Impact of Serum Calcium Levels on Alzheimer’s Disease: A Mendelian Randomization Study

Abstract

Background:

Altered calcium homeostasis is hypothesized to underlie Alzheimer’s disease (AD). However, it remains unclear whether serum calcium levels are genetically associated with AD risk.

Objective:

To develop effective therapies, we should establish the causal link between serum calcium levels and AD.

Methods:

Here, we performed a Mendelian randomization study to investigate the causal association of increased serum calcium levels with AD risk using the genetic variants from a large-scale serum calcium genome-wide association study (GWAS) dataset (61,079 individuals of European descent) and a large-scale AD GWAS dataset (54,162 individuals including 17,008 AD cases and 37,154 controls of European descent). Here, we selected the inverse-variance weighted (IVW) as the main analysis method. Meanwhile, we selected other three sensitivity analysis methods to examine the robustness of the IVW estimate.

Results:

IVW analysis showed that the increased serum calcium level (per 1 standard deviation (SD) increase 0.5 mg/dL) was significantly associated with a reduced AD risk (OR = 0.57, 95% CI 0.35–0.95, p = 0.031). Meanwhile, all the estimates from other sensitivity analysis methods were consistent with the IVW estimate in terms of direction and magnitude.

Conclusion:

In summary, we provided evidence that increased serum calcium levels could reduce the risk of AD. Meanwhile, randomized controlled study should be conducted to clarify whether diet calcium intake or calcium supplement, or both could reduce the risk of AD.

Keywords

Alzheimer’s disease genome-wide association study inverse-variance weighted Mendelian randomization serum calcium

INTRODUCTION

Alzheimer’s disease (AD) is the most common neurodegenerative disorder [1, 2]. Calcium is involved in many biological processes including the many neural processes in the body [3, 4]. Altered calcium homeostasis is widely hypothesized to underlie cognitive deficits in normal aging subjects, and many neurodegenerative diseases including AD [3, 4]. A recent article updated the calcium hypothesis of AD on the basis of emerging evidence since 1994 [5]. This updated hypothesis provided a framework for integrating new evidence such as aging, genetic, and environmental factors into a comprehensive theory of pathogenesis [5]. Until now, the observational studies on the association between serum calcium levels and AD risk are potentially biased by some known confounders (age, sex, socioeconomic position, diet, smoking, and alcohol intake, and some unknown confounders), reverse causation (when the AD process influences the calcium levels, rather than vice versa), and a variety of other biases [6, 7]. Hence, the observational studies often reported inconsistent findings.

Some observational studies reported the protective role of high serum calcium levels in AD. Landfield et al. found that AD cases were characterized by lower serum calcium levels compared with age-matched controls or demented subjects with mild indications of vascular contributions [8]. The serum chemistry tests further showed that moderately low values of phosphate, calcium, or both identified 74% of AD cases and 100% of early onset AD cases, compared with only 46% of mixed/vascular dementia cases and 31% of normal age-matched control samples [8]. Compared with mildly affected individuals, the severely demented patients had lower serum calcium levels [9]. Meanwhile, compared with normal age-matched controls, AD patients showed a decrease in serum calcium level, increase in serum parathyroid hormone level, increase in urinary calcium excretion, and a decrease in serum 1,25 dihydroxyvitamin D concentration [10]. In 2019, Sato et al. identified that lower serum calcium could increase the risk of mild cognitive impairment conversion to AD in Japanese cohorts [11].

Other observational studies reported harmful role of high serum calcium levels in AD. In 2016, Kern et al. conducted a longitudinal population-based study [12]. They found that women with calcium supplements (n = 98) had increased risk of dementia and stroke-related dementia than women without supplementation (n = 602) [12]. However, their conclusion has been questioned by two experts [13, 14]. Rosenberg’s concern was the steep increase in serum calcium levels caused by the calcium supplements, as Kern et al. mentioned [14]. Rosenberg considered that a small increase in serum calcium levels over several hours could not be termed steep. Rosenberg further suggested that the potential mechanism, if it existed, would be outside of the realm of calcium’s pharmacokinetics [14]. Beale’s concern was the low number of women taking calcium supplements [13]. In the original study, 84 out of 98 women had calcium and vitamin D supplements [12, 13]. Hence, only 14 individuals were relevant to the study hypothesis [12, 13]. In 2019, Jadiya et al. found that high calcium levels in mitochondria could cause neuronal death in AD mice [15]. Evidence showed that calcium intake (diet and supplements) could increase serum calcium levels [16 –18]. Calcium supplements even could acutely increase serum calcium levels to a modest degree [16 –18].

In order to develop effective therapies, we should evaluate the causal link between serum calcium levels and AD. However, it remains unclear whether serum calcium levels are causally associated with AD risk. In recent years, the Mendelian randomization (MR) approach is gaining popularity [19 –25]. MR is instrumental variable analysis method using genetic variant as an instrumental variable to estimate the effect of endogenous variables on outcomes [26]. MR is based on the premise that the human genetic variants have a natural, random assortment during meiosis, and yield a random distribution in a population [27]. These genetic variants are largely not associated with these confounders such as age, sex, socioeconomic position, diet, smoking, and alcohol intake [28]. Hence, MR could apply the genetic variants to determine whether there is a causal association between a risk factor and an outcome [27]. Until now, large-scale genome-wide association study (GWAS) promptly identified common genetic variants and provided insight into the genetics of serum calcium and AD [29, 30]. Here, we performed a MR study to investigate the causal association using a large-scale serum calcium GWAS dataset and a large-scale AD GWAS dataset.

MATERIALS AND METHODS

Study design

MR is based on three principal assumptions. The first assumption, genetic variants as the instrumental variables should be significantly associated with the exposure (serum calcium levels) (assumption 1) [19]. The second assumption, these genetic variants should not be associated with the confounders of an outcome (AD) (assumption 2) [19]. The third assumption, genetic variants should affect the risk of the outcome (AD) only through the exposure (serum calcium levels) (assumption 3) [19]. The second and third assumptions are collectively known as independence from pleiotropy [19]. This study is based on publicly available, large-scale GWAS summary datasets. All participants gave informed consent in all these corresponding original studies. Our recent study has provided more detailed information about the study design [19].

