Type 2 diabetes mellitus and cardiovascular health: Evidence of causal relationships in a European ancestry population

Abstract

Aims

Type 2 diabetes mellitus (T2DM) is associated with increased cardiovascular disease (CVD) risk, but whether T2DM directly causes adverse cardiac remodelling is uncertain. We performed a comprehensive Mendelian randomization (MR) analysis to investigate the causal relevance of T2DM to CVD outcomes and cardiac structure/function.

Methods and results

Bidirectional two‐sample MR was conducted using summary‐level data from European‐ancestry genome‐wide association studies. The T2DM GWAS data included 80 154 cases and 853 816 controls from the DIAGRAM consortium. Outcomes included coronary artery disease (CAD), myocardial infarction (MI), stroke, heart failure, atrial fibrillation, and various quantitative cardiac imaging traits assessed by magnetic resonance imaging. MR analysis revealed causal associations between genetic predisposition to T2DM and increased risk of CAD (odds ratio [OR] 1.104, 95% confidence interval [CI] 1.078–1.130, P = 2.59e‐16), MI (OR 1.129, 95% CI 1.094–1.166, P = 6.02e‐14) and stroke (OR 1.086, 95% CI 1.064–1.109, P = 1.02e‐14). These associations were validated in the FinnGen cohort (CAD: OR 1.117, 95% CI 1.075–1.158, P = 1.56e‐9; MI: OR 1.132, 95% CI 1.083–1.184, P = 4.27e‐8; stroke: OR 1.138, 95% CI 1.107–1.170, P = 3.52e‐20). Multivariable MR show consistent findings (CAD: OR 1.063, 95% CI 1.031–1.097, P = 1.11e‐4; MI: OR 1.088, 95% CI 1.042–1.135, P = 1.12e‐4; stroke: OR 1.066, 95% CI 1.032–1.101, P = 1.18e‐4) after adjusting for cardiometabolic traits. T2DM was causally associated with higher left ventricular mass index (β = 0.473, 95% CI 0.193 to 0.752, P = 0.001), lower indexed right atrial minimum (β = −0.048, 95% CI −0.073 to −0.022, P = 2.1e‐5), and maximum (β = −0.042, 95% CI −0.065 to −0.019, P = 4.12e‐5) areas. The effects on right atrial size remained significant after adjusting for risk factors (minimum area: β = −0.041, 95% CI −0.072 to −0.010, P = 0.009; maximum area: β = −0.039, 95% CI −0.069 to −0.008, P = 0.012). Both apolipoprotein A1 and SBP are important mediators in the causal relationship between T2DM and left ventricular mass index. No reverse causal associations were identified.

Conclusions

Our MR study demonstrates that genetic liability to T2DM plays causal roles in CAD, MI, stroke, and cardiac structure changes including left ventricular hypertrophy and reduced right atrial dimensions. These findings provide genetic evidence supporting glycaemic control in T2DM to mitigate cardiovascular complications and adverse cardiac remodelling.

Introduction

Type 2 diabetes mellitus (T2DM) is a complex metabolic disorder, primarily marked by chronic hyperglycemia due to progressive insulin resistance and a decline in pancreatic β‐cell function. ¹ Over the past few decades, the prevalence of T2DM has dramatically increased worldwide, impacting nearly half a billion adults globally. ² , ³ T2DM significantly increases the risk of cardiovascular disease (CVD), two to four times, which remains the leading cause of death and serious health complications in this population. ⁴

Emerging evidence suggests that T2DM may directly induce adverse structural and functional changes in the myocardium, contributing to an increased susceptibility to heart failure (HF), arrhythmias, and adverse cardiovascular events. ⁵ Imaging studies have demonstrated subtle changes in cardiac structure and function among asymptomatic patients with T2DM, including left ventricular hypertrophy and diastolic dysfunction. ⁶ , ⁷ More advanced myocardial disease characterized by diffuse fibrosis and systolic impairment is common in those with established T2DM. ⁸ These myocardial abnormalities occur early in the natural history of T2DM, often preceding the onset of clinical CVD. ⁷ Whether such changes are causal or merely represent epiphenomena is uncertain.

Randomized controlled trials (RCTs) are the gold standard for establishing causality but are often hindered by ethical, practical, or financial constraints. Therefore, past studies exploring the causality between diabetes and cardiovascular outcomes have predominantly been observational. Diabetic patients frequently present comorbidities like hypertension and hyperlipidaemia, introducing confounding factors that can bias results. Assessing causality between T2DM and cardiovascular complications in observational studies is challenging due to potential residual confounding and reverse causation.

Moreover, few studies have specifically investigated the causal relevance of T2DM to myocardial structure and function. Cardiac magnetic resonance (CMR) imaging enables detailed phenotyping of cardiac structure and function. CMR is the gold standard for accurately quantifying ventricular size, morphology, performance, and tissue characteristics. ⁹ It has proven valuable in enhancing CVD risk stratification and assessing subclinical disease severity. ⁸ , ⁹ Large‐scale population studies using CMR are shedding light on the genomic factors influencing cardiac traits. ¹⁰ However, research into the causal effects of T2DM on myocardial structure and function remains limited, primarily due to substantial financial constraints.

