Abstract
The correlation between type 2 diabetes mellitus (T2DM) and heart failure has been extensively studied; however, the causal nature of this relationship remains a topic of debate. Our study employs a two-sample Mendelian randomization (MR) approach to rigorously examine the causal relationship between T2DM and heart failure. By utilizing summary statistics from genome-wide association studies (GWAS) involving 655,666 individuals for T2DM (including 61,714 cases and 1178 controls) and 977,323 individuals for heart failure (including 47,309 cases and 930,014 controls), we aimed to provide strong evidence for a causal association. The analysis revealed a significant causal relationship, with individuals with T2DM exhibiting a 7% increased odds of heart failure (inverse variance weighting [IVW]: OR = 1.07, 95% CI = 1.04–1.10, P = 1.07 × 10−6). This finding highlights the significance of T2DM as an independent risk factor for heart failure and suggests that MR analysis could be a valuable tool for the early identification of heart failure risk in diabetic populations. Our results emphasize the need for enhanced screening and preventive measures for heart failure in patients with T2DM, potentially leading to improved patient outcomes through early diagnosis and intervention.
Keywords: adverse cardiac events, causality, heart failure, Mendelian randomization, type 2 diabetes mellitus
1. Introduction
The number of individuals with diabetes worldwide has reached 537 million, with an unchecked growth trend. An estimated 12.1% of global deaths in 2021 were associated with diabetes; diabetes and its complications resulted in US $966 billion in global economic losses.[1] In China, the prevalence of diabetes has increased rapidly. According to the latest epidemiological survey data, the prevalence of diabetes in China is 12.4%, but the blood glucose control rate remains below 50%.[2] Approximately 90% of all diabetes cases are type 2 diabetes mellitus (T2DM). The number of affected individuals is increasing rapidly, with concerning trends observed in children and young people. Heart failure is a prevalent cardiovascular disease and represents the final stage of various heart conditions. It is characterized by structural or functional abnormalities of the heart, leading to impaired ventricular filling or contraction, which results in a complex clinical syndrome with symptoms such as dyspnea, edema, chest tightness, shortness of breath, and fatigue.[3] The global prevalence of heart failure in adults is currently estimated at 1% to 2%. In China, the estimated number of individuals with heart failure is approximately 12.05 million, with around 2.97 million new cases annually. The prevalence of heart failure is approximately 1.38% in individuals aged 35 and older, 3.09% in those aged 60 to 79, and 7.55% in individuals aged 80 and older.[4] A growing body of research suggests that diabetes mellitus may be an independent risk factor for heart failure, with the incidence of heart failure in patients with diabetes mellitus significantly higher than in nondiabetic patients.
Although a growing number of studies suggest a correlation between T2DM and heart failure, traditional observational studies are prone to confounding variables and reverse causality, limiting their ability to clarify the causal relationship between the 2. Randomized controlled trials (RCTs) represent the most rigorous study design for inferring causality. However, the implementation of RCTs is often constrained by various factors, including ethical concerns and the need for substantial human and material resources. Mendelian randomization (MR) studies are a method for assessing the causal relationship between various exposures and disease outcomes, using genetic variants as instrumental variables (IVs).[5] This analysis draws causal inferences by using genetic variants as modifiable risk factors or proxies for health outcomes.[6] Since genes are randomly assigned at conception, genetic variation is largely independent of confounding factors.[7] MR studies must satisfy 3 basic assumptions[8]: the association hypothesis: genetic variants are associated with risk factors; the independence assumption: genetic variation is not associated with any known or unknown confounding factors; the exclusion restriction hypothesis: genetic variation affects outcomes only through risk factors.
Two-sample MR involves 2 distinct study populations. For example, T2DM data are measured in 1 sample, and heart failure data are measured in the other. This design offers 2 advantages: first, MR studies do not require the collection of risk factors and outcomes; second, the sample sizes can be very large (typically >50,000), enabling the use of pooled results from genome-wide association studies (GWAS), thus enhancing the precision and statistical efficiency of the results. Additionally, it can address the challenges of difficult or costly sample collection.[9]
In this study, we used an MR approach to explore the potential causal relationship between genetic variants associated with T2DM and heart failure.
