Genetically predicted androgenic profiles and adverse cardiac markers: a sex‐specific Mendelian randomization study

Abstract

Aims

Observational evidence suggests associations between sex hormone levels and heart failure (HF). We used sex‐specific genetic variants associated with androgenic sex hormone profiles to investigate the causal relevance of androgenic sex hormone profiles on cardiac structure and function and HF using Mendelian randomization (MR).

Methods and results

Sex‐specific uncorrelated genome‐wide significant (P < 5 × 10⁻⁸) variants predicting sex hormone‐binding globulin (SHBG), total testosterone, and bioavailable testosterone were extracted from summary statistics of genome‐wide association study (GWAS) on 425 097 participants in the UK Biobank. Sex‐specific gene–outcome association estimates were computed for left ventricular ejection fraction (LVEF), left ventricular end‐diastolic and end‐systolic volumes (LVEDV and LVESV, respectively), left ventricular stroke volume (LVSV), cardiac index, and cardiac output in 11 528 female and 14 356 male UK Biobank Imaging Study participants and for incident or prevalent HF in an external cohort of 47 309 cases and 930 014 controls. Inverse‐variance weighted MR was the primary analysis method. In females, higher genetically predicted bioavailable testosterone was associated with lower LVEDV [β per nmol/L = −0.11 (−0.19 to −0.03), P = 0.006], lower LVESV [β = −0.09 (−0.17 to −0.01), P = 0.022], lower LVSV [β = −0.11 (−0.18 to −0.03), P = 0.005], lower cardiac output [β = −0.08 (−0.16 to 0.00), P = 0.046], and lower cardiac index [β = −0.08 (−0.16 to −0.01), P = 0.034] and a higher risk of HF [odds ratio 1.10 (1.01–1.19), P = 0.026] on external validation analysis in larger scale, sex‐adjusted GWAS data. Higher genetically predicted SHBG was associated with higher LVEDV [β per nmol/L = 0.17 (0.08–0.25), P = 2 × 10⁻⁴], higher LVESV [β = 0.13 (0.05–0.22), P = 0.003], and higher LVSV [β = 0.18 (0.08–0.28), P = 2 × 10⁻⁴]. In males, higher genetically predicted total and bioavailable testosterone was associated with lower LVESV [β = −0.07 (−0.12 to −0.02), P = 0.007] and LVEF [β = −0.11 (−0.18 to −0.04), P = 0.003], respectively.

Conclusions

This study supports a causal effect of pro‐androgenic sex hormone profiles in females on adverse markers of left ventricular structure and function typically associated with HF with preserved ejection fraction and with HF. There was weaker evidence of association in males.

Introduction

Heart failure (HF) is a major global health threat, affecting 64.3 million worldwide.^1,2 Trends in lifetime HF risk vary substantially between sexes, suggesting a potential role of sex hormones in its pathogenesis.

Hyperandrogenic states, characterized by increased levels of bioavailable testosterone, commonly occur in both males and females due to physiological, pathological, and iatrogenic factors. Physiologically, females are known to switch to a proportionally more pro-androgenic hormonal profile after the menopause,³ and a transient physiological increase in testosterone levels is known to occur after exercise in both males and females.⁴ Pathologically, hyperandrogenic states might occur with adrenal tumours and, for females, in the much more common polycystic ovarian syndrome (PCOS).⁵ Finally, iatrogenic increases might occur secondary to the use of testosterone-based hormonal therapies. In males, the use of testosterone replacement therapy peaked as high as 3.2% among adults in the United States in 2011, reducing to 1.3% in 2016 following research linking its use to heightened cardiovascular risk.^6,7 In birth-assigned females, the use of testosterone-based gender-affirming hormonal therapy is also on the rise. While up-to-date figures regarding prescription of these therapies are currently lacking, population-based estimates of gender incongruence for birth-assigned female persons range from 0.4% to 1.2%.⁸ Additionally, multiple hormone replacement therapy (HRT) preparations used during the peri-menopause or post-menopause contain testosterone.⁹

The cardiovascular associations of hyperandrogenism vary between males and females. In males, an inverse association between androgen levels and cardiovascular risk has been reported,¹⁰ and concordantly, androgen deprivation therapies have been associated with improved cardiometabolic risk profiles.¹¹ However, prior studies have reported a paradoxically higher rate of major cardiovascular events.⁷ Additionally, randomized studies of active testosterone administration among at-risk males have been generally null,^12,13 and some even describe harm.^14,15 The relationship is more complex in females. Observational studies in pre-menopausal females describe an association between pro-androgenic hormonal profiles and cardiovascular risk,¹⁶ and PCOS is known to be associated with adverse cardiometabolic profile and higher cardiovascular event rates.¹⁷ In post-menopausal females, results are conflicted, with some observational studies reporting a protective effect of higher androgen levels^18,19 and others a detrimental one.^20,21 Cohort studies in transgender men undergoing testosterone-based hormone therapy have reported a higher risk of cardiovascular disease and adverse cardiometabolic profiles.^22,23

The observational nature of currently available evidence is a major limitation to causal inference because observational associations are liable to residual confounding and bias. Mendelian randomization (MR) is a method that leverages ‘natural’ randomization to high or low genetic predisposition to risk factors within an instrumental variable framework to explore the causal relevance of risk factors on an outcome. As the allocation to genetic risk is random, this is akin to randomization in a clinical trial and thus less liable to influence by observational confounding.

Improving current understanding regarding sex hormones and cardiovascular risk is a key public health priority. First, it might lead to improved life course risk stratification relating to physiological, pathological, or iatrogenic hormonal changes. Second, it may better inform care for people receiving testosterone-based hormonal therapy by improving knowledge of cardiovascular effects.

This study aims to leverage genetic variants associated with bioavailable testosterone, as well the closely related measures of sex hormone-binding globulin (SHBG) and total testosterone, to evaluate their causal relevance on multiple cardiac magnetic resonance (CMR) imaging markers of cardiac structure and function, and HF risk.

Methods

Ethics and data access

The study is reported using the methodological framework from the Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STROBE-MR) guidelines²⁴ and is in line with recommendations outlined in the Sex and Gender Equity in Research guidelines.²⁵ For the purpose of sex-specific analyses, sex was defined as that which was genetically inferred.