Mendelian randomization (MR) has been increasingly applied to investigate the causal role of various exposures on complex diseases such as CVD. By employing genetic variants associated with exposures as instrumental variables (IVs), MR can assess whether exposures is causally related to cardiovascular outcomes and cardiac traits, overcoming many of the limitations of traditional observational studies. MR relies on the principle that genetic variants are randomly allocated at conception, mirroring the randomization process in controlled trials. ¹¹ These genetic variants, which are associated with the exposure of interest, can serve as unconfounded proxies for the exposure, allowing for an unbiased assessment of the causal effect on the outcome. This approach can help inform prevention and treatment strategies by identifying causal risk factors and potential therapeutic targets.

In this study, we conducted a comprehensive MR analysis to explore the causal relationship between T2DM and a range of CVD endpoints. Additionally, we investigated the influence of T2DM on cardiac structure and function. This investigation aimed to provide novel insights into the causal relevance of T2DM on cardiovascular outcomes and cardiac remodelling. Our findings will help inform a precision medicine approach to CVD management in T2DM.

Methods

Study design

In this study, we utilized bidirectional MR analyses to explore the potential causal relationships between T2DM and cardiovascular outcomes, as well as its impact on cardiac structure and function. To ensure robustness, we employed multiple MR methods and conducted reverse MR to assess any reverse causality. To minimize confounding factors, we adjusted for related cardiometabolic traits using multivariable MR when examining the causal effects of T2DM on cardiovascular outcomes. For external validation of these cardiovascular endpoints, we used data from the FinnGen database. Our data sources are outlined in Table 1 . This research adheres to the STROBE‐MR Guideline. ²² The procedural outline of our study, created with BioRender.com, is illustrated in Figure 1 .

Table 1.

Summary of GWAS data sources used in the study

Phenotype	Study or Biobank	Ancestry	Cases/controls	Disease definition	Units	PMID
Exposure
T2DM	Mahajan et al. 12	EUR	80 154/853 816	Type 2 diabetes mellitus	Log (OR)	35 551 307
Cardiac outcomes
Atrial fibrillation	Nielsen et al. 13	EUR	60 620/970 216	Clinically diagnosed atrial fibrillation or flutter	Log (OR)	30 061 737
Coronary artery disease	Aragam et al. 14	EUR	181 522/984 168	documented history of angina pectoris, myocardial infarction or other forms of coronary heart disease, as well as a history of revascularization procedures	Log (OR)	36 474 045
Myocardial infarction	Hartiala et al. 15	EUR	61 000/578 000	1.Doctor‐diagnosed MI; 2.ICD10 I21, I22, I23, I25.2; 3.Self‐reported MI	Log (OR)	33 532 862
Heart failure	Shah et al. 16	EUR	47 309/930 014	All‐cause heart failure	Log (OR)	31 919 418
Stroke	Mishra et al. 17	EUR	73 652/1 234 808	Any stroke	Log (OR)	36 180 795
Validation cardiac outcomes
Atrial fibrillation	FinnGen R9 data release	EUR	45 766/191 924	ICD‐10 I48	Log (OR)	NA
Heart failure	FinnGen R9 data release	EUR	26 872/349 361	All‐cause heart failure	Log (OR)	NA
Myocardial infarction	FinnGen R9 data release	EUR	24 185/313 400	ICD‐10 I21, I22	Log (OR)	NA
Coronary artery disease	FinnGen R9 data release	EUR	47 550/313 400	ICD‐10 I24, I25, T82.2, Z95.1	Log (OR)	NA
Ischaemic stroke	FinnGen R9 data release	EUR	25 398/339 920	ICD‐10 I61, I63, I64	Log (OR)	NA
Cardiac structure and function
Left ventricle, aorta, pulmonary artery, and right heart	Pirruccello et al. 10	EUR	45 504	NA	1 SD	35 697 867
Left atrial	Ahlberg et al. 18	EUR	35 658	NA	1 SD	34 338 756
Left ventricular mass	Khurshid et al. 19	EUR	43 230	NA	1 SD	36 944 631
Myocardial interstitial fibrosis	Nauffal et al. 20	EUR	41 505	NA	1 SD	37 081 215
Risk factors
HDL cholesterol	UK Biobank	EUR	432 009	NA	1 SD	NA
LDL cholesterol	UK Biobank	EUR	469 869	NA	1 SD	NA
Triglycerides	UK Biobank	EUR	470 337	NA	1 SD	NA
Apolipoprotein A1	UK Biobank	EUR	429 666	NA	1 SD	NA
Apolipoprotein B	UK Biobank	EUR	468 375	NA	1 SD	NA
Systolic blood pressure	UK Biobank	EUR	475 939	NA	1 mmHg	NA
Diastolic blood pressure	UK Biobank	EUR	475 944	NA	1 mmHg	NA
Body mass index	Pulit et al. 21	EUR	694 649	NA	1 SD	30 239 722

T2DM, type 2 diabetes mellitus.

