Preeclampsia, gestational hypertension, and cardiovascular disease risk: a genetic epidemiological study

Abstract

Background and Aims

Observational studies show that women with hypertensive disorders of pregnancy have greater risk of cardiovascular disease later in life. Whether these associations reflect causal pathways is uncertain. This study used genetic epidemiology to explore the causal relevance of preeclampsia and gestational hypertension on cardiovascular disease.

Methods

Two-sample Mendelian randomization (MR) analyses were conducted using summary-level data from FinnGen and from the to date largest consortia for each outcome. One-sample MR analyses were performed using individual-level data from 202 876 White British women from the UK Biobank. Genetic instruments for preeclampsia and gestational hypertension were from the most updated genome-wide association study (n = 20 064 preeclampsia cases; n = 703 117 controls; n = 11 027 gestational hypertension cases; n = 412 788 controls).

Results

In two-sample MR analyses, higher genetic predisposition to preeclampsia was associated with greater risk of ischaemic heart disease [odds ratio 1.20 (95% confidence interval 1.06–1.35)], myocardial infarction [1.29 (1.13–1.47)], stroke [1.23 (1.12–1.35)], ischaemic stroke [1.21 (1.10–1.33)], atrial fibrillation [1.13 (1.01–1.25)], and heart failure [1.11 (1.04–1.20)]. For higher genetic predisposition to gestational hypertension, corresponding odds ratios were 1.21 (1.10–1.33), 1.26 (1.16–1.36), 1.30 (1.23–1.37), 1.24 (1.17–1.32), 1.29 (1.17–1.42), and 1.09 (1.03–1.15). The MR-Egger results did not suggest pleiotropy. One-sample analyses were broadly consistent with the main findings.

Conclusions

Genetic predisposition to hypertensive disorders of pregnancy was associated with greater risk of cardiovascular disease later in life, highlighting the importance of enhanced cardiovascular surveillance in this population.

Structured Graphical Abstract

Structured Graphical Abstract.

An illustration of the aim, methods, and findings of the study. BMI, body mass index; CI, confidence interval; MR, Mendelian randomization.

See the editorial comment for this article ‘Genetics, pre-eclampsia, and later life cardiovascular disease: nature, nurture, inevitable consequence?’, by M. Ardissino et al., https://doi.org/10.1093/eurheartj/ehaf665.

Introduction

Atherosclerotic cardiovascular disease is the most frequent cause of mortality globally with ischaemic heart disease and stroke as main drivers of the cardiovascular disease burden.^1,2 Despite improvements in treatment and prevention, mortality due to atherosclerotic cardiovascular disease is increasing^3,4 especially in European and American middle-aged women.⁵ Hormonal and physiological changes during pregnancy and menopause influence future risk of cardiovascular disease for women.^6,7 Recently, a systematic review and meta-analysis, including 13 million women, revealed that preeclampsia was associated with a two-fold higher risk of cardiovascular disease,⁸ and studies on gestational hypertension reported similar associations.^9,10 Whether these observational associations mark causal pathways warrants further scrutiny.

Hypertensive disorders of pregnancy, including chronic hypertension, gestational hypertension, and preeclampsia, affect around 10% of all pregnancies¹¹ and are the second most common cause of maternal death globally.¹² Hypertensive disorders of pregnancy share risk factors with cardiovascular disease,⁶ making it difficult to determine if they are causal or merely intermediate phenotypes of other cardiovascular risk factors. In the Mendelian randomization (MR) approach, individuals are grouped by genetic variants randomly assorted at conception, thereby minimizing residual observational confounding and reverse causation.¹³ Two previous studies have applied the MR approach to investigate the association between hypertensive disorders of pregnancy and cardiovascular disease by using summary-level¹⁴ and individual-level¹⁵ data. One of these studies used a combined cardiovascular disease endpoint,¹⁵ and the other¹⁴ was conducted before publication of the to date most powerful genome-wide association study (GWAS).¹⁶ As hypertensive disorders of pregnancy pose a threat to women’s health,^8–10 it is timely and necessary to address the causal nature of this female exposure for cardiovascular diseases—several of which are underdiagnosed in women with poorer prognoses than in men.⁵

Therefore, we investigated the association between preeclampsia and gestational hypertension with six cardiovascular outcomes including ischaemic heart disease, myocardial infarction, stroke, ischaemic stroke, atrial fibrillation, and heart failure using the largest currently available GWAS as instruments for one- and two-sample MR analyses and applying a series of sensitivity analyses to address major sources of bias.

