Large GWAS of CVD in the context of T2D have investigated the presence of overlapping loci by (1) conducting GWAS of individual traits and looking for overlap, analysing CVD and T2D in a combined GWAS, and analysing CVD stratified by T2D status. Based on individual GWAS studies of T2D [
6•], CAD [
7], large artery stroke [
10] and PAD [
9], there are overlapping signals of association in the
CDKN2A/B locus (Table
1). Variants in this locus have been associated with CAD severity in subjects with T2D, but this study was small [
18], and larger studies of CAD stratified by T2D status have shown no difference in allelic effects by T2D status in this region, indicating that this locus is associated with CAD irrespective of T2D status [
19]. A combined GWAS of T2D and CAD identified a single variant associated with both T2D and CAD at genome-wide significance (
p ≤ 5 × 10
−8, a threshold used in GWAS to identify robust associations) near
IRS1 (Table
1). The study describes eight additional lead variants from eight loci that are associated with T2D and CAD, but these are not genome-wide significant for both diseases. They share the same risk allele for seven lead variants and opposite risk alleles for
APOE [
20•]. This highlights that even in large studies there are few overlapping CVD and T2D loci, that few variants contribute jointly to CVD and T2D, and that these variants do not always share the same risk allele for T2D and CVD.
Table 1
There are two loci with evidence of association signal overlap amongst cardiovascular disease and type 2 diabetes based on large genome-wide association studies
9:22043612 | rs1412830 (CDKN2A/B) | Type 2 diabetes | C (0.63) | 1.04 (1.02–1.05) | 9.1 × 10−8 |
Coronary artery disease | C (0.68) | 1.12 (1.10–1.15) | 2.6 × 10−30 |
Peripheral artery disease | C (0.63) | 1.06 (1.03–1.09) | 7.6 × 10−4 |
Large vessel stroke | C (0.63) | 1.16 (1.08–1.26) | 1.5 × 10−4 |
2:227020653 | rs7578326 (IRS1) | Type 2 diabetes | A (0.65) | 1.07 (1.05–1.09) | 2.3 × 10−13 |
Coronary artery disease | A (0.65) | 1.05 (1.04–1.07) | 4.7 × 10−10 |
Smaller studies have reported loci associated with CVD in the context of T2D. A study of CAD in subjects with T2D reported an association near
GLUL with CAD that showed some evidence for interaction with T2D status [
21]. This finding was not supported by a larger study of CAD in subjects with T2D that found no variants specifically associated with CAD in the context of T2D [
19]. The Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial reported two variants associated with fatal cardiovascular events in subjects under intense glycaemic control. Despite large-scale efforts to identify overlapping genomic regions, there are few robustly identified loci. However, most T2D risk-raising alleles are also risk-raising alleles for CAD [
20•] and MR studies have shown a causal link between T2D and CAD [
20•].
A Causal Relationship Between Type 2 Diabetes and Cardiovascular Disease
MR analyses deconvolute the causal relationship between two traits where the relationship may be confounded by environmental effects. In contrast to genetic correlation, which assesses the shared genetic background of two traits across all variants in a GWAS, MR uses genetic risk scores (GRS) of SNPs as a genetic instrument to test the causal effects of one trait on another. Since unlinked genotypes are randomly allocated at birth, the association between genetic variation determining one trait and the genetic variation determining another is free from environmental confounding. Bidirectional MR analyses can be used to distinguish between biomarkers that are on the causal pathway to a disease from those that are a consequence of a disease—otherwise known as reverse causation, such as the relationship between C-reactive protein (CRP) and CAD [
22]. There are several assumptions that must be met for MR analyses to be valid. One is that the genetic instrument should only represent the effect of the trait being tested. Variants can have pleiotropic effects, where they influence more than one trait, for example CRP and low-density lipoprotein cholesterol (LDL-C) levels [
23]. If this pleiotropy is not taken into account, this can lead to spurious associations between a trait GRS and an outcome [
24].
Variants associated with T2D have been shown to contribute to the development of the disease through different mechanisms, such as beta cell function, obesity, insulin secretion, obesity and hyperlipidaemia [
6•,
15]. MR analyses of T2D and CVD have been conducted during different epochs of T2D locus discovery. Early studies, that did not take variant pleiotropy (associations with other cardiometabolic traits) into account, found associations between a T2D-GRS and CAD [
25,
26] and separately with ischaemic stroke (large artery and small vessel stroke only) [
27]. However, the average CAD risk per T2D allele was lower than expected [
25] indicating that the GRS did not account for the all the risk of CAD observed in subjects with T2D. More recent MR studies have generated GRS based on more T2D-associated variants and have leveraged pleiotropic variant associations to construct T2D-GRS that contain variants associated with other similar traits. T2D-GRS constructed of non-pleiotropic T2D-associated variants was associated with CAD [
28], but those constructed from variants pleiotropic for established CAD risk factors and for glycaemic traits have shown variable associations with CAD.