2. Methods
2.1. Study design
In this study, a two-sample MR approach was employed, with T2DM as the exposure, single nucleotide polymorphisms (SNPs) as the IVs, and heart failure as the outcome, to verify the causal relationship between genetically predicted T2DM and heart failure. MR analysis must satisfy 3 core assumptions[10]: the association assumption requires that the SNPs used as IVs must be closely related to the exposure factor of interest. The independence assumption assumes that the selected genetic variant is independent of other confounders that may affect exposure and outcome. The exclusivity assumption stipulates that the effect of SNPs on the outcome can only be mediated through the exposure and not via any other pathway (Fig. 1).
Figure 1.
Schematic of MR core assumptions. MR = Mendelian randomization.
2.2. Data sources
Since there are multiple datasets for T2DM and heart failure in the IEU Open GWAS, we selected the dataset with the largest sample size for analysis to ensure the reliability of the results. Data for T2DM were obtained from pooled GWAS data of 655,666 European individuals, including 61,714 patients and 1178 controls, with 5030,727 SNPs. Summary data were retrieved from https://gwas.mrcieu.ac.uk/datasets/ebi-a-GCST006867/, which is publicly accessible. Data on heart failure were obtained from pooled GWAS data of 977,323 European individuals, including 47,309 heart failure patients and 930,014 controls. The pooled GWAS results included 7773,021 SNPs, available at https://gwas.mrcieu.ac.uk/datasets/ebi-a-GCST009541/. All participants were of European ancestry.
No additional ethical approval was required for this MR study, as the data used were from a publicly available database. The sample information for the exposure factor (T2DM) and the outcome (heart failure) used in this study is detailed in Table 1.
Table 1.
Details of samples included in MR study.
| Variables of interest | Sample Size | Case group | Control group | Source of samples | Year of publication | Website |
|---|---|---|---|---|---|---|
| Diabetes mellitus type 2 | 655,666 | 61,714 | 1178 | Europe | 2018 | https://gwas.mrcieu.ac.uk/datasets/ebi-a-GCST006867/ |
| Heart failure | 977,323 | 47,309 | 930,014 | Europe | 2020 | https://gwas.mrcieu.ac.uk/datasets/ebi-a-GCST009541/ |
2.3. Selection of IVs
SNPs were required to meet the following criteria: the significance threshold for SNPs at the genome-wide level had to be <1.0 × 10−6, indicating a significant association with T2DM; the explained variation (R2) of SNPs had to be at least 0.01, or the genetic distance had to be no >5000 kb. SNPs with R2 values >0.01 within 5000 kb of the most significant SNPs were removed, excluding those in linkage disequilibrium. SNPs with palindromic structures that could affect the accuracy of the analysis were also excluded. SNPs associated with potential confounders (e.g., hypertension, obesity, body mass index, and educational attainment) were excluded. SNPs with an F-statistic of <10 were excluded to filter out potentially weak IVs. Weak IVs are genetic variants that, despite their association with the exposure under investigation, have limited explanatory power for the exposure. Due to their small effect on the exposure factor, these variants provide minimal statistical power for testing the research hypothesis, potentially leading to impaired precision in causal effect estimates. Weak IVs may lead to problems such as an increased risk of false positive errors (i.e., type I errors).[9] The strength of IVs can be assessed by the F-statistic, which for a single SNP can be calculated using the formula: F = beta2/se2,[11] which is less affected by weak instrumental bias when F > 10. Based on GWAS results for T2DM and heart failure, SNPs with identical alleles were harmonized to align the effect estimates for exposure and outcome.[12]
2.4. Statistical analysis
In this study, the inverse variance weighted (IVW) method was employed as the primary analytical approach to assess the potential causal relationship between T2DM and the risk of heart failure. This method is most appropriate when the genetic instruments are strongly associated with the exposure and there is minimal heterogeneity, allowing for the generation of unbiased estimates. The intercept test from the MR-Egger regression was conducted to detect the presence of horizontal pleiotropy among the IVs. A P value > .05 was considered indicative of the absence of horizontal pleiotropy.[13] Additionally, a leave-one-out analysis was conducted to evaluate the influence of individual SNPs on the overall causal effect estimate. Cochran Q test was employed to assess heterogeneity among the IVs. A P value < .05 indicates significant heterogeneity, in which case the IVW method with a random-effects model is recommended.[14] For sensitivity analysis, a leave-one-out approach was applied, in which each SNP was sequentially removed and the IVW method was re-applied to estimate the causal effect based on the remaining variants. This method aims to evaluate the robustness of the causal estimates by assessing the potential influence of individual SNPs on the overall MR results. However, in the presence of heterogeneity or pleiotropy, the IVW method may yield biased estimates. Therefore, to complement the IVW method, additional sensitivity analyses were conducted using the MR-Egger method, weighted median estimator (WME), simple mode method (SM), and weighted mode method (WM).