Methods

A flowchart of the study design is shown in Figure 1.

Figure 1.

Flowchart of study design. Flowchart of the study design for two- and one-sample Mendelian randomization analyses conducted in this study. AF, atrial fibrillation; GH, gestational hypertension; GRS, genetic risk score; GWAS, genome-wide association study; HF, heart failure; IHD, ischaemic heart disease; MI, myocardial infarction; MR, Mendelian randomization; PE, preeclampsia; PRESSO, Pleiotropy RESidual Sum and Outlier

Selection of genetic instruments

The selected genetic instrumental variables were single-nucleotide polymorphisms (SNPs) at genome-wide significance level (P < 5 × 10⁻⁸) from meta-analyses for preeclampsia/eclampsia and gestational hypertension in the to date largest GWAS of hypertensive disorders during pregnancy.¹⁶ The GWAS comprised data from FinnGen, Estonia Biobank, Genes & Health, Michigan Genomics Initiative, Mass General Brigham Biobank, BioBank Japan, BioMe, InterPregGen consortium, HUNT, Penn Biobank, UK Biobank, and nuMoM2b (n = 20 064 preeclampsia cases; n = 703 117 controls; n = 11 027 gestational hypertension cases; n = 412 788 controls) of primarily European ancestry (>80%).¹⁶ The number of identified variants in the GWAS was 14 for preeclampsia and 8 for gestational hypertension (see Supplementary data online, Table S1). Information on genes, a detailed description of the phenotypes associated with the variants, and a summary of these phenotypic traits are given in Supplementary data online, Tables S2–S4. One variant (rs9855086), associated with both preeclampsia and gestational hypertension, was palindromic and excluded from all analyses. A second preeclampsia variant (rs1421085—a variant in the FTO gene with a well-established association with body mass index¹⁷) was significantly associated with potential confounding factors, including body mass index and diabetes, and therefore excluded from all analyses. Hence, the number of genetic variants suitable for inclusion was 12 for preeclampsia and 7 for gestational hypertension. The final number of included SNPs in each analysis further varied according to data availability (see Supplementary data online, Table S5). All analyses were first conducted with all available SNPs. Hereafter, analyses were repeated with a restricted genetic instrument, where SNPs associated with the outcome of interest were excluded to avoid possible horizontal pleiotropy (see Supplementary data online, Table S6).

The selected variants had F-statistics (beta/se)² ranging from 24 to 58 (see Supplementary data online, Table S1).

Two-sample Mendelian randomization

Publicly available summary-level data for the six cardiovascular outcomes from FinnGen (https://www.finngen.fi/en/access_results) and from the to date largest consortia were used, including ischaemic heart disease,¹⁸ myocardial infarction,¹⁹ stroke,²⁰ ischaemic stroke,²⁰ atrial fibrillation and flutter,²¹ and heart failure²² (see Supplementary data online, Table S7 for a summarized description). Since the summary-level data include both women and men, comparisons between β-coefficient estimates in a women-only and a combined-sex dataset were conducted (see Supplementary data online, Table S8).

Firstly, estimates from the effect alleles of the instrumental variables on the exposures and outcomes were harmonized and palindromic instrumental variables were removed. The primary analysis was the inverse-variance weighted (IVW) conducted in FinnGen and in each consortium individually and subsequently compiled in a fixed-effect meta-analysis for each outcome. Furthermore, several sensitivity analyses were performed to evaluate the MR assumptions and the strength of the estimates. The weighted median estimator was performed, which can provide a reliable estimate if at least half of the instrumental variables are valid. The MR-Egger method was conducted, where the slope represents the estimate of the causal effect of the exposure on the outcome and the intercept represents the magnitude of the pleiotropic effects across the instrumental variables. The Mendelian Randomization Pleiotropy RESidual Sum and Outlier (MR-PRESSO) method was performed for outlier detection and thereby corrected for horizontal pleiotropy. Leave-one-out analyses were conducted to investigate if results were driven by a single genetic variant.

