The primary literature search was conducted on August 27, 2021, and two updated searches on February 2 and July 28, 2023, using four electronic databases: PubMed, Scopus, Web of Science and Embase via OVID. Databases were searched using the title, abstract and full-text fields to find all significant articles. The Patient, Population or Problem, Intervention, Comparison and Outcome (PICO) search strategy was used to categorise the relevant keywords (Supplementary Table S2) (33). The search keywords used included “cardiovascular disease”, “heart arrhythmias”, “stroke”, “cardiometabolic risk”, “hypertension”, “dyslipidaemia”, “cardiomyopathy”, “atherosclerotic heart diseases”, “rheumatic heart disease”, “cerebrovascular disease”, “coronary artery disease”, “heart failure”, “heart valve disease”, “ischaemic heart disease”, “inflammatory heart disease”, “heart disorder”, “cardiac arrest”, “hypertensive heart disease”, “carditis”, “peripheral artery disease”, “myocardial infarction”, “acute coronary syndrome”, “cardiac failure”, “left ventricular systolic dysfunction”, “preeclampsia”, “toxaemia”, “maternal syndrome”, “pregnancy complication”, “pregnancy-specific disorder”, “pregnancy-induced hypertension”, “maternal hypertension”, “eclampsia”, “HELPP syndrome”, “genetic”, “candidate gene studies”, “association analysis”, “linkage studies”, “Genome-wide association studies”, “Genome-wide linkage studies”, “gene-gene interactions”, “gene-environmental interactions”, “epistasis”, “heritability”, “DNA methylation”, “epigenetics”, “microRNA”, “histone modification”, “chromatin modification”, “epigenetic modification”, “Epigenome-wide association studies”, “posttranslational regulation”, “transcriptional gene silencing”, “nucleosome remodelling”, “non-coding RNA regulation” and “RNA editing”. Detailed search keywords are mentioned in the data Supplementary Tables S2–S5.

This systematic review included original articles published in English from-1980 until July 28, 2023, in case-control, cohort, or cross-sectional study designs. We chose post-1980s due to significant advancements in genomic research and mapping techniques. Publications relevant to progression of PE to CVD endpoints (e.g., coronary artery disease, stroke, etc) or CVD risk factors (e.g., systolic blood pressure, cholesterol level) in women based on genetic, epigenetic or transcriptomic factors were included. PE was defined either based on the clinical outcome presented during pregnancy or as a history of PE from self-reported questionnaires answered by the participants. PE is clinically defined by the presence of new-onset hypertension, with blood pressure readings of 140/90 mmHg or higher, taken at least twice, four hours apart, after 20 weeks of gestation. The diagnosis of PE can be further confirmed with the presence of additional criteria such as proteinuria, identified by an excess of proteins in the urine, or maternal organ dysfunction. Maternal organ dysfunction encompasses liver dysfunction, renal insufficiency, neurological complications, or haematological complications (8). The self-reported responses obtained through questionnaires were crosschecked with the medical records of the participants by an Obstetrician/Gynaecologist to ensure the accuracy of the data.

The articles were restricted to humans and peer-reviewed empirical studies. Studies were excluded if: (i) CVD was not evaluated as the outcome; (ii) they were on behavioural CVD risk factors, including diet, physical activity, alcohol consumption, and tobacco use; (iii) they were solely aimed at other pregnancy complications other than PE, such as gestational hypertension, gestational diabetes, stillbirth, small-for-gestational-age; (iv) focused on offspring rather than mothers; (v) based on animal models; (vi) there was a lack of genetic, epigenetic or transcriptomic evidence; (vii) they were systematic reviews, discussion papers, case reports, case series, editorials, or conference abstracts; (viii) focused on a topic not related to CVD or PE; and (ix) different women or different cohorts of CVD and PE were used to study the association between both diseases.

The research articles from all four databases were imported to EndNote 20 to remove duplicates (34). The remaining articles were exported to Rayyan Systematic Review software for further assessment, screening, selection and extraction of data (35). Three reviewers (G.K., T.N., and N.T.) collaborated to screen and select relevant studies. First, reviewers independently conducted a blind assessment of the titles and abstracts of all studies. The eligible articles were then subjected to full-text screening based on the inclusion and exclusion criteria. Finally, disagreements or inconsistencies between the three reviewers were addressed through discussion and consultation with a fourth reviewer (P.M.). The reasons for exclusion were clearly stated at the end of each screening stage.

Data were retrieved from the included studies and recorded on a data extraction table. The significant extracted characteristics included study designs, gene identification approaches, number of variants from genes, polymorphisms or miRNAs, participant demographic data, CVD follow-up time after first index pregnancy, and the definition of PE and CVD outcome measures.

