Healthy lifestyle reduces cardiovascular risk in women with genetic predisposition to hypertensive disorders of pregnancy

Ethics approval

This study complies with all relevant ethical regulations for research involving human participants and was conducted in accordance with the criteria set by the Declaration of Helsinki. The UKBB study was approved by the National Research Ethics Committee (June 17, 2011 [RES Reference¹¹ /NW/0382] and was extended on May 10, 2016 [RES Reference¹⁶ /NW/0274]). The collection, storage, and analyses of biospecimens, genetic data, and data derived from electronic health records as part of the PMBB study is approved under the University of Pennsylvania IRB protocol #813913. Participants from the UKBB and PMBB cohorts provided written informed consent allowing for the use of their samples and data for medical research purposes. The use of de-identified data from these biobanks for this specific research was covered under the existing approvals. All data was handled in accordance with relevant data protection and privacy regulations. No additional ethical approval was required for this specific analysis of existing approved datasets. This study adhered to the requirements of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.

Study population

The UKBB is a large prospective observational cohort study that recruited over 500,000 adults from 22 centers across the United Kingdom between 2006 and 2010. Participants aged 40–69 years were enrolled and have been followed for subsequent health events. The UKBB collected extensive baseline data, including demographics, lifestyle factors, and physical measurements, alongside biological samples for genotyping and biomarker analysis. For the present study, we included women of European descent who had at least one live birth and available genetic data. The full protocol of the UKBB study is publicly available for reference³³.

The PMBB study (i.e., a large-scale academic medical biobank) non-selectively recruited participants from outpatient settings. These participants provided access to their electronic health record data and generated genomic and biomarker data³⁴. All International Classification of Diseases (ICD)-9 and ICD-10 diagnostic codes, clinical imaging, and laboratory measurements up to July 2020 were extracted from the electronic health records. The workflows underlying this study are illustrated in Supplementary Figs. 1 and 2.

Hypertensive disorders of pregnancy and comorbidities

In the UKBB study, female participants provided details about their reproductive history, including parity, during a baseline survey. HDP was defined as gestational hypertension, preeclampsia, eclampsia, or superimposed preeclampsia. This identification was based on self-reports at enrollment or the corresponding ICD codes obtained from primary care or hospital records (Supplementary Table 14). Similarly, in the PMBB study, HDP was defined using relevant ICD codes.

To evaluate HDP risk according to the HDP-PRS, age at first pregnancy and the presence of a disease conferring high risk for HDP before pregnancy were selected as covariates, consistent with the clinical guidelines for high-risk factors for HDP^35,36. According to these guidelines, high HDP-risk diseases include hypertension, diabetes mellitus, and dyslipidemia. The presence of a high-risk HDP disease before pregnancy was determined by either self-reporting or diagnosis with relevant ICD codes for each disease (Supplementary Table 15) that occurred before the first live birth.

Cardiovascular outcomes

To analyze incident ASCVD and its association with HDP-PRS, participants with congenital heart disease were excluded to eliminate the possible association between congenital heart disease and CV outcomes (Supplementary Methods contains the relevant diagnosis codes). Prevalent metabolic comorbidities, including hypertension, diabetes mellitus, and dyslipidemia, were used as adjusting covariates and were ascertained either by self-reporting at enrollment or by ICD codes, as described in Supplementary Methods.

Incident ASCVD was defined as a diagnosis of coronary artery disease, myocardial infarction, ischemic stroke, peripheral artery disease, or aortic aneurysm after enrollment of participants without preexisting CVD. In addition, myocardial infarction was algorithmically defined using the UKBB data. For each new-onset ASCVD considered, participants with a preexisting disease at enrollment were excluded from the analysis. For example, participants with preexisting coronary artery disease at enrollment were excluded from the analysis of new-onset coronary artery disease, which ensured that recurrent coronary artery disease was not erroneously counted as new-onset coronary artery disease.

Variables

During the enrollment process in the UKBB study, participants provided information about their sociodemographic characteristics, health/medical history, and lifestyle/environmental factors through a self-administered touchscreen questionnaire and in-person baseline interviews.