Figure 1.

Schematic representation of mendelian randomization analysis. BMI, body mass index.

Data sources for type 2 diabetes mellitus

The summary data for our analysis were sourced from the Diabetes Genetics Replication And Meta‐analysis (DIAGRAM) consortium. ¹² This consortium compiled association summary statistics from 122 genome‐wide association studies (GWAS), encompassing 180 834 T2DM cases and 1 159 055 controls across five ancestry groups. For our study, we specifically utilized GWAS data from European ancestry, which included 80 154 T2DM cases and 853 816 controls.

Data sources for cardiac outcomes

In this study, our cardiovascular outcomes, including atrial fibrillation (AF), ¹³ coronary artery disease (CAD), ¹⁴ myocardial infarction (MI), ¹⁵ HF ¹⁶ and stroke, ¹⁷ were derived from large‐scale GWAS meta‐analyses using summary data predominantly from individuals of European ancestry. For external validation of these cardiovascular outcomes, we utilized summary data from a GWAS meta‐analysis conducted by the FinnGen consortium, employing the FinnGen R9 data release. ²³ The sources of our data are comprehensively detailed in Table 1 .

Data sources for cardiac structure and function

Estimates of genetic associations with key CMR parameters were comprehensively gathered from several studies. Parameters including left ventricular end‐systolic and end‐diastolic volumes, ejection fraction, and dimensions of the right atrium and ventricle were reported in the study by Pirruccello et al., ¹⁰ utilizing data from the UK Biobank Imaging Cohort. The research conducted by Ahlberg et al. ¹⁸ shed light on left atrial maximum volume and total ejection fraction, while Khurshid et al. ¹⁹ provided estimates for left ventricular mass. Additionally, Nauffal et al. ²⁰ focused on myocardial interstitial fibrosis in their study. These diverse investigations, drawing on data from the UK Biobank Imaging Cohort, collectively deepened our understanding of cardiac structure and function. To ensure comparability across different body sizes, all cardiac imaging measures were indexed to body surface area, except for dimensionless metrics like total ejection fraction and fractional area change.

Data sources for cardiometabolic traits

In this study, we accounted for several confounding factors: body mass index (BMI), high‐density lipoprotein (HDL) cholesterol, low‐density lipoprotein (LDL) cholesterol, triglycerides (TG), apolipoprotein A1, apolipoprotein B, systolic blood pressure (SBP), and diastolic blood pressure (DBP). The GWAS data for BMI was obtained from a comprehensive meta‐analysis of European ancestry individuals, combining data from the UK Biobank and the Genetic Investigation of ANthropometric Traits (GIANT) consortium. ²¹ The data for the other cardiometabolic traits were exclusively sourced from the UK Biobank. ²⁴

Instrumental variable selection

In our study, we carefully adhered to the three essential MR assumptions for selecting IVs ¹² : (1) the IVs were associated with the exposure; (2) the IVs were not associated with any confounders; and (3) the IVs influenced the outcome solely through their impact on the exposure. Our selection of IVs involved rigorously chosen, independent single nucleotide polymorphisms (SNPs) that were genome‐wide significant, with linkage disequilibrium r ² < 0.001 and a P < 5e‐8. We refined our SNP selection by excluding those with minor allele frequencies under 0.01 and ensuring their inclusion in both the exposure and outcome GWAS datasets. For reverse MR involving SNPs related to cardiac MRI structure and function, we employed a threshold of P < 5e‐7, acknowledging the potential bias due to a limited number of SNPs (less than three in some traits). Additionally, we ascertained the strength of each SNP as an instrument by verifying an F‐statistic greater than 10, reflecting a substantial proportion of variance in the exposure explained by the SNP, as inferred from its effect size and standard error in the exposure GWAS regression analysis. This approach aimed to ensure the reliability of our IVs, minimizing the risk of violating MR assumptions. Details of the selected IVs are provided in Table S1 .

Main Mendelian randomization analyses

In our primary MR analyses, we employed inverse‐variance weighted (IVW) method, which provides an overall estimate of the causal effect by combining the ratio estimates from each genetic variant in a meta‐analysis model. ²⁵ We used a multiplicative random‐effects model to account for potential heterogeneity between the causal estimates from each variant. To account for multiple testing across the various outcomes analysed, we applied a conservative Bonferroni correction to the significance thresholds, minimizing the risk of false positive findings. Specifically, we set the significance level at P < 0.01 (0.05/5) for the CVD endpoints and P < 2.77e‐3 (0.05/18) for the cardiac imaging traits.

