Article Text
Abstract
Objective To determine the association between recurrent pre-eclampsia and long-term cardiovascular hospitalisation.
Methods This study identified cardiovascular hospitalisations up to 25 years after pregnancy for all women who delivered between 1989 and 2013 in Québec, Canada. Exposures included recurrent and non-recurrent pre-eclampsia in women with two deliveries or more (N=606 820), and pre-eclampsia in women with only one delivery (N=501 761). Incidence, timing and risk of cardiovascular complications were calculated using accelerated failure time models adjusted for age, pre-existing disease, socioeconomic deprivation and period. Outcomes included a range of cardiovascular hospitalisations and procedures.
Results Women with recurrent pre-eclampsia had higher incidence of cardiovascular hospitalisation (281.4 per 1000) than women with non-recurrent (167.7 per 1000) or no pre-eclampsia (72.6 per 1000). Mean time to cardiovascular hospitalisation was 10.5 years for recurrent, 11.6 years for non-recurrent and 12.7 years for no pre-eclampsia, a difference of 17.3% for recurrent and 8.7% for non-recurrent relative to no pre-eclampsia. Compared with no pre-eclampsia, recurrent pre-eclampsia was associated with 2 times the risk of heart disease (95% CI 1.69 to 2.29) and 3 times the risk of cerebrovascular disease (95% CI 2.25 to 4.05). Pre-eclampsia in women with one delivery was associated with 3 times greater risk of cardiovascular hospitalisation compared with no pre-eclampsia in women with two deliveries or more (95% CI 2.96 to 3.25).
Conclusions Recurrent pre-eclampsia is associated with higher risk of future cardiovascular hospitalisation compared with no pre-eclampsia, and significantly shorter time to first cardiovascular event. Cardiovascular screening should be performed earlier for women with recurrent pre-eclampsia.
Statistics from Altmetric.com
Introduction
Pre-eclampsia ranks as a major obstetric disease, due to significant morbidity and mortality, and widespread involvement of the cardiovascular system.1 Pre-eclampsia affects 2%–5% of pregnancies,1 and recurs in 15% of women.2 Recurrent pre-eclampsia is particularly concerning due to potential for more extensive cardiovascular damage. In the long term, women with pre-eclampsia, particularly pre-eclampsia with onset early in gestation, have greater risk of a range of cardiovascular end points years after the first pregnancy.3–9 Yet the impact of recurrent pre-eclampsia on risk is poorly understood.
A high proportion of women have cardiovascular disease at some point in life.9 ,10 Cardiovascular disease is the leading cause of death in women, and is associated with significant morbidity and costs to society.10 The few studies that investigated the association between recurrent pre-eclampsia and later risk of cardiovascular morbidity focused on a limited number of end points11–13 did not evaluate timing of disease onset, or consider the possible contribution of unknown sources of bias. In particular, index event or selection bias is common in research on recurrent events, with potential to affect risk estimates.14 Recurrent pre-eclampsia may be associated with greater risk of long-term cardiovascular disease than previously thought.
The objective of this study was to determine whether recurrent pre-eclampsia was a stronger risk factor for cardiovascular disorders later in life than non-recurrent pre-eclampsia, using a large population cohort of women followed for 25 years. To minimise selection bias, both women with two or more deliveries as well as only one delivery were analysed. In addition, to identify when to ideally begin cardiovascular screening, time to first cardiovascular hospitalisation was estimated for recurrent and non-recurrent pre-eclampsia.
Methods
Data and study design
This longitudinal follow-up study included all women who delivered a live or stillborn infant between 1989 and 2013 in hospitals in the province of Québec, Canada. The cohort was followed over time to identify later cardiac hospitalisations that occurred before 31 March 2014. Data were drawn from discharge abstracts in the Maintenance and Use of Data for the Study of Hospital Clientele registry, a compilation of all hospitalisations in the province. Québec represents nearly a quarter of the Canadian population and over 99% of women deliver in hospital.15 Because pre-eclampsia cannot be diagnosed before 20 weeks gestation,1 only women with pregnancies extending past 20 weeks were included.
