Association between circulating 25-hydroxyvitamin D and incident type 2 diabetes: a mendelian randomisation study

Summary Background Low circulating concentrations of 25-hydroxyvitamin D (25[OH]D), a marker of vitamin D status, are associated with an increased risk of type 2 diabetes, but whether this association is causal remains unclear. We aimed to estimate the unconfounded, causal association between 25(OH)D concentration and risk of type 2 diabetes using a mendelian randomisation approach. Methods Using several data sources from populations of European descent, including type 2 diabetes cases and non-cases, we did a mendelian randomisation analysis using single nucleotide polymorphisms (SNPs) within or near four genes related to 25(OH)D synthesis and metabolism: DHCR7 (related to vitamin D synthesis), CYP2R1 (hepatic 25-hydroxylation), DBP (also known as GC; transport), and CYP24A1 (catabolism). We assessed each SNP for an association with circulating 25(OH)D concentration (5449 non-cases; two studies), risk of type 2 diabetes (28 144 cases, 76 344 non-cases; five studies), and glycaemic traits (concentrations of fasting glucose, 2-h glucose, fasting insulin, and HbA1c; 46 368 non-cases; study consortium). We combined these associations in a likelihood-based mendelian randomisation analysis to estimate the causal association of 25(OH)D concentration with type 2 diabetes and the glycaemic traits, and compared them with that from a meta-analysis of data from observational studies (8492 cases, 89 698 non-cases; 22 studies) that assessed the association between 25(OH)D concentration and type 2 diabetes. Findings All four SNPs were associated with 25(OH)D concentrations (p<10−6). The mendelian randomisation-derived unconfounded odds ratio for type 2 diabetes was 1·01 (95% CI 0·75–1·36; p=0·94) per 25·0 nmol/L (1 SD) lower 25(OH)D concentration. The corresponding (potentially confounded) relative risk from the meta-analysis of data from observational studies was 1·21 (1·16–1·27; p=7·3 × 10−19). The mendelian randomisation-derived estimates for glycaemic traits were not significant (p>0·25). Interpretation The association between 25(OH)D concentration and type 2 diabetes is unlikely to be causal. Efforts to increase 25(OH)D concentrations might not reduce the risk of type 2 diabetes as would be expected on the basis of observational evidence. These findings warrant further investigations to identify causal factors that might increase 25(OH)D concentration and also reduce the risk of type 2 diabetes. Funding UK Medical Research Council Epidemiology Unit and European Union Sixth Framework Programme.


Descriptions of studies contributing to mendelian randomisation analysis
Studies contributing to mendelian randomisation (MR) analysis are shown in Supplementary Table 1 and Supplementary Figure 1. In our MR analysis, there was some overlap of study participants who contributed to the multiple datasets, as highlighted by asterisk (*) in text and as conducted in previous MR analysis on type 2 diabetes (T2D). 1 We evaluated them in an appropriate fashion to avoid double-counting. Sample sizes for genetic analysis varied across single nucleotide polymorphisms (SNPs) and cohorts, due to a small number of missing genotyping data in each of our datasets (<4·5% for each SNP). We excluded individuals with missing information on genes and missing information on 25-hydroxy vitamin D biomarkers, 25(OH)D, or diagnosis of T2D, depending on the outcome in analysis. The numbers of individuals included in MR analysis are shown in Supplementary Figures 1, 3, and 4.

Studies contributing to estimates of associations between SNPs and concentrations of 25(OH)D.
The Ely study*: The Ely study is a population-based cohort of white European men and women in Ely, Cambridge, UK. [2][3][4][5] European-origin adults, aged 40-69 years and free of known diabetes and registered with a single practice serving Ely, were randomly selected. The baseline examination took place between 1990 and 1992. In the ADDITION-Ely case-control study (described below), controls were selected from the non-diabetic adults in Ely 1 , where oral-glucose tolerance test was implemented at baseline and during the follow-up. Among 1,608 non-diabetic adults with genotype information, 25(OH)D measurements were undertaken among 684 adults. 5 All participants gave written informed consent.