For binary outcomes, like CAD, our results are expressed as odds ratios, providing a measure of association between genetic variants and the likelihood of the outcome. In contrast, for continuous outcomes, notably cardiac imaging parameters, we report the findings as regression coefficients. These coefficients represent the change in outcome per one standard deviation increase in the genetically predicted values, thereby quantifying the effect size of genetic predispositions on these parameters.

Sensitivity analyses

To ensure the robustness of our findings, we performed several sensitivity analyses using alternative MR methods with different assumptions regarding instrument validity and pleiotropy. These included MR‐Egger regression, which allows for the presence of directional pleiotropy and provides a test for the presence of such pleiotropy; the weighted median estimator, which provides consistent estimates even when up to 50% of the information comes from invalid IVs; and the simple mode and weighted mode methods, which are robust to the presence of outliers and can provide consistent estimates even when the majority of instruments are invalid. We assessed heterogeneity across the individual variant estimates using Cochran's Q statistic and adopted a random‐effects model if significant heterogeneity was detected (P < 0.05). The presence of horizontal pleiotropy, which occurs when the genetic variants influence the outcome through pathways independent of the exposure, was evaluated using the MR‐Egger intercept test and the MR‐PRESSO global test. ²⁶ MR‐PRESSO also identifies and corrects for potential outliers, providing adjusted estimates after removing outlier variants.

For establishing significant causal relationships, we sought consistency across IVW, MR‐Egger, WME, SM, and WMo methods, with MR‐Egger additionally confirming the absence of horizontal pleiotropy. We also performed reverse MR analysis, swapping the exposure and outcome, to explore any potential reverse causality.

Replication analysis

In our replication analysis for cardiovascular outcomes, we used data from the FinnGen cohort. Successful replication required statistical significance (P < 0.05) in the IVW method and consistent effect estimates across all methods.

Multivariable Mendelian randomization analysis

In the multivariable MR analysis, we adjusted for potential confounding factors, such as cardiovascular risk factors, to strengthen the causal inference between T2DM and cardiovascular outcomes. Multivariable MR uses genetic variants associated with the confounding factors as additional IVs in the MR model, allowing for the estimation of the direct causal effect of T2DM on the outcomes, independent of the effects of the confounders.

The genetic variants used as IVs for the confounding factors were obtained from previously published GWAS that identified robust associations between these variants and the respective risk factors in large populations.

The multivariable MR analysis was performed using the IVW method, an extension of the standard two‐sample MR analysis. In this method, the genetic associations with the exposure (T2DM) and the confounding factors were combined in a single multivariable model. The causal effect estimates for each genetic variant were obtained by regressing the variant‐outcome associations on the variant‐exposure and variant‐confounder associations. These estimates were then combined using a multiplicative random‐effects meta‐analysis model to obtain an overall causal effect estimate adjusted for the confounding factors. ²⁷

Analysis software

All statistical analyses were conducted using R version 4.2.3 (R Foundation for Statistical Computing, Vienna, Austria). For MR analyses, we utilized the TwosampleMR (version 0.5.7) and MR‐PRESSO (version 1.0) ²⁶ R packages.

Results

Type 2 diabetes mellitus and cardiovascular outcomes

The MR analysis demonstrated significant causal associations between genetic predisposition to T2DM and increased risk of CAD (OR 1.104, 95% CI 1.078–1.130, P = 2.59e‐16), stroke (OR 1.086, 95% CI 1.064–1.109, P = 1.02e‐14) and MI (OR 1.129, 95% CI 1.094–1.166, P = 6.02e‐14).

In the FinnGen replication cohort, T2DM showed causal relationships with stroke (OR 1.138, 95% CI 1.107–1.170, P = 3.52e‐20), CAD (OR 1.117, 95% CI 1.075–1.158, P = 1.56e‐9), and MI (OR 1.132, 95% CI 1.083–1.184, P = 4.27e‐8). The detailed results of these analyses are presented in Figure 2 and Table S2 . The forest plot and scatter plots visually represent this causal relationship, as depicted in Figure 3 A and Figure S1 .

Figure 2.

Causal effects of T2DM on cardiovascular diseases. T2DM, type 2 diabetes mellitus; IVW, inverse variance weighted; SM, simple mode; WME, weighted median estimator; WMo, weighted mode.

Figure 3.

The results of an analysis using the inverse variance weighted method, illustrating the causal effects of T2DM on cardiac outcomes, structure and function. (A) Causal effects of T2DM on cardiac outcomes. (B) Causal effects of T2DM on cardiac structure and function. Asc Aorta Diam, ascending aorta diameter; LA max, left atrial maximum volume; LATEF, left atrial total ejection fraction; LVEDV, left ventricular end diastolic volume; LVEF, left ventricular ejection fraction; LVESV, left ventricular end systolic volume; LV mass, left ventricular mass; LVSV, left ventricular stroke volume; Prox PA Diam, proximal pulmonary artery diameter; RA FAC, right atrial fractional area change; RA max, right atrial maximum area; RA min, right atrial minimum area; RVEDV, right ventricular end diastolic volume; RVESV, right ventricular end systolic volume; RVSV, right ventricular systolic volume; T2DM, type 2 diabetes mellitus.

