Diabetes, glycaemic traits and cardiovascular disease in females and males: Observational and Mendelian randomisation analyses in the UK Biobank

Abstract

Introduction: Observational studies have shown that the association between type 2 diabetes and cardiovascular disease (CVD) is stronger in females than in males. It remains unclear whether the causal effects of diabetes and glycaemic traits on CVD are also different between females and males.

Methods: We performed sex-stratified observational and Mendelian randomisation (MR) analyses in the UK Biobank to investigate the sex-specific associations of type 2 diabetes and HbA1c with CVD outcomes (combined CVD, coronary heart disease [CHD], myocardial infarction, stroke, ischaemic stroke, intracerebral haemorrhage and subarachnoid haemorrhage). As secondary analyses, we performed sex-stratified MR for the association of genetically proxied fasting glucose and insulin with CVD outcomes.

Results: In observational analysis, diabetes was associated with a greater excess risk for CHD in females than in males (female-to-male ratio of hazard ratios 1.11 [95% CI 1.03, 1.21]). The association of HbA1c with CVD outcomes was similar in both sexes. In MR, the relationship between genetic liability to diabetes and CHD was similar in females and males (female-to-male ratio of odds ratios 0.98 [95% CI 0.91, 1.05]). No sex differences were found for the association between diabetes and stroke in both observational and MR analyses. Moreover, MR results on HbA1c, fasting glucose and fasting insulin were similar for females and males.

Conclusion: This study suggests that causal effects of diabetes and glycaemic traits on CVD are similar in females and males.

Keywords: Mendelian randomisation; cardiovascular disease; diabetes; glycaemic traits; risk factors; sex differences.

FIGURE 1

Cox regression estimates and Mendelian randomisation estimates of the association between type 2 diabetes and cardiovascular disease outcomes in females and males. MR estimates are from inverse‐variance weighted MR, and ORs can be interpreted as the effect per unit increase in log odds of genetic liability to diabetes. MR analyses were performed in 337 386 UK Biobank participants. Cox regressions were performed in 468 838 UK Biobank participants and were adjusted for sex, Townsend deprivation index (an area‐based measure of socioeconomic status), systolic blood pressure, total cholesterol levels, smoking status, body mass index, use of lipid‐lowering medication and use of antihypertensives, including an interaction term between each of these adjustment variables and sex. RHRs present the female‐to‐male ratios of HRs as obtained from an interaction term of diabetes with sex, and RORs present the female‐to‐male ratios of ORs as obtained from two separate MR analyses. CI, confidence interval; HR, hazard ratio; MR, Mendelian randomisation; OR, odds ratio; RHR, ratio of hazard ratios; ROR, ratio of odds ratios.

FIGURE 2

Cox regression estimates and Mendelian randomisation estimates of the association between HbA1c and cardiovascular disease outcomes in females and males. MR estimates are from inverse‐variance weighted MR and odds ratios (ORs) can be interpreted as the effect per 1% increase in genetically predicted HbA1c level. MR analyses were performed in 337 386 UK Biobank participants. Cox regressions were performed in 468 838 UK Biobank participants and adjusted for sex, type 2 diabetes status, Townsend deprivation index (an area‐based measure of socioeconomic status), systolic blood pressure, total cholesterol levels, smoking status, body mass index, use of lipid‐lowering medication and use of antihypertensives, including an interaction term between each of these adjustment variables and sex. RHRs present the female‐to‐male ratios of HRs as obtained from an interaction term of HbA1c and sex, and RORs present the female‐to‐male ratios of ORs as obtained from two separate MR analyses. CI, confidence interval; HR, hazard ratio; MR, Mendelian randomisation; OR, odds ratio; RHR, ratio of hazard ratios; ROR, ratio of odds ratios.

FIGURE 3

Mendelian randomisation estimates of the association between fasting glucose and cardiovascular disease outcomes in females and males. Estimates are from inverse‐variance weighted MR. The odds ratios (ORs) can be interpreted as the effect per 1 mmol/L increase in genetically predicted fasting glucose levels. MR analyses were performed in 337 386 UK Biobank participants. RORs present the female‐to‐male ratios of ORs as obtained from two separate MR analyses. CI, confidence interval; OR, odds ratio; ROR, ratio of odds ratios.

FIGURE 4

Mendelian randomisation estimates of the association between fasting insulin and cardiovascular disease outcomes in females and males. Estimates are from inverse‐variance weighted MR. The odds ratios (ORs) can be interpreted as the effect per one log unit increase in genetically predicted fasting insulin level. MR analyses were performed in 337 386 UK Biobank participants. RORs present the female‐to‐male ratios of ORs as obtained from two separate MR analyses. CI, confidence interval; OR, odds ratio; ROR, ratio of odds ratios.