Adverse Pregnancy Outcomes and Incident Heart Failure in the Women’s Health Initiative

Key Points

Question  Are adverse pregnancy outcomes independently associated with the development of heart failure among postmenopausal women?

Findings  In this cohort study including 10 292 Women’s Health Initiative participants, hypertensive disorders of pregnancy were independently associated with incident heart failure, particularly heart failure with preserved ejection fraction, in postmenopausal women.

Meaning  These findings suggest that hypertensive disorders of pregnancy are sex-specific factors associated with risk of heart failure, particularly heart failure with preserved ejection fraction.

Importance  Some prior evidence suggests that adverse pregnancy outcomes (APOs) may be associated with heart failure (HF). Identifying unique factors associated with the risk of HF and studying HF subtypes are important next steps.

Objective  To investigate the association of APOs with incident HF overall and stratified by HF subtype (preserved vs reduced ejection fraction) among postmenopausal women in the Women’s Health Initiative (WHI).

Design, Setting, and Participants  In 2017, an APO history survey was administered in the WHI study, a large multiethnic cohort of postmenopausal women. The associations of 5 APOs (gestational diabetes, hypertensive disorders of pregnancy [HDP], low birth weight, high birth weight, and preterm delivery) with incident adjudicated HF were analyzed. In this cohort study, the association of each APO with HF was assessed using logistic regression models and with HF subtypes using multinomial regression, adjusting for age, sociodemographic characteristics, smoking, randomization status, reproductive history, and other APOs. Data analysis was performed from January 2020 to September 2021.

Exposures  APOs (gestational diabetes, HDP, low birth weight, high birth weight, and preterm delivery).

Main Outcomes and Measures  All confirmed cases of women hospitalized with HF and HF subtype were adjudicated by trained physicians using standardized methods.

Results  Of 10 292 women (median [IQR] age, 60 [55-64] years), 3185 (31.0%) reported 1 or more APO and 336 (3.3%) had a diagnosis of HF. Women with a history of any APO had a higher prevalence of hypertension, diabetes, coronary heart disease, or smoking. Of the APOs studied, only HDP was significantly associated with HF with a fully adjusted odds ratio (OR) of 1.75 (95% CI, 1.22-2.50), and with HF with preserved ejection fraction in fully adjusted models (OR, 2.06; 95% CI, 1.29-3.27). In mediation analyses, hypertension explained 24% (95% CI, 12%-73%), coronary heart disease 23% (95% CI, 11%-68%), and body mass index 20% (95% CI, 10%-64%) of the association between HDP and HF.

Conclusions and Relevance  In this large cohort of postmenopausal women, HDP was independently associated with incident HF, particularly HF with preserved ejection fraction, and this association was mediated by subsequent hypertension, coronary heart disease, and obesity. These findings suggest that monitoring and modifying these factors early in women presenting with HDP may be associated with reduced long-term risk of HF.

Women account for most cases of heart failure (HF) with preserved ejection fraction (HFpEF).¹ Approximately 85% of US women experience pregnancy and childbirth, and up to 30% of pregnancies are complicated by 1 or more adverse pregnancy outcomes (APOs).² Several APOs have been associated with a higher risk of developing cardiovascular disease (CVD), including gestational diabetes (GD), hypertensive disorders of pregnancy (HDP), preterm delivery (PTD), low birth weight (LBW), and high birth weight (HBW).^3-6

Prior studies^7-10 suggest that preeclampsia, gestational hypertension, and GD may be associated with an increased risk of developing HF. However, prior studies have neither jointly considered the associations of multiple APOs with HF nor distinguished between HFpEF and HF with reduced ejection fraction (HFrEF) because of limited HF phenotyping.^7,8,11,12 In addition, mediators of these associations have not been robustly explored to date. Given the availability of both reproductive data and adjudicated HF outcomes, the Women’s Health Initiative (WHI) is a unique resource to test the individual and joint associations between APOs and HF.