Our sensitivity analyses revealed heterogeneity (P < 0.05), necessitating the adoption of a random‐effects model. The MR‐PRESSO global test identified outliers (P < 0.05), but even after correction for these outliers, the causal estimates remained significant and directionally consistent (refer to Table S3 ). MR‐Egger regression indicated the presence of horizontal pleiotropy (P < 0.05), suggesting that the SNPs were associated with confounding factors. To account for this and explore the direct causal effects, we employed multivariable MR, adjusting for cardiovascular risk factors.

Furthermore, in the reverse MR analysis, causal relationships were observed between stroke (OR 1.131, 95% CI 1.010–1.266, P = 0.03) and CAD (OR 1.051, 95% CI 1.006–1.098, P = 0.03) on T2DM. However, after applying the Bonferroni correction, no reverse causal associations were found between these cardiovascular outcomes and T2DM. The detailed results are presented in Table S4 .

Type 2 diabetes mellitus and cardiac structure and function

The MR analysis revealed T2DM causally influences right atrial minimum area indexed (β = −0.048, 95% CI −0.073 to −0.022, P = 2.1e‐5), right atrial maximum area indexed (β = −0.042, 95% CI −0.065 to −0.019, P = 4.12e‐5) and left ventricular mass indexed (β = 0.473, 95% CI 0.193 to 0.752, P = 0.001). The detailed results of these analyses are presented in Figure S2 and Table S5 . The forest plot and scatter plots visually represent this causal relationship, as depicted in Figure 3 B and Figure S1 .

Sensitivity analyses revealed significant heterogeneity (P < 0.05), necessitating the implementation of a random‐effects model. MR‐Egger regression showed no evidence of horizontal pleiotropy (P > 0.05). Crucially, the MR‐PRESSO global test detected outliers (P < 0.05), yet the causal effects persisted as significant even after adjusting for these outliers. Furthermore, our analyses did not uncover any reverse causal relationships between the CMR parameters and T2DM. The detailed results are presented Table  S4 .

Multivariable Mendelian randomization and mediation analysis

In the multivariable MR analysis, we adjusted for several cardiovascular risk factors, including BMI, HDL, LDL, TG, apolipoprotein A1, apolipoprotein B, SBP, and DBP, to estimate the direct causal effects of genetically predicted T2DM on CAD, MI, and stroke. After adjusting for these risk factors, we found that T2DM remained significantly associated with an increased risk of CAD (OR 1.063, 95% CI 1.031–1.097, P = 1.11e‐04), MI (OR 1.088, 95% CI 1.042–1.135, P = 1.12e‐04), and stroke (OR 1.066, 95% CI 1.032–1.101, P = 1.18e‐04). Detailed results can be found in Figure 4 and Table S6 .

Figure 4.

Multivariable MR analysis of the causal effects of T2DM on caridac outcomes after adjusting for cardiovascular risk factors. (A) The causal effect of T2DM on Stroke after adjusting for risk factors. (B) The causal effect of T2DM on Stroke in Finngen cohort after adjusting for risk factors. (C) The causal effect of T2DM on Coronary artery disease after adjusting for risk factors. (D) The causal effect of T2DM on Coronary artery disease in Finngen cohort after adjusting for risk factors. (E) The causal effect of T2DM on myocardial infarction after adjusting for risk factors. (F) The causal effect of T2DM on myocardial infarction in Finngen cohort after adjusting for risk factors. Apo A1, apolipoprotein A1; Apo B, apolipoprotein B; BMI, body mass index; DBP, diastolic blood pressure; HDL, high‐density lipoprotein cholesterol; LDL, low‐density lipoprotein cholesterol; LV mass, left ventricular mass; SBP, systolic blood pressure; T2DM, type 2 diabetes mellitus; TG, triglycerides.

Additionally, we conducted multivariable MR analyses to assess the effects of T2DM on cardiac structure while adjusting for the aforementioned cardiovascular risk factors. After these adjustments, the negative associations with right atrial maximum area indexed (β = −0.039, 95% CI −0.069 to −0.008, P = 0.012) and right atrial minimum area indexed (β = −0.041, 95% CI −0.072 to −0.010, P = 0.009) remained significant. Conversely, the positive association between T2DM and left ventricular mass indexed was attenuated after adjusting for apolipoprotein A1 and SBP. Specifically, the association decreased from β = 0.473 (P = 0.001) to β = 0.219 (P = 0.111) when adjusted for apolipoprotein A1, and to β = 0.270 (P = 0.089) when adjusted for SBP. We further explored the mediating roles of apolipoprotein A1 and SBP in the relationship between T2DM and left ventricular mass indexed using a two‐step MR approach. T2DM was causally associated with both apolipoprotein A1 and SBP, which in turn were causally associated with left ventricular mass index. This finding is consistent with the total effect, indicating that apolipoprotein A1 and SBP are important mediators in the causal pathway between T2DM and left ventricular mass indexed. The detailed results of these analyses are presented in Figure 5 , Figure S3 and Table S7‐S9.