Serum calcium GWAS dataset

Here, we selected genetic variants that influence serum calcium levels as the instrumental variables based on the GWAS dataset of serum calcium concentration [29]. This GWAS included 39,400 individuals from 17 population-based cohorts in discovery stage and 21,679 individuals in replication stage (61,079 individuals of European descent) [29]. The average age at examination is 57.32, and the percent of women is 56.7% [29]. An additive genetic effect was used to evaluate the association of each genetic variant with serum calcium levels by adjusting for some key covariates including age, sex, principal components, and study center [29]. The discovery stage and the meta-analysis of the discovery and replication stage identified 8 genetic variants to be associated with serum calcium levels with the genome-wide significance (p < 5.00E-08) [29]. All these 8 genetic variants were located in different genes and were not in linkage disequilibrium (LD) (Table 1) [29]. We provided more detailed information about the methods to measure serum calcium levels in the Supplementary Table 1.

Table 1

Main characteristics of 6 genetic variants in serum calcium and AD GWAS datasets

SNP	Chromosome	Nearby Genes	EA^a	NEA	EAF^b	Serum calcium GWAS			AD GWAS
						Beta^c	SE^c	p ^c	Beta^d	SE^d	p ^d
rs1550532*	2	DGKD	C	G	0.31	0.018	0.003	8.20E-11	-0.0172	0.0167	0.3034
rs1801725	3	CASR	T	G	0.15	0.071	0.004	8.90E-86	-0.0358	0.0226	0.1132
rs10491003	10	GATA3	T	C	0.09	0.027	0.005	4.80E-09	-0.0158	0.0276	0.5673
rs7336933	13	DGKH/KIAA0564	G	A	0.85	0.022	0.004	9.10E-10	-0.0048	0.0221	0.8273
rs1570669	20	CYP24A1	G	A	0.34	0.018	0.003	9.10E-12	-0.0059	0.017	0.7269
rs17711722*	7	VKORC1L1	T	C	0.47	0.021	0.003	2.80E-11	-0.0275	0.0231	0.2342

SNP, single-nucleotide polymorphism; EA, effect allele; NEA, non-effect allele; EAF, effect allele frequency; AD, Alzheimer’s disease; GWAS, genome-wide association studies; SE, standard error. ^aSerum calcium raising allele (effect allele). ^bFrequency of the serum calcium raising allele in the serum calcium GWAS dataset including up to 61,079 individuals of European ancestry [29]. ^cSummary statistics (beta coefficient, standard error, and p value) were obtained from a serum calcium GWAS dataset including up to 61,079 individuals of European ancestry [29]. Beta is the regression coefficient based on the serum calcium raising allele (effect allele). Beta > 0 and Beta < 0 means that this effect allele regulates increased and reduced serum calcium levels, respectively. ^dSummary statistics (beta coefficient, standard error, and p value) were obtained from International Genomics of Alzheimer’s Project including 17,008 cases and 37,154 controls of European descent [30]. Beta is the regression coefficient based on the serum calcium raising allele (effect allele). Beta > 0 and Beta < 0 means that this effect allele regulates increased and reduced AD risk, respectively. Beta is the overall estimated effect size for the effect allele, beta = ln(odd ratio); *rs17711722 is not available in AD GWAS dataset. We selected rs1829942, which showed high LD with rs17711722 (r² = 0.91 and D’ = 0.96) using the HaploReg v4.1 based on LD information in 1000 Genomes Project (CEU) [37]. rs1550532 (C/G, the C allele frequency = 0.31 in serum calcium GWAS) is an ambiguous palindromic variant (i.e., with alleles either A/T or C/G). Hence, we selected its proxy rs838708 (G/A), which showed high LD with rs1550532 (r² = 0.93 and D’ = 0.99) in AD GWAS dataset using the HaploReg v4.1 based on LD information in 1000 Genomes Project (CEU) [37].

In addition to statistical association, all these 8 genetic variants could map to genes implicated in calcium pathways or related phenotypic traits [29]. In brief, GCKR, DGKD, CASR, GATA3, DGKH/KIAA0564, CYP24A1, and CARS were associated with bone metabolism and endocrine control of calcium [29]. More detailed information has been described in previous study [29]. The physiological function of VKORC1L1, a paralogous enzyme sharing about 50% protein identity with VKORC1, is unknown. However previous findings showed that VKORC1 could inhibit calcium oxalate salt crystallization, adhesion, and aggregation [31].

AD GWAS dataset

The AD GWAS dataset is from the large-scale meta-analysis, which is performed by the International Genomics of Alzheimer’s Project (IGAP) [30]. IGAP is a large two-stage study based upon GWAS on individuals of European ancestry [30]. In stage 1, IGAP used genotyped and imputed data on 7,055,881 single nucleotide polymorphisms (SNPs) to meta-analyze four previously-published GWAS datasets consisting of 17,008 AD cases and 37,154 controls (The European AD Initiative, the Alzheimer Disease Genetics Consortium, The Cohorts for Heart and Aging Research in Genomic Epidemiology consortium, and The Genetic and Environmental Risk in AD consortium) [30]. All patients were late-onset AD (LOAD), and were defined by the NINCDS-ADRDA criteria or DSM-IV guidelines [30]. The average age at onset for these 17,008 AD cases is 74.2, and the average age at examination for these 37,154 controls is 69.8. The percents of women in AD cases and controls are 61.3% and 57.3%, respectively [30]. In each of these four GWAS datasets, a logistic regression model was used to evaluate the association of each genetic variant with AD by adjusting for some key covariates including age, sex,s and principal components to avoid possible population stratification [30].

Pleiotropy analysis

In MR analysis, one important issue is potential violation of MR assumptions 2 and 3. Here, we performed an assessment for pleiotropy to assure that the selected genetic variants do not exert effects on AD risk through biological pathways independent of serum calcium levels. A number of steps were taken to reduce the risk of pleiotropy. In 2017, a major review identified nine potentially modifiable risk factors linked to dementia including low levels of education, midlife hearing loss, physical inactivity, high blood pressure, type 2 diabetes, obesity, smoking, depression, and social isolation [32]. This review suggests that around 35% of dementia is attributable to a combination of these nine risk factors [32]. Meanwhile, alcohol drinking also plays an important role in AD risk [32]. In stage 1, we evaluated the potential pleiotropy using these known confounders including type 2 diabetes, systolic blood pressure and diastolic blood pressure, smoking behavior (cigarettes smoked per day), depression, obesity including body mass index (BMI), hearing loss, social isolation and alcohol drinking. In stage 2, we conducted a systematic literature search to further explore sources of pleiotropy in previous studies, which evaluated the association of serum calcium with other human diseases or traits [3, 33].