One-sample Mendelian randomization

One-sample MR analyses were additionally performed to validate the results from the two-sample MR analyses. We included 202 876 non-related White British women from the UK Biobank in the one-sample MR study. Preeclampsia or gestational hypertension and the six cardiovascular outcomes were defined according to International Classification of Diseases codes (see Supplementary data online, Tables S9 and S10). For each participant, a genetic risk score (GRS), weighted by the associations of the genetic instrumental variables from the GWAS,¹⁶ was calculated. To test the association between the GRSs and the exposures, the GRSs were categorized in three groups by tertiles, and frequencies of preeclampsia and gestational hypertension were calculated within each tertile. The continuous GRSs were then used to calculate the GRS–exposure and GRS–outcome associations using logistic regressions adjusted for age at baseline, genotyping batch, and the 10 first principal components. Subsequently, the MR estimates were obtained using the Wald ratio method dividing GRS–outcome associations with GRS–exposure associations. Since data from the UK Biobank were used in the GWAS to additionally identify SNPs associated with preeclampsia and gestational hypertension, the one-sample MR analyses were repeated with exclusion of these SNPs from the GRS to avoid sample overlap. Furthermore, the main analyses were repeated only including women who experienced a normal pregnancy in the control group.

Power calculations

The variations explained by exposures (R²) were calculated using the Mangrove package in R,²³ setting the prevalence of 3% for preeclampsia^11,24 and 5% for gestational hypertension,^11,24 respectively. Statistical power was calculated using all available SNPs via the tool for binary outcomes for MR studies (https://github.com/kn3in/mRnd),²⁵ where the alpha level was set to 0.05.

In the two-sample settings, we have sufficient power (80%) to detect a very small effect [odds ratio (OR) < 1.15] of either preeclampsia or gestational hypertension on all outcomes, owing to the large sample sizes from the largest genomic consortia and the strength of instruments (see Supplementary data online, Figure S1). In the one-sample setting, we had 80% power to detect an effect ranging from 1.25 to 1.47 of preeclampsia and 1.24 to 1.45 of gestational hypertension across different outcomes, which was expected as two-sample MR gives stronger statistical power (see Supplementary data online, Figure S2).

Mediation analyses

We performed genetic mediation analyses using the individual-level data from the UK Biobank to investigate potential mediation effects by systolic blood pressure, body mass index, and Type 2 diabetes in the association of preeclampsia and gestational hypertension with ischaemic heart disease. We first conducted multivariable MR including potential mediators and compared the adjusted association of the exposure on the outcome with the unadjusted association. An attenuation of the association following adjustment was interpreted as evidence that the included variable may act as a mediator. To quantify the mediation effect, we then applied the product of coefficients method in a two-step MR framework using two-stage least squares regression.²⁶ In the first step, a univariable MR was conducted to estimate the effect of the exposure on the mediator (${\beta }_{X\mathrm{\_}M}$), and in the second step, the effect of the mediator on the outcome (${\beta }_{M\mathrm{\_}Y}$) was estimated using multivariable MR including GRS for both the exposure and the mediator in the first and second stage regressions. The indirect effect was then calculated by multiplying the effect estimates from the two steps (${\beta }_{X\mathrm{\_}M}\times {\beta }_{M\mathrm{\_}Y}$), and the proportion mediated by the mediator of interest was retrieved by dividing the indirect effect by the total effect $\left(\frac{{\beta }_{X\mathrm{\_}M}\times {\beta }_{M\mathrm{\_}Y}}{{\beta }_{X\mathrm{\_}Y}}\right)$. Finally, 95% confidence intervals (CIs) of the mediated proportion were calculated using bootstrapping with 1000 repeats.