EPIC-Norfolk study*:
The EPIC (European Prospective Investigation of Cancer)-Norfolk study is a prospective cohort of 25,639 men and women, aged 40-79 years, who were recruited between 1993 and 1997 in Norfolk, UK. 5,6 A detailed description of the study design and cohort characteristics has previously been published. 5 In the current analyses, 6785 participants free of known diabetes at baseline or during follow-up were randomly selected for genotyping. Among them, 4765 participants had 25(OH)D data. 1 For 25(OH)D concentrations , ultra-performance liquid chromatography-tandem mass spectrometry was used to measure 25(OH) vitamin D 2 (present in very small concentration) and 25(OH) vitamin D 3 , which were summed to provide a measure of total 25(OH) vitamin D. 5 Ethical approval for the study was obtained from the Norfolk and Norwich Hospital Ethics Committee.

Studies contributing to estimates of associations between SNPs and risk of T2D
Cambridgeshire case-control study (CCCS): the Cambridgeshire case-control study is a population based study of type 2 diabetes (T2D). 1,7 Diabetes cases aged 45-76 years were randomly selected from general practitioner (GP) diabetes registers in Cambridgeshire, UK, and T2D was defined as onset of diabetes after the age of 30 years and without insulin use in the first year after diagnosis. Controls were recruited at random from the same population sampling frame and individually matched to cases for age (years), sex and site of GP registry. 1 Controls were screened as non-diabetic by medical records and glycated haemoglobin not more than 6%. We excluded participants with missing information on body mass index (BMI) (n=15). In the current analyses, we included 552 T2D cases and 534 controls that had DNA available and information on BMI. The study received ethical approval from the Cambridge Local Research Ethics Committee, and participants provided informed consent.

ADDITION-Ely case-control study:
Cases of this study was derived from the ADDITION-Cambridge study, a multi-centre intervention study which focuses on effectiveness of stepwise screening on morbidity and mortality among people with new onset T2D. 1,8 People aged 40-69 years at high-risk of undiagnosed diabetes in East Angelia region participated in the study. Written informed consent was obtained for all participants at the time of the diabetes screening appointment and subsequent diagnostic test. Among adults aged 40 to 69 year participating in the UK Cambridge arm of the ADDITION study, new onset T2D cases were identified via a population-based stepwise screening strategy, including casual glucose, plasma glucose, glycated haemoglobin, and oral-glucose tolerance test. 8 All T2D cases were confirmed by 75g oral-glucose tolerance test. The controls were identified from Ely Study through matching by age, sex, and GP registry (as seen above*). 1 We excluded participants with missing data on BMI (n=11), or on age and/or sex (n=3), or non-white participants (n=30). Current analyses included 896 T2D cases and 1,608 controls of white European-origin men and women who had DNA available and information on BMI. The Cambridge Research Ethics Committee approved both studies.
Norfolk Diabetes case-control study: The Norfolk Diabetes case-control study is an ongoing study of white European men and women with T2D in Norfolk. 5,9 All diabetes patients identified through GP diabetes registers in Norfolk, local hospital diabetes clinic and retinal screening programme patient registers were invited to participate. European cases aged 30 years or older were included as cases in this study. T2D was defined by not treated with insulin during the first year of diagnosis. Those with cystic fibrosis, chronic pancreatitis or long term steroid use were excluded from the study. After excluding participants with missing data on sex (n=18), or non-white participants (n=277), a total of 6359 cases were included in the current analyses. Control participants (n = 6785) free of known diabetes at baseline or during follow-up were randomly selected from among EPIC-Norfolk participants (as seen above*). Because 830 participants from the EPIC-Norfolk study were also part of the EPIC-InterAct study (see below), we excluded those participants from the EPIC-Norfolk study to avoid double counting. The Norfolk study was approved by the Norwich Local Research Ethics Committee.