According to the AHA, four factors primarily define lifestyle behaviors; these include current smoking status, obesity, physical activity, and eating habits^37,38. Smoking status was classified as current smoker or non-smoker. Obesity was defined as a BMI ≥ 30 kg/m² according to the World Health Organization international classification. With respect to physical activity, participants were classified as having a healthy lifestyle if they reported more than five days per week of moderate or vigorous activity. Eating habits were defined following the recommendations on dietary priorities for CV health, which categorized common dietary components as fruits, vegetables, whole grains, fish, dairy, refined grains, processed meats, and unprocessed meats. Eating habits were considered healthy if participants adhered to at least half of the dietary recommendations for CV health, as assessed using a food frequency questionnaire³⁹. Collectively, lifestyle behaviors were categorized into three groups—unfavorable (0–1 healthy lifestyle factor)⁴⁰, intermediate (2 healthy lifestyle factors), and favorable ( ≥ 3 healthy lifestyle factors). More detailed descriptions and definitions of the variables considered in lifestyle behaviors can be found in Supplementary Methods.

Metabolic health status was identified according to the presence of the five components of MetS based on criteria from the IDF consensus report⁴¹. Metabolic health status was categorized into three groups—ideal (0–1 MetS factor), intermediate (2–3 MetS factors), and poor (≥ 4 MetS factors). Detailed descriptions and definitions of the variables considered in MetS can be found in Supplementary Table 16 and Supplementary Methods.

In the PMBB cohort, smoking status and obesity (BMI ≥ 30) were restrictively used as variables for the replication analysis.

Genotype data quality control and imputation

Genotyping and quality control (QC) procedures and imputation followed standard practices and were performed using a cohort genotyping platform pair. Further details are provided in Supplementary Methods.

UK Biobank

UKBB samples (version 3; March 2018) were genotyped for more than 800,000 SNPs using either the Affymetrix UK BiLEVE Axiom array or the Affymetrix UKBB Axiom array. After QC and imputation, 164,500 European (White British) female participants were found to be eligible for validation genetic analyses.

Penn Medicine Biobank

The PMBB data consisted of 43,623 samples that were genotyped using a GSA genotyping array. After exclusion, 982 parous women participants with European (non-Hispanic White) ancestry and 1019 parous women participants with African American (non-Hispanic Black) ancestry were considered eligible for the replication analysis.

Polygenic risk score

The HDP-PRS was generated using summary statistics of a large-scale HDP GWAS (13,071 cases and 177,808 controls) from the FinnGen Consortium (Data Freeze R8v4)⁴². The PRS was calculated using the Bayesian polygenic prediction method PRS-CS⁴³. Individual PRSs were determined by applying PLINK version 1.90 using the –score command and were computed from beta coefficients as the weighted sum of the risk alleles⁴⁴. The details of PRS analysis are described in Supplementary Methods.

Statistical analysis

Demographic and clinical characteristics are presented as mean ± SD or number (percentage). Continuous variables were compared using Student’s t test, one-way ANOVA, or the Mann-Whitney U test, as appropriate. Categorical variables were compared using the chi-squared test or Fisher’s exact test, as appropriate.

To evaluate the risk of HDP according to the HDP-PRS, we used a multivariate logistic regression model to evaluate the association between HDP-PRS and HDP. We calculated the OR and 95% CI after adjusting for age at first live birth, BMI, smoking status, first ten PCs of ancestry, and genotyping array type in the multivariate logistic regression model.

In the primary analysis, the association between HDP-PRS and new-onset CV outcomes was examined using the multivariate Cox regression analysis. Adjustments were made for a history of HDP, age at first live birth, BMI, smoking status, first ten PCs of ancestry, and genotype array to calculate the HR and 95% CIs. The ORs and HRs of the PRSs for HDP were used as quantitative variables reported per SD, and the categorical variables were defined as follows: low (< 20%), intermediate (20–80%), high (80–99%), and very high (>99%). Subsequently, we conducted joint association analyses to investigate the interplay between genetic risk, lifestyle, and MetS status. In addition, we performed sensitivity analyses based on the ASCVD subtypes.

For the replication analysis, the impact of maintaining a favorable lifestyle across different genetic risk groups was studied using the chi-square test and Cox regression analysis in an independent PMBB cohort. Among 2001 parous women in the PMBB cohort, cases with new-onset ASCVD were few, but conditional Cox regression analysis was feasible for incident hypertension, considering smoking status and obesity (BMI ≥ 30) as lifestyle variables.

All statistical tests were two-sided, and statistical significance was set at P < 0.05. All statistical analyses were conducted using the R Statistical Software (version 4.1.0; R Foundation for Statistical Computing, Vienna, Austria) and PLINK version 1.90⁴⁴. The details of statistical analyses are described in Supplementary Methods.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.