The WHI is a longitudinal study of ethnically diverse postmenopausal women aged 50 to 79 years at entry, recruited from 40 US clinical centers between 1993 and 1998, and followed prospectively since enrollment for multiple outcomes. Details of recruitment, baseline questionnaires, and examinations performed have been described elsewhere.^13,14 Briefly, women participated in 1 or more of 3 clinical trials (of hormone therapy, dietary modification, and calcium or vitamin D supplementation) or enrolled in an observational study.

This cohort study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines and was approved by the University of California San Francisco institutional review board. All participants gave written informed consent to participate in WHI and its extension study.

Of the 161 808 women in the WHI cohort, a subset of 44 174 participants were included in the incident HF physician adjudication subcohort. The present study was based on women in the HF subcohort who completed the APO survey and were free of HF at entry into WHI (baseline) (Figure 1).

In 2017, a follow-up survey was sent to all surviving WHI participants.³ The survey included 6 questions on APOs during any pregnancy, with possible responses of no, yes, and do not know (eFigure 1 in the Supplement). APOs surveyed included GD, preeclampsia, gestational hypertension, PTD (<37 weeks gestation), LBW (<5.5 lb [2500 g]) or HBW (>9 lb 14 oz [4500 g]). More than 1 APO may have occurred in the same woman, but not necessarily during the same pregnancy. Any history of APO was defined as a participant reporting 1 or more APOs. Preeclampsia or eclampsia and gestational hypertension were combined into a single HDP variable because of substantial overlap in responses and similar point estimates with respect to their association with the outcome (ie, HF), as in prior published analyses.³

Participants in the HF subcohort were included in our study population if they completed the APO survey, had a history of pregnancy lasting for more than 6 months, and were alive and still participating in the WHI at the time of the survey. Nonresponders were defined as eligible WHI participants in the HF subcohort who did not answer the survey, had a history of pregnancy lasting for more than 6 months, and were alive and still participating in the WHI at the time of the survey (eFigure 2 in the Supplement).

The primary outcome was development of an HF diagnosis during the WHI study follow-up period through 2018. Secondary outcomes included the development of HF subtypes HFrEF and HFpEF. All confirmed cases of HF hospitalization and patient-reported development of HF, angina, or CVD during hospitalization were adjudicated by trained physicians using standardized methods.^15-17 Briefly, hospital records of suspected HF were abstracted to include evidence of new onset of symptoms, history of HF, general medical history, physical examination signs and symptoms, diagnostic tests, biomarkers (brain natriuretic peptide, N-terminal prohormone brain natriuretic peptide, and cardiac troponins), and medications. Physician adjudicators reviewed this information for evidence of HF. Subtypes of HF were classified as HFrEF for patients with EF less than 50% and as HFpEF for those with EF 50% or higher, consistent with American and European clinical practice guidelines^18-20

Baseline characteristics were obtained by interviews and questionnaires at WHI study enrollment.^14,21 Factors associated with risk included age at enrollment and pack-years of smoking. Reproductive factors included a history of breastfeeding (defined as breastfeeding for at least 1 month over the woman’s entire reproductive period), number of live births, stillbirths (defined as number of stillbirths from a pregnancy lasting ≥6 months), miscarriages (defined as number of spontaneous miscarriages), age at first term pregnancy, menstrual cycle irregularity, age at menopause, and history of oral contraceptive use. Sociodemographic factors included income (defined as annual household income), education level, and race and ethnicity determined from the questionnaire (non-Hispanic White, Black or African American, Hispanic or Latino, and other, which refers to American Indian or Alaskan Native, Asian or Pacific Islander [Chinese, Indo-Chinese, Korean, Japanese, Pacific Islander, Vietnamese], any other race or ethnicity, or not reported). Race and ethnicity were assessed in this study to ensure the generalizability of findings and to be able to detect any racial or ethnic disparities in associations. Randomization status indicates whether a participant has been randomized to 1 or more of the clinical trial components. Potential mediators included coronary heart disease (CHD) (defined by self-report of physician-diagnosed cardiovascular disease, or adjudicated first occurrence of clinical myocardial infarction, definite silent myocardial infarction, or coronary revascularization before or at the same time as HF outcome was diagnosed), hypertension (defined by self-report of physician-diagnosed hypertension, or systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg at WHI study enrollment), diabetes (defined as not pregnancy related, self-reported physician diagnosis, or use of diabetes medication at study enrollment), and body mass index (BMI; defined as weight in kilograms divided by height in meters squared and measured by trained clinic staff at enrollment).