Summary-level data from publicly available genome-wide association study (GWAS) summary statistics were used and are available to download freely from the cited publications. Relevant ethical approval and participant consent were obtained in each individual study. Individual-level data in the UK Biobank were accessed under Application 13784. Participants provided written informed consent for data release for research purposes within approved applications. Further details on the protocol of the UK Biobank can be found in previous publications.^26–28

All statistical analyses were performed using R Version 4.1.1 (2021-02-15)²⁹ using the MendelianRandomization³⁰ and TwoSampleMR packages.³⁰

Instrumental variable selection

Sex-specific instrumental variants were extracted from Ruth et al.'s GWAS on 425 097 UK Biobank participants of predominantly European ancestry.³¹ The primary exposure considered in this study was bioavailable testosterone (males n = 178 782, females n = 188 507, unit = nmol/L). As testosterone exists in a physiological equilibrium between its bioavailable form and its bound form to SHBG, we additionally studied the secondary exposures of SHBG (males n = 180 726, females n = 189 473, unit = nmol/L) and total testosterone (males n = 194 453, females n = 230 454, unit = nmol/L). Given their equilibrium state, it would be expected that any direct association observed for the exposure of bioavailable testosterone would be reflected by an inverse association for SHBG and vice versa. Sex-specific, uncorrelated (r² < 0.05) single nucleotide polymorphisms (SNPs) associated with each exposure at genome-wide significance (P < 5 × 10⁻⁸) were selected as instrumental variants. The list of instrumental variants for each exposure, with respective effect sizes and effect alleles, is provided in Supporting Information, Tables S1–S6. A summary of the study methodology and data sources is presented in Figure 1.

[IMAGE OMITTED. SEE PDF]

Study outcomes

Outcomes included CMR measures of cardiac structure and function, including left ventricular (LV) end-systolic volume (LVESV), LV end-diastolic volume (LVEDV), LV ejection fraction (LVEF), LV stroke volume (LVSV), cardiac output (CO), and cardiac index (CI), and clinical diagnosis of HF.

The methods used to derive measures of cardiac structure and function in the UK Biobank Imaging Study have been previously described.³² The definition of HF in the UK Biobank cohort is presented in Supporting Information, Table S7, which includes diagnoses recorded across all hospital inpatient records. Sex-specific gene–outcome association estimates were extracted from the UK Biobank and UK Biobank Imaging Study. In females, sample sizes were as follows: 11 528 participants for LVEF, LVEDV, LVESV, LVSV, and CO and 11 525 participants for CI. In males, sample sizes were as follows: 14 358 participants for LVEF, LVEDV, LVESV, LVSV, and CO and 14 356 participants for CI. Genetic association estimates for the outcome of HF were extracted from the results of Shah et al.'s GWAS on 47 309 HF cases and 930 014 controls.³³ This GWAS was not sex-specific, but the association estimates were adjusted for sex. Validation analysis was performed on sex-specific data from the UK Biobank to look for consistency of estimates, including 4128 cases and 194 687 controls for females and 7713 cases and 161 014 controls for males.

Mendelian randomization

Inverse-variance weighted (IVW) MR with multiplicative random effects³⁴ was used for primary analysis for all models.³⁵ Results are presented as odds ratios (ORs) with 95% confidence intervals for each unit increment in genetically predicted exposure on HF risk. For the CMR outcomes, results are presented as beta coefficients (β) and 95% confidence intervals, indicating the expected unit change in CMR parameter for each nmol/L increment in the genetically predicted exposure.

Sensitivity analyses

Sensitivity analyses were carried out to ascertain potential violations of the instrumental variable assumptions. These include that: (i) the instruments associate with exposures of interest (relevance); (ii) do not have confounders associated with the outcome (independence); and (iii) no horizontal/direct pleiotropic effects exist (exclusion restriction).

To assess the relevance assumption, instrument strength was quantified using F-statistics, reported in Supporting Information, Table S8. To assess the exclusion restriction assumption, sensitivity analysis was carried out using weighted median MR and MR-Egger.³⁶ The weighted mean estimate has been shown to provide consistent results in simulations when up to 50% of instrumental variants included in the analysis violate this assumption.³⁷ The MR-Egger test detects evidence of directional pleiotropy under a weaker assumption that the instrument strength is independent of direct effects (InSIDE assumption)³⁶ through the introduction of an intercept term.

Where evidence of potential pleiotropy was discovered, instrumental variants were examined in more detail through a phenome-wide association search using PhenoScanner (Supporting Information, Table S9). Among phenotypes associated with instrumental variants, candidate phenotypes to adjust for in multivariable MR (MVMR³⁸) analyses were chosen based on prior knowledge of causal associations of the phenotypes with HF. Association estimates between genetically predicted sex hormones and the outcome in question were adjusted by the associations of the instrumental SNPs with body mass index (BMI) and systolic blood pressure (SBP), extracted respectively from Pulit et al.'s³⁹ (n = 806 834 participants) and Evangelou et al.'s⁴⁰ (n = 738 168 participants) GWAS summary statistics. The strength of instruments for the sex hormone trait in the modified multivariable analyses was quantified using conditional F-statistics.⁴¹

Results

Females

Measures of cardiac structure

In females, higher genetically predicted bioavailable testosterone was associated with lower LVEDV [β per nmol/L = −0.11 (−0.19 to −0.03), P = 0.006], lower LVESV [β = −0.09 (−0.17 to −0.01), P = 0.022], and lower LVSV [β = −0.11 (−0.18 to −0.03), P = 0.005]. As would be expected, the associations for SHBG were in the opposite direction: genetically predicted SHBG was associated with higher LVEDV [β per nmol/L = 0.17 (0.08–0.25), P = 2 × 10⁻⁴], higher LVESV [β = 0.13 (0.05–0.22), P = 0.003], and higher LVSV [β = 0.18 (0.08–0.28), P = 2 × 10⁻⁴]. Higher genetically predicted total testosterone was associated with higher LVEDV [β per nmol/L = 0.06 (0.00–0.11), P = 0.036], but no significant difference in LVESV or LVSV.

The results are reported in Figure 2 and Table 1. Sensitivity analyses revealed evidence of potential directional pleiotropy in the association between bioavailable testosterone and LVSV [MR-Egger intercept test P = 0.043; MR-Egger β = 0.01 (−0.13 to 0.15), P = 0.933], as displayed in Table 2. Adjustment for genetically predicted SBP on multivariable analysis did not attenuate the association between bioavailable testosterone and LVSV [β = −0.12 (−0.21 to −0.03), P = 0.0.013, conditional F-statistic = 65.4]; however, adjustment for BMI did [β = 0.01 (−0.12 to 0.10), P = 0.843, conditional F-statistic = 37.7], indicating likely pleiotropy through BMI.