Figure 5.

Multivariable MR analysis of the causal effects of T2DM on cardiac structure after adjusting for cardiovascular risk factors. (A) The causal effect of T2DM on RA Max Indexed after adjusting for risk factors. (B) The causal effect of T2DM on RA Min Indexed in Finngen cohort after adjusting for risk factors. (C) The causal effect of T2DM on LV mass indexed after adjusting for risk factors. Apo A1, apolipoprotein A1; Apo B, apolipoprotein B; BMI, body mass index; DBP, diastolic blood pressure; HDL, high‐density lipoprotein cholesterol; LDL, low‐density lipoprotein cholesterol; LV mass, left ventricular mass; SBP, systolic blood pressure; T2DM, type 2 diabetes mellitus; TG, triglycerides.

Discussion

In our MR study, we demonstrate a causal link between genetic predisposition to T2DM and an increased risk of CAD, MI, and stroke. These relationships remained robust after adjusting for cardiovascular risk factors, including blood pressure, lipid levels, and obesity. Furthermore, our research findings suggest that a greater genetic predisposition to T2DM correlates with smaller indexed areas for both the minimum and maximum sizes of the right atrium, alongside an elevation in the left ventricular mass index. However, our findings do not demonstrate a causal relationship between T2DM and other cardiac outcomes or cardiac structure and function. The external validation and sensitivity analyses of our results confirm their consistency across various populations.

Type 2 diabetes mellitus and cardiovascular outcomes

Our study provides evidence that genetic susceptibility to T2DM is causally associated with an increased risk of CAD, MI, and stroke. These findings aligns with previous observational studies showing T2DM as an independent risk factor for atherosclerotic CVD and stroke. ³ , ⁴ , ⁵ , ²⁸

Several pathophysiological mechanisms are contribute to the cardiovascular complications observed in T2DM. Hyperglycemia plays a pivotal role in the development of microvascular and macrovascular complications in diabetes. ²⁹ Chronic hyperglycaemia induces the overproduction of superoxide by the mitochondrial electron transport chain, resulting in oxidative stress. ³⁰ This generates reactive oxygen species that activate several pathways implicated in the pathogenesis of diabetic vascular disease, including augmented advanced glycation end product formation, protein kinase C isoform activation, and increased flux through the hexosamine and polyol pathways. ³¹ In addition, hyperglycaemia causes abnormal endothelial function by reducing the bioavailability of nitric oxide. Nitric oxide helps maintain vascular homeostasis through its anti‐inflammatory, anti‐atherosclerotic and antithrombotic properties. ³²

Endothelial dysfunction, a condition exacerbated by factors like hyperglycaemias, insulin resistance, inflammation, and oxidative stress, is considered an early marker of atherosclerotic disease and a predictor of long‐term cardiovascular risk. ³³ , ³⁴ This dysfunction is characterized by a diminished availability of vasodilators like nitric oxide, along with an upsurge in vasoconstrictor secretion. These alterations lead to both microvascular and macrovascular damages. They manifest as changes in vascular permeability and stiffness, as well as altered vascular reactivity. Such changes further exacerbate the situation by promoting leukocyte adhesion, smooth muscle cell proliferation, platelet aggregation, and thrombosis. ³³ , ³⁴ , ³⁵

Hyperglycemia also increases the production of advanced glycation end products that trigger sustained low‐grade inflammation via nuclear factor kappa B (NF‐kB) and alter extracellular matrix proteins through non‐enzymatic glycosylation. ³⁴ , ³⁵ Together, these perturbations result in accelerated atherosclerosis, plaque instability, and heightened risk of occlusive vascular events like MI and ischaemic stroke in T2DM.

Additionally, several studies have reported a higher prevalence of traditional cardiovascular risk factors such as hypertension and obesity among patients with T2DM. ³⁶ , ³⁷ However, the excessive cardiovascular risk associated with T2DM appears to extend beyond that attributable to conventional risk factors alone. Our multivariable MR analysis adjusted for lipid traits, blood pressure, and obesity. The persistence of causal links between T2DM and cardiovascular outcomes despite such adjustments indicates that the observed causal relationships between genetic predisposition to T2DM and CAD, MI, and stroke cannot be fully explained by mediating effects of BMI, lipids or blood pressure. This implies that T2DM may directly induce adverse vascular changes through mechanisms like hyperglycaemias‐induced oxidative stress and advanced glycation end‐product formation.