In addition to the known confounders above, there may also be some unknown confounders. In stage 3, we selected a statistical method MR-Egger intercept test to evaluate the potential pleiotropic associations of these 8 genetic variants with known and unknown confounders. MR-Egger intercept test could provide an assessment of the validity of the instrumental variable assumptions, and provide a statistical test the presence of potential pleiotropy [19]. In stage 4, we selected a newly developed statistical method named MR pleiotropy residual sum and outlier (MR-PRESSO) test to identify the horizontal pleiotropic outliers [34]. In stage 5, we selected the Cochran’s Q test (together with the I² statistic), which is a useful tool to explore the presence of heterogeneity due to pleiotropy or other causes, especially in MR studies with large sample sizes based on summary data [35]. Cochran’s Q approximately follows a χ² distribution with k-1 degrees of freedom (k stands for the number of studies for analysis) [36]. I² = (Q-(k-1)) / - Q × 100 %, which ranges from 0 to 100%, was also used [36]. Low, moderate, large, and extreme heterogeneity corresponded to 0–25%, 25–50%, 50–75%, and 75–100%, respectively [36]. In stage 1–5, the significance threshold for the association of these 8 variants with these known and unknown confounders is p < 0.05.

MR main analysis methods

Here, we have successfully extracted the summary results about the associations of each genetic variant G_j (j = 1, …, 8) with serum calcium levels and AD including the beta coefficients $({\overset{\land}{β}}_{Xj}, {\overset{\land}{β}}_{Yj})$ and their standard errors ( $se ({\overset{\land}{β}}_{Xj})$ , and $se ({\overset{\land}{β}}_{Yj})$ ). Meanwhile, we have adjusted the effect alleles of these 8 genetic variants to be associated with increased serum calcium levels in Table 1. We then aligned the effect alleles of these 8 genetic variants in other GWAS datasets. We transferred the effect alleles of these 8 genetic variants in other GWAS datasets to be consistent with the effect alleles of these 8 genetic variants in serum calcium GWAS dataset. Of these 8 independent genetic variants, rs1550532 (C/G, C with the minor allele frequency (MAF) = 0.31 in serum calcium GWAS) is a ambiguous palindromic variant (i.e., with alleles either A/T or C/G). Hence, we selected its proxy rs838708 (G/A), which showed high LD with rs1550532 (r² = 0.93 and D’ = 0.99) in AD GWAS dataset using the HaploReg v4.1 based on LD information in 1000 Genomes Project (CEU) [37].

For a given genetic variant meeting the instrumental variable assumptions, the causal effect of serum calcium levels on AD can be consistently estimated as a simple ratio of association estimate $θ_{j}^{\land} = \frac{{\overset{\land}{β}}_{Yj}}{{\overset{\land}{β}}_{Xj}}$ and its approximate variance $v_{j} = \frac{se {({\overset{\land}{β}}_{Yj})}^{2}}{{\overset{\land}{β}}_{Xj}^{2}}$ . Here, we combined the variant-specific estimates to get the overall estimate using the inverse-variance weighted meta-analysis (IVW) [19].

MR sensitivity analysis methods

The selection of multiple MR methods could examine the robustness of the estimate with each other. Here, we selected the weighted median regression method and MR-Egger method. The weighted median regression method could derive consistent estimates when up to 50% of instruments are not valid [28]. MR-Egger could provide a statistical test the presence of potential pleiotropy, and account for this potential pleiotropy [38 –42]. Under the InSIDE assumption, the MR-Egger slope parameter provides a test for a causal effect, and a consistent estimate of the causal effect even if the intercept differs from zero [42]. Meanwhile, we selected the leave-one-out permutation analysis [19] and the newly developed MR-PRESSO test [34]. The leave-one-out permutation analysis could sequentially remove each genetic variant from the MR analysis, and evaluate the influence of single genetic variant on the genetic estimate [19]. MR-PRESSO could detect horizontal pleiotropy (the MR-PRESSO global test), correct for horizontal pleiotropy via outlier removal (the MR-PRESSO outlier test), and test significant differences in the causal estimates before and after correction for outliers (the MR-PRESSO distortion test) [34].

Statistical analyses

The odds ratio (OR) as well as 95% confidence interval (CI) of AD corresponds to per 0.5 mg/dL increase (about 1 standard deviation (SD)) in serum calcium levels. All statistical analyses were conducted using the statistical program R (version 3.2.4), R package ‘MendelianRandomization’ based on the multiplicative random effects model [42], and R Package for meta-analysis ‘meta’ (version 3.1-0). The threshold of statistical significance for the potential causal association between serum calcium levels and AD risk was p < 0.05.

Power analysis

The proportion of serum calcium variance (R²) explained by the instrumental variables can be estimated as the following formula $R^{2} = \sum_{i = 1}^{K} \frac{β_{i}^{2} * 2 * MA F_{SN P_{i}} (1 - MA F_{SN P_{i}})}{var (X)}$

Where β_i is the effect size (beta coefficient) associated with the serum calcium levels for SNP_i, MAF_{SNP
_i} is the minor allele frequency for SNP_i, K is the number of genetic variants, and var (X) is the variance of the serum calcium levels (var (X) = SD², and 1 SD = 0.5 mg/dL) [3]. The strength of the instrumental variables was evaluated using the first-stage F-statistic. A common threshold is F > 10 to avoid bias in MR studies [43]. Here, we calculated the F-statistic and statistical power to estimate the minimum detectable magnitudes of association using the web-based tool mRnd and a two-sided type-I error rate α of 0.05 [44]. In brief, the proportion of variance of the R², the total number of individuals in the analysis, and the proportion of cases in the study, were used to detect the minimum detectable OR [44]. More detailed information has been provided at the mRnd website (https://cnsgenomics.shinyapps.io/mRnd/) [44].

RESULTS

Association of serum calcium variants with AD

Of these 8 genetic variants associated with serum calcium levels, we extracted the summary statistics for 7 variants in stage 1 dataset of AD GWAS. The rs17711722 is not available in stage 1 dataset of AD GWAS. We then selected its proxy rs1829942, which showed high LD with rs17711722 (r² = 0.91 and D’ = 0.96) using the HaploReg v4.1 based on LD information in 1000 Genomes Project (CEU) with r²≥0.8 [37]. None of these 8 genetic variants was significantly associated with AD risk at the significance threshold p < 0.05 (Supplementary Table 2).