All statistical analyses were conducted in R statistical software, version 4.3.0 (The R Foundation for Statistical Computing, Vienna, Austria).

Results

Two-sample Mendelian randomization

Odds ratios and 95% CI for associations between preeclampsia or gestational hypertension and six cardiovascular outcomes using the IVW method are shown in Figure 2. Higher genetic predisposition to preeclampsia was associated with greater risk of ischaemic heart disease [OR 1.20 (95% CI 1.06–1.35)], myocardial infarction [OR 1.29 (95% CI 1.13–1.47)], stroke [OR 1.23 (95% CI 1.12–1.35)], ischaemic stroke [OR 1.21 (95% CI 1.10–1.33)], atrial fibrillation [OR 1.13 (95% CI 1.01–1.25)], and heart failure [OR 1.11 (95% CI 1.04–1.20)]. Correspondingly, higher genetic predisposition to gestational hypertension was associated with ischaemic heart disease [OR 1.21 (95% CI 1.10–1.33)], myocardial infarction [OR 1.26 (95% CI 1.16–1.36)], stroke [OR 1.30 (95% CI 1.23–1.37)], ischaemic stroke [OR 1.24 (95% CI 1.17–1.32)], atrial fibrillation [OR 1.29 (95% CI 1.17–1.42)], and heart failure [OR 1.09 (95% CI 1.03–1.15)]. The results of the analyses after stricter inclusion of genetic variants are shown in Figure 3. Associations remained significant between preeclampsia and myocardial infarction, ischaemic stroke, and heart failure and between gestational hypertension and all cardiovascular outcomes. Scatterplots of SNP–outcome and SNP–exposure associations for preeclampsia and gestational hypertension in the consortia and FinnGen data are shown in Supplementary data online, Figures S3–S6. Results from sensitivity analyses for preeclampsia and gestational hypertension are given in Supplementary data online, Tables S11 and S12, respectively. Estimates from the weighted median estimator and MR-PRESSO method did not differ substantially from the primary IVW analyses. Furthermore, no significant pleiotropic effects were detected from the MR-Egger analyses (P-values from MR-Egger intercept ≥ 0.1). Leave-one-out analyses for preeclampsia and gestational hypertension in consortia and FinnGen data are shown in Supplementary data online, Figures S7–S10, suggesting that the overall associations were not driven by a particular SNP.

Figure 2.

Two-sample Mendelian randomization analyses for preeclampsia and gestational hypertension. Forest plot showing odds ratios (95% confidence intervals) for associations between preeclampsia or gestational hypertension and six cardiovascular outcomes. The genetic variant rs1421085 was excluded from the analyses to avoid confounding of results. Analyses were conducted using publicly available summary-level data. CI, confidence interval; OR, odds ratio; SNP, single nucleotide polymorphism

Figure 3.

Two-sample Mendelian randomization analyses for preeclampsia and gestational hypertension ‘strict version’. Forest plot showing odds ratios (95% confidence intervals) for associations between preeclampsia or gestational hypertension and six cardiovascular outcomes. All genetic instruments associated with the outcome were excluded from the analyses to avoid any horizontal pleiotropic effect. Analyses were conducted using publicly available summary-level data. CI, confidence interval; OR, odds ratio; SNP, single nucleotide polymorphism