[IMAGE OMITTED. SEE PDF]

Table 1 Mendelian randomization estimates for the effects of androgenic sex hormones on cardiovascular outcomes, using an inverse-variance weighted model

Exposure	Outcome	SNPs	β/(*OR)	Lower 95% CI	Upper 95% CI	P value
Female bioavailable testosterone (per 1-standard deviation increase)	LVEF	175	0.04	−0.04	0.12	0.358
LVEDV	175	−0.11	−0.19	−0.03	0.006
LVESV	175	−0.09	−0.17	−0.01	0.022
LVSV	175	−0.11	−0.18	−0.03	0.005
Cardiac output	175	−0.08	−0.16	0.00	0.046
Cardiac index	175	−0.08	−0.16	−0.01	0.034
Heart failurea	175	1.13	0.96	1.34	0.147
Female total testosterone (per 1-standard deviation increase)	LVEF	243	0.00	−0.06	0.05	0.903
LVEDV	243	0.06	0.00	0.11	0.036
LVESV	243	0.05	0.00	0.11	0.053
LVSV	242	0.04	−0.01	0.10	0.135
Cardiac output	243	0.02	−0.04	0.08	0.568
Cardiac index	243	0.01	−0.04	0.07	0.617
Heart failurea	243	1.08	0.96	1.21	0.199
Female sex hormone-binding globulin (per 1-standard deviation increase)	LVEF	348	−0.06	−0.15	0.02	0.149
LVEDV	347	0.17	0.08	0.25	2 × 10−4
LVESV	348	0.13	0.05	0.22	0.003
LVSV	347	0.18	0.08	0.28	2 × 10−4
Cardiac output	347	0.08	−0.01	0.17	0.097
Cardiac index	347	0.06	−0.03	0.15	0.212
Heart failurea	348	0.95	0.80	1.13	0.544
Male bioavailable testosterone (per 1-standard deviation increase)	LVEF	121	−0.11	−0.18	−0.04	0.003
LVEDV	121	−0.03	−0.1	0.03	0.322
LVESV	121	−0.03	−0.09	0.03	0.367
LVSV	121	−0.02	−0.09	0.04	0.462
Cardiac output	121	−0.01	−0.08	0.05	0.681
Cardiac index	121	−0.01	−0.08	0.05	0.719
Heart failurea	103	0.94	0.82	1.06	0.314
Male total testosterone (per 1-standard deviation increase)	LVEF	226	−0.01	−0.06	0.04	0.766
LVEDV	225	−0.04	−0.09	0.01	0.090
LVESV	225	−0.07	−0.12	−0.02	0.007
LVSV	226	−0.01	−0.06	0.03	0.567
Cardiac output	226	−0.01	−0.06	0.04	0.732
Cardiac index	226	−0.01	−0.06	0.04	0.786
Heart failurea	216	0.94	0.86	1.02	0.125
Male sex hormone-binding globulin (per 1-standard deviation increase)	LVEF	353	0.04	−0.06	0.13	0.446
LVEDV	352	0.00	−0.09	0.09	0.964
LVESV	352	−0.02	−0.11	0.07	0.724
LVSV	353	0.01	−0.08	0.10	0.880
Cardiac output	353	0.00	−0.09	0.09	0.981
Cardiac index	353	−0.01	−0.10	0.08	0.873
Heart failurea	338	1.05	0.90	1.21	0.563

Table 2 Mendelian randomization (MR) sensitivity analyses for the effects of reproductive factors on cardiovascular outcomes, using weighted median MR and MR-Egger models