Logistic regression tested the association of each APO with incident HF, adjusting for potential confounders. In this model, each APO was coded using 3 categories (yes, no, or do not know). The responses yes and do not know were compared with the reference category (no). In a secondary analyses, we tested whether results were changed in models in which the do not know responses were treated as missing and then imputed and a model where they were combined with the yes responses, as done previously³ (eTable 1 in the Supplement). For simplicity, we present these secondary analyses only in the supplement because findings were similar to those in the primary analysis.

The first model was unadjusted, and subsequent models were also adjusted for (1) age; (2) sociodemographic factors (race and ethnicity, education, and income), smoking, and randomization status; (3) other APOs and reproductive history (GD, HDP, LBW, HBW, PTD, live births, stillbirths, miscarriages, history of any or ever breastfeeding, age at first birth, menstrual cycle irregularity, age at menopause, and history of oral contraceptive use); and (4) a fully adjusted model including all covariates in the preceding models. In addition to examining HF overall, we performed multinomial regression on HF subtypes (no HF vs HFpEF vs HFrEF) using the same adjustment models. All models were estimated using multivariate imputation by chained equations, pooling results from 10 data sets using standard methods to capture the inflation of SEs by the imputation²² to create 10 data sets.

We used the mediation package in R²³ to test potential mediation by hypertension, CHD, diabetes, and BMI for the association of APOs with HF. In brief, this approach uses nested models to estimate the proportion of the total adjusted association of an exposure explained by its indirect association via the mediator, with 95% CIs estimated using a nonparametric bootstrapping method. Each mediation analysis model was run using 1000 simulations. We applied full covariate adjustment to both the mediator and outcome model for consistency across mediation analyses.

To make the questionnaire respondents more representative of the overall HF subcohort (eTable 2 in the Supplement) and to address potential survival bias, we performed sensitivity analysis using inverse probability of inclusion weights²⁴ based on a logistic model for the association of baseline WHI covariates with inclusion in the questionnaire sample. We assessed modification of the associations of APOs with HF by race and ethnicity, age, BMI, hypertension, diabetes, CHD, history of breastfeeding, and by other APOs. Finally, we did a sensitivity analysis excluding women with CHD.

We considered 2-sided P < .05 to be statistically significant. Analyses were performed using R statistical software version 3.5.1 (R Project for Statistical Computing) and SAS Enterprise statistical software version 9.4 (SAS Institute). Data analysis was performed from January 2020 to September 2021.

Of 44 174 women in the WHI HF subcohort, 27 204 had a history of pregnancy lasting for more than 6 months, were alive, and were still participating in the WHI at the time of the survey; 10 292 responded and formed the study population (eFigure 2 in the Supplement). The median (IQR) age of participants was 60 (55-64) years. In this study population, 3185 women (31.0%) reported a history of 1 or more APOs. The most frequently reported APO was PTD in 1509 women (14.7%), followed by LBW in 1424 (13.8%), HDP in 759 (7.4%), HBW in 644 (6.3%), and GD in 260 (2.5%) (Table 1). The most common combination of APOs in the study population was PTD and LBW and was reported by 732 women (7.1%).

