Commentary: Role of vitamin D in disease through the lens of Mendelian randomization—Evidence from Mendelian randomization challenges the benefits of vitamin D supplementation for disease prevention

A decrease in vitamin D levels, as determined through measurement of 25-hydroxyvitamin D (25OHD) has been associated with >100 human traits and diseases, among which many are extra-skeletal outcomes.1 Moreover, approximately 40% of the general adult population in the USA has vitamin D insufficiency,2 which can be diagnosed by a simple blood draw and can be corrected safely and inexpensively with oral vitamin D supplements. Notably, a 60-fold increase in vitamin D supplement use has been observed between 2000 and 2014 in the USA, where 18% of the population take at least 1000 IU of vitamin D daily.3 Yet, the evidence from the literature on the causal role of low 25OHD in disease is often contradictory. Since levels of 25OHD are confounded by known drivers of disease, such as smoking, obesity and the healthy-user effect,4 the most prudent method to interrogate the effect of vitamin D on disease risk would be through large-scale randomized controlled trials. This is because observational studies may be biased by the aforementioned confounders. However, such trials require large numbers of participants and long follow-up, and would likely be funded by the public purse, since vitamin D is not patentable. In the present and a recent (December 2018) issue of the International Journal of Epidemiology, Meng et al.5 and Jiang et al.6 in two separate papers provide new and clinically relevant evidence on the relationship between 25OHD and disease in this context, and in the absence of high-quality trial data. In both papers, a study design called Mendelian randomization was applied in order to improve causal inference of the effect of 25OHD on disease risk. Mendelian randomization uses genetic determinants of 25OHD levels as instrumental variables to decrease potential bias due to confounding.7 This bias is reduced because genetic variants are randomized at conception and this randomization process generally breaks the association with potentially confounding factors. Therefore, this method can be used to reduce the confounding that would be expected in a study of the association of 25OHD level with disease risk. Also, since the genetic variants are assigned at conception and remain stable over the lifetime, they inform us on a causal effect of a lifetime of lowered 25OHD level. A limitation of Mendelian randomization is the assumption that the genetic polymorphisms affect the studied outcomes only through the exposure, here the 25OHD level. In the case of 25OHD, bias from this assumption is less likely since all its related genetic variants are in or near genes directly involved in vitamin D synthesis or metabolism.8 In the two Mendelian randomization studies, Meng and colleagues and Jiang and colleagues used six common genetic alleles strongly associated with lowered 25OHD level, identified in a cohort of 79 366 individuals of European ancestry.9 Bycroft et al. then examined if these same alleles were associated with 920 disease outcomes in 339 256 White British individuals from UK Biobank.10 Jiang and colleagues did the same for breast and prostate cancer, in a large epidemiologic cohort, comprising 122 977 breast cancer cases and 79 148 prostate cancer cases.11 Both studies had sufficient statistical power to detect moderate effects (minimum odds ratios of 1.20 for a genetically determined 1 standard deviation change in log-transformed 25OHD in UK Biobank, or a minimum odds ratio of 1.08 per 25 nmol/L decrease in 25OHD in the two cancer cohorts). Both studies did not provide evidence of a causal association between reduced circulating vitamin D level and the studied outcomes, and these null results persisted despite sensitivity analyses. The major strength of the two studies is that they took advantage of the largest to date genome-wide association studies on vitamin D levels and various outcomes to test the role of low vitamin D levels, an approach that greatly reduces confounding. Also, they were well-powered to show moderate effects on the outcomes per changes in genetically determined 25OHD levels comparable to the effect of taking vitamin D supplements. Indeed, most multivitamin preparations contain 400 IU of vitamin D, which confer an average increase in 25OHD levels of 21.2 nmol/L.12 However, both studies are limited by their insufficient power to detect smaller effects of vitamin D on the outcomes. Further, the variable sample sizes for disease outcomes in UK Biobank influence power for different outcomes. The populations that were studied were not selected to have high, or low, 25OHD levels. This means that the results are applicable only to the effects of changes in 25OHD levels in the general population and should not be extrapolated to 25OHD deficiency. For example, it is known that the effects of vitamin D on fracture risk become more clinically apparent at extremely low 25OHD levels, but evidence from Mendelian randomization suggests that increasing 25OHD levels in the general population do not prevent fracture.13 Also, both papers tested linear effects of vitamin D level but do not provide insight into non-linear effects—in other words whether correcting a more profoundly low 25OHD could prevent disease. Similar to all published Mendelian randomization analyses for 25OHD, these studies are limited in their ability to elucidate causal effects of the biologically active form of vitamin D, 1,25-dihydroxyvitamin D (1,25-OH2D). Although genetically lowered total 25OHD levels do not appear to be associated with increased risk of disease, these studies still leave open the possibility that reduced lifelong 1,25-OH2D levels could actually affect disease risk. In this respect, concentrations of total 25OHD and circulating or intracellular 1,25-OH2D are weakly correlated.14 Despite these limitations, the studies by Meng et al.5 and Jiang et al. 6 are important, since they provide insight on genetic effects that are equivalent to those of vitamin D supplementation in individuals with generally normal vitamin D levels. The aforementioned two studies add to the increasing repertoire of Mendelian randomization papers exploring the causal role of vitamin D on a variety of health outcomes. Indeed, among 62 Mendelian randomization papers for vitamin D published over the past 8 years, evidence supporting a causal role of vitamin D was provided only for a limited number of outcomes (14 in total, including multiple sclerosis8 Alzheimer’s disease,15 delirium,16 lipid levels,17 hypertension,18 all-cause mortality,19 cancer mortality,20 adiponectin levels,21 ovarian cancer22 and type 2 diabetes23). Interestingly, the study by Meng et al., using larger sample sizes from UK Biobank, challenges the positive results of previous Mendelian randomization studies testing the role of vitamin D in hypertension, all-cause mortality and type 2 diabetes. Nonetheless, since small effects and effects of frank vitamin D deficiency cannot currently be tested with Mendelian randomization, full testing of these hypotheses would still require large-scale randomized controlled trials. Taken together, this evidence suggests that, although we cannot rule out small beneficial effects of vitamin D supplementation for certain diseases, we can exclude large effects of vitamin D on the majority of studied outcomes, and that most of its attributed causal associations are likely driven by confounding. These findings are also concordant with results of the recent VITamin D and OmegA-3 TriaL (VITAL) study that showed a lack of effect of vitamin D supplementation on cancer and major cardiovascular events.24 With the emergence of larger scale genome-wide association studies for 25OHD levels, future Mendelian randomization studies will take advantage of an expanded number of vitamin D variants, explaining a larger portion of the variance of 25OHD levels. Yet at present, Mendelian randomization studies are providing increasingly clear insights into the role of 25OHD in risk of disease and helping to refine the list of diseases that may be influenced by low 25OHD levels. The Richards lab is supported by the Canadian Institutes of Health Research (CIHR), the Canadian Foundation for Innovation and the Fonds de Recherche Santé Québec (FRQS). Dr Richards is supported by a FRQS Clinical Research Scholarship. Despoina Manousaki is supported by the Juvenile Diabetes Research Foundation (JDRF) (Award number: 3-PDF-2017-370-A-N). Conflict of interest: None declared.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.