Exposure	Outcome	Method	SNPs	β/OR	Lower 95% CI	Upper 95% CI	P value
Female bioavailable testosterone (per 1-standard deviation increase)	LVEF	Weighted median	175	0.00	−0.13	0.14	0.952
MR-Egger	175	0.09	−0.05	0.23	0.211
				Intercept	0.373
LVEDV	Weighted median	175	−0.07	−0.20	0.06	0.283
MR-Egger	175	−0.04	−0.18	0.10	0.604
				Intercept	0.212
LVESV	Weighted median	175	−0.09	−0.22	0.04	0.186
MR-Egger	175	−0.05	−0.19	0.09	0.511
				Intercept	0.437
LVSV	Weighted median	175	0.01	−0.12	0.14	0.867
MR-Egger	175	0.01	−0.13	0.14	0.933
				Intercept	0.043
Cardiac output	Weighted median	175	0.02	−0.11	0.14	0.815
MR-Egger	175	0.10	−0.03	0.24	0.143
				Intercept	0.002
Cardiac index	Weighted median	175	−0.01	−0.14	0.11	0.821
MR-Egger	175	0.08	−0.06	0.21	0.249
				Intercept	0.005
Heart failure	Weighted median	175	1.06	0.79	1.44	0.687
MR-Egger	175	1.21	0.90	1.62	0.215
				Intercept	0.589
Female total testosterone (per 1-standard deviation increase)	LVEF	Weighted median	243	−0.04	−0.15	0.07	0.472
MR-Egger	243	0.03	−0.07	0.13	0.575
				Intercept	0.450
LVEDV	Weighted median	243	0.09	−0.02	0.20	0.108
MR-Egger	243	0.04	−0.06	0.14	0.439
				Intercept	0.644
LVESV	Weighted median	243	0.10	−0.01	0.21	0.069
MR-Egger	243	0.04	−0.06	0.14	0.435
				Intercept	0.725
LVSV	Weighted median	242	0.03	−0.09	0.14	0.651
MR-Egger	242	0.02	−0.08	0.13	0.696
				Intercept	0.599
Cardiac output	Weighted median	243	0.00	−0.11	0.12	0.942
MR-Egger	243	0.01	−0.10	0.11	0.905
				Intercept	0.813
Cardiac index	Weighted median	243	−0.04	−0.14	0.07	0.478
MR-Egger	243	−0.02	−0.12	0.09	0.779
				Intercept	0.502
Heart failure	Weighted median	243	1.28	1.06	1.54	0.009
MR-Egger	243	1.33	1.08	1.64	0.007
				Intercept	0.017
Female sex hormone-binding globulin (per 1-standard deviation increase)	LVEF	Weighted median	348	−0.06	−0.21	0.08	0.411
MR-Egger	348	−0.08	−0.21	0.05	0.219
				Intercept	0.735
LVEDV	Weighted median	347	0.06	−0.09	0.22	0.440
MR-Egger	347	0.13	0.01	0.26	0.041
				Intercept	0.513
LVESV	Weighted median	348	0.07	−0.07	0.22	0.323
MR-Egger	348	0.13	0.00	0.25	0.044
				Intercept	0.971
LVSV	Weighted median	347	−0.01	−0.17	0.16	0.925
MR-Egger	347	0.09	−0.05	0.23	0.224
				Intercept	0.075
Cardiac output	Weighted median	347	−0.13	−0.29	0.04	0.133
MR-Egger	347	−0.04	−0.17	0.09	0.553
				Intercept	0.018
Cardiac index	Weighted median	347	−0.13	−0.29	0.03	0.106
MR-Egger	347	−0.05	−0.18	0.09	0.492
				Intercept	0.035
Heart failure	Weighted median	348	0.95	0.70	1.29	0.750
MR-Egger	348	0.90	0.69	1.16	0.419
				Intercept	0.588
Male bioavailable testosterone (per 1-standard deviation increase)	LVEF	Weighted median	121	−0.18	−0.28	−0.07	0.001
MR-Egger	121	−0.18	−0.29	−0.07	0.002
				Intercept	0.118
LVEDV	Weighted median	121	−0.07	−0.17	0.04	0.206
MR-Egger	121	−0.10	−0.20	0.00	0.065
				Intercept	0.112
LVESV	Weighted median	121	−0.08	−0.19	0.02	0.128
MR-Egger	121	−0.06	−0.17	0.04	0.220
				Intercept	0.393
LVSV	Weighted median	121	−0.06	−0.17	0.04	0.256
MR-Egger	121	−0.09	−0.19	0.02	0.098
				Intercept	0.124
Cardiac output	Weighted median	121	−0.03	−0.14	0.07	0.517
MR-Egger	121	−0.07	−0.17	0.03	0.168
				Intercept	0.148
Cardiac index	Weighted median	121	−0.04	−0.15	0.06	0.414
MR-Egger	121	−0.08	−0.18	0.03	0.146
				Intercept	0.114
Heart failure	Weighted median	103	0.84	0.70	1.01	0.064
MR-Egger	103	0.96	0.76	1.20	0.711
				Intercept	0.814
Male total testosterone (per 1-standard deviation increase)	LVEF	Weighted median	226	0.04	−0.06	0.13	0.424
MR-Egger	226	0.03	−0.05	0.11	0.452
				Intercept	0.225
LVEDV	Weighted median	225	−0.09	−0.18	−0.01	0.033
MR-Egger	225	−0.03	−0.11	0.05	0.463
				Intercept	0.655
LVESV	Weighted median	225	−0.12	−0.20	−0.03	0.007
MR-Egger	225	−0.05	−0.13	0.03	0.198
				Intercept	0.592
LVSV	Weighted median	226	−0.06	−0.14	0.03	0.194
MR-Egger	226	−0.01	−0.08	0.07	0.857
				Intercept	0.812
Cardiac output	Weighted median	226	−0.05	−0.14	0.03	0.231
MR-Egger	226	0.00	−0.08	0.08	0.994
				Intercept	0.770
Cardiac index	Weighted median	226	−0.05	−0.13	0.04	0.290
MR-Egger	226	0.00	−0.08	0.07	0.949
				Intercept	0.887
Heart failure	Weighted median	216	0.90	0.78	1.03	0.127
MR-Egger	216	1.03	0.90	1.18	0.673
				Intercept	0.071
Male sex hormone-binding globulin (per 1-standard deviation increase)	LVEF	Weighted median	353	0.17	0.01	0.33	0.037
MR-Egger	353	0.13	−0.01	0.26	0.060
				Intercept	0.057
LVEDV	Weighted median	352	0.00	−0.15	0.15	0.986
MR-Egger	352	−0.03	−0.16	0.09	0.615
				Intercept	0.499
LVESV	Weighted median	352	−0.14	−0.30	0.01	0.074
MR-Egger	352	−0.09	−0.22	0.03	0.154
				Intercept	0.093
LVSV	Weighted median	353	−0.03	−0.17	0.11	0.634
MR-Egger	353	0.01	−0.11	0.14	0.831
				Intercept	0.879
Cardiac output	Weighted median	353	−0.06	−0.22	0.09	0.406
MR-Egger	353	0.02	−0.10	0.14	0.768
				Intercept	0.654
Cardiac index	Weighted median	353	−0.08	−0.22	0.06	0.257
MR-Egger	353	0.02	−0.11	0.14	0.786
				Intercept	0.579
Heart failure	Weighted median	338	1.20	0.95	1.51	0.122
MR-Egger	338	0.97	0.78	1.20	0.760
				Intercept	0.310

Measures of cardiac function

In females, higher genetically predicted bioavailable testosterone was associated with a lower CO [β = −0.08 (−0.16 to 0.00), P = 0.046] and a lower CI [β = −0.08 (−0.16 to −0.01), P = 0.034]. No associations were evident for the secondary exposures of genetically predicted SHBG and genetically predicted total testosterone with either outcome, and none of the exposures were associated with LVEF. The results are reported in Figure 2 and Table 1.

Sensitivity analyses revealed potential directional pleiotropy in the association of bioavailable testosterone with CO (MR-Egger intercept test P = 0.002) and CI (MR-Egger intercept test P = 0.005), as displayed in Table 2. Adjustment for genetically predicted SBP on multivariable analysis mildly attenuated the association between bioavailable testosterone and CO [β = −0.08 (0.01 to −0.17), P = 0.062, conditional F-statistic = 65.4] and did not attenuate the association with CI [β = −0.10 (−0.18 to −0.01), P = 0.021, conditional F-statistic = 65.4]. However, adjustment for BMI attenuated the association with both CO [β = 0.09 (−0.02 to 0.19), P = 0.112, conditional F-statistic = 37.7] and CI [β = 0.06 (−0.05 to 0.17), P = 0.021, conditional F-statistic = 37.7], indicating likely pleiotropy through BMI.

Heart failure

Higher genetically predicted bioavailable testosterone in females was associated with higher odds of HF [OR 1.10 (1.01–1.19), P = 0.026]. In the validation analysis on the smaller UK Biobank cohort, this association was not significant though the association estimate remained consistent in both direction and magnitude [OR 1.13 (0.96–1.34), P = 0.147].

There was no significant association with HF for genetically predicted SHBG and genetically predicted total testosterone in both the primary and validation analyses, as reported in Tables 1 and 3 and Figure 3.