Pre-eclampsia
Pregnancies complicated by pre-eclampsia were identified using International Classification of Disease (ICD) codes for mild (642.3, 642.4, O13), severe (642.5, 642.6, O14, O15) or superimposed (642.7, O11) disease. Superimposed pre-eclampsia consists of de novo proteinuria with pre-existing hypertension. Mild pre-eclampsia included gestational hypertension, following the 10th revision of the ICD which groups both diagnoses together.
The focus was on recurrent pre-eclampsia in women with two or more pregnancies. Therefore, recurrent pre-eclampsia was identified among the 606 820 women who delivered at least twice in the cohort. Women who delivered only once (N=501 761) were analysed to minimise selection bias. Among women who delivered twice or more, pre-eclampsia was defined as recurrent, non-recurrent or no pre-eclampsia. Among women who delivered only once, exposure was defined as pre-eclampsia or no pre-eclampsia.
Cardiovascular outcomes
A comprehensive set of cardiovascular hospitalisation end points was analysed, including cardiovascular disease overall, heart disease (heart failure, ischaemia, infarction, angina, arrest, inflammation, conduction, valve, cardiomyopathy), pulmonary heart disease (pulmonary embolism, other), cerebrovascular disease (ischaemic stroke, cerebral haemorrhage), hypertension, atherosclerosis, aneurysm, aortic dissection and deep vein thrombosis. Procedures involving the heart (angioplasty, coronary artery bypass graft, cardiopulmonary resuscitation, open heart resuscitation, pacemaker, valve surgery), vessels (angiography, aorta surgery, intracranial vessel surgery) and coronary care unit admission were also analysed.
Cardiovascular diagnoses were identified using ICD codes (see online supplementary table S1), and procedures using the Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures and the Canadian Classification of Health Interventions (see online supplementary table S2). Admission to coronary care units was documented directly on discharge abstracts.
Supplementary tables
Covariates
Baseline covariates included age (<20, 20–24, 25–29, 30–34, 35–39, ≥40 years), pre-existing gestational or non-gestational diabetes (ICD 648.0, 648.8, V12.21, O24), pre-existing cardiovascular disease (ICD 401-445, 447-453, 642.0-642.2, 642.9, 646.2, I10-I82, O10), socioeconomic deprivation (poorest fifth of the population, or not)16 and time period (1989–1996, 1997–2004, 2005–2013).
Data analysis
Cumulative incidence curves were plotted using the cumulative incidence function with death as a competing event.17 A time-to-event approach was employed using a parametric accelerated failure time model with Weibull distribution to obtain HRs for pre-eclampsia relative to no pre-eclampsia, including 95% CIs.18 The Weibull distribution allows the shape of the hazard function to vary over time.18 The time scale was defined as number of years since first delivery, and women were censored if they died or had no event by 31 March 2014, the last day of follow-up.17 Non-pre-eclamptic women with two or more deliveries were used as the reference. This reference was selected because women with only one delivery have greater risk of long-term cardiovascular disease.13 The models further estimated the time at which each woman's cumulative probability of cardiovascular outcomes was 1 per 1000, with the median across women summarising the time to reach an incidence of 1/1000. All models were adjusted for baseline age, pre-existing diabetes, pre-existing cardiovascular disease, socioeconomic deprivation and time period. Model fit was assessed using a likelihood-ratio test19 and the proportional hazards assumption using plots of the survival function.
Compared with Cox proportional hazards regression, accelerated failure time models are ideal for hazards that plateau with time, and exposures such as recurrent pre-eclampsia that potentially affects the time to event more than the absolute risk.20 In addition to HRs, the Weibull model yields acceleration factors, or the ratio of the time to event for recurrent and non-recurrent pre-eclampsia relative to no pre-eclampsia. An acceleration factor of 1 indicates no difference in the time to event, <1 a shorter time to event and >1 a longer time to event.
In sensitivity analyses, the impact of women who delivered only once but had potentially not completed childbearing was examined, because recurrent pre-eclampsia may be misclassified as non-recurrent in women with longer interpregnancy intervals towards the end of the study. To assess this possibility, the analysis was restricted to women with at least 7 years of follow-up after first pregnancy, providing enough time for women with incomplete childbearing to have their second child.13 In the dataset, 95% of women with at least two deliveries had their second child within 7 years. In addition, women with pre-existing cardiovascular disease were excluded to determine the influence on results. Finally, Cox proportional hazards regression was used to verify the results.