Baseline characteristics of women at entry into the WHI differed by the presence and type of APO (eTable 3 in the Supplement). Women with a history of any APO had a higher prevalence of hypertension, diabetes, CHD, smoking (>20 pack-years), and stillbirth; lower levels of education and household income; younger age at first birth and older age at menopause; higher BMI; were less likely to have reported a history of breastfeeding, miscarriage, and menstrual cycle irregularity; and reported fewer live births. HDP was more prevalent among Black women compared with White women. Hypertension at baseline was more prevalent among women who reported HDP (68 women [62%]) than among women reporting any APO (1339 women [42%]) or no APO (2066 women [33%]). Diabetes at baseline was more prevalent among women reporting previous GD (56 women [22%]) than among women reporting any other APO (168 women [5.3%]) or no APO (160 women [2.5%]). Women who developed HF during follow-up were more likely to have hypertension, diabetes, and a higher BMI than those who did not experience HF (Table 2).

Of our cohort of 10 292 participants, 336 (3.3%) had a diagnosis of HF, 180 (1.8%) had HFpEF, and 111 (1.1%) had HFrEF. Women with a history of APO had a higher rate of HF than those without a history of APO (121 women [3.8%] vs 184 women [2.9%]). Women with HDP had the highest rate of HF (39 women [5.1%]), and women with GD had the lowest rate of HF (8 women [3.1%]) (eFigure 3 in the Supplement). HDP was the only APO with a significant association with HF in univariate models, with an odds ratio (OR) of 1.77 (95% CI, 1.25 to 2.50). HDP remained significantly associated with HF after adjusting for age (OR, 1.87; 95% CI, 1.32 to 2.65), sociodemographic factors, smoking and randomization status (OR, 1.76; 95% CI, 1.25 to 2.50), other subsequent APOs and reproductive history (OR, 1.70; 95% CI, 1.19 to 2.42), and in a model adjusting for all these factors (OR, 1.75; 95% CI, 1.22 to 2.50) (Figure 1). In analyses of HF subtypes, only HDP was significantly associated with HFpEF in a fully adjusted model (OR, 2.06; 95% CI, 1.29 to 3.27), but not with HFrEF (OR, 1.17; 95% CI, 0.59 to 2.30) (Figure 2). In mediation analysis, hypertension explained 24% (95% CI, 12% to 73%) of the association of HDP with HF, BMI explained 20% (95% CI, 10% to 64%), diabetes explained 1% (95% CI, −3.6% to 7.3%), and CHD explained 23% (95% CI, 11% to 68%) (Figure 3).

All results from the analysis were virtually unchanged in the complete case analysis (eTable 1 in the Supplement) and in models with inverse probability of inclusion weights (eTable 4 in the Supplement). We found no significant modification of the association of HDP with HF by covariates including other APOs. Upon excluding women with CHD, the association between HDP and HF was similar (OR, 1.67; 95% CI, 1.01-2.76).

In this large cohort study of postmenopausal women, a history of HDP was independently associated with a 1.75-fold odds of developing subsequent heart failure. This association was significant after adjustment for multiple confounding factors, including other APOs, without evidence of modification by sociodemographic or reproductive factors or comorbidities. HDP was significantly associated with developing HFpEF, but not HFrEF, among women in late midlife. Furthermore, the association of HDP with HF was partially mediated by hypertension, BMI, and CHD. Our findings highlight that a subset of women with HDP will not develop hypertension before developing HF.

Several mechanisms may explain the association between HDP and HF. They share several risk factors, such as hypertension and obesity, which underlie this association.⁷ HDP is associated with the development of hypertension,²⁵ and having both is associated with persistent left ventricular remodeling.²⁶ A meta-analysis revealed that HDP and HFpEF share several biomarkers, including immune activation, myocardial stress, and autonomic function.²⁷ Women with a history of HDP have persistent structural differences in the heart and microvasculature more than 25 years after their pregnancy.^28,29 Women with a history of HDP have increased echocardiographic diastolic parameters, including left ventricular mass index, increased relative wall thickness, lower impaired relaxation (ie, lower transmitral Doppler E/A ratio), and higher left ventricular filling pressure (ie, E/e′ ratio) in comparison with those with a history of normotensive pregnancies, all of which are associated with HF.^28,30,31 Women with HDP also have impaired coronary flow reserve, which is a key feature of HFpEF.³² Persistence of these changes may contribute preferentially to risk of HFpEF (vs HFrEF).³³