Table 3 Mendelian randomization estimates for the effects of reproductive factors on heart failure, using an inverse-variance weighted model

Exposure	Outcome	Number of SNPs	Odds ratio	Lower 95% CI	Upper 95% CI	P value
Female bioavailable testosterone	Heart failure	132	1.10	1.01	1.19	0.026
Female total testosterone	185	1.02	0.96	1.08	0.492
Female sex hormone-binding globulin	275	0.92	0.83	1.01	0.081
Male bioavailable testosterone	82	1.04	0.96	1.11	0.359
Male total testosterone	154	0.95	0.90	1.01	0.095
Male sex hormone-binding globulin	256	0.97	0.88	1.06	0.504

[IMAGE OMITTED. SEE PDF]

Males

Measures of cardiac structure

Higher genetically predicted total testosterone was associated with a lower LVESV in males [β = −0.07 (−0.12 to −0.02), P = 0.007], but no associations were observed for genetically predicted bioavailable testosterone and SHBG, as reported in Figure 4. Sensitivity analyses were consistent with the primary analyses, as displayed in Table 2. In males, none of the exposures were associated with either LVEDV or LVSV. The results are reported in Figure 3.

[IMAGE OMITTED. SEE PDF]

Measures of cardiac function

In males, higher genetically predicted bioavailable testosterone was associated with a lower LVEF [β = −0.11 (−0.18 to −0.04), P = 0.003]. No associations with LVEF were observed for genetically predicted SHBG and total testosterone. Additionally, as displayed in Figure 4, none of the genetically predicted exposures were associated with CO or CI.

Heart failure

In the analysis on the external cohort, higher genetically predicted bioavailable testosterone was not associated with HF [OR 1.04 (0.96–1.11), P = 0.359], and neither were genetically predicted total testosterone or SHBG (Table 3 and Figure 4). The null results were consistent in the validation analysis in the UK Biobank, as reported in Table 1.

Discussion

In this study, we explored associations between androgenic sex hormone profiles and multiple markers of cardiac structure and function. A pro-androgenic profile was associated with adverse LV remodelling with features of greater cardiac ageing broadly reflective of an HF with preserved ejection fraction (HFpEF) phenotype. These relationships appeared more prominent in females, in whom higher bioavailable testosterone (and lower SHBG) was associated with a smaller structure (lower LVEDV, LVESV, and LVSV) and function (lower LVSV, CO, and CI). A pro-androgenic profile was associated with a greater likelihood of clinical HF in females, but not in males. Thus, our results demonstrate androgenic hormones as a proponent of poorer cardiovascular health, in particular HF, which appears more important in females than males.

Several observational studies have explored the association between androgenic sex hormone profiles and cardiovascular risk in females, linking to both menopausal and pathological conditions. Post-menopausal females exhibit proportionally more androgenic sex hormone profiles, which has been, albeit inconsistently,^21,42,43 suggested to associate with increased cardiovascular risk.^19,20 This pattern is reflected in females with PCOS, who exhibit higher absolute androgen levels and have also been reported to suffer higher risk of cardiovascular disease. However, drawing causal inference from observational studies is limited due to the residual confounding. The present study provides genetic evidence to support a causal role of higher androgen levels in females on adverse cardiac remodelling. This study adds to current literature as the first to employ MR to explore this question, limiting the influence of observational confounding and therefore providing more reliable evidence of causal relationships.

There are several potential mechanisms behind the findings of this study. Females have a 20-fold to 25-fold lower circulating androgen concentration compared with males.⁴⁴ Long-term androgen receptor activation in females has been shown to predispose to insulin resistance, combined with up-regulation of genes involved in HDL catabolism leading to dyslipidaemia.¹⁶ Indeed, Ruth et al. have previously described an association between higher androgen levels and worse cardiometabolic profiles in females using MR.³¹ Additionally, hormonal changes occurring in the peri-menopause, when androgenic transient hormones increase and then decline to 15% of pre-menopausal levels post-menopause,⁴⁴ have been shown to accelerate adverse arterial remodelling through proliferation of vascular smooth muscle cells,^45,46 leading to greater arterial stiffness and thus accelerated age-related hypertension, contributing to augmented cardiovascular risk. Physiological levels of testosterone increase nitric oxide synthesis through gene induction, which is not seen with high levels of testosterone.⁴⁶ Finally, the menopause transition and its concurrent proportionally androgenic state have been associated with significant changes in body composition,⁴⁷ including increased deposition of visceral fat. Increased visceral fat, especially in the pericardial compartment⁴⁸ and abdominal area,⁴⁶ might further augment cardiovascular risk through both endocrine and inflammatory paracrine mechanisms. It is important to note that the above mechanisms form a complex picture and existing research is contradictory alluding to multifactorial explanations linking androgens to cardiovascular risk.

In males, higher androgen levels were associated with reduced LVEF and LVESV. These results are contradictory with a prior MR analysis⁴⁹ and also with available observational evidence. In the prior MR analysis, genetically predicted testosterone was associated with HF.⁴⁹ The different findings could be attributed to the different health statuses in their study population compared with the UK Biobank, comprising a relatively healthy cohort. Furthermore, there were low numbers of participants with HF used in the prior study⁴⁹ and survival bias could have played a part in the UK Biobank. In observational studies, androgen deprivation use as a treatment for prostate cancer⁵⁰ or due to hypogonadism,^7,51,52 have been associated with cardiovascular risk. A potential explanation for this apparent equivocality might relate to non-linear effects, whereby those with both extremely low and high androgen levels suffer increased risk. Overall, given the important inconsistencies in available evidence, cardiac pharmacovigilance in the context of both androgen-depriving and androgen-enhancing therapies in males remains an important priority, and further research is needed to explore potential non-linear effects or establish effect modifiers.

In this study, we identified potential pleiotropic effects in the association between bioavailable testosterone and cardiac function in females. This finding requires further discussion. First, this only involves one of the exposures, bioavailable testosterone, and the specific outcomes of LVSV, CO, and CI (not LVESV, LVEDV, and HF). Given the biologically plausible and consistent results observed across other outcomes and exposures, the potential pleiotropy discovered in some specific associations does not substantially alter the broader interpretation of study findings. Second, we further explored this by highlighting potential culprit phenotypes of BMI and SBP and demonstrated a more likely role of BMI in the pleiotropic effects. This suggests that BMI might either directly or indirectly be responsible for at least part of the observed associations, highlighting it as a key target for surveillance and risk modification.