SAS V.9.3 (SAS Institute, Cary, North Carolina, USA) was used for data analysis, including two-sided hypothesis tests with statistical significance at p=0.05. Data were de-identified, and the University of Montreal Hospital Centre institutional review board waived ethical review.
Results
There were 6066 women with recurrent (0.5%) and 33 493 with non-recurrent (3.0%) pre-eclampsia in the study (table 1). In addition, 24 799 women with only one delivery had pre-eclampsia (2.2%). There were in total 16 122 579 person-years of follow-up, with 58% of the follow-up time to women with two deliveries or more. The maximum length of follow-up was 25.2 years, and the median 15.5 years. Less than 1% of women died during the study.
Women with two deliveries or more
Among women with two deliveries or more, the incidence of cardiovascular hospitalisation at end of follow-up was higher for recurrent (281.4 per 1000) than non-recurrent (167.7 per 1000) or no pre-eclampsia (72.6 per 1000) (table 2). Incidence was greater for recurrent pre-eclampsia regardless of the cardiovascular event or procedure, but was especially elevated for heart failure, atherosclerosis and cerebrovascular disease compared with non-recurrent pre-eclampsia. Incidence of procedures such as coronary artery bypass graft, cardiopulmonary resuscitation and pacemaker insertion was higher for recurrent pre-eclampsia. For most outcomes, the time to reach an incidence of 1 per 1000 was shorter for recurrent pre-eclampsia. For instance, the time for angioplasty was 10.7 years for recurrent, 12.4 years for non-recurrent and 17.6 years for no pre-eclampsia.
Incidence plots showed that cardiovascular disorders and procedures cumulated to higher and steeper levels for recurrent pre-eclampsia than non-recurrent or no pre-eclampsia (figure 1). For recurrent pre-eclampsia, the cumulative incidence of cardiovascular outcomes diverged from non-recurrent and no pre-eclampsia after only 5–10 years of follow-up. Incidence of outcomes such as myocardial infarction, angioplasty, coronary artery bypass graft and aorta surgery differed little between recurrent, non-recurrent and no pre-eclampsia at 3–15 years of follow-up, but then noticeably diverged. Grey's test for equality of cumulative incidence functions resulted in p values of <0.05 for all curves, suggesting statistically significant differences in incidence between groups regardless of the outcome.
In adjusted regression models, HRs for most outcomes were significantly elevated for recurrent pre-eclampsia compared with no pre-eclampsia (table 3). Relative to no pre-eclampsia, women with recurrent pre-eclampsia had nearly 4 times the risk of any cardiovascular hospitalisation (95% CI 3.57 to 4.17). HRs for recurrent pre-eclampsia were disproportionately greater than non-recurrent pre-eclampsia, particularly for heart failure, ischaemic heart disease, cardiomyopathy, atherosclerosis, hypertension, any cerebrovascular disease and intracranial vessel surgery. There was no outcome where risk was significantly higher for non-recurrent than recurrent pre-eclampsia.
The adjusted acceleration factors suggested that time to first cardiovascular hospitalisation was 0.39 times shorter for recurrent than no pre-eclampsia (95% CI 0.37 to 0.41). Time to cardiovascular hospitalisations and procedures was shorter for recurrent pre-eclampsia for nearly all outcomes. Time to first angiography, for example, was 0.18 times shorter for women with recurrent pre-eclampsia than no pre-eclampsia, an 82% reduction in time. In addition, compared with no pre-eclampsia, recurrent pre-eclampsia was associated with shorter cardiovascular event times than non-recurrent pre-eclampsia.
Women with only one delivery
Overall, women with one delivery had higher incidence of most types of cardiovascular hospitalisations and procedures compared with women with two deliveries or more (table 4). The incidence of cardiovascular events in women with one delivery was greater for pre-eclampsia than no pre-eclampsia. Incidence of cardiovascular hospitalisation for pre-eclamptic women with one delivery was situated between recurrent and non-recurrent pre-eclampsia in women with two or more previous deliveries. Among pre-eclamptic women with one delivery, the time to reach an incidence of 1 per 1000 was similar to recurrent pre-eclampsia in women with two deliveries or more.
Cumulative incidence curves for pre-eclamptic women with one delivery resembled recurrent pre-eclampsia for most outcomes, or were between recurrent and non-recurrent pre-eclampsia (figure 2). Rates tended to diverge from no pre-eclampsia as early as 2 years after delivery.