One-sample Mendelian randomization

Baseline characteristics from participants in the UK Biobank are shown in Table 1. The median age was 37 years at exposure diagnosis for both preeclampsia and gestational hypertension and ranged from 66 to 72 years for diagnosis of any of the six outcomes. Supplementary data online, Table S13 shows median ages for time at exposure and outcome diagnoses stratified by GRS tertiles. Frequency plots showing the proportion of exposure cases according to GRS tertiles are displayed in Supplementary data online, Figure S11. Odds ratios (95% CI) for associations between exposures and outcomes using the Wald ratios are shown in Figure 4. In Model 1, all eligible SNPs are included in the analyses, and in Model 2, SNPs with possible horizontal pleiotropic effects are excluded from the GRS calculations. Higher genetic predisposition to preeclampsia was associated with greater risk of ischaemic heart disease [OR 1.26 (95% CI 1.11–1.43)], myocardial infarction [OR 1.26 (95% CI 0.99–1.60)], stroke [OR 1.28 (95% CI 1.04–1.57)], ischaemic stroke [OR 1.37 (95% CI 1.08–1.74)], atrial fibrillation [OR 1.05 (95% CI 0.89–1.22)], and heart failure [OR 1.12 (95% CI 0.89–1.42)]. Correspondingly, higher genetic predisposition to gestational hypertension was associated with ischaemic heart disease [OR 1.17 (95% CI 1.07–1.28)], myocardial infarction [OR 1.16 (95% CI 0.98–1.36)], stroke [OR 1.35 (95% CI 1.17–1.56)], ischaemic stroke [OR 1.36 (95% CI 1.15–1.60)], atrial fibrillation [OR 1.16 (95% CI 1.04–1.29)], and heart failure [OR 1.22 (95% CI 1.04–1.43)]. When excluding all potentially pleiotropic SNPs from the GRS, associations for preeclampsia were null. Results for gestational hypertension remained similar and were only slightly attenuated. Results from analyses without sample overlap from the discovery GWAS are shown in Supplementary data online, Figure S12. These results are similar to the primary analyses and strengthen the estimates for most outcomes. Finally, exclusion of the never pregnant women from the control group (n = 166 421) gave similar results as the main analyses (see Supplementary data online, Table S14).

Table 1.

Baseline characteristics for participants in the UK Biobank

	No hypertensive disorder	Gestational hypertension	Preeclampsia
N	202 298	374	257
Age at recruitment, years	59 (51–63)	43 (42–46)	44 (42–46)
Body mass index, kg/m2	26 (23–29)	27 (24–31)	26 (23–30)
Education, years	13 (10–20)	13 (10–20)	15 (10–20)
Number of live births	2 (1–2)	2 (1–2)	2 (1–2)
Age at first live birth	25 (22–28)	32 (29–35)	33 (29–35)
Menopause	128 479 (0.6)	22 (0.06)	17 (0.07)
Smoking status
Current smoker	17 156 (8)	20 (5)	15 (6)
Previous smoker	64 643 (32)	97 (26)	75 (29)
Never smoker	119 797 (59)	257 (69)	167 (65)
Alcohol consumption
Daily/almost daily	34 474 (17)	40 (11)	32 (12)
1–4 times per week	96 640 (48)	192 (51)	137 (53)
<1 time per week	54 669 (27)	123 (33)	75 (29)
Never	16 380 (8)	18 (5)	13 (5)
Physical activity
Low	28 933 (18)	76 (23)	43 (20)
Moderate	67 328 (43)	147 (44)	95 (44)
High	60 403 (39)	109 (33)	78 (36)
Systolic blood pressure, mmHg	134 (122–148)	134 (121–147)	129 (118–140)
Diastolic blood pressure, mmHg	81 (74–87)	85 (78–92)	83 (76–90)
LDL cholesterol, mmol/L	3.6 (3.0–4.2)	3.3 (2.8–3.8)	3.3 (2.9–3.7)
Triglycerides, mmol/L	1.36 (0.98–1.92)	1.21 (0.84–1.71)	1.18 (0.83–1.71)
Apolipoprotein B, g/L	1.03 (0.87–1.19)	0.96 (0.82–1.10)	0.93 (0.84–1.08)
Diabetes mellitus	7173 (0.04)	18 (0.06)	10 (0.04)
Chronic hypertension	98 249 (49)	178 (48)	100 (39)
Antihypertensive therapy	36 579 (18)	49 (13)	32 (12)

Categorical variables are numbers (%), and continuous variables are median (25th–75th percentiles).

Figure 4.