The quality level of each study that reached the final screening stage was independently reviewed by three reviewers (G.K., T.N., and N.T.). The evaluation was conducted using the Newcastle-Ottawa Quality Assessment Scale for case-control, cohort and cross-sectional studies (36). The articles were ranked on a scale of 0–10 and have been further customised for this review (See Supplementary Tables S6–S8). The quality assessment criteria were based on case selection, comparability between cases and controls and outcome or exposure. Most studies had more than one study population with different study designs. Hence, the quality of each of these cohorts has been individually assessed, as they could not be categorised into one study design. The reasons for excluding studies despite having a good quality score were mentioned in the Quality Assessment Supplementary Tables (see Supplementary Tables S6–S11). Any difference of opinion between the three reviewers was resolved through discussion with the fourth reviewer (P.M.).

Three reviewers (G.K., T.N., and N.T.) performed the data extraction using Microsoft Word based on the Cochrane handbook for systematic reviews (2nd edition) (31). Included articles were thoroughly read and classified into two study types: CVD endpoints and CVD risk factors. However, the study design, methodologies, results, and the CVD outcomes analysed differed in each included study. Hence, a meta-analysis was not performed due to cross study heterogeneity.

The Systematic Review Flowchart provides a detailed overview of the selection process (Figure 1). The primary search generated 13,469 records including the articles from the updated search. A total of 9,231 studies were then subjected to the title/abstract screening after eliminating duplicates (N = 4,238). Further, 9,009 additional studies were removed after the eligibility criteria were narrowed to be more specific to the study aim, excluding studies that: (i) were not focusing on both PE and CVD; (ii) were on animal studies; (iii) were systematic and literature reviews; (iv) did not evaluate CVD as the outcome; and (v) lacking genomic evidence. This resulted in the full-text screening of 222 studies. Following the full-text screening, 14 articles were included for quality assessment, excluding 208 studies with any of the above reasons. In addition, the articles which studied other hypertensive disorders of pregnancy and not PE, studies on offspring and those which were not original research were also excluded in the full-text screening stage. After further refinement, eight articles were removed for the following reasons: (i) two studies for using different cohorts for CVD and PE (20, 37); (ii) one article for investigating different women with CVD and a history of PE, although from the same cohort (38); (iii) two for making conclusions on the shared risk of PE and CVD using genes predisposed to CVD from another study (39, 40) and (iv) for not testing for CVD risk (39, 40); (v) two for studying CVD risk in women with PE during delivery, but not conducting any follow-up research on CVD risk (41, 42); and (vi) one study for not mentioning the number of women with PE (19). The included studies consisted of two studies on CVD endpoints following PE (35, 36) and four on CVD risk factors following PE (37–40) (see Supplementary Tables S6–S11). Ultimately, six articles met the study selection criteria.

The quality scores of six included studies (43–48) and the excluded eight studies (19, 20, 37–42) that reached the final screening stage are shown in Supplementary Tables S6–S8. The six studies that passed the final screening were all case-control design. Hence, the New-Castle Ottawa scale for case-control studies was used (maximum score: nine); three studies scored 7.5 (44, 45, 48), one received 8.5 (43) and another obtained 5.5 (47). The final included study consisted of two case-control cohorts, and both were assessed for quality separately (46). Among these, cohort 1 scored 6 and cohort 2 scored 6.5 (46). Four studies had reliable methods of ascertaining PE from medical records or databases (43–45, 48). However, two studies based their PE diagnosis on self-reported questionnaires (46, 47). All studies investigated a variety of genomic factors involved in the progression to CVD or CVD risk factors following PE in the same women from the same cohorts of PE and CVD.

The study characteristics for the CVD endpoints and CVD risk factors studies are summarised in Tables 1, 2, respectively. The articles included consisted of three candidate gene studies (44, 45, 48), two miRNA studies (46, 47), and one epigenetic study investigating whole-genome bisulphite sequencing (43).