Adjusted HRs confirmed that women with one delivery had greater risk of most cardiovascular outcomes regardless of pre-eclampsia (table 5). Compared with non-pre-eclamptic women with two or more deliveries, pre-eclamptic women with one delivery had 3.11 times the risk of cardiovascular disorders (95% CI 2.96 to 3.25), and non-pre-eclamptic women with one delivery had 1.26 times the risk (95% CI 1.23 to 1.29). For many end points, HRs for pre-eclamptic women with one delivery resembled or were greater than those with recurrent pre-eclampsia. This was also the case for the acceleration factors, which indicated significantly shorter onset times for women with one delivery regardless of pre-eclampsia.
Sensitivity analysis
Excluding women with potentially incomplete childbearing from the cohort had little impact on the results. Excluding women with pre-existing cardiovascular disease slightly strengthened the estimated HRs and acceleration factors, but not significantly. There was no evidence that hazards were non-proportional in plots of the survival function.
Discussion
In this longitudinal study with 25 years of follow-up on a cohort of 1 108 581 Canadian women, recurrent pre-eclampsia was associated with significantly greater risk of long-term cardiovascular hospitalisation compared with no pre-eclampsia in women with at least two deliveries. The associations were present over a wide range of cardiovascular outcomes. Non-recurrent pre-eclampsia was not as extensively associated with cardiovascular outcomes as recurrent pre-eclampsia. In addition, the time to first cardiovascular event was significantly shorter for recurrent pre-eclampsia, indicating that affected women experience cardiovascular problems significantly earlier in life than those with non-recurrent or no pre-eclampsia. Furthermore, pre-eclamptic women with one delivery had long-term risks of cardiovascular hospitalisation similar to recurrent pre-eclampsia, an interesting finding considering that only a proportion of these women would have had recurrent pre-eclampsia had they had a second delivery. However, women with one delivery are a mixed group and may be disproportionately older. Overall, the results suggest that screening for cardiovascular disease should begin as early as post partum or even during the second pregnancy with recurrent pre-eclampsia. Pre-eclamptic women with only one delivery over the lifetime may also benefit from earlier screening.
These findings align with studies showing a relationship between pre-eclampsia and later risk of cardiovascular disease,7 ,9 ,21 ,22 regardless of recurrence. The limited studies on recurrent pre-eclampsia suggest a dose-response relationship between recurrent, non-recurrent and no pre-eclampsia with later cardiovascular disease, but for fewer outcomes.11–13 To date, outcomes have only included cardiovascular mortality,12 ,13 ischaemic heart disease11 ,12 and risk factors such as hypertension and diabetes.11 ,23 This study demonstrates that recurrent pre-eclampsia is associated with higher risk of cardiovascular hospitalisation over a wide range of end points. Overall, the risks were greatest for hypertension and procedures such as cardiopulmonary resuscitation or aorta surgery, and lowest for deep vein thrombosis.
Despite knowledge of a link with later cardiovascular disease since the early 1990s, the American Heart Association only added pre-eclampsia as a cardiovascular risk factor in 2011, followed by the European Society of Cardiology in 2012.3 ,7 ,24 Furthermore, recommendations for cardiovascular screening in women with a history of pre-eclampsia are not consistent across the disciplines of cardiology and obstetrics,24 with little guidance on when screening should begin.7 Consequently, there is low awareness of the benefits of screening by gynaecologists and internists.25 This is particularly concerning in light of results which show that women with recurrent pre-eclampsia have numerous cardiovascular disorders earlier than women with non-recurrent or no pre-eclampsia. While more research is needed to confirm optimal screening windows, the results suggest that women with recurrent pre-eclampsia may benefit from screening soon after or even during their second pre-eclamptic pregnancy. In fact, the time needed to reach an incidence of hypertension of 1 per 1000 was only 1.6 years after the first delivery, which means women with recurrent pre-eclampsia may even have hypertension by the time of the second delivery. Organisations issuing recommendations for screening should consider including recurrent pre-eclampsia as a particularly strong risk factor for cardiovascular disease, as well as the option to begin screening as early as post partum or during pregnancy.24
Exact reasons why recurrent pre-eclampsia is associated with stronger risks of cardiovascular disease are not known, but are likely tied to a cascade of systemic events linked to angiogenic or inflammatory biomarkers that initially manifest at pregnancy. Pre-eclampsia, particularly severe or recurrent pre-eclampsia, is preceded by an imbalance in angiogenic biomarkers such as soluble fms-like tyrosine kinase and placental growth factor, which lead to maternal endothelial dysfunction and inflammation,26 with evidence that angiogenic dysregulation persists after pregnancy.24 ,27 Increased levels of inflammatory markers including C-reactive protein and fibrinogen have also been documented in women with a history of recurrent pre-eclampsia.28 Markers of systemic inflammation, including C-reactive protein, are known to be associated with chronic disease,29 especially atherosclerosis and cardiovascular mortality.24 ,30 These studies suggest that biomarker imbalance in recurrent pre-eclampsia may be involved as mediators in the cardiovascular cascade throughout life.