APOs may also be associated with the development of CVD risk factors themselves, which could mediate associations between APOs and future CVD.³⁴ Whether an APO uncovers a predisposition to CVD, exacerbates a preexisting subclinical condition, or initiates a pathway that results in CVD is unclear, although recent data may suggest all may be true.^33,34

Our mediation analysis showed large but incomplete mediations of the associations from hypertension (24%), BMI (20%), and CHD (23%). These findings align with known factors associated with the risk of HFpEF.³⁵ However, this also implies that CVD-related factors may not fully explain the association between HDP and HF. Two previous studies^4,36 found that hypertension mediated 49% of the association of HDP with HF and cardiomyopathy.

HFpEF disproportionately affects elderly women, and hypertension is an important factor associated with risk of HFpEF among women.³⁷ Furthermore, HFrEF is less common in women than in men.³⁸ Our findings that HDP is independently associated with HFpEF and that hypertension is a mediator of the association between HDP and HF suggest that both hypertension and novel pathways likely explain the HDP and HFpEF association we found in our study. It has been unclear whether some APOs are more closely related to HF and CVD risk, because different APOs may share some elements of underlying pathophysiology.^2,39 Our findings show that only a history of HDP was independently associated with HF in postmenopausal women and suggests that this APO merits particular scrutiny in future epidemiological and mechanistic studies.

Although prior studies demonstrated an association between GD and HF,⁸ we did not confirm this association. The women in our study, however, reported a relatively low prevalence of GD (2.5%, vs the current US prevalence of 8%).⁴⁰ This difference may reflect evolving practices for GD screening, which was not widely and routinely implemented in the US until the early 1980s, after many women in the present data set would have completed their pregnancies, as well as an increasing prevalence of factors associated with the risk of GD among reproductive-aged women (eg, obesity or family history of diabetes).⁴¹ Therefore, we may have both underestimated the prevalence of GD in our study and had a cohort of women less likely to have GD. Contemporary studies assessing the association between GD and HF would be less likely to have these limitations.^42,43

History of HDP represents an opportunity for early, aggressive, preventive interventions for HF and other CVD, possibly before development of the traditional risk factors (hypertension, diabetes, and obesity). In fact, the recent Postnatal Enalapril to Improve Cardiovascular Function Following Preterm Preeclampsia Study⁴⁴ showed improved diastolic function and left ventricular remodeling after 6 months of postnatal enalapril treatment for women with preterm preeclampsia. The concept of the fourth trimester has been introduced by the obstetric-gynecologic community to highlight the need to retain focus on optimizing maternal health beyond the standard 6-week postdelivery window to enable more targeted and aggressive risk factor modification in women with APOs,⁴⁵ with lifestyle interventions, earlier monitoring, and tighter control of traditional risk factors, such as hypertension, diabetes, and obesity. Long-term studies are needed to assess to what extent earlier cardiovascular prevention techniques will prove effective in women with a history of HDP.⁴⁶