These results carry important clinical implications. First, the results support a potentially adverse mechanistic role of androgens in the cardiovascular association of multiple pathological states characterized by higher androgen levels in females, such as PCOS, and cardiac disease. The results therefore encourage further research to elucidate whether androgens are an important or primary mediator of this association, as they represent therapeutic targets amenable to intervention. Second, they support further research exploring the long-term cardiovascular health of individuals taking testosterone-based hormone treatments. These include post-menopausal females taking HRT preparations that contain testosterone, as well as individuals undergoing gender-affirming hormone treatments. Gender-affirming hormone treatment use has increased significantly in recent years, and creating a high-quality evidence base regarding long-term cardiovascular health should be a key priority to ensure high standards of care. Evidence among existing studies is conflicting, with some suggesting an increased risk of cardiovascular disease in transgender males undergoing hormone therapy^53,54 yet others suggesting no effect.^22,55,56 Importantly, higher risk of hypertension has been reported, suggesting mechanistic effects on the vascular system consistent with the HFpEF imaging phenotype observed in this study.

There are several limitations to discuss. First, magnitude of effect estimates for reported associations cannot be interpreted intuitively as real-life unit changes in outcome expected per unit change in the exposure through pharmacological modification. Measures of exposures in MR studies reflect a lifetime exposure, rather than a time-limited modification in the exposure secondary to pharmacotherapy. Second, we were unable to expand the analysis to other sex hormones (e.g. oestradiol and progesterone) and additional measures of cardiac structure and function (e.g. LV mass) due to the lack of sufficiently powered data available. This should be an important target for future research. Finally, in this study, we utilized two-sample MR methods in a one-sample setting as both gene–exposure and gene–outcome association data were derived from the UK Biobank. However, this has been shown to be unlikely to significantly bias analyses in large-scale biobanks.⁵⁷

Conclusions

In conclusion, these data support a causal association between bioavailable testosterone in females and adverse measures of cardiac structure and function in a pattern broadly consistent with HFpEF, as well as a higher risk of HF. In males, higher androgen levels were associated with reduced LVEF and LVESV. The results call for further investigation of the prognostic value of sex-specific risk factors typically associated with pro-androgenic states (e.g. PCOS) for cardiovascular risk stratification of females. By highlighting a likely causal role of testosterone, they also encourage exploration of whether this might be a modifiable mediator of the increased cardiovascular risk that is known to occur in PCOS. Additionally, they encourage pharmacovigilance and ongoing research to characterize the long-term cardiovascular health of individuals undergoing testosterone-based therapies for both gender affirmation and hormone replacement.

Acknowledgements

The authors would like to acknowledge the participants and investigators of the UK Biobank. The authors would also like to thank Dr Beata Wojciak-Stothard and Dr Adam Byrne for reading and checking their manuscript. The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.

Conflict of interest

None declared.

Funding

This work was supported by core funding from the British Heart Foundation (BHF) (RG/13/13/30194 and RG/18/13/33946), BHF Cambridge Centre of Research Excellence (RE/18/1/34212), and NIHR Cambridge Biomedical Research Centre (BRC-1215-20014). This work was also supported by Health Data Research UK, which is funded by the UK Medical Research Council, Engineering and Physical Sciences Research Council, Economic and Social Research Council, Department of Health and Social Care (England), Chief Scientist Office of the Scottish Government Health and Social Care Directorates, Health and Social Care Research and Development Division (Welsh Government), Public Health Agency (Northern Ireland), BHF, and Wellcome. M.A. was supported by the NIHR Academic Clinical Fellowship. J.Y.C. and R.K.R. were supported by the NIHR Imperial Biomedical Research Centre HEE Specialised Foundation post. A.M.M. was supported by the EU/EFPIA Innovative Medicines Initiative Joint Undertaking BigData@Heart (Grant No. 116074). Z.R.-E. was supported by the BHF Clinical Research Training Fellowship (Grant No. FS/17/81/33318). S.B. was supported by the Sir Henry Dale Fellowship jointly funded by the Wellcome Trust and the Royal Society (Grant No. 204623/Z/16/Z). E.D.A. has no relevant funding. F.S.N. was supported by the NIHR Imperial Biomedical Research Centre funding and the BHF (RG/16/3/32175).

References

Forouzanfar MH, Alexander L, Bachman VF, Biryukov S, Brauer M, Casey D, et al. Global, regional, and national comparative risk assessment of 79 behavioural, environmental and occupational, and metabolic risks or clusters of risks in 188 countries, 1990–2013: A systematic analysis for the Global Burden of Disease Study 2013. Lancet 2015;386:2287‐2323. doi:10.1016/S0140‐6736(15)00128‐2

James SL, Abate D, Abate KH, Abay SM, Abbafati C, Abbasi N, et al. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990–2017: A systematic analysis for the Global Burden of Disease Study 2017. Lancet 2018;392:1789‐1858. doi:10.1016/S0140‐6736(18)32279‐7

El Khoudary SR, Aggarwal B, Beckie TM, Hodis HN, Johnson AE, Langer RD, et al. Menopause transition and cardiovascular disease risk: Implications for timing of early prevention: A scientific statement from the American Heart Association. Circulation 2020;142:E506‐E532. doi:10.1161/CIR.0000000000000912

Riachy R, McKinney K, Tuvdendorj DR. Various factors may modulate the effect of exercise on testosterone levels in men. J Funct Morphol Kinesiol 2020;5:81. doi:10.3390/jfmk5040081

Sheehan MT. Polycystic ovarian syndrome: Diagnosis and management. Clin Med Res 2004;2:13‐27. doi:10.3121/cmr.2.1.13

Baillargeon J, Kuo Y‐F, Westra JR, Urban RJ, Goodwin JS. Testosterone prescribing in the United States, 2002‐2016. JAMA 2018;320:200‐202. doi:10.1001/jama.2018.7999

Vigen R, O'Donnell CI, Barón AE, Grunwald GK, Maddox TM, Bradley SM, et al. Association of testosterone therapy with mortality, myocardial infarction, and stroke in men with low testosterone levels. JAMA 2013;310:1829‐1836. doi:10.1001/jama.2013.280386

Winter S, Diamond M, Green J, Karasic D, Reed T, Whittle S, et al. Transgender people: Health at the margins of society. Lancet 2016;388:390‐400. doi:10.1016/S0140‐6736(16)00683‐8

Scott A, Newson L. Should we be prescribing testosterone to perimenopausal and menopausal women? A guide to prescribing testosterone for women in primary care. Br J Gen Pract 2020;70:203‐204. doi:10.3399/bjgp20X709265

Laughlin GA, Barrett‐Connor E, Bergstrom J. Low serum testosterone and mortality in older men. J Clin Endocrinol Metab 2008;93:68‐75. doi:10.1210/jc.2007‐1792