All analyses were conducted in the R statistical environment (version 4.3.3), with statistical analyses and data visualization performed using the “TwoSampleMR” package (version 0.5.6). Given its superior efficiency in detecting causal effects, the IVW method was selected as the primary approach for causal effect estimation in this study. To account for the relaxed threshold and minimize the risk of type I error, additional measures were taken to assess the potential impact of multiple testing on the study findings.
3. Results
3.1. Determination of IVs
SNPs associated with T2DM at genome-wide significance (P < .05) were selected using R software for further analysis. After excluding SNPs in linkage disequilibrium, a total of 228 independent SNPs were retained as IVs. The F-statistics of the individual SNPs ranged from 29.94 to 1578.20, indicating a strong association between the genetic variants and the exposure, and suggesting that the MR estimates are unlikely to be biased by weak IVs. (see Table S1, Supplemental Digital Content, https://links.lww.com/MD/P63).
3.2. Causal relationship between T2DM and heart failure
MR analysis showed that IVW: OR = 1.07, 95% CI = 1.04–1.10, P = 1.07 × 10−6; MR Egger: OR = 1.00, 95% CI = 0.94–1.07, P = .95; WME: OR = 1.05, 95% CI = 0.97–1.31, P = .01; Simple mode: OR = 0.98, 95% CI = 0.81–1.15, P = .78; Weighted mode: OR = 1.04, 95% CI = 0.99–1.09, P = .15. The direction of causality obtained by the 5 methods was consistent, showing that T2DM patients would increase the risk of heart failure; IVW showed statistical significance (P < .05, see Table 2 and Fig. 2).
Table 2.
Results of MR analysis of the relationship between EC and BC.
| Factors of exposure | Methods | OR | 95% CI | P |
|---|---|---|---|---|
| T2DM | MR Egger | 1.00 | 0.94–1.07 | 0.95 |
| WME | 1.05 | 0.97–1.31 | 0.01 | |
| IVW | 1.07 | 1.04–1.10 | 1.07 × 10−6 | |
| Simple mode | 0.98 | 0.81–1.15 | 0.78 | |
| Weighted mode | 1.04 | 0.99–1.09 | 0.15 |
CI = confidence interval, IVW = inverse variance weighting, MR = Mendelian randomization, OR = odds ratio, SM = simple mode, WM = weighted mode, WME = weighted median.
Figure 2.
Scatter plot of MR analysis results for both samples. MR = Mendelian randomization.
3.3. Results of heterogeneity test and pleiotropy test
Cochran Q test indicated that a P value below .01 signifies significant heterogeneity among the IVs. IVW yielded a heterogeneity P value of 1.10 × 10−15, while MR-Egger regression showed a P value of .01, indicating significant heterogeneity among the 227 SNPs (Table 3). The MR-Egger regression showed no evidence of horizontal pleiotropy, as indicated by an intercept P value > .05 (Table 3).
Table 3.