Pleiotropy analysis

In stage 1, rs780094 and rs7481584 were significantly associated with known confounders at the significance threshold p < 0.05, as described in Supplementary Table 3. In brief, rs780094 variant was significantly associated with type II diabetes (p = 1.00E-05) and BMI (p = 3.22E-04). The rs7481584 variant was significantly associated with type II diabetes (p = 1.50E-02) and BMI (p = 4.39E-03). In stage 2, we found evidence that rs780094 had pleiotropic associations with lipids, glycemic traits, type 2 diabetes, and measures of adiposity, which have been identified in a recent study [3]. In summary, the pleiotropy analysis in stage 1–2 clearly showed that rs780094 and rs7481584 may have potential violation of MR assumptions. Hence, we excluded both rs780094 and rs7481584 variants, and selected the remaining six genetic variants in the following analysis, as provided in Table 1.

In stage 3, MR-Egger intercept test indicated no evidence of pleiotropy with intercept = –0.005, and p = 0.765. In stage 4, MR-PRESSO test did not identify any horizontal pleiotropic outliers with p = 0.974. In stage 5, the heterogeneity test using 8 serum calcium variants showed no significant heterogeneity with Cochran’s Q = 5.26, p = 0.628, and I^∧2 = 0%, but with wide 95% CI [0%, 56.9%]. Interestingly, the heterogeneity test using six serum calcium variants (excluding rs780094 and rs7481584 variants) showed no significant heterogeneity with Cochran’s Q = 0.82, p = 0.9735, and I^∧2 = 0% as well as 95% CI [0%, 0%]. Hence, the heterogeneity test further supported the findings from stage 1–4.

Association of serum calcium levels with AD risk

IVW method showed that the increased serum calcium levels (per 1 SD increase 0.5 mg/dL) were significantly associated with the reduced AD risk (OR = 0.57, 95% CI: 0.35–0.95, p = 0.031) (Table 2). Interestingly, all the estimates from other three sensitivity analysis methods were consistent with the IVW estimate in terms of direction and magnitude (Table 2). Importantly, MR-PRESSO showed significant association of increased serum calcium levels with the reduced AD risk with p < 0.05 (Table 2). The leave-one-out permutation further showed that the direction and precision of the genetic estimates between increased serum calcium levels and reduced risk of AD remained largely unchanged using all these four MR methods. Figure 1 shows individual causal estimate from each of 6 serum calcium genetic variants using IVW method.

Table 2

Association between increased serum calcium levels (0.5 mg/dL) and AD

Method	OR	95% CI	p
Weighted median	0.6	0.34-1.06	0.079
IVW	0.57	0.35-0.95	0.031
MR-Egger	0.64	0.26-1.63	0.353
MR-PRESSO	0.57	0.46-0.71	0.0035

OR < 1 mean that high serum calcium levels could reduce AD risk. OR, odds ratio; CI, confidence interval; IVW, Inverse-variance weighted; The association between serum calcium levels and AD was at the significance level p < 0.05.

Fig. 1

Individual and pooled causal estimate from each of 6 serum calcium genetic variants using IVW method. Serum calcium effects on odds of AD including the individual and inverse variance weighted (IVW) results.

Power analysis

Here, all these six genetic variants could explain about 0.81% of the serum calcium variance (R² = 0.81%). The first-stage F-statistic for the instrument including these six genetic variants was 443.29 > 10, so weak instrument bias is unlikely. The actual N for AD GWAS is 54,162, and the proportion of cases is 0.314. Power analysis suggested that our MR study had 80% power to detect an OR of 0.74 or lower per SD (0.5 mg/dL) increase in serum calcium levels for AD. The powers are 95% and 100% to detect the causal association between increased serum calcium levels and reduced AD risk with OR = 0.67 and OR = 0.57, respectively. The required sample sizes for 80% power to detect the OR = 0.67 and 0.57 are 32,223 and 17,235, respectively. In all these power analyses, the two-sided type-I error rate is α= 0.05.

DISCUSSION

Until now, it has been a long time since the establishment of the calcium hypothesis of AD based on emerging evidence beginning in 1994 [5]. However, it remains unclear whether serum calcium level (diet and supplements) is genetically associated with AD risk. In recent years, the existing large-scale serum calcium and AD GWAS datasets prompts us to investigate the potential causal association between serum calcium and AD risk using multiple MR methods. Our results indicated genetically increased serum calcium levels were significantly associated with reduced AD risk. Interestingly, our results are consistent with previous studies, which reported reduced serum calcium levels in AD [8 –10]. Meanwhile, a recent study also reported the reduced serum calcium levels in Parkinson’s disease [45]. The levels of serum calcium in the Parkinson’s disease group with dementia were significantly lower than the normal control group (p < 0.001) [45]. There is also a correlation between serum calcium levels and cognitive impairment [45]. Lower serum calcium levels could predict worse cognitive scores [45]. Importantly, two recent MR studies provided clear evidence that increased vitamin D could reduce the risk of AD and another neurodegenerative disease, multiple sclerosis [46, 47]. Hence, our results are comparable to these recent findings.

Until now, many clinical trials of therapies for AD have failed, especially the double-blind, placebo-controlled, phase III trial involving patients with mild dementia due to AD [48, 49]. Hence, the causal association between serum calcium levels and AD risk may have clinical and public health implications. The 2016 AD facts and figures showed an estimated 5.4 million people to have AD in the United States [50]. By 2050, the number of people living with AD will be 13.8 million in the United States [50]. Until now, two ways could increase the serum calcium levels. One is diet calcium intake. Meanwhile, it remains unclear whether calcium intake from dietary sources has health advantages over supplements [51]. Until now, the recommended daily calcium intake is 1000 to 1200 mg [12]. It is difficult to get this recommended amount through diet alone, so calcium supplements are widely used [12]. In the United States, about 43% of people, including about 70% of older women, take calcium supplements [52].

Our findings provide genetic evidence that diet calcium intake or calcium supplement, or both, may reduce the risk of AD. However, these serum calcium genetic variants just represent lifelong exposure and as such do not directly equate with an intervention [53]. Until now, there is no randomized clinical trial for calcium supplementation and dementia/AD. The Women’s Health Initiative Calcium/Vitamin D supplementation study (WHI CaD) had investigated the relationship of both calcium and vitamin D supplementation with the risk of cognitive impairment using a randomized double-blinded placebo-controlled trial [54]. WHI CaD found no association of calcium and vitamin D supplementation with incident cognitive impairment [54]. WHI CaD suggested that further studies should investigate the effects of calcium and vitamin D supplementation separately, in men, in other ages, other ethnic groups, or with other doses [54]. In order to translate these genetic findings into clinical and public health implications, the potential mechanisms underlying this causal association remain to be thoroughly evaluated. Meanwhile, randomized controlled study should be conducted to assess the effect of serum calcium levels on AD risk, and further clarify whether diet calcium intake or calcium supplement, or both, could reduce the risk of AD.