The study populations were primarily of European ancestry, with the total number of participants ranging from 4 to 2,945 and maternal ages ranging from 27 to 49.5 years. The follow-up time after the index pregnancy varied from 12 weeks to 22 years. The International Society for the Study of Hypertension in Pregnancy (49), the American College of Obstetricians and Gynaecologists (50) and the Royal College of Obstetricians and Gynaecologists' (51, 52) diagnostic criteria were used in four studies to identify women with PE. The CVD outcome measures included premature acute coronary syndrome (ACS) (46, 47), angina pectoris (48), myocardial infarction (48), cerebral stroke (48) and cardiometabolic risk traits such as blood pressure (43, 45, 48), glucose levels (43, 45, 48), weight (45), BMI (43, 48), waist circumference (43, 45), hip circumference (43, 45), left ventricular ejection fraction (LVEF) (43), left ventricular mass index (43), diastolic function (43), apolipoprotein A (Apo A) (45), apolipoprotein B (Apo B) (45), LDL-C (43, 45), HDL-C (43, 45, 48), total cholesterol (43, 45, 48), triglyceride levels (43) and lipoprotein (a) [Lp(a)] levels (44).

Two studies were included on the progression of PE to CVD endpoints (Tables 1, 3) (46, 47). Both studies focused on circulating miRNAs, their association with PE, and premature ACS, the outcome of the studies (46, 47).

In the first study, Dayan et al. investigated 372 miRNAs in women who developed premature ACS 14.2 years after PE (N = 30) vs. those who had normotensive pregnancies (N = 146) (47). This led to the identification of 16 differentially expressed miRNAs in PE, which consisted of: (i) an increase of 10 miRNAs with a fold change of 1.3–2.0; and (ii) a decrease of 6 miRNAs with a fold change of 1.3–2.8. However, of the 16 miRNAs, only three miRNAs that met the eligibility requirements for evaluations in larger validation cohorts were selected in the study. Moreover, the three chosen miRNAs were also associated with various biological mechanisms involved in CVD risk (47, 53). The mir-1225p was previously associated with hepatic lipid metabolism, miR-126-3p with angiogenesis and miR-146a-5p with anti-inflammation (47, 53). However, in this study, all three miRNAs (miR-122-5; miR-126-3p; and miR-146a-5p) significantly lowered in women with premature ACS and a history of PE, even after adjusted for chronic hypertension (47) (See Tables 1, 3 for more details).

In the second study, 2,578 miRNAs were screened comparing circulating miRNA levels between four cohorts: (i) an ACS cohort with a history of PE (N = 18) vs. normotensive (N = 17); (ii) a non-ACS cohort with a history of PE (N = 20) vs. normotensive (N = 20); (iii) an ACS vs. non-ACS cohort; and (iv) women with PE vs. normotensive without ACS (46). The development of premature ACS varied from 16- and 19-years post PE across the ACS and non-ACS cohorts respectively. Among the 2,578 miRNAs screened, only one miRNA (miR-206) was altered in all four cohorts. However, a history of PE was linked to approximately ten-fold lower plasma levels of miR-206 in women with ACS compared to a history of normotensive pregnancy. This was confirmed in a second cohort of women without ACS, but the change was more moderate at 1.8-fold. Moreover, through miRNA pathway enrichment analysis, Wnt-signalling was identified as the most significantly modified pathway common to PE and ACS. Besides, the most interacting genes with miR-206 in the gene target interaction network were identified as Nuclear Factor of Activated T-cells 5 (NFAT5), Cyclin D2 (CCND2) and Mothers Against Decapentaplegic homolog 2 (SMAD2) (46) (See Tables 1, 3 for more details).

Four case-control studies were included that investigated CVD risk factors (Tables 2, 3) (43–45, 48). Two were candidate gene studies on PE and later-life hypertension (45, 48). The later-life hypertension was considered as the outcome in both studies.

In the first study, four genetic variants from four genes: Aquaporin-3 (AQP3; rs2231231), Aquaporin-7 (AQP7; rs2989924), Nitric oxide synthase 3 (NOS3; 4B/A intron) and Cytochrome B-245 Alpha Chain (CYBA; rs4673) were investigated for their association with later-life hypertension, in a cohort of women with prior PE (N = 48) or who had a previous normotensive pregnancy (N = 98) (45). Previous studies have identified that all four genes included in the analysis play an essential role in redox homeostasis and oxidative stress, which are major components of a metabolic syndrome (a group of risk factors specific to CVD) (45, 54, 55). However, among the four single nucleotide polymorphisms (SNPs), only one intronic SNP rs2231231 from the AQP3 gene, dominant and recessive model of A allele [AA + AC] and [AA] respectively, was associated with PE and the development of hypertension in women 2–16 years post-pregnancy. Moreover, in the study, the AQP3 (rs2231231), [AA + AC] was also linked to a greater risk of CVD in women with a history of PE, as it showed different genotype distributions in four different groups of women. Group 1 included preeclamptic and hypertensive women, group 2 included preeclamptic women and those with normal blood pressure post-pregnancy, group 3 included women with normal blood pressure during pregnancy and hypertensive women, and group 4 included women with normal blood pressure during and post-pregnancy(45) (See Tables 2, 3 for more details).