In contrast, our MR study did not demonstrate causal associations of T2DM with atrial fibrillation or HF. These results appear inconsistent with previous observational studies demonstrating positive associations between T2DM and these outcomes. ³⁸ We found that T2DM was causally associated with smaller right atrial size but not left atrial size. Left atrial enlargement, rather than right atrial changes, is more strongly linked to AF development. Left atrial dilation reflects elevated left ventricular filling pressures and is a key substrate for AF. Our findings indicate that T2DM does not directly cause left atrial structural remodelling, potentially explaining the lack of a causal association with AF risk. Additionally, previous studies proposed that T2DM may impair the heart through structural changes including hypertrophy and diffuse fibrosis. However, these structural alterations may not necessarily further progress to functional impairment and overt HF, especially in the absence of intermediary conditions like hypertension or CAD. ³⁹ Residual confounding likely influenced earlier observational results, considering that T2DM often co‐occurs with other cardiovascular risk factors such as hypertension and obesity. Additionally, reverse causation remains an inherent limitation in observational research.

Type 2 diabetes mellitus and cardiac structure/function

Our study investigated the influence of T2DM on quantitative cardiac traits using CMR imaging, revealing evidence of T2DM's causal role in altering cardiac structure and function. Key findings indicate that a greater genetic predisposition to T2DM correlates with changes in both the maximum and minimum indexed areas of the right atrium. Furthermore, we found that genetic risk factors for T2DM are associated with an increase in left ventricular mass index, suggesting left ventricular hypertrophy (LVH). However, our study did not establish definitive causal links between T2DM and several other cardiac parameters we assessed.

Our observation that T2DM lowers right atrial size conforms with Linssen et al.'s research indicating atrial remodelling in diabetes. ⁴⁰ Their study demonstrated that right atrial (RA) volume index was significantly lower in both prediabetes and T2DM compared with individuals with normal glucose metabolism (P < 0.01), and the differences remained significant after adjusting for potential confounding factors. Specifically, the study observed a decreasing trend in RA volume index across the three groups: normal glucose metabolism, prediabetes, and T2DM (median values of 25.2, 22.1, and 21.7 mL/m², respectively; P < 0.01). In the multivariable adjusted model, prediabetes and T2DM were associated with a 0.26 and 0.29 standard deviation decrease in RA volume index, respectively, compared with normal glucose metabolism (P < 0.01). The mechanisms underlying atrial remodelling in T2DM are likely multifactorial. Autonomic dysfunction is highly prevalent in diabetic patients and can impact atrial size through alternations in sympathetic and parasympathetic neural activity. Additionally, atrial interstitial fibrosis secondary to glycation end‐product deposition and oxidation may cause atrial stiffening and contractile dysfunction. Atrial fibrosis can lead to chamber stiffness, impaired contractility, and reduced compliance, ultimately resulting in a decrease in atrial dimensions. A reduction in right atrial size may lead to impaired right atrial reservoir function, reduced cardiac output, and altered venous return. Furthermore, microvascular ischaemia resulting from the metabolic derangements of diabetes could also contribute. ³⁵ Our study reveals that T2DM may adversely affect the right atrium even before overt CVD becomes apparent. Assessing RA size may help identify early cardiac structural changes in patients with diabetes. Further prospective studies are needed to investigate the underlying mechanisms and clinical implications of right atrial remodelling in T2DM.

The elevation in left ventricular mass index among those genetically predisposed to T2DM aligns with previous research indicating LVH is common in diabetes. ⁴¹ The development of LVH is associated with various haemodynamic and non‐haemodynamic factors. Hypertension and increased haemodynamic load are the main triggers for LVH. However, in T2DM, hyperinsulinaemia and hyperglycaemia caused by insulin resistance may directly stimulate the hypertrophic growth of cardiomyocytes. Studies have found that insulin and insulin‐like growth factor‐1 can activate the serine/threonine kinase Akt and mTOR pathways in cardiomyocytes, promoting protein synthesis and cell growth. ⁴² Chronic hyperglycaemia‐induced glycation of cardiac proteins and lipids can stimulate cardiac fibroblast proliferation and enhance collagen deposition. ¹ , ³⁴ Oxidative stress and low‐grade inflammation induced by T2DM can also trigger cardiac remodelling.

Interestingly, our results indicate that both apolipoprotein A1 and SBP are important mediators in the causal relationship between T2DM and left ventricular mass index. This implies that by controlling these mediating factors, it may be possible to attenuate the increase in LV mass caused by T2DM. Strategies to control blood pressure and increase apolipoprotein A1 levels may help to mitigate cardiac remodelling and LVH.