Our findings showed that high serum calcium levels may reduce the risk of AD. However, high serum calcium levels are not always better to other human disease such cardiovascular diseases [55]. Higher level of serum calcium is often assumed as a risk factor for cardiovascular diseases [55]. Take coronary artery disease (CAD) for example; the relationships between calcium intake or serum calcium levels and CAD identified by the observational studies, MR studies, and randomized clinical trials remain controversial. In 2017, two independent MR studies showed that increased serum calcium levels could increase the risk of coronary artery disease [3, 33]. In 2016, Chung et al. performed an updated systematic review and meta-analysis of 15 observational studies evaluating the calcium intake and CAD risk [56]. Their results showed that calcium intake within tolerable upper intake levels (2000 to 2500 mg/d) was not associated with CAD risk in generally healthy adults [56]. In 2018, the US Preventive Services Task Force (USPSTF) updated the evidence for benefits and harms of vitamin D, calcium, or combined supplementation for the primary prevention of fractures in community-dwelling adults [57, 58]. USPSTF analyzed eleven randomized clinical trials (N = 51,419) in adults 50 years and older conducted over 2 to 7 years, and reported that supplementation with calcium alone or with vitamin D had no significant effect on all-cause mortality or incident cardiovascular disease [57, 58]. Hence, future studies should consider the effects of calcium more holistically.

This MR study may have several strengths. First, this study may benefit from the large-scale serum calcium GWAS dataset (N = 61,079 individuals) and AD GWAS dataset (N = 54,162 individuals), which provide ample power (100%) to detect the causal association between serum calcium levels and AD risk. Second, both the serum calcium and AD GWAS datasets are from subjects of European descent, which may reduce the influence on the potential association caused by the population stratification. Third, multiple independent genetic variants are taken as instruments, which may reduce the influence on the potential association caused by the linkage disequilibrium. Fourth, we selected a total of 13 MR methods, which increase the precision of the estimate. Fifth, we performed a comprehensive pleiotropy analysis to reduce the risk of pleiotropy and meet the MR assumptions. Hence, the MR assumptions did not seem to be violated.

This MR study may have some limitations. First, we only selected one AD GWAS dataset. We think that a replication data may be necessary. However, a replication dataset with a similar large-scale of AD GWAS dataset was not available. Second, multiple steps were taken to reduce the risk of pleiotropy, as done in our pleiotropy analysis. However, we could not completely rule out additional confounders. Until now, it is almost impossible to fully rule out pleiotropy present in any MR study [3 , 59]. Third, it could not be completely ruled out that population stratification may have had some influence on the estimate. In order to reduce the effect of population stratification, our study was restricted to individuals of European ancestry. Fourth, the causal association between serum calcium levels and AD risk may differ by ethnicity or genetic ancestry. This causal association should be further evaluated in other ancestries. Fifth, we have performed additional analysis to rule out selection bias including the beneficial effects of blood pressure and smoking on AD [59]. However, we could not exclude the possibility of other selection biases because the AD case and control participants in IGAP GWAS dataset are quite old. These participants may have survived other exposures in order to be able to get AD. Hence, the MR study may generate unconfounded estimates, but not necessarily unbiased estimates.

In summary, we demonstrate that the increased serum calcium levels could reduce AD risk in people of European descent using large-scale GWAS datasets. These findings provide rationale for further investigating the clinical and public health implications of high serum calcium levels by diet or supplement, or both, in preventing the progression of AD.

Ethics approval and consent to participate

This article contains human participants collected by several studies performed by previous studies. All participants gave informed consent in all the corresponding original studies, as described in the Materials and methods. Here, our study is based on the publicly available, large-scale GWAS summary datasets, and not the individual-level data. Hence, ethical approval was not sought.

Data availability

All relevant data are within the paper and its Supplementary Material. The authors confirm that all data underlying the findings are either fully available without restriction through consortia websites, or may be made available from consortia upon request. IGAP consortium data are available at http://web.pasteur-lille.fr/en/recherche/u744/igap/igap_download.php. NIAGADS Consortium: https://www.niagads.org/datasets. Social Science Genetic Association Consortium (SSGAC): https://www.thessgac.org/data; type 2 diabetes from DIAbetes Genetics Replication and Meta-analysis (DIAGRAM) Consortium: http://diagram-consortium.org/downloads.html; Ukbiobank: http://www.ukbiobank.ac.uk/scientists-3/genetic-data/.

Footnotes

ACKNOWLEDGMENTS

We thank the International Genomics of Alzheimer’s Project (IGAP) for providing summary results data for these analyses. The investigators within IGAP contributed to the design and implementation of IGAP and/or provided data but did not participate in analysis or writing of this report. IGAP was made possible by the generous participation of the control subjects, the patients, and their families. The i-Select chips was funded by the French National Foundation on AD and related disorders. EADI was supported by the LABEX (laboratory of excellence program investment for the future) DISTALZ grant, Inserm, Institut Pasteur de Lille, Université de Lille 2 and the Lille University Hospital. GERAD was supported by the Medical Research Council (Grant n° 503480), Alzheimer’s Research UK (Grant n° 503176), the Wellcome Trust (Grant n° 082604/2/07/Z) and German Federal Ministry of Education and Research (BMBF): Competence Network Dementia (CND) grant n° 01GI0102, 01GI0711, 01GI0420. CHARGE was partly supported by the NIH/NIA grant R01 AG033193 and the NIA AG081220 and AGES contract N01-AG-12100, the NHLBI grant R01 HL105756, the Icelandic Heart Association, and the Erasmus Medical Center and Erasmus University. ADGC was supported by the NIH/NIA grants: U01 AG032984, U24 AG021886, U01 AG016976, and the Alzheimer’s Association grant ADGC-10-196728. We also thank the Social Science Genetic Association Consortium (SSGAC), type 2 diabetes from DIAbetes Genetics Replication and Meta-analysis (DIAGRAM) Consortium, Genetic Investigation of ANthropometric Traits (GIANT) consortium, International Consortium of Blood Pressure (ICBP) consortium, Tobacco and Genetics Consortium (TGC), depression from Psychiatric Genomics Consortium (PGC), NIAGADS Consortium, and Ukbiobank for other GWAS datasets. We thank Prof. CM Schooling and Prof. Sara Hagg for careful review of our manuscript and giving us these useful suggestions.

