the heel using ultrasound instead of DXA (26). The finding is partially in line with our univariable MR analysis using 7 genetic instruments that null causality was observed for serum calcium level on whole-body BMD. Unlike our study, which demonstrated that serum calcium may exert its causal effects on whole-body BMD independent of serum PTH, 25(OH)D, and phosphate levels, Cerani et al. (26) did not evaluate the causal effect of serum calcium per se on eBMD. Furthermore, whole-body BMD is more relevant in the clinical diagnosis of osteoporosis than eBMD and we described their discrepancies elsewhere (27). Briefly, whole-body BMD and eBMD measured different skeletal sites and often derived discordant results (28). Whole-body BMD is strongly and positively correlated with DXA-measured BMD at sites prone to fracture (femoral neck and lumbar spine) (r > 0.9) (18). Whereas, eBMD is just moderately correlated with DXA-measured BMD (r = 0.5–0.6) (29) and it is a weak predictor of hip fracture when compared to BMD measured at the femoral neck (30). In view of the discrepancies between eBMD and DXA-measured BMD, cautious interpretation is required to claim null causal association between serum calcium level on BMD when eBMD was adopted as the only outcome in MR analyses.

Dietary intake is one of the most important means of regulations in calcium homeostasis. Calcium supplementation can increase serum calcium levels (2, 31). In an interventional study, serum total calcium was increased by ~0.33 mg/dl 4 hours after taking 1000 mg of calcium citrate or carbonate (2). Moreover, vitamin D supplementation can also increase serum calcium levels, especially in high doses (3, 31). A vitamin D supplementation trial in vitamin D-deficient adults with elevated risk for cardiovascular disease showed that supplementation of 10 000 IU vitamin D3 increased serum calcium by 0.12 mg/dL, while change in serum calcium, but not 25(OH)D, after vitamin D repletion was significantly associated with increase in serum low-density lipoprotein cholesterol (3), suggesting that vitamin D supplementation can affect clinical outcome via changes in serum calcium. Similarly, a recent case report showed that use of a high dose of vitamin D led to hypercalcemia and hypercalcemia-associated acute renal failure in an old man (32). These studies showed that not only calcium supplementation, but vitamin D supplementation, could increase serum calcium levels and lead to calcium-related clinical outcomes. Notably, high-dose vitamin D supplementation could reduce bone mass. A recent RCT, the Calgary vitamin D study, showed a dose–response increase in hypercalcemia after vitamin D supplementation (400, 4000, and 10 000 IU per day) over 3 years (31). Moreover, the Calgary vitamin D study demonstrated that total volumetric BMD at both the radius and the tibia, and trabecular number in the radius measured by high-resolution peripheral quantitative computed tomography (HR-pQCT) were significantly reduced in a dose-dependent manner with higher intake of vitamin D supplementation (from 400 to 10 000 IU) (31). Given that high vitamin D intake is associated with elevated serum calcium (31), the reduction in BMD and trabecular number could be potentially due to the elevated serum calcium levels, according to the present findings. Nevertheless, this might also be attributed to the fact that supraphysiological dose of vitamin D might stimulate bone resorption by synthesis of receptor activator of Nuclear Factor Kappa B (NF‐κB) ligand (33). In addition, it should be noted that the effect of vitamin D and calcium supplementation in human is complex, and is affected by PTH and calcium balance (2). Age, antiosteoporosis medication, exercise, bone metabolism, and gut and kidney function are all known to affect calcium balance and hence serum calcium levels. The complexity of the calcium–vitamin D–PTH axis is demonstrated by the use of teriparatide, an antiosteoporosis medication in the form of recombinant PTH, while hyperparathyroidism is a known secondary cause of osteoporosis. The once-daily injection of teriparatide increased serum calcium level 4 to 6 hours after dosing (34). It increased bone mass by stimulating bone formation that exceeded bone resorption by a large extent (34). Based on findings of the current study, the negative effect of serum calcium on bone may be masked by the stronger effect of bone formation upon the use of teriparatide. In view of above, cautious interpretation is required for this study, as there is insufficient evidence to demonstrate the increased serum calcium level caused by supplementation would have the same negative effect on bone, which warrants further investigation.