Hu JR, Duncan MS, Morgans AK, Brown JD, Meijers WC, Freiberg MS, et al. Cardiovascular effects of androgen deprivation therapy in prostate cancer: Contemporary meta‐analyses. Arterioscler Thromb Vasc Biol 2020;40:E55‐E64. doi:10.1161/ATVBAHA.119.313046

Budoff MJ, Ellenberg SS, Lewis CE, Mohler ER, Wenger NK, Bhasin S, et al. Testosterone treatment and coronary artery plaque volume in older men with low testosterone. JAMA 2017;317:708‐716. doi:10.1001/jama.2016.21043

Hudson J, Cruickshank M, Quinton R, Aucott L, Aceves‐Martins M, Gillies K, et al. Adverse cardiovascular events and mortality in men during testosterone treatment: An individual patient and aggregate data meta‐analysis. Lancet Heal Longev 2022;3:e381‐e393. doi:10.1016/S2666‐7568(22)00096‐4

Finkle WD, Greenland S, Ridgeway GK, Adams JL, Frasco MA, Cook MB, et al. Increased risk of non‐fatal myocardial infarction following testosterone therapy prescription in men. PLoS ONE 2014;9: [eLocator: e85805]. doi:10.1371/journal.pone.0085805

Basaria S, Coviello AD, Travison TG, Storer TW, Farwell WR, Jette AM, et al. Adverse events associated with testosterone administration. N Engl J Med 2010;363:109‐122. doi:10.1056/NEJMoa1000485

Osibogun O, Ogunmoroti O, Michos ED. Polycystic ovary syndrome and cardiometabolic risk: Opportunities for cardiovascular disease prevention. Trends Cardiovasc Med 2020;30:399‐404. doi:10.1016/j.tcm.2019.08.010

Okoth K, Chandan JS, Marshall T, Thangaratinam S, Thomas GN, Nirantharakumar K, et al. Association between the reproductive health of young women and cardiovascular disease in later life: Umbrella review. BMJ 2020;371: [eLocator: m3502]. doi:10.1136/bmj.m3502

Islam RM, Bell RJ, Handelsman DJ, Robinson PJ, Wolfe R, Davis SR. Longitudinal changes over three years in sex steroid hormone levels in women aged 70 years and over. Clin Endocrinol (Oxf) 2021;94:443‐448. doi:10.1111/cen.14401

Sievers C, Klotsche J, Pieper L, Schneider HJ, März W, Wittchen HU, et al. Low testosterone levels predict all‐cause mortality and cardiovascular events in women: A prospective cohort study in German primary care patients. Eur J Endocrinol 2010;163:699‐708. doi:10.1530/EJE‐10‐0307

Zhao D, Guallar E, Ouyang P, Subramanya V, Vaidya D, Ndumele CE, et al. Endogenous sex hormones and incident cardiovascular disease in post‐menopausal women. J Am Coll Cardiol 2018;71:2555‐2566. doi:10.1016/j.jacc.2018.01.083

Benn M, Voss SS, Holmegard HN, Jensen GB, Tybjærg‐Hansen A, Nordestgaard BG. Extreme concentrations of endogenous sex hormones, ischemic heart disease, and death in women. Arterioscler Thromb Vasc Biol 2015;35:471‐477. doi:10.1161/ATVBAHA.114.304821

Wierckx K, Elaut E, Declercq E, Heylens G, De Cuypere G, Taes Y, et al. Prevalence of cardiovascular disease and cancer during cross‐sex hormone therapy in a large cohort of trans persons: A case–control study. Eur J Endocrinol 2013;169:471‐478. doi:10.1530/EJE‐13‐0493

Meyer IH, Brown TNT, Herman JL, Reisner SL, Bockting WO. Demographic characteristics and health status of transgender adults in select US regions: Behavioral Risk Factor Surveillance System, 2014. Am J Public Health 2017;107:582‐589. doi:10.2105/AJPH.2016.303648

Skrivankova VW, Richmond RC, Woolf BAR, Yarmolinsky J, Davies NM, Swanson SA, et al. Strengthening the Reporting of Observational Studies in Epidemiology Using Mendelian Randomization: The STROBE‐MR statement. JAMA 2021;326:1614‐1621. doi:10.1001/jama.2021.18236

Heidari S, Babor TF, De Castro P, Tort S, Curno M. Sex and Gender Equity in Research: Rationale for the SAGER guidelines and recommended use. Res Integr Peer Rev 2016;1:1‐9. doi:10.1136/bmj.m3502

Bycroft C, Freeman C, Petkova D, Band G, Elliott LT, Sharp K, et al. The UK Biobank resource with deep phenotyping and genomic data. NATNEWS 2018;562:203‐209. doi:10.1177/00732753221087558

Suinesiaputra A, Sanghvi MM, Aung N, Paiva JM, Zemrak F, Fung K, et al. Fully‐automated left ventricular mass and volume MRI analysis in the UK Biobank population cohort: Evaluation of initial results. Int J Cardiovasc Imaging 2018;34:281‐291. doi:10.1007/s10554‐017‐1225‐9

Petersen SE, Matthews PM, Francis JM, Robson MD, Zemrak F, Boubertakh R, et al. UK Biobank's cardiovascular magnetic resonance protocol. J Cardiovasc Magn Reson 2016;18:1‐7. doi:10.1186/s12968‐016‐0227‐4

R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna; 2021.

Yavorska OO, Burgess S. MendelianRandomization: An R package for performing Mendelian randomization analyses using summarized data. Int J Epidemiol 2017;46:1734‐1739. doi:10.1093/ije/dyx034

Ruth KS, Day FR, Tyrrell J, Thompson DJ, Wood AR, Mahajan A, et al. Using human genetics to understand the disease impacts of testosterone in men and women. Nat Med 2020;26:252‐258. doi:10.1038/s41591‐020‐0751‐5

Petersen SE, Aung N, Sanghvi MM, Zemrak F, Fung K, Paiva JM, et al. Reference ranges for cardiac structure and function using cardiovascular magnetic resonance (CMR) in Caucasians from the UK Biobank population cohort. J Cardiovasc Magn Reson 2017;19:1‐19. doi:10.1186/s12968‐017‐0327‐9

Shah S, Henry A, Roselli C, Lin H, Sveinbjörnsson G, Fatemifar G, et al. Genome‐wide association and Mendelian randomisation analysis provide insights into the pathogenesis of heart failure. Nat Commun 2020;11:163. doi:10.1038/s41467‐019‐13690‐5