This work was supported by the Scientific Research Project of Tianjin Education Commission (2016YD06), Tianjin Natural Science Foundation (18JCQNJC79700), and the National Natural Science Foundation of China (81701177). This work was partially supported by funding from the National Key R & D Program of China (2016YFC1202302 and 2017YFSF090117), Natural Science Foundation of Heilongjiang Province (F2015006), the National Natural Science Foundation of China (Grant No. 61822108 and 61571152), and the Fundamental Research Funds for the Central Universities (AUGA5710001716). This work is supported by the Fund of Academic Promotion Program of Shandong First Medical University & Shandong Academy of Medical Sciences (No.2019QL016, No.2019PT007), Fund of Taishan Scholar Project, National Natural Science Foundation of China (No. 81870938), and Natural Science Foundation of Shandong Province (No. ZR2019ZD32).

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Liu

, Zhang

, Wang

, Xu

, Chen

, Bao

, Hu

, Jin

, Tian

, Bai

, Zhou

, Wang

, Han

, Zong

, Jiang

(2018) Alzheimer’s disease rs11767557 variant regulates EPHA1 gene expression specifically in human whole blood. J Alzheimers Dis 61, 1077–1088.

Liu

, Xu

, Jiang

, Zhang

, Feng

, Jiang

(2017) PICALM rs3851179 variant confers susceptibility to Alzheimer’s disease in Chinese population. Mol Neurobiol 54, 3131–3136.

Larsson

, Burgess

, Michaelsson

(2017) Association of genetic variants related to serum calcium levels with coronary artery disease and myocardial infarction. JAMA 318, 371–380.

LaFerla

(2002) Calcium dyshomeostasis and intracellular signalling in Alzheimer’s disease. Nat Rev Neurosci 3, 862–872.

(2017) Calcium Hypothesis of Alzheimer’s disease and brain aging: A framework for integrating new evidence into a comprehensive theory of pathogenesis. Alzheimers Dement 13, 178–182 e117.

Jepsen

, Johnsen

, Gillman

, Sorensen

(2004) Interpretation of observational studies. Heart 90, 956–960.

Davey Smith

, Paternoster

, Relton

(2017) When will mendelian randomization become relevant for clinical practice and public health? JAMA 317, 589–591.

Landfield

, Applegate

, Schmitzer-Osborne

, Naylor

(1991) Phosphate/calcium alterations in the first stages of Alzheimer’s disease: Implications for etiology and pathogenesis. J Neurol Sci 106, 221–229.

Deary

, Hendrickson

, Burns

(1987) Serum calcium levels in Alzheimer’s disease: A finding and an aetiological hypothesis. Pers Individ Dif 8, 75–80.

10.

Conley

, Mukherjee

, Kammerer

, DeKosky

, Kamboh

, Finegold

, Ferrell

(2009) Evidence supporting a role for the calcium-sensing receptor in Alzheimer disease. Am J Med Genet B Neuropsychiatr Genet 150B, 703–709.

11.

Sato

, Mano

, Ihara

, Suzuki

, Tomita

, Arai

, Ishii

, Senda

, Ito

, Ikeuchi

, Kuwano

, Matsuda

, Iwatsubo

, Toda

, Iwata

(2019) Lower serum calcium as a potentially associated factor for conversion of mild cognitive impairment to early Alzheimer’s disease in the Japanese Alzheimer’s Disease Neuroimaging Initiative. J Alzheimers Dis 68, 777–788.

12.

Kern

, Kern

, Blennow

, Zetterberg

, Waern

, Guo

, Borjesson-Hanson

, Skoog

, Ostling

(2016) Calcium supplementation and risk of dementia in women with cerebrovascular disease. Neurology 87, 1674–1680.

13.

Beale

(2017) Letter re: Calcium supplementation and risk of dementia in women with cerebrovascular disease. Neurology 88, 812.

14.

Rosenberg

(2017) Letter re: Calcium supplementation and risk of dementia in women with cerebrovascular disease. Neurology 88, 812.

15.

Jadiya

, Kolmetzky

, Tomar

, Di Meco

, Lombardi

, Lambert

, Luongo

, Ludtmann

, Pratico

, Elrod

(2019) Impaired mitochondrial calcium efflux contributes to disease progression in models of Alzheimer’s disease. Nat Commun 10, 3885.

16.

Bolland

, Avenell

, Baron

, Grey

, MacLennan

, Gamble

, Reid

(2010) Effect of calcium supplements on risk of myocardial infarction and cardiovascular events: meta-analysis. BMJ 341, c3691.

17.

Bolland

, Grey

, Avenell

, Gamble

, Reid

(2011) Calcium supplements with or without vitamin D and risk of cardiovascular events: reanalysis of the Women’s Health Initiative limited access dataset and meta-analysis. BMJ 342, d2040.

18.

Anderson

, Kruszka

, Delaney

, He

, Burke

, Alonso

, Bild

, Budoff

, Michos

(2016) Calcium intake from diet and supplements and the risk of coronary artery calcification and its progression among older adults: 10-year follow-up of the Multi-Ethnic Study of Atherosclerosis (MESA). J Am Heart Assoc 5, e003815.

19.

Liu

, Zhao

, Jin

, Hu

, Wang

, Tian

, Han

, Xu

, Jiang

(2018) Circulating vitamin E levels and Alzheimer’s disease: a Mendelian randomization study. Neurobiol Aging 72, 189e181–189e189.

20.

Zhao

, Schooling

(2019) Effect of linoleic acid on ischemic heart disease and its risk factors: a Mendelian randomization study. BMC Med 17, 61.

21.

Zhao

, Kwok

, Schooling

(2019) Effect of glutamate and aspartate on ischemic heart disease, blood pressure, and diabetes: a Mendelian randomization study. Am J Clin Nutr 109, 1197–1206.

22.

Luo

, Au Yeung

, Zhao

, Burgess

, Schooling

(2019) Association of genetically predicted testosterone with thromboembolism, heart failure, and myocardial infarction: Mendelian randomisation study in UK Biobank. BMJ 364, l476.

23.

Liu

, Jin

, Jiang

(2019) Interleukin-6 receptor and inflammatory bowel disease: A Mendelian randomization study. Gastroenterology 156, 823–824.

24.

Wang

, Qiao

, Zhang

, Hua

, Jin

, Liu

(2020) Circulating vitamin D levels and Alzheimer’s disease: A Mendelian randomization study in the IGAP and UK Biobank. J Alzheimers Dis 73, 609–618.

25.