The second study investigated a single genetic variant (rs4606) in the Regulator of G Protein Signaling 2 (RGS2) gene to check its association with later-life hypertension 15-years following PE. This analysis was conducted in Norwegian PE cases (N = 934) and controls (N = 2,011) (48). Past studies identified that numerous vasoconstrictors are negatively regulated by the regulator of G protein signaling 2 (56). Moreover, CG or GG genotypes of rs4606 in the RGS2 gene have been previously found to be associated with PE women (57). However, this study identified the association of rs4606 [CG, CG + GG] polymorphism with the risk of later-life hypertension and a history of early-onset PE after adjusting for age and BMI. In addition, associations were also identified with the SNP, rs4606 [GG], later-life hypertension and a history of PE and between rs4606[GG, CG + GG], later-life stage 2 hypertension and a history of PE (48) (See Tables 2, 3 for more details).

The third CVD risk factor study is a candidate gene study on PE and Lp(a) levels (44). The Lp(a) levels were considered the study's outcome. Lp(a), which is associated with the plasminogen-like glycoprotein, is a significant risk factor for atherosclerotic CVD, mainly in those with LDL-C or HDL-C (58, 59). Moreover, from previous studies, Lp(a) levels were also observed to be increased in women with a history of PE (60, 61). However, this research was conducted in a cohort with a history of various placenta-mediated pregnancy complications (N = 360), including PE (N = 154), stillbirth (N = 121), and small-for-gestational-age (N = 85) as cases. Healthy women with no history of vascular disorders (conditions that affects blood vessels, e.g., venous thrombosis, Aneurysm) and pregnancy complications were included as control groups. The study investigated the involvement of the two polymorphisms (rs1853021: + 93C > T and rs1800769: 121G > A) in the Lipoprotein A (LPA) gene in modifying Lp(a) levels and placenta-mediated pregnancy complications risk. The Lp(a) levels were analysed 12 weeks post-pregnancy, and it was observed that women with a history of PE and stillbirth had elevated Lp(a) levels. Moreover, as the unfavourable allelic burden of LPA gene elevated, Lp(a) concentrations gradually increased. A similar association of increased risk with PE and stillbirth with Lp(a) levels was identified, although it was not significant (P = 0.06) (See Tables 2, 3 for more details).

The last included article used whole-genome bisulphite DNA methylation study with two identical twin sister pairs discordant for PE (43). This study investigated the epigenomic alterations associated with CVD risk following PE. The twin sister pairs were examined for epigenetic risk two years post-pregnancy. Furthermore, 18 CVD markers were tested between the twin sisters to understand the phenotypic risk 6–12 years post-pregnancy; no differences were observed. However, a genome-wide methylC-sequencing of 22,732 differentially methylated regions (DMRs) revealed 107 DMRs significantly altered in all individuals. Among these, 12 DMRs were found to be shared by the affected twin sisters, with at least half a difference in their methylation percentage and having the same up or down-regulation. These findings were vastly different from those of their unaffected twins (Table 3). These 12 DMRs may be potentially linked to CVD risk following PE, and the genes associated with the regions can be found in Table 3. One of these genes DHX58 was found to be associated with coronary artery disease in another study (62). However, the remaining genes associated with DMRs were linked to CVD mainly in animal studies (63, 64). The authors of this epigenetic study concluded that the ongoing long-term CVD risk in the affected twin sister might be due to the changes in her DNA methylation caused by PE.

In this systematic review, we surveyed the published literature to identify common evidence on shared genomic factors associated with CVD endpoints or risk factors following PE available as of July 28, 2023. Following quality control and screening, we identified six case-control studies longitudinally testing for CVD endpoints or CVD risk factors following PE (43–48). All the included studies were of case-control design and European ancestry. Both studies investigating CVD endpoints focussed on premature ACS following PE using miRNA markers (46, 47). These studies identified four miRNAs (miR-122-5p, -126-3p, -146a-5p, -206) differentially expressed in women with premature ACS following PE and concluded that these findings might offer better insights into biological mechanisms that could be responsible for the elevated risk of CVD post-PE (46, 47).