The causal relationship between T2DM and increased left ventricular mass may partially explain the elevated risk of coronary heart disease (CHD), MI, and stroke observed in our study. Bluemke et al.’s study found that, after adjusting for cardiovascular risk factors, the endpoints of incident CHD and stroke were positively associated with increased left ventricular mass‐to‐volume ratio (CHD, hazard ratio [HR] 2.1 per g/mL, P = 0.02; stroke, HR 4.2 per g/mL, P = 0.005). ⁴³ These findings are consistent with our MR analysis results and underscore the prognostic importance of diabetes‐related LVH. The LV structural changes of increased mass may impair myocardial perfusion, reduce coronary flow reserve, and promote myocardial fibrosis and scarring, thereby increasing the vulnerability to ischaemia and infarction. The especially strong link between LV hypertrophy and stroke risk may reflect the impact of LVH on systemic vascular function and cerebral blood flow regulation.

Therefore, for patients with T2DM, monitoring and controlling the progression of LVH may help improve cardiac structure and function and reduce the risk of cardiovascular events. It should be noted that LVH is just one manifestation of diabetic cardiomyopathy, and its relationship with the risk of CHD, myocardial infarction and stroke is also influenced by various factors such as cardiac function and metabolic disorders. Future prospective studies are needed to validate the predictive value and clinical significance of LVH in diabetes‐related cardiovascular complications.

We did not uncover clear evidence that a genetic predisposition to T2DM influences other measures of cardiac structure and function. Prior studies, including those utilizing CMR, have reported that T2DM associates with subtle changes in left and right ventricular morphology and performance, even among those free of clinically‐apparent CVD. ⁸ , ⁴¹ Additionally, imaging analyses indicate T2DM alters myocardial tissue characteristics, including interstitial fibrosis, infiltration, edema, and infarction. ⁴⁴ Our failure to demonstrate causal effects of T2DM on several of these cardiac parameters may relate to the myocardial changes reported in earlier observational analyses may not be a direct result of T2DM but rather related to comorbid factors including hypertension, vascular disease, dyslipidaemia, or lifestyle factors. Additional adequately powered studies are warranted to clarify the causal relevance of T2DM to other subtle changes in cardiac structure and function beyond right atrial size and LV mass.

Limitation

Our study presents several limitations. Firstly, the participant was solely composed of individuals of European descent, which constrains the generalizability of our findings to other ethnic groups. This highlights the necessity for additional research to verify our results across varied ethnic backgrounds. Additionally, despite choosing genetic variants based on strict criteria, we cannot entirely rule out the influence of pleiotropy and residual confounding factors. Another limitation of our study is that we lacked access to suitable summary data on blood glucose‐lowering medications and detailed medication intake information for participants in the UK Biobank and FinnGen cohorts. Although hyperglycaemia is a key mechanism in the development of diabetic cardiovascular complications, we were unable to directly assess the impact of pharmacologic glycaemic control on the observed causal relationships.

Conclusions

In conclusion, our comprehensive MR study demonstrates causal effects of genetic liability to T2DM on increasing the risk of CAD, MI, and stroke in a European ancestry population. Furthermore, our findings reveal that genetic liability to T2DM is causally associated with LV hypertrophy and reduced right atrial dimensions, providing evidence for the significant role of diabetes in adverse cardiac remodelling. These findings provide genetic support for targeting glycaemic control in T2DM to prevent or alleviate cardiovascular complications and adverse cardiac remodelling.

Conflict of Interest

The authors declare that the study was conducted without any commercial or financial relationships that could be construed as a potential conflict of interest.

Funding

None.

Supporting information

Figure S1. Scatter plots illustrating the causal effects of T2DM on cardiac outcomes, structure and function.

Figure S2. Causal effects of Type 2 diabetes mellitus on cardiac structure and function.

T2DM, Type 2 diabetes mellitus.

Figure S3. A. Causal effects of T2DM on cardiometabolic traits. B. Causal effects of cardiometabolic traits on LV Mass Indexed.

Table S1. Instrumental variables of T2DM used for MR analysis.

Table S2. Effect estimates of the T2DM on cardiovascular diseases.

Table S3. Test results for pleiotropy and heterogeneity ‐ causal effects of the T2DM on cardiovascular diseases, structure and function.

Table S4. Effect estimates of the cardiovascular diseases on T2DM using reverse MR.

Table S5. Effect estimates of the T2DM on cardiovascular magnetic resonance imaging parameters of cardiac structure and function.

Table S6. Multivariable MR analysis of the causal effects of T2DM on caridac outcomes after adjusting for cardiovascular risk factors.

Table S7. Multivariable MR analysis of the causal effects of T2DM on caridac structure after adjusting for cardiovascular risk factors.

Table S8. Effect estimates of the T2DM on cardiovascular risk factors.

Table S9. Effect estimates of the cardiovascular risk factors on T2DM.

Ruan, W. , Zhou, X. , Li, J. , Liu, H. , Wang, T. , Zhang, G. , and Lin, K. (2024) Type 2 diabetes mellitus and cardiovascular health: Evidence of causal relationships in a European ancestry population. ESC Heart Failure, 11: 3105–3119. 10.1002/ehf2.14877.

Data availability statement

Publicly available genome‐wide association study (GWAS) summary data were utilized for the analyses. These are available to download at cited sources.