Zhang

, Wang

, Han

, Wang

, Zhang

, Liu

(2020) Impact of vitamin D binding protein levels on Alzheimer’s disease: a Mendelian randomization study. J Alzheimers Dis 74, 991–998.

26.

Swerdlow

, Kuchenbaecker

, Shah

, Sofat

, Holmes

, White

, Mindell

, Kivimaki

, Brunner

, Whittaker

, Casas

, Hingorani

(2016) Selecting instruments for Mendelian randomization in the wake of genome-wide association studies. Int J Epidemiol 45, 1600–1616.

27.

Emdin

, Khera

, Kathiresan

(2017) Mendelian randomization. JAMA 318, 1925–1926.

28.

Emdin

, Khera

, Natarajan

, Klarin

, Zekavat

, Hsiao

, Kathiresan

(2017) Genetic association of waist-to-hip ratio with cardiometabolic traits, type 2 diabetes, and coronary heart disease. JAMA 317, 626–634.

29.

O’Seaghdha

, Wu

, Yang

, Kapur

, Guessous

, Zuber

, Kottgen

, Stoudmann

, Teumer

, Kutalik

, Mangino

, Dehghan

, Zhang

, Eiriksdottir

, Li

, Tanaka

, Portas

, Lopez

, Hayward

, Lohman

, Matsuda

, Padmanabhan

, Firsov

, Sorice

, Ulivi

, Brockhaus

, Kleber

, Mahajan

, Ernst

, Gudnason

, Launer

, Mace

, Boerwinckle

, Arking

, Tanikawa

, Nakamura

, Brown

, Gaspoz

, Theler

, Siscovick

, Psaty

, Bergmann

, Vollenweider

, Vitart

, Wright

, Zemunik

, Boban

, Kolcic

, Navarro

, Brown

, Estrada

, Ding

, Harris

, Bandinelli

, Hernandez

, Singleton

, Girotto

, Ruggiero

, d’Adamo

, Robino

, Meitinger

, Meisinger

, Davies

, Starr

, Chambers

, Boehm

, Winkelmann

, Huang

, Murgia

, Wild

, Campbell

, Morris

, Franco

, Hofman

, Uitterlinden

, Rivadeneira

, Volker

, Hannemann

, Biffar

, Hoffmann

, Shin

, Lescuyer

, Henry

, Schurmann

, Munroe

, Gasparini

, Pirastu

, Ciullo

, Gieger

, Marz

, Lind

, Spector

, Smith

, Rudan

, Wilson

, Polasek

, Deary

, Pirastu

, Ferrucci

, Liu

, Kestenbaum

, Kooner

, Witteman

, Nauck

, Kao

, Wallaschofski

, Bonny

, Fox

, Bochud

(2013) Meta-analysis of genome-wide association studies identifies six new Loci for serum calcium concentrations. PLoS Genet 9, e1003796.

30.

Lambert

, Ibrahim-Verbaas

, Harold

, Naj

, Sims

, Bellenguez

, DeStafano

, Bis

, Beecham

, Grenier-Boley

, Russo

, Thorton-Wells

, Jones

, Smith

, Chouraki

, Thomas

, Ikram

, Zelenika

, Vardarajan

, Kamatani

, Lin

, Gerrish

, Schmidt

, Kunkle

, Dunstan

, Ruiz

, Bihoreau

, Choi

, Reitz

, Pasquier

, Cruchaga

, Craig

, Amin

, Berr

, Lopez

, De Jager

, Deramecourt

, Johnston

, Evans

, Lovestone

, Letenneur

, Moron

, Rubinsztein

, Eiriksdottir

, Sleegers

, Goate

, Fievet

, Huentelman

, Gill

, Brown

, Kamboh

, Keller

, Barberger-Gateau

, McGuiness

, Larson

, Green

, Myers

, Dufouil

, Todd

, Wallon

, Love

, Rogaeva

, Gallacher

, St George-Hyslop

, Clarimon

, Lleo

, Bayer

, Tsuang

, Yu

, Tsolaki

, Bossu

, Spalletta

, Proitsi

, Collinge

, Sorbi

, Sanchez-Garcia

, Fox

, Hardy

, Deniz Naranjo

, Bosco

, Clarke

, Brayne

, Galimberti

, Mancuso

, Matthews

, Moebus

, Mecocci

, Del Zompo

, Maier

, Hampel

, Pilotto

, Bullido

, Panza

, Caffarra

, Nacmias

, Gilbert

, Mayhaus

, Lannefelt

, Hakonarson

, Pichler

, Carrasquillo

, Ingelsson

, Beekly

, Alvarez

, Zou

, Valladares

, Younkin

, Coto

, Hamilton-Nelson

, Gu

, Razquin

, Pastor

, Mateo

, Owen

, Faber

, Jonsson

, Combarros

, O’Donovan

, Cantwell

, Soininen

, Blacker

, Mead

, Mosley

Jr. , Bennett

, Harris

, Fratiglioni

, Holmes

, de Bruijn

, Passmore

, Montine

, Bettens

, Rotter

, Brice

, Morgan

, Foroud

, Kukull

, Hannequin

, Powell

, Nalls

, Ritchie

, Lunetta

, Kauwe

, Boerwinkle

, Riemenschneider

, Boada

, Hiltuenen

, Martin

, Schmidt

, Rujescu

, Wang

, Dartigues

, Mayeux

, Tzourio

, Hofman

, Nothen

, Graff

, Psaty

, Jones

, Haines

, Holmans

, Lathrop

, Pericak-Vance

, Launer

, Farrer

, van Duijn

, Van Broeckhoven

, Moskvina

, Seshadri

, Williams

, Schellenberg

, Amouyel

(2013) Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer’s disease. Nat Genet 45, 1452–1458.

31.

, Wu

, Ma

, Wu

, Lu

, Shi

(2014) Involvement of VKORC1 in the inhibition of calcium oxalate crystal formation in HK-2 cells. J Huazhong Univ Sci Technolog Med Sci 34, 376–381.

32.

Livingston

, Sommerlad

, Orgeta

, Costafreda

, Huntley

, Ames

, Ballard

, Banerjee

, Burns

, Cohen-Mansfield

, Cooper

, Fox

, Gitlin

, Howard

, Kales

, Larson

, Ritchie

, Rockwood

, Sampson

, Samus

, Schneider

, Selbaek

, Teri

, Mukadam

(2017) Dementia prevention, intervention, and care. Lancet 390, 2673–2734.

33.

, Lin

, Schooling

(2017) A Mendelian randomization study of the effect of calcium on coronary artery disease, myocardial infarction and their risk factors. Sci Rep 7, 42691.