One-sample Mendelian randomization analyses for preeclampsia and gestational hypertension. Forest plot showing odds ratios (95% confidence intervals) for associations between preeclampsia or gestational hypertension and six cardiovascular outcomes. Analyses were conducted using individual-level data from the UK Biobank. In Model 1, rs1421085 was excluded from all analyses to avoid confounding of results, and in Model 2, genetic variants directly associated with the outcome were further excluded from all analyses to avoid any horizontal pleiotropic effects. The number of cases in the analyses was 257 for preeclampsia and 374 for gestational hypertension. CI, confidence interval; OR, odds ratio

Mediation analyses

Results from genetic mediation analyses using the UK Biobank data are shown in Figure 5. We applied two different strategies, a multivariable MR analysis, where attenuation of the overall association following adjustment was indicative of mediation through the adjustment factor, as well as the product of coefficients methods in a two-step MR framework to quantify the mediated fraction. In the multivariable MR analysis, the association between genetic predisposition to preeclampsia and risk of ischaemic heart disease adjusted for genetically predicted systolic blood pressure was null, whereas adjusting for genetically predicted body mass index or Type 2 diabetes attenuated the effect size. From the two-step MR framework, the proportion mediated by systolic blood pressure, body mass index and Type 2 diabetes were 77% (95% CI 6%–347%), 13% (95% CI 2%–39%), and 16% (95% CI −6% to 68%), respectively. Results for gestational hypertension from the multivariable MR analysis were similarly null when adjusting for genetically predicted systolic blood pressure and did not attenuate when adjusting for genetically predicted body mass index or Type 2 diabetes. Results for gestational hypertension from the two-step MR framework were 120% (95% CI 47%–393%) for systolic blood pressure, −2% (95% CI −11% to 5%) for body mass index, and 6% (95% CI −5% to 21%) for Type 2 diabetes.

Figure 5.

Genetic mediation analyses for preeclampsia and gestational hypertension and the risk of ischaemic heart disease. Forest plot showing odds ratios (95% confidence intervals) from the multivariable Mendelian randomization analyses. Effect estimates are given for the association between the exposure and outcome in crude estimates, only adjusted for age at baseline and the 10 first principal components, and estimates with adjustment for systolic blood pressure, body mass index, and Type 2 diabetes, respectively. The proportion mediated is calculated from the two-step Mendelian randomization framework and given in percentages. All analyses were conducted using individual-level data from the UK Biobank. CI, confidence intervals; MR, Mendelian randomization; OR, odds ratio; SD, standard deviation

Discussion

The principal finding of this study was that genetic predisposition to preeclampsia or gestational hypertension was associated with greater risk of developing cardiovascular disease, providing evidence to support causal relevance of this association (Structured Graphical Abstract).

A comprehensive meta-analysis of 13 million women recently demonstrated the strong observational association between hypertensive disorders of pregnancy and cardiovascular disease⁸ and a nationwide registry study reported an association with vascular dementia.²⁷ Observational mediation analyses have suggested that a large proportion of these associations are explained by later development of established cardiovascular risk factors including chronic hypertension, hypercholesterolaemia, changes in body mass index, and Type 2 diabetes, especially in women with previous gestational hypertension.²⁸ On the other hand, parts of the association between hypertensive disorders of pregnancy and cardiovascular disease appear to be independent of traditional cardiovascular risk factors.²⁹ Observational associations and observational mediation analyses will, however, always be prone to bias by confounding and reverse causation, in contrast to genetic association and MR studies, where such issues are largely mitigated.