Burgess S, Bowden J, Fall T, Ingelsson E, Thompson SG. Sensitivity analyses for robust causal inference from Mendelian randomization analyses with multiple genetic variants. Epidemiology 2017;28:30‐42. doi:10.1097/EDE.0000000000000559

Burgess S, Davey Smith G, Davies NM, Dudbridge F, Gill D, Glymour MM, et al. Guidelines for performing Mendelian randomization investigations. Wellcome Open Res 2020;4:186. doi:10.12688/wellcomeopenres.15555.2

Burgess S, Thompson SG. Interpreting findings from Mendelian randomization using the MR‐Egger method. Eur J Epidemiol 2017;32:377‐389. doi:10.1007/s10654‐017‐0255‐x

Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 2016;40:304‐314. doi:10.1002/gepi.21965

Sanderson E, Spiller W, Bowden J. Testing and correcting for weak and pleiotropic instruments in two‐sample multivariable Mendelian randomization. Stat Med 2021;40:5434‐5452. doi:10.1002/sim.9133

Pulit SL, Stoneman C, Morris AP, Wood AR, Glastonbury CA, Tyrrell J, et al. Meta‐analysis of genome‐wide association studies for body fat distribution in 694 649 individuals of European ancestry. Hum Mol Genet 2019;28:166‐174. doi:10.1093/hmg/ddy327

Evangelou E, Warren HR, Mosen‐Ansorena D, Mifsud B, Pazoki R, Gao H, et al. Genetic analysis of over 1 million people identifies 535 new loci associated with blood pressure traits. Nat Genet 2018;50:1412‐1425. doi:10.1038/s41588‐018‐0205‐x

Sanderson E, Windmeijer F. A weak instrument F‐test in linear IV models with multiple endogenous variables. J Econom 2016;190:212‐221. doi:10.1016/j.jeconom.2015.06.004

Ouyang P, Vaidya D, Dobs A, Golden SH, Szklo M, Heckbert SR, et al. Sex hormone levels and subclinical atherosclerosis in postmenopausal women: The Multi‐Ethnic Study of Atherosclerosis. Atherosclerosis 2009;204:255. doi:10.1016/j.atherosclerosis.2008.08.037

Rexrode KM, Manson JAE, Lee IM, Ridker PM, Sluss PM, Cook NR, et al. Sex hormone levels and risk of cardiovascular events in postmenopausal women. Circulation 2003;108:1688‐1693. doi:10.1161/01.CIR.0000091114.36254.F3

Horstman AM, Dillon EL, Urban RJ, Sheffield‐Moore M. The role of androgens and estrogens on healthy aging and longevity. Journals Gerontol Ser A 2012;67:1140‐1152. doi:10.1093/gerona/gls068

Samargandy S, Matthews KA, Brooks MM, Barinas‐Mitchell E, Magnani JW, Janssen I, et al. Arterial stiffness accelerates within 1 year of the final menstrual period: The SWAN Heart study. Arterioscler Thromb Vasc Biol 2020;40:1001‐1008. doi:10.1161/ATVBAHA.119.313622

Salerni S, Di Francescomarino S, Cadeddu C, Acquistapace F, Maffei S, Gallina S. The different role of sex hormones on female cardiovascular physiology and function: Not only oestrogens. Eur J Clin Invest 2015;45:634‐645. doi:10.1111/eci.12447

Abildgaard J, Ploug T, Al‐Saoudi E, Wagner T, Thomsen C, Ewertsen C, et al. Changes in abdominal subcutaneous adipose tissue phenotype following menopause is associated with increased visceral fat mass. Sci Rep 2021;11:14750. doi:10.1038/s41598‐021‐94189‐2

Ardissino M, McCracken C, Bard A, Antoniades C, Neubauer S, Harvey NC, et al. Pericardial adiposity is independently linked to adverse cardiovascular phenotypes: A CMR study of 42 598 UK Biobank participants. Eur Heart J Cardiovasc Imaging 2022;23:1471‐1481. doi:10.1093/ehjci/jeac101

Luo S, Au Yeung SL, Zhao JV, Burgess S, Schooling CM. Association of genetically predicted testosterone with thromboembolism, heart failure, and myocardial infarction: Mendelian randomisation study in UK Biobank. BMJ 2019;364:476. doi:10.1136/bmj.l476

Lester JF, Mason MD. Cardiovascular effects of hormone therapy for prostate cancer. Drug Healthc Patient Saf 2015;7:129. doi:10.2147/DHPS.S50549

Pecori Giraldi F, Toja PM, Filippini B, Michailidis J, Scacchi M, Stramba Badiale M, et al. Increased prevalence of prolonged QT interval in males with primary or secondary hypogonadism: A pilot study. Int J Androl 2010;33:e132‐e138. doi:10.1111/j.1365‐2605.2009.00985.x

Herring MJ, Oskui PM, Hale SL, Kloner RA. Testosterone and the cardiovascular system: A comprehensive review of the basic science literature. J Am Hear Assoc Cardiovasc Cerebrovasc Dis 2013;2:271. doi:10.1161/JAHA.113.000271

Gagliano‐Jucá T, Basaria S. Testosterone replacement therapy and cardiovascular risk. Nat Rev Cardiol 2019;16:555‐574. doi:10.1056/NEJMc2213461

Wallis CJD, Lo K, Lee Y, Krakowsky Y, Garbens A, Satkunasivam R, et al. Survival and cardiovascular events in men treated with testosterone replacement therapy: An intention‐to‐treat observational cohort study. Lancet Diabetes Endocrinol 2016;4:498‐506. doi:10.1016/S2213‐8587(16)00112‐1

Getahun D, Nash R, Flanders WD, Baird TC, Becerra‐Culqui TA, Cromwell L, et al. Cross‐sex hormones and acute cardiovascular events in transgender persons: A cohort study. Ann Intern Med 2018;169:205‐213. doi:10.7326/M17‐2785

Nokoff NJ, Scarbro S, Juarez‐Colunga E, Moreau KL, Kempe A. Health and cardiometabolic disease in transgender adults in the United States: Behavioral Risk Factor Surveillance System 2015. J Endocr Soc 2018;2:349‐360. doi:10.1210/js.2017‐00465

Minelli C, Del Greco MF, van der Plaat DA, Bowden J, Sheehan NA, Thompson J. The use of two‐sample methods for Mendelian randomization analyses on single large datasets. Int J Epidemiol 2021;50:1651‐1659. doi:10.1093/ije/dyab084