The CVD risk factors category consisted of: (i) two candidate gene studies (45, 48) on PE and later-life hypertension (ii) one candidate gene study(44) on PE and Lp(a) levels and (iii) an epigenetic study on PE and later-life CVD risk factors (43). The first candidate gene study identified one SNP rs2231231 from the AQP3 gene associated with PE and later-life hypertension (45). The study concluded that as the AQP3 gene was only associated with hypertension post-pregnancy, the role of the gene might be linked to later-life hypertension risk factors including oxidative stress and metabolic syndrome (45). The second candidate gene study also identified another SNP rs4606 from the RGS2 gene associated with PE and later-life hypertension (48). In this analysis, even after accounting for rs4606 SNP and other CVD risk factors, PE continued to be a standalone risk factor for future hypertension (48). Moreover, from another candidate gene study on PE and lipoprotein(a) [Lp(a)] levels, two polymorphisms (rs1853021: + 93C > T and rs1800769: 121G > A) in the LPA gene were observed in modifying Lp(a) levels and placenta-mediated pregnancy complications risk (44). The study detected that, those women with a history of PE and stillbirth had an increased concentration of Lp(a). This research helped in confirming the relationship between pregnancy complications and the atherothrombotic marker, Lp(a) (44). Finally, from the epigenetic study, 12 DMRs associated with CVD risk following PE were identified (43). The study concluded that the whole-genome bisulfite DNA methylation sequencing approach used in this study would help in identifying biomarkers that can be used for early CVD risk stratification for women after a complicated pregnancy (43).

None of the included studies reached the maximum score on the Newcastle-Ottawa Scale, as most studies did not include their non-response rate (Supplementary Tables S6–S11). Thus, the non-response bias that may have existed in the studies could not be calculated. The limited amount of primary research demonstrates a critical gap in the literature that needs to be addressed. Hence, despite the increasing number of publications on PE and its relationship with later-life CVD (19, 20, 65–67), more empirical research is required to identify the genomic factors involved in the progression to CVD after a history of PE using a longitudinal study framework.

While suggestive evidence has also been found using MR and PRS with PE and later-life CVD, more comprehensive research is required to fully understand the causal genetics (30, 68, 69). A genome-wide association study using MR found that hypertensive disorders of pregnancy were associated with an increased risk of coronary artery disease and stroke (68). Conversely, the causal genetic variants associated with PE and later-life CVD remain unclear in the study. In addition, a recent investigation using PRS found a heightened risk of atherosclerotic CVD associated with significant genetic susceptibility for hypertensive disorders during pregnancy. This study highlights the benefit of PRS in predicting long-term CVD outcomes in later life for hypertensive disorders during pregnancy (30). However, this study did not solely focus on PE and its associated genetic risk with later-life CVD.

Several systematic reviews and meta-analyses discuss CVD risk following PE (70–73), but only a small number consider the shared genomic risk factors associated with both diseases (74, 75). In addition, many of these previous systematic reviews make conclusions combining the results between animal and human studies (75). While animal models on PE are well established across different species, the focus on CVD in vivo models is heterogeneous in nature and does not always capture the complexity of how biological processes may evolve. Hence, a standard animal model does not work for CVD, and several animal models or a personalised model would be required for a better understanding, increasing the complexity (76). Thus, this systematic review focused on including studies on genetic, epigenetic or transcriptomic factors associated with women with a history of PE and later-life CVD risk.

A comprehensive analysis was conducted to systematically identify up-to-date studies on genomic loci associated with the progression of PE to CVD. Only studies conducted using the same cohorts of women with CVD and PE, and articles on CVD risk following PE, were included. This would help better identify the overlapping genomic risk loci of both diseases over time without any bias. The scores were given based on the checklist relevant to their study design. Moreover, the criteria for ascertainment of exposure in the Newcastle-Ottawa scale were tailored to incorporate genetic or epigenetic assessments. These customisations in quality assessments have helped thoroughly analyse the studies, irrespective of their study type or design.

This study has certain drawbacks that need to be addressed. First, each included literature had different methodologies, study designs, findings, and analyses of the CVD outcomes. Thus, a meta-analysis could not be performed because of the significant heterogeneity between studies. Second, the sample size of each included study was comparatively small. Third, only articles published in English were included. Fourth, as all included studies were of European ancestry, there was no data relevant to other ethnicities. Finally, we recognise that there are various other women-specific CVD risk factors, such as foetal growth restriction, polycystic ovarian syndrome, menopause, and premature ovarian failure, that can influence the development of CVD and may contribute to the onset of metabolic syndrome (77). However, the primary focus of our review is to analyse the genomic variations associated with the progression of CVD in later life following an incidence of PE. While we acknowledge the significance of other contributing factors, our current review focuses solely on this relationship. Our goal is to provide a comprehensive analysis that delves into the genomic risk association between PE and its progression to later-life CVD.