The present study is the to date largest to investigate genetic associations between hypertensive disorders of pregnancy and cardiovascular disease to dissect potential causal pathways. In summary-level data two-sample MR analyses, we observed significant associations for both preeclampsia and gestational hypertension with all six cardiovascular outcomes. These results were attenuated but remained robust in all sensitivity analyses and even after very conservative exclusions of potential pleiotropic genetic variants. In individual-level data one-sample analyses, these associations were confirmed for gestational hypertension; however, for preeclampsia, associations were only significant with the less strict selection of genetic instruments, most likely due to lack of statistical power. Our findings are in alignment with a recent individual-level data MR study, which showed associations for preeclampsia and gestational hypertension with a combined endpoint of myocardial infarction, stroke, and ischaemic stroke,¹⁵ and which further observed that men and nulligravida women carrying the genetic variants associated with hypertensive disorders of pregnancy also had increased risk of cardiovascular disease, despite lack of the exposure.¹⁵ A sex-stratified phenome-wide association analysis between preeclampsia and gestational hypertension polygenic risk scores and hypertension, hypercholesterolaemia, Type 2 diabetes, and obesity in both men and women further supported that cardiovascular risk factors are major contributors to the cardiovascular disease burden with or without hypertensive disorders of pregnancy.¹⁶ We now present robust results in both individual-level (one-sample) and summary-level (two-sample) data MR analyses and further show associations with atrial fibrillation and heart failure. We additionally tested vascular dementia as an outcome in one-sample individual-level data; however, likely due to limited statistical power, no genetic association was observed, in contrast to previous robust observational data.²⁷ One other previous two-sample MR study, with less statistical power than the present analysis, also reported significant genetic associations for preeclampsia and gestational hypertension with coronary heart disease and for the combined hypertensive disorders of pregnancy with stroke.¹⁴ Compared with this study, we included more genetic variants for gestational hypertension, all our selected variants were at genome-wide significance level, six specific cardiovascular disease endpoints were analysed, and we investigated the associations using both one- and two-sample MR, including robust sensitivity analyses adequately addressing horizontal pleiotropy and weak instrument and outlier bias. The mediation analyses collectively indicate that the associations between preeclampsia and gestational hypertension with ischaemic heart disease are largely driven by hypertension, whereas body mass index and Type 2 diabetes also seem to mediate substantial parts of the association between preeclampsia and ischaemic heart disease. A previous study supported our findings on systolic blood pressure and Type 2 diabetes using a multivariable MR analysis approach.¹⁴ We chose to apply two complementary genetic mediation strategies to ensure that the potential mediating factors, systolic blood pressure, Type 2 diabetes, and body mass index all could be evaluated, either by attenuation of the overall association by adjusting for the mediating factors or by quantifying the proportion mediated. We suggest this two-pronged mediation strategy as the most robust setting in the present study because the proportion mediated strategy for systolic blood pressure likely would be grossly overestimated as systolic blood pressure also is a criterion in defining the exposure.

While preeclampsia pathogenesis is not fully understood, it is well described that the placenta plays an important pathophysiological role. Inadequate invasion of trophoblast cells in the uterine wall leads to formation of dysfunctional spiral arteries in the placenta, which causes reduced uteroplacental blood flow.^30,31 Downstream of the stressed placenta toxic factors, such as anti-angiogenic factors and cytokines, is released to the maternal circulation, which leads to inflammation, endothelial dysfunction, and systemic manifestation in the pregnant woman.^30,31 Some of the genetic variants used in the present MR study underline this pathophysiology, as the genes, they are located in, are related to angiogenesis (ZNF831, FLT1) and immune function (SH2B3, MICA).¹⁶ Other genetic variants are located in genes regulating natriuretic peptides (NPR3, MTHFR-CLCN6) indicating impaired signalling as a pathophysiological reason for hypertensive disorders of pregnancy.¹⁶ Not surprisingly, many of the genetic variants are associated with blood pressure since elevated blood pressure is a central part of the manifestation of both preeclampsia and gestational hypertension. However, since elevated blood pressure has both been described as a risk factor for developing preeclampsia or gestational hypertension during pregnancy³² and as a long-term effect of preeclampsia and gestational hypertension,^14,15,33 it is difficult to decide whether it should be regarded as a causal risk factor, mediator, or consequence. Since elevated blood pressure is part of the definition of hypertensive disorders of pregnancy, we found it most plausible to view it as vertical pleiotropy, i.e. being part of the biological pathway, in our analyses. Furthermore, our sensitivity analyses showed no indications of horizontal pleiotropy—a form of bias that can confound MR studies by going through other unmeasured exposures than the one in question. Throughout the analyses, we excluded the rs1421085 variant, located in the FTO gene, which associate with preeclampsia. This genetic variant is associated with blood pressure, increased body mass index, obesity, and Type 2 diabetes.³² Although these factors have been described as risk factors for developing preeclampsia and therefore could be on the causal pathway to cardiovascular disease, the variant was excluded to avoid any potential horizontal pleiotropy.