Detail table of SNP tool variables in two-sample Mendelian randomization analysis.
| Exposure | Test of heterogeneity | Testing for pleiotropy | ||
|---|---|---|---|---|
| MR Egger Q P value |
Inverse variance weighting method Q P value |
MR Egger intercept value | MR Egger Intercept P value | |
| T2DM | 9.55 × 10−15 | 1.10 × 10−15 | 0.01 | 0.23 |
MR = Mendelian randomization, Q = Cochran Q value. The intercept is an intercept.
3.4. Sensitivity analysis
Leave-one-out sensitivity analysis revealed that sequential exclusion of individual SNPs yielded combined OR values closely aligned with those from the IVW method (P > .05), and no single SNP exerted a disproportionate influence on the overall results (Fig. 3). These findings suggest that the effect estimates derived from the IVW method are robust, and that the MR results are not driven by any individual SNP.
Figure 3.
T2DM with Heart failure by leave-one-out method. T2DM = type 2 diabetes mellitus.
4. Discussion
Leveraging large-scale pooled GWAS summary statistics, we conducted two-sample MR analyses involving 61,714 patients with T2DM and 1178 controls, and 47,309 patients with heart failure and 930,014 controls, to investigate the potential causal relationship between T2DM and heart failure risk. The causal association was further supported by IVW, MR-Egger regression, the WME, SM, and WM methods. Leave-one-out sensitivity analysis indicated that no single SNP exerted an undue influence on the causal estimates, thereby enhancing the robustness of the findings. Overall, our findings provide evidence of a positive causal association between genetically predicted T2DM and the risk of heart failure.
This two-sample MR study is of significant clinical importance. Firstly, as the first MR study using T2DM as an exposure to estimate its causal effect on heart failure risk, it demonstrates a major strength in its two-sample MR design, effectively mitigating the influence of potential confounding factors. It addresses the issue that traditional observational studies are unable to establish the relationship between T2DM and heart Failure. Secondly, the sample of this study comes from the GWAS database,[15] which makes the study an effective causal inference with high statistical power. Thirdly, the results were robust through rigorous quality control conditions and a series of sensitivity analyses. After removing SNPs associated with potential confounders, MR studies suggest a causal relationship between genetically predicted T2DM and the risk of heart failure, with an independent association between the 2.
Nevertheless, this study has some limitations. On the on hand, this study can only preliminarily conclude a causal relationship between T2DM and heart failure. We were unable to perform stratified analyses by important factors such as sex and age, and we could not distinguish between HFrEF and HFpEF to determine whether T2DM has differential effects. On the other hand, the population in this study primarily consisted of individuals of European descent, which may limit the generalizability of our findings to other populations.
The Framingham Heart Study in the United States shows that diabetes can increase the risk of heart failure by 2.4 times in men and 5 times in women, and this increase is not related to hypertension or coronary artery disease.[16] A retrospective cohort study from Estonia found that newly diagnosed T2DM patients exhibited a higher mortality rate, while pre-T2DM was not significantly different from non-T2DM patients in all age group.[17] The risk of heart failure in patients with diabetes is significantly elevated, primarily due to the increased likelihood of atherosclerosis, which is often accompanied by diffuse, severe arterial stenosis. This leads to myocardial ischemia, contributing to the development of ischemic cardiomyopathy and, ultimately, severe heart failure. Furthermore, even in cases of mild coronary artery disease in diabetic patients, chronic hyperglycemia can result in autonomic neuropathy, microcirculatory dysfunction, and disturbances in metabolism and energy homeostasis.