MR studies of shared risk factors for T2D and CVD have shown variable association with either T2D and CVD. There is variable evidence to support a role for high-density lipoprotein cholesterol (HDL-C), LDL-C, triglycerides, obesity, adiponectin and hyperglycaemia in the development of T2D and CVD [
29‐
33]. Genetic variability increasing levels of LDL-C and triglycerides have been associated with increased risk of CAD but decreased the risk of T2D [
34]. This is juxtaposed to the literature that correlates increased serum levels of LDL-C and triglycerides with increased T2D risk [
35] Genetic variability increasing HDL-C levels was shown to be protective of CAD and T2D [
34,
36], but other studies that excluded pleiotropic variants found no effect of HDL-C on CAD or T2D risk [
37,
38].
Genetic determinants of glycaemic traits have been studied in healthy populations, which do not necessarily represent the genetic variability that determines glycaemic traits in subjects with T2D [
8]. Some MR studies have tried to address this through careful selection of variants used to build genetic instruments for glycaemic traits. Genetic instruments consisting of T2D-associated variants that were also associated with any other glycaemic trait were not associated with CAD [
20•]. However, GRS constructed from T2D associated variants also associated with HbA1c, beta cell function and insulin resistance were respectively associated with increased CAD risk but not those associated with fasting glucose and T2D [
26]. On the other hand, instruments for fasting glucose that exclude T2D-associated variants were associated with increased CAD risk [
28]. This suggests, perhaps unsurprisingly, that genetic mechanisms and underlying pathways that increase the risk of T2D do not uniformly influence CAD risk.
Results from MR studies can vary despite testing the relationship between the same traits and outcomes. This could reflect between-study differences, such as GRS strength, choice of variants included in the GRS, the sample size of the study, how subjects were recruited and the study design. Study design and participation can introduce collider bias creating false causal relationships between two traits [
39,
40]. In cross-sectional studies, this can be introduced by selection bias. If subjects with the highest genetic risk of a disease are less likely to participate in a study then this may cause an inverse association between GRS for known risk factors and the disease outcome [
39]. In MR studies of disease progression in cases only, factors that are causal for disease onset may also be associated with disease progression through association with confounders of disease incidence and progression. Overall, MR studies have highlighted the complex relationships amongst T2D, CVD and their shared risk factors and how these are challenging to deconvolute.
Epigenetic Changes and Hyperglycaemia
Epigenetic changes to gene expression usually involve histone modifications and DNA methylation in response to a stimulus, such as disease state or environment. There is some experimental evidence of epigenetic changes that modify the risk of CVD, induced by shared risk factors for CVD and T2D, including hyperglycaemia [
41]. In the Diabetes Control and Complications Trial (DCCT) and follow-up Epidemiology of Diabetic Complications and Interventions Trial (EDIC), intensive glycaemic control was shown to reduce the progression of complications (including cardiovascular) in subjects with T1D, but not completely abrogate them [
42]. The concept of ‘metabolic memory’ was coined to account for the observation, which has been shown to occur because of epigenetic changes induced by hyperglycaemia [
43]. Hyperglycaemia has been shown to play a causal role in the development of T2D and CVD [
30], although trials of intensive glycaemic control in subjects with T2D have demonstrated variable effects on rates of CVD and all-cause mortality.
A large meta-analysis of glycaemic intervention trials (34,533 subjects) showed a small reduction in non-fatal myocardial infarction in the glucose-lowering group, but no overall effect on all-cause mortality or CVD death [
44]. Two of the largest glycaemic intervention trials were conducted for 5 years or less [
45,
46], and may have been too short to observe effects on cardiovascular outcomes and all-cause mortality [
47]. After 5 years of follow-up in the Action in Diabetes and Vascular Disease (ADVANCE) trial, there was a non-significant trend of reduced CVD and all-cause mortality in the treatment arm, and in a 10-year follow-up of the United Kingdom Prospective Diabetes Study (UKPDS), there were significantly reduced rates of CVD and all-cause mortality in the treatment intervention groups [
47]. A reduction in myocardial infarction and all-cause death was observed in the 10-year follow-up, despite the loss of glycaemic differences after a year [
47]. The ‘legacy effect’ was used to explain this observation and may correspond to ‘metabolic memory’ in subjects with T1D [
43]. While the exact mechanism of the ‘legacy effect’ is unknown, there is experimental evidence from vascular cells that supports a role for epigenetic changes in the risk of CVD induced by hyperglycaemia [
48]. Epigenetic modifications change the expression of genes and pathways associated with endothelial dysfunction (a key step in atherogenesis), and genes involved in metabolic and cardiovascular disease [
48]. These epigenetic changes could explain the mechanism by which hyperglycaemia could increase the risk of CVD in subjects with T2D.