There are biological explanations for the current study. Previous in vitro (35, 36) and animal (37) studies showed that addition of high calcium content or beyond the optimal levels in osteoblast cultures may lead to reduced osteoblast differentiation and mineralization. These findings highlight that optimal calcium within a narrow range is required for optimal bone growth, while elevated calcium levels may lead to suboptimal growth of bone. Calcium supplementation is known to increase the risk of hypercalcemia, and some patients have higher serum calcium levels within the “normal” physiological range after receiving calcium supplementation. In fact, several trials have indicated that calcium loading with oral calcium salts significantly reduced BMD (38) and bone attenuation (39) in hemodialysis patients, despite reductions in serum PTH. Whereas, in the 2009 to 2010 Korea National Health and Nutrition Examination Survey, where low calcium intake is prevalent among participants, calcium intake is still required to suppress PTH levels and improve BMD (40). The above evidence suggested that the baseline serum calcium level, and the dose of calcium supplementation, might determine the effects of supplementation on BMD.

As shown in the Calgary vitamin D RCT, calcium supplementation with vitamin D supplementation increases serum calcium in a dose-dependent manner while the reduction in BMD also followed the dose-dependent manner (31). Thus, when a high dose of vitamin D and calcium is used, the conventional concept of calcium and vitamin D supplementation being good for bone may not be applicable. Moreover, elevated serum calcium has been shown to be associated with various clinical outcomes, such as cardiovascular disease (41), diabetes (10), and cancer (42). Therefore, the current study has an important clinical impact: despite the need to prevent deficiency of nutrients in older adults, careful use of calcium and vitamin D supplementations is required to prevent excessive dosage, which might lead to unexpected reduction of BMD, and other adverse clinical outcomes.

There are several strengths in the current study. In evaluating the causality of the serum calcium level on BMD, the effects of other mineral markers were taken into account by either adjusting or excluding them. The similar results obtained suggested the findings are robust and other confounding factors were unlikely to explain the observed associations. Regarding the observational analyses, we performed the data analysis in 2 independent cohorts for comparison of different genetic background: Mexican Americans, Hispanic, and non-Hispanic in the NHANES, and southern Chinese in the HKOS. The adjustment for various confounders in the analysis provided an extra layer of information for the relationship between serum calcium and BMD.

Nevertheless, there are limitations. Firstly, it is possible that horizontal pleiotropy may be present, that is, calcium loci may affect BMD through an extracalcium pathway. We did 3 sensitivity analyses to address this. The MR–Egger intercept analysis showed no evidence for unbalanced horizontal pleiotropy; calcium loci affecting other mineral metabolisms were excluded; and similar results were observed in multivariable IVW analyses, suggesting that the presence of horizontal pleiotropy is unlikely. Secondly, approximately 20% of the sample in the GWAS of serum calcium and whole-body BMD were derived from the same cohorts https://osf.io/m35bv/ (13), which could result in overfitting. Thirdly, this MR study did not utilize the DXA-derived BMD data measured at sites prone to fracture (femoral neck and lumbar spine). The reason for not using such data is that the small sample size of the available GWAS meta-analysis (n = 32 965) (43) would lead to underpowering of subsequent MR analysis. Fourthly, the sample size of the GWAS of serum calcium and whole-body BMD with 39 400 and 66 628 individuals were still relatively small, resulting in lower power in detecting genuine genetic loci for the traits. The MR analysis may be revisited when GWASs of larger sample size become available. Lastly, whole-body BMD data are not available in the HKOS baseline cohort, thus whether serum calcium is associated with reduced whole-body BMD in Chinese people is unknown.

In conclusion, lifelong genetic exposure to increased serum calcium is inversely associated with whole-body BMD, independent of genetically predicted serum PTH, phosphate, and vitamin D levels. However, there is insufficient evidence to demonstrate the increased serum calcium level caused by supplementation would have the same negative effect on bone, which warrants further investigation. In addition to the possible adverse clinical outcomes (such as diabetes and cancer), careful use of high dose of calcium and/or vitamin D supplementations may be necessary.

Glossary

Abbreviations

BMD

bone mineral density

DXA

dual-energy X-ray absorptiometry

CI

confidence interval

GWAS

genome-wide association study

HKOS

Hong Kong Osteoporosis Study

IVW

inverse variance weighted

MR

Mendelian randomization

NHANES

National Health and Nutrition Examination Survey

PTH

parathyroid hormone

RCT

randomized controlled trial

SD

standard deviation

SNP

single nucleotide polymorphism

Additional Information

Disclosure Summary: I certify that neither I nor my co-authors have a conflict of interest as described above that is relevant to the subject matter or materials included in this Work. The authors declare no conflict of interest.

Ethics Approval: The Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster.

Data Availability: Restrictions apply to the availability of data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.