Genetically decreased vitamin D and risk of Alzheimer disease

Abstract

Objective: To test whether genetically decreased vitamin D levels are associated with Alzheimer disease (AD) using mendelian randomization (MR), a method that minimizes bias due to confounding or reverse causation.

Methods: We selected single nucleotide polymorphisms (SNPs) that are strongly associated with 25-hydroxyvitamin D (25OHD) levels (p < 5 × 10^-8) from the Study of Underlying Genetic Determinants of Vitamin D and Highly Related Traits (SUNLIGHT) Consortium (N = 33,996) to act as instrumental variables for the MR study. We measured the effect of each of these SNPs on 25OHD levels in the Canadian Multicentre Osteoporosis Study (CaMos; N = 2,347) and obtained the corresponding effect estimates for each SNP on AD risk from the International Genomics of Alzheimer's Project (N = 17,008 AD cases and 37,154 controls). To produce MR estimates, we weighted the effect of each SNP on AD by its effect on 25OHD and meta-analyzed these estimates using a fixed-effects model to provide a summary effect estimate.

Results: The SUNLIGHT Consortium identified 4 SNPs to be genome-wide significant for 25OHD, which described 2.44% of the variance in 25OHD in CaMos. All 4 SNPs map to genes within the vitamin D metabolic pathway. MR analyses demonstrated that a 1-SD decrease in natural log-transformed 25OHD increased AD risk by 25% (odds ratio 1.25, 95% confidence interval 1.03-1.51, p = 0.021). After sensitivity analysis in which we removed SNPs possibly influenced by pleiotropy and population stratification, the results were largely unchanged.

Conclusions: Our results provide evidence supporting 25OHD as a causal risk factor for AD. These findings provide further rationale to understand the effect of vitamin D supplementation on cognition and AD risk in randomized controlled trials.

Figure 1. Mendelian randomization (MR) design using directed acyclic graphs

Red arrows show the potential sources of bias that may influence estimates derived from observational studies. Because it is difficult to fully protect observational studies from confounding, residual confounding from lifestyle factors that determine both vitamin D status and Alzheimer disease (AD) risk is possible. Green arrow denotes potential reverse causation, which is of concern in the investigation of a disease such as AD with a late-life onset. Thus, in observational epidemiology, it is difficult to assess whether decreased vitamin D preceded disease onset. With the MR design, bias due to reverse causation or confounding is greatly reduced because the single nucleotide polymorphisms are assigned before disease onset and are not associated with confounders as a result of the process of randomization at conception.

Figure 2. Vitamin D pathway

Here we show the vitamin D pathway. Red indicates the gene products with which our single nucleotide polymorphisms (SNPs) have been associated. Note that all 4 SNPs lie in or near genes involved in vitamin D synthesis, transport, or metabolism. UVB = ultraviolet B. Adapted from Mokry et al.

Figure 3. Forest plot of main results

Forest plot of our main mendelian randomization analysis. Boxes and error bars represent the effect of an individual single nucleotide polymorphism (SNP) on Alzheimer disease (AD) weighted by its effect on natural log 25-hydroxyvitamin D (25OHD) in SD units. Red diamond represents the summary estimate for the effect of a 1-SD decrease in genetically determined natural log 25OHD on the odds of AD. Chr = chromosome; CI = confidence interval; OR = odds ratio.

Figure 4. Forest plot of sensitivity analyses

Forest plot of our sensitivity and stratified mendelian randomization analyses plotted against our main analysis for comparison. AD = Alzheimer disease; CI = confidence interval; OR = odds ratio; 25OHD = 25-hydroxyvitamin D.