A 10-year longitudinal cohort study showed that a high hemoglobin glycation index was independently associated with cardiovascular disease events in patients with T2DM during 11.1 years of follow-up. Patients with a high hemoglobin glycation index at baseline have an inherent higher risk of developing cardiovascular disease.[18] Cheng et al[19] found that a high hemoglobin glycation index was directly related to the reduction of cardiovascular mortality but not related to major adverse cardiac events. In addition, Chen Wei team conducted a study based on the MIMIC-IV database that showed the hemoglobin glycation index was significantly associated with all-cause mortality in patients with critical coronary artery disease, especially in patients with low hemoglobin glycation index. The hemoglobin glycation index can be used as a potential indicator to evaluate the short-term and long-term mortality risk of such patients.[20] The hemoglobin glycation index is an independent predictor of major adverse cardiovascular events in patients with T2DM and coronary heart disease.[21] To deal with diabetes complicated with heart failure, we must first do a good job of prevention at ordinary times, actively improve living habits, reduce blood sugar and lipids, and control weight. Second, attention should also be paid to “hidden” heart failure. For example, when a diabetic patient has poor physical strength, shortness of breath, lying down at night and often awakened, and cold hands and feet, systemic edema, sudden apathy, or personality change, it is necessary to go to the hospital as soon as possible to check and diagnose. For diabetic patients with coronary stenosis, even if there are no symptoms of angina pectoris, it is recommended to deal with stenosis through coronary angiography as soon as possible. If the condition of a patient with diabetes and heart failure is stable, coronary angiography should also be performed to identify the presence or absence of ischemic cardiomyopathy.
Although heart failure is the end stage of heart disease, it does not mean that it will be life threatening in a short time. If we can achieve early diagnosis and treatment, choose appropriate hypoglycemic drugs and anti-heart failure drugs, and pay attention to improving living habits, monitoring, and inspection, it can also delay the disease and improve the quality of life. Therefore, patients with diabetes mellitus complicated with heart failure should try to choose drugs with cardioprotective effects during hypoglycemic therapy and actively cooperate with doctors to adjust the medication plan to avoid the adverse effects of drugs. Routine echocardiographic evaluation and biomarkers of cardiac stress (e.g., NT-proBNP) could be considered for T2DM patients, particularly those with additional risk factors. Moreover, MR-derived insights may support the development of personalized risk stratification tools based on genetic profiles, thereby facilitating earlier identification and intervention in individuals predisposed to heart failure. Such strategies may ultimately contribute to reducing heart failure incidence and improving long-term outcomes in the diabetic population.
In conclusion, this study explored the positive causal relationship between T2DM exposure and the outcome of heart failure, but the mechanism and details need to be further studied so that clinical workers can strengthen the screening of heart failure for T2DM patients, aiming to achieve early detection, early diagnosis, and early treatment.
Author contributions
Conceptualization: Xiaoyu Yang.
Data curation: Xiaoyu Yang, Binlu Wang.
Formal analysis: Xiaoyu Yang.
Methodology: Xiaoyu Yang, Binlu Wang.
Project administration: Xiaoyu Yang, Qin Wan.
Resources: Xiaoyu Yang.
Software: Xiaoyu Yang.
Supervision: Qin Wan.
Validation: Xiaoyu Yang.
Visualization: Xiaoyu Yang.
Writing – original draft: Xiaoyu Yang.
Writing – review & editing: Xiaoyu Yang.
Supplementary Material
Abbreviations:
- CI
- confidence interval
- GWAS
- genome-wide association studies
- IV
- instrumental variables
- IVW
- inverse variance weighting
- MIMIC-IV
- Medical Information Mart for Intensive Care IV
- MR
- Mendelian randomization
- MR-Egger
- Mendelian randomization Egger regression
- OR
- odds ratio
- SM
- simple mode method
- SNPs
- single nucleotide polymorphisms
- T2DM
- type 2 diabetes mellitus
- WM
- weighted mode method
- WME
- weighted median method
The authors have no conflicts of interest to disclose.
The datasets generated during and/or analyzed during the current study are publicly available.
Supplemental Digital Content is available for this article.
How to cite this article: Yang X, Wang B, Wan Q. Genetically predicted causal association between type 2 diabetes and heart failure: A two-sample Mendelian randomization study. Medicine 2026;105:17(e42638).