Strengths of this study include the unique availability of comprehensive baseline characteristics, reproductive history, and adjudicated HF outcomes, which allowed us to comprehensively evaluate the association between APOs and HF outcomes with long-term follow-up. This study also has limitations. Survivorship bias among the included sample resulted from participants needing to survive until the APO survey in 2017, be free of HF at baseline, and women with HDP (especially severe and recurrent preeclampsia) having increased mortality compared with other women.^12,47 Therefore, our estimates would have been biased toward the null rather than leading to spuriously high estimates. However, our sensitivity analysis using inverse probability of inclusion weights demonstrated that the findings were similar. The WHI cohort is known to be representative regarding race and ethnicity,⁴⁸ but our substudy had fewer racial and ethnic minority women than the overall WHI. The obstetric records were unavailable to us for validating the APO information, providing more specific APO phenotyping such as HDP severity. Therefore, we cannot confirm or deny the possibility that recall bias affected our results. WHI did not have information on prepregnancy risk factors such as BMI or gestational weight gain. We do not have information on when this cohort of postmenopausal women had their deliveries, which presents a potential recall bias, and women’s ability to recall APOs has only been validated previously in short-term studies, but not in long-term studies.⁴⁹ We were unable to differentiate whether women had multiple APOs in the same pregnancy or in recurrent pregnancies, nor did we have information on size for gestational age.

In this study, a history of HDP was associated with HF, particularly HFpEF, among postmenopausal women, independently of conventional HF risk factors, other APOs, and sociodemographic and reproductive factors. Close clinical monitoring of women with a history of HDP may provide opportunities for early prevention of HF and other CVD. Hypertension, BMI, and CHD played partial, mediating roles in the associations demonstrated. Further research is needed to better understand the potential mechanisms that link HDP with later development of HF. Dedicated studies are needed to establish effective interventions to mitigate long-term risk of HF and other CVD in women with APOs.

Accepted for Publication: October 13, 2021.

Published: December 9, 2021. doi:10.1001/jamanetworkopen.2021.38071

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2021 Hansen AL et al. JAMA Network Open.

Corresponding Author: Nisha I. Parikh, MD, MPH, Department of Medicine, Division of Cardiology, University of California San Francisco, 505 Parnassus Ave, Moffitt Bldg, San Francisco, CA 94143 (nisha.parikh@ucsf.edu).

Author Contributions: Dr Parikh had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Hansen, Manson, Van Horn, S. Liu, Parikh.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Hansen, Wild, Parikh.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Hansen, Søndergaard, Vittinghoff, Nah, Stefanick, Mongraw-Chaffin, S. Liu, Parikh.

Obtained funding: Eaton.

Administrative, technical, or material support: Hansen, Manson, Farland, Van Horn, S. Liu, Parikh.

Supervision: Søndergaard, Manson, Farland, Van Horn, Wild, Allison, Honigberg, Parikh.

Conflict of Interest Disclosures: Dr Manson reported receiving grants from National Institutes of Health (NIH) during the conduct of the study. Dr Gunderson reported receiving grants from the NIH for primary data collection for the Division of Reasearch Women’s Health Initiative site for participants and data collection during the conduct of the study, the National Institute of Diabetes and Digestive and Digestive and Kidney Diseases (NIDDK R01 SWIFT Women Follow Up, NIDDK R01 SWIFT Study in Youth, and NIDDK R01 CARDIA Pregnancy RF and T2D), the National Heart, Lung, and Blood Institute (NHLBI R01 Pregnancy Blood Pressure and CVD and NHLBI R21 CARDIA Pregnancy SMM), the American Heart Association Project on Gestational Diabetes Mellitus and GrimAge, and Janssen Metabolomics for research on type 2 diabetes after gestational diabetes outside the submitted work; in addition, Dr Gunderson had a patent for metabolites and type 2 diabetes pending. No other disclosures were reported.

Funding/Support: The Women’s Health Initiative program was funded by the NHLBI, NIH, and US Department of Health and Human Services through contracts HHSN268201100046C, HHSN268201100001C, HHSN268201100002C, HHSN268201100003C, and HHSN268201100004C. Mr Hansen was supported by a Danish American Research Exchange award from the Lundbeck Foundation, Fonden til Lægevidenskabens Fremme from the A. P. Moller Foundation, and travel grants from University of Southern Denmark and Odense University Hospital. Drs Parikh and Gunderson are supported by NIH grant R56HL148260. Dr Honigberg is supported by NHLBI grant T32HL094301-07.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank the WHI investigators, staff, and the trial participants for their outstanding dedication and commitment.