34.

Verbanck

, Chen

, Neale

, Do

(2018) Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 50, 693–698.

35.

Greco

, Minelli

, Sheehan

, Thompson

(2015) Detecting pleiotropy in Mendelian randomisation studies with summary data and a continuous outcome. Stat Med 34, 2926–2940.

36.

Liu

, Zhang

, Cai

, Ma

, Zhang

, Jiang

, Feng

, Liao

, Chen

, Zhao

, Li

(2013) PICALM gene rs3851179 polymorphism contributes to Alzheimer’s disease in an Asian population. Neuromolecular Med 15, 384–388.

37.

Ward

, Kellis

(2012) HaploReg: a resource for exploring chromatin states, conservation, and regulatory motif alterations within sets of genetically linked variants. Nucleic Acids Res 40, D930–934.

38.

Tillmann

, Vaucher

, Okbay

, Pikhart

, Peasey

, Kubinova

, Pajak

, Tamosiunas

, Malyutina

, Hartwig

, Fischer

, Veronesi

, Palmer

, Bowden

, Davey Smith

, Bobak

, Holmes

(2017) Education and coronary heart disease: mendelian randomisation study. BMJ 358, j3542.

39.

Dale

, Fatemifar

, Palmer

, White

, Prieto-Merino

, Zabaneh

, Engmann

JEL

, Shah

, Wong

, Warren

, McLachlan

, Trompet

, Moldovan

, Morris

, Sofat

, Kumari

, Hypponen

, Jefferis

, Gaunt

, Ben-Shlomo

, Zhou

, Gentry-Maharaj

, Ryan

, Mutsert

, Noordam

, Caulfield

, Jukema

, Worrall

, Munroe

, Menon

, Power

, Kuh

, Lawlor

, Humphries

, Mook-Kanamori

, Sattar

, Kivimaki

, Price

, Davey Smith

, Dudbridge

, Hingorani

, Holmes

, Casas

(2017) Causal associations of adiposity and body fat distribution with coronary heart disease, stroke subtypes, and type 2 diabetes mellitus: a Mendelian randomization analysis. Circulation 135, 2373–2388.

40.

Bowden

, Davey Smith

, Haycock

, Burgess

(2016) Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 40, 304–314.

41.

Bowden

, Davey Smith

, Burgess

(2015) Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol 44, 512–525.

42.

Yavorska

, Burgess

(2017) MendelianRandomization: an R package for performing Mendelian randomization analyses using summarized data. Int J Epidemiol 46, 1734–1739.

43.

Burgess

, Thompson

(2011) Avoiding bias from weak instruments in Mendelian randomization studies. Int J Epidemiol 40, 755–764.

44.

Brion

, Shakhbazov

, Visscher

(2013) Calculating statistical power in Mendelian randomization studies. Int J Epidemiol 42, 1497–1501.

45.

Liu

, Zhou

, Zhang

, Liu

, Luo

(2016) [Correlation analysis of serum calcium level and cognition in the patients with Parkinson’s disease]. Zhonghua Yi Xue Za Zhi 96, 3284–3288.

46.

Mokry

, Ross

, Morris

, Manousaki

, Forgetta

, Richards

(2016) Genetically decreased vitamin D and risk of Alzheimer disease. Neurology 87, 2567–2574.

47.

Mokry

, Ross

, Ahmad

, Forgetta

, Smith

, Goltzman

, Leong

, Greenwood

, Thanassoulis

, Richards

(2015) Vitamin D and risk of multiple sclerosis: a Mendelian randomization study. PLoS Med 12, e1001866.

48.

Honig

, Vellas

, Woodward

, Boada

, Bullock

, Borrie

, Hager

, Andreasen

, Scarpini

, Liu-Seifert

, Case

, Dean

, Hake

, Sundell

, Poole Hoffmann

, Carlson

, Khanna

, Mintun

, DeMattos

, Selzler

, Siemers

(2018) Trial of solanezumab for mild dementia due to Alzheimer’s disease. N Engl J Med 378, 321–330.

49.

Anderson

, Hadjichrysanthou

, Evans

, Wong

(2017) Why do so many clinical trials of therapies for Alzheimer’s disease fail? Lancet 390, 2327–2329.

50.

Alzheimer’s Association (2016) 2016 Alzheimer’s disease facts and figures. Alzheimers Dement 12, 459–509.

51.

Abrahamsen

(2017) The calcium and vitamin D controversy. Ther Adv Musculoskelet Dis 9, 107–114.

52.

Bailey

, Dodd

, Goldman

, Gahche

, Dwyer

, Moshfegh

, Sempos

, Picciano

(2010) Estimation of total usual calcium and vitamin D intakes in the United States. J Nutr 140, 817–822.

53.

Boef

, le Cessie

, Dekkers

(2015) Mendelian randomization studies in the elderly. Epidemiology 26, e15–16.

54.

Rossom

, Espeland

, Manson

, Dysken

, Johnson

, Lane

, LeBlanc

, Lederle

, Masaki

, Margolis

(2012) Calcium and vitamin D supplementation and cognitive impairment in the women’s health initiative. J Am Geriatr Soc 60, 2197–2205.

55.

Reid

, Gamble

, Bolland

(2016) Circulating calcium concentrations, vascular disease and mortality: a systematic review. J Intern Med 279, 524–540.

56.

Chung

, Tang

, Fu

, Wang

, Newberry

(2016) Calcium intake and cardiovascular disease risk: An updated systematic review and meta-analysis. Ann Intern Med 165, 856–866.

57.

Grossman

, Curry

, Owens

, Barry

, Caughey

, Davidson

, Doubeni

, Epling

Jr. , Kemper

, Krist

, Kubik

, Landefeld

, Mangione

, Silverstein

, Simon

, Tseng

(2018) Vitamin D, calcium, or combined supplementation for the primary prevention of fractures in community-dwelling adults: US Preventive Services Task Force Recommendation Statement. JAMA 319, 1592–1599.

58.

Kahwati

, Weber

, Pan

, Gourlay

, LeBlanc

, Coker-Schwimmer

, Viswanathan

(2018) Vitamin D, calcium, or combined supplementation for the primary prevention of fractures in community-dwelling adults: evidence report and systematic review for the US Preventive Services Task Force. JAMA 319, 1600–1612.

59.

Larsson

, Traylor

, Malik

, Dichgans

, Burgess

, Markus

(2017) Modifiable pathways in Alzheimer’s disease: Mendelian randomisation analysis. BMJ 359, j5375.