References
- [1].Sun H, Saeedi P, Karuranga S, et al. IDF Diabetes Atlas: Global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res Clin Pract. 2022;183:109119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [2].Wang L, Gao P, Zhang M, et al. Prevalence and ethnic pattern of diabetes and prediabetes in China in 2013. JAMA. 2017;317:2515–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Wattanachayakul P, Kittipibul V, Fudim M. Radiofrequency-based wearable sensor patch in heart failure management: a focused review [published online ahead of print September 2025]. Heart Fail Rev. doi: 10.1007/s10741-025-10511-4. [DOI] [PubMed] [Google Scholar]
- [4].Wang H, Chai K, Du M, et al. Prevalence and incidence of heart failure among urban patients in China: a national population-based analysis. Circ Heart Fail. 2021;14:e008406. [DOI] [PubMed] [Google Scholar]
- [5].Weng H, Li H, Zhang Z, et al. Association between uric acid and risk of venous thromboembolism in East Asian populations: a cohort and Mendelian randomization study. Lancet Reg Health West Pac. 2023;39:100848. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Davey Smith G, Hemani G. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet. 2014;23:R89–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Emdin CA, Khera AV, Kathiresan S. Mendelian randomization. JAMA. 2017;318:1925–6. [DOI] [PubMed] [Google Scholar]
- [8].Davies NM, Holmes MV, Davey Smith G. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. BMJ. 2018;362:k601. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Pierce BL, Burgess S. Efficient design for Mendelian randomization studies: subsample and 2-sample instrumental variable estimators. Am J Epidemiol. 2013;178:1177–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Davey Smith G, Holmes MV, Davies NM, Ebrahim S. Mendel’s laws, Mendelian randomization and causal inference in observational data: substantive and nomenclatural issues. Eur J Epidemiol. 2020;35:99–111. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Chen L, Yang H, Li H, He C, Yang L, Lv G. Insights into modifiable risk factors of cholelithiasis: a Mendelian randomization study. Hepatology. 2022;75:785–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Hartwig FP, Davies NM, Hemani G, Davey Smith G. Two-sample Mendelian randomization: avoiding the downsides of a powerful, widely applicable but potentially fallible technique. Int J Epidemiol. 2016;45:1717–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Chen X, Kong J, Diao X, et al. Depression and prostate cancer risk: a Mendelian randomization study. Cancer Med. 2020;9:9160–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Wang X, Wang X, Wang H, Yang M, Dong W, Shao D. Association between psoriasis and lung cancer: two-sample Mendelian randomization analyses. BMC Pulm Med. 2023;23:4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Ishigaki K, Akiyama M, Kanai M, et al. Large-scale genome-wide association study in a Japanese population identifies novel susceptibility loci across different diseases. Nat Genet. 2020;52:669–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Yang Y, Yang B, Zhao S, et al. A lower eGFR(cystatin C)/eGFR(creatinine) ratio is associated with greater cardiovascular risk (higher Framingham Risk Score) in Chinese patients with newly diagnosed type 2 diabetes mellitus. Ren Fail. 2024;46:2346267. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Asser P, Fischer K, Ainla T, et al. Examining the impact of renal dysfunction and diabetes on post-myocardial infarction mortality: insights from a comprehensive retrospective cohort study across different age groups. Scand Cardiovasc J. 2024;58:2395875. [DOI] [PubMed] [Google Scholar]
- [18].Kim MK, Jeong JS, Yun JS, et al. Hemoglobin glycation index predicts cardiovascular disease in people with type 2 diabetes mellitus: a 10-year longitudinal cohort study. J Diabetes Complications. 2018;32:906–10. [DOI] [PubMed] [Google Scholar]
- [19].Cheng W, Huang R, Pu Y, et al. Association between the haemoglobin glycation index (HGI) and clinical outcomes in patients with acute decompensated heart failure. Ann Med. 2024;56:2330615. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Wei X, Chen X, Zhang Z, et al. Risk analysis of the association between different hemoglobin glycation index and poor prognosis in critical patients with coronary heart disease-a study based on the MIMIC-IV database. Cardiovasc Diabetol. 2024;23:113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Xu S, Qin Z, Yuan R, et al. The hemoglobin glycation index predicts the risk of adverse cardiovascular events in coronary heart disease patients with type 2 diabetes mellitus. Front Cardiovasc Med. 2022;9:992252. [DOI] [PMC free article] [PubMed] [Google Scholar]