It is important to point out that women with only one delivery also had higher risk of cardiovascular disease, regardless of pre-eclampsia. The relationship between low parity and future risk of cardiovascular disease has been previously documented,13 ,24 but the causal pathways are still unclear. Pre-eclampsia led to even greater risks, suggesting women with only one delivery may be less healthy than women with higher parity. Low parity may therefore reflect poorer health in general, predisposing to reduced fertility and possibly higher risk of later cardiovascular disease.24 However, many factors contribute to a woman's decision to have only one child, making it unclear how parity, pre-eclampsia and cardiovascular disease overlap.
An important strength of this study was the use of all pregnant women in the analysis, rather than only women with two or more deliveries. Studies restricted to women with two or more deliveries may be subject to index event bias by selecting study subjects based on history of pregnancy.14 In the data, pre-eclamptic women with only one delivery had risks of cardiovascular hospitalisation that were nearly identical to those with recurrent pre-eclampsia, a finding that would have been missed had these women not been included. The results would have incorrectly suggested that recurrent pre-eclampsia was the riskiest exposure, when in fact pre-eclamptic women with only one delivery deserve as much attention. However, index event bias due to excluding women who never had children or whose pregnancies never reached 20 weeks gestation could not be ruled out. Risk of cardiovascular hospitalisation may vary in such women.
This study has several limitations. Major confounders were adjusted for, but residual confounding can still be present. In particular, it was not possible to adjust for ethnicity, smoking or metabolic risk factors such as obesity as such information was absent from the dataset. Coding errors could not be ruled out. Accelerated failure models have been shown to be advantageous for long-term follow-up studies,18 ,20 but the possibility of time-dependent confounding cannot be excluded. A small proportion of women who delivered at home or in birthing centres were not included, but all women with pre-eclampsia would be referred to hospital in Québec. Cardiovascular events that did not result in hospitalisation could not be analysed. Finally, Canada has universal public healthcare, and the generalisability of the results to settings with other forms of healthcare delivery is to be determined.
Conclusion
In this study, recurrent pre-eclampsia was associated with more than 2 times the risk of future cardiovascular hospitalisation compared with no pre-eclampsia, and significantly shorter time to first cardiovascular event in women with two deliveries or more. Risks were just as elevated for pre-eclamptic women with only one delivery. Cardiovascular screening should be performed earlier for women with recurrent pre-eclampsia, and pre-eclamptic women with only one delivery.
Key messages
What is already known on this subject?
Pre-eclampsia is associated with later risk of cardiovascular disease, but the association with recurrent pre-eclampsia is not understood.
What might this study add?
Women with recurrent pre-eclampsia have higher risk of several cardiovascular outcomes compared with non-recurrent or no pre-eclampsia.
Recurrent pre-eclampsia significantly decreases the time to first cardiovascular event compared with no pre-eclampsia.
How might this impact on clinical practice?
Cardiovascular screening should begin earlier for women with recurrent pre-eclampsia, shortly after or even during pregnancy.
References
Footnotes
Contributors NA conceived and designed the study with JH-P. JH-P analysed the data under guidance of NA. MS provided statistical expertise. WDF, LL and GP helped interpret the results. NA drafted the manuscript with JH-P, and all authors revised the article for important intellectual content. All authors approved the final version. NA is guarantor.
Funding This study was funded by the Canadian Institutes of Health Research (MOP-130452). NA received a career award from the Fonds de recherche du Québec-Santé (grant 25128).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.