The strengths of this study include the use of both individual-level data, with detailed information on each participant, and summary-level data, where statistical power is high due to large genetic consortia, and lastly the use of robust sensitivity analyses to validate the findings. There are, however, important limitations to the study. The one-sample MR analyses were restricted to participants of European ancestry to minimize confounding from population stratification. Therefore, the findings might not apply to populations of other ethnicities. The prevalence of hypertensive disorders of pregnancy varies substantially globally with the highest prevalence in Africa and lowest prevalence in the Western pacific (335 per 100 000 vs 16 per 100 000).³⁴ Both genetic and socioeconomic differences contribute to the differences in prevalences, and more research is needed in non-European ancestries to entangle these differences. When conducting the one-sample MR analyses on the restricted panel of genetic variants to avoid pleiotropy, associations for preeclampsia became null, whereas associations for gestational hypertension were only attenuated and for most outcomes remained significant. It must however be emphasized that in the two-sample restricted panel MR analysis—the analysis with the highest statistical power—the associations remained significant for myocardial infarction, ischaemic stroke, and heart failure for both preeclampsia and gestational hypertension. Furthermore, the two-sample analyses are conducted on summary-level data including both women and men, and therefore analyses could not be restricted to women only nor to pregnant women. However, since β-coefficient estimates for the genetic instruments are similar for men and women, and since results were similar in the one-sample analyses including only women, the findings seem robust. One-sample analyses restricted to women with a previous pregnancy compared with analyses including all women showed similar results. Additionally, there is sample overlap in the two-sample MR analyses since data from FinnGen, UK Biobank, and HUNT were included in both the preeclampsia GWAS and in some genetic consortia endpoint data. In the one-sample MR analyses, removing SNPs with potential bias due to sample overlap did not alter the results—if anything the estimates were strengthened.³⁵ Finally, the mediation analyses might be affected by non-collapsibility since the proportion mediated is calculated based on ORs. The association between both exposures and ischaemic heart disease disappeared in multivariable MR mediation analyses adjusting for systolic blood pressure, suggesting that systolic blood pressure is driving the associations. The calculated mediated portion supported this; however, the exact percentages of the proportion mediated for systolic blood pressure should be interpreted with caution.

In conclusion, this study indicates that genetic predisposition to preeclampsia and gestational hypertension is associated with greater risk of cardiovascular diseases later in life, providing evidence to support the causal relevance of hypertensive disorders of pregnancy on cardiovascular disease. These findings suggest a requirement for heightened cardiovascular disease surveillance in women with a history of hypertensive disorders of pregnancy, the timing and format of which will require investigation in interventional studies.

Supplementary data

Supplementary data are available at European Heart Journal  online.

Declarations

Disclosure of Interest

S.T.N., J.L., A.T.-H., K.I., and J.Q.T. have nothing to declare. H.B. reports talks sponsored by Amgen, Sanofi, MSD, and BMS. R.F.-S. reports consultancies or talks sponsored by Novo Nordisk.

Data Availability

Data from the UK Biobank could be obtained upon request to the board. Summary-level data from FinnGen are available on https://www.finngen.fi/en/access_results and from consortia on https://www.ebi.ac.uk/gwas/.

Funding

This work was supported by Snedkermester Sophus Jacobsen og Hustru Astrid Jacobsens Fond to S.T.N. and the Lundbeck Foundation, the Leducq Foundation, and the Research Fund at Sygeforsikringen Danmark, all to R.F.-S. The funding organizations played no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the report; or in the decision to submit the report for publication.

Ethical Approval

The UK Biobank has approval from the North West Multi-centre Research Ethics Committee (MREC).

Pre-registered Clinical Trial Number

None supplied.