EPIC-InterAct study:
The EPIC-InterAct study is a large prospective case-cohort study nested within the Europe-wide EPIC study, involving 27,779 individuals from eight European countries. 10 The EPIC-InterAct study comprises 12,403 incident cases of T2D derived from a total cohort of 340,234 individuals comprising 3·99 million person-years of follow-up and includes a representative sub-cohort (n=16,154) that also includes 778 of the 12,403 incident T2D cases (as per the design of a case-cohort study). 10 Ascertainment of incident T2D involved a review of the existing EPIC datasets at each centre using multiple sources of evidence. None of the cases were ascertained solely by self-reported diagnosis. Follow-up was censored at the date of diagnosis, the 31st of December 2007, or the date of death, whichever came first. After excluding participants with missing data on BMI (n=197), in the current analyses, we included 8166 incident cases of T2D and 10,555 diabetes free participants. All participants gave written informed consent, and the study was approved by the local ethics committees in the participating centres and the Internal Review Board of the International Agency for Research on Cancer.

DIAGRAM (DIAbetes Genetics Replication And Meta-analysis):
The DIAGRAM consortium represents the effort of groups of international researchers working on the genetics of T2D, with a principal focus on sample of European descent. The details are available in a prior publication. 11 Summary data from DIAGRAM consortium (12,171 T2D cases and 56,862 controls) are available at www.diagram-consortium.org.

Studies contributing to estimates of associations between SNPs and glycaemic traits.
MAGIC (the Meta-Analyses of Glucose and Insulin-related traits Consortium): MAGIC represents a collaborative effort to combine data from multiple GWAS to identify additional loci that impact on glycaemic and metabolic traits. The details are available in prior publications. [12][13][14] Summary data from MAGIC (n=46,391) are available at www.magicinvestigators.org.

Searching strategy and meta-analysis of prospective studies on circulating 25-hydroxy vitamin D and incident type 2 diabetes
We updated our previous meta-analysis of the association between circulating 25-hydroxy vitamin D, 25(OH)D, and incident T2D. 5 We identified all the additional studies published between 31 January 2012 and 17 June 2014. Electronic searches were performed for publications in PubMed. We also reviewed reference lists of articles identified during the search. The search terms were related to vitamin D concentrations ("25-hydroxy vitamin D" or "25(OH)D" or "vitamin D"), and diabetes risk ("diabetes" or "glucose" or "metabolic syndrome" or "hyperglycaemia") without limits on publication date or language.
We included prospective studies that measured circulating 25(OH)D and used standard definitions of T2D based on World Health Organization criteria. 5 Since MR analyses only included participants of European descent, the meta-analysis of prospective studies was also restricted to European individuals.
For each eligible study newly identified (not in our previous meta-analysis 5 ), information was extracted based on a pre-specified protocol. We considered odds ratios, risk ratios, and hazard ratios as estimates of the relative risk. The prospective association of 25(OH)D concentrations with T2D was summarized using random effects meta-analysis. The overall effect was estimated for per 1 SD lower 25(OH)D concentration. Heterogeneity was assessed using the Q-statistic test 15 [12][13][14] MAGIC provided information on associations of SNP with glycaemic traits among non-diabetic adults, not evaluating T2D cases.

Supplementary Figure 1. Association of four single nucleotide polymorphisms with 25-hydroxy vitamin D concentrations, their genetic scores, and various characteristics assessed in non-diabetic adults in four studies: control participants from Cambridgeshire case-control study, non-diabetic adults from the Ely study, EPIC-Norfolk study, and the EPIC-InterAct study.
Estimates were based on random-effects metaanalysis of study specific associations of each SNP with potential confounders. Four SNPs related to vitamin D synthesis and metabolism: CYP2R1 rs10741657, DHCR7 rs12785878, DBP rs4588, and CYP24A1 rs17217119. * SD changes for continuous variables and change in levels (based on polynomial logistic regression) for family history of diabetes (yes or no); smoking status (never smoked, ever smoker, or current smoker); or physical activity (inactive, moderately inactive, moderately active, and active). Abbreviations: 25-hydroxy vitamin D, 25(OH)D; single nucleotide polymorphism, SNP; BMI, body-mass index; CI, confidence interval; HDL, high-density lipoprotein; LDL, low-density lipoprotein; SD, standard deviation.