Hypertensive disorders of pregnancy and risk of stroke in U.S. Black women

Abstract

Background:

Black women have a disproportionately higher burden of both preeclamptic pregnancy and stroke compared with White women, but virtually all existing evidence on this possible association has been generated from women of European ancestry.

Methods:

In the Black Women’s Health Study (BWHS), a prospective cohort of U.S. Black women who enrolled in 1995, 42,924 participants were parous and free of cardiovascular disease at baseline. Biennial questionnaires included questions on preeclampsia, gestational hypertension, and stroke. Medical records were sought for participants who reported a stroke and were reviewed by study neurologists blinded to reproductive history. Cox proportional hazard models, with control for potential confounders, were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs).

Results:

Over a median of 22 years of follow-up, there were 1,555 incident strokes, including 310 among 4,938 women with a history of hypertensive disorders of pregnancy (HDOP). The multivariable HR for stroke for women with any HDOP compared with those who had never experienced HDOP was 1.66 (95% CI 1.46-1.89). Comparable HRs were 1.53 (95% CI 1.29-1.82) for preeclampsia and 1.81 (95% CI 1.53-2.13) for gestational hypertension only. Associations were similar among women under age 55 and those aged 55 and older.

Conclusions:

In this prospective study of Black women, a history of HDOP was associated with an estimated 66% increased long-term risk of stroke. This association may contribute to the disproportionately higher stroke incidence in Black women, given the higher prevalence of HDOP in this population.

Introduction

In the United States, 55,000 more women than men have a stroke each year^1,2. Black women are particularly vulnerable and experience stroke and stroke-related mortality at higher rates^3,4 and earlier onset than other racial groups^5,6. Preeclampsia is a pregnancy-specific multi-system syndrome⁷ affecting approximately 2-8% of pregnancies⁸ and is the second leading cause of maternal mortality worldwide⁹. The prevalence of preeclampsia/eclampsia among Black women in the U.S. is 60% higher than among White women (70 per 1,000 deliveries in 2014 for Black women vs. 43 per 1,000 for White women)^10,11. In recent years, rates of severe preeclampsia have been increasing for Black women¹².

The American Heart Association (AHA) guidelines recently added pregnancy complications as risk factors for stroke. However, the recommendations were based largely on data from women of Northern European ancestry^13–15. The AHA has called for more research on risk of stroke in women, especially women from underrepresented populations. To address the knowledge gap in Black women, we examined the relation of hypertensive disorders of pregnancy (HDOP), and preeclampsia in particular, to long-term stroke risk in a large prospective cohort of self-identified Black women, the Black Women’s Health Study (BWHS).

Methods

Study population:

The BWHS is a prospective cohort study of 59,000 Black women enrolled from across the continental United States in 1995¹⁶. All but 5% of BWHS participants were born in the U.S. At baseline in 1995, participants were 21 to 69 years of age, with a median age of 38. Participants have provided information on biennial questionnaires on demographic characteristics, socioeconomic factors, medical conditions, and lifestyle factors. Follow-up has been successful for 85% of potential person-years through 11 completed biennial rounds of follow-up.

We restricted the analysis to the 44,298 women who had had at least one birth (either live birth or still birth), thereby excluding nulliparous women. Most participants were already parous at the time of BWHS enrollment in 1995, but 14% had their first birth during follow-up and they entered this analysis at the year of that first birth. We further excluded women who had a history of cardiovascular disease before entry into the analysis. The final analytic sample comprised 42,924 parous Black women (Figure 1). The Boston University Medical Campus Institutional Review Board approved the study. Study participants were informed in each questionnaire that taking part was entirely voluntary and that at any time they could choose not to participate either by not completing the questionnaire or by asking for their name to be removed from the study mailing list.

Figure 1. Flow diagram of the analytical study cohort.

Figure shows exclusions and inclusions for analytical dataset and stroke outcomes

Ascertainment of HDOP: History of HDOP was defined as a self-report of gestational hypertension and/or preeclampsia/toxemia. In 1997 and 1999, all participants were asked about history of gestational hypertension (“high blood pressure during pregnancy”). In 1997 and again in 2009, participants were asked if they had ever had a pregnancy with toxemia/preeclampsia. In addition, detailed pregnancy characteristics were queried for women who had births between 1995 and 2003; women who reported having a baby born at least 3 weeks early were asked about the reason for the preterm birth, including the option “labor was induced or had c-section because blood pressure was too high (preeclampsia/toxemia)”. Those who checked that option were classified as positive for a history of preeclampsia. For analyses of specific types of HDOP, women who reported both preeclampsia and gestational hypertension were classified as having had “preeclampsia” and women who reported only gestational hypertension were classified as having had “gestational hypertension”.

Ascertainment of stroke:

Stroke cases were ascertained through biennial BWHS questionnaires, which asked participants to self-report occurrences of physician-diagnosed stroke. Fatal stroke was ascertained through linkage with the National Death Index (NDI). Validity of self-reported stroke was assessed through medical record review by an independent team of neurologists using the World Health Organization (WHO) definition:¹⁷ a new focal (or at times global) neurological impairment of sudden onset, lasting more than 24 hours (or leading to death) and of presumed vascular origin¹⁷. Events not meeting this definition but with neuroimaging evidence (by report) consistent with acute ischemia or hemorrhage were also classified as stroke events, as were strokes that had been aborted by thrombolysis or mechanical thrombectomy and transient ischemic attacks (TIA) with demonstration of an ischemic lesion on magnetic resonance diffusion weighted imaging. Appropriate medical records were obtained for 618 (31%) of participants who reported stroke; for the remainder, either the participants were unwilling to sign a medical records release or the hospital did not provide the appropriate records (Figure S1). Among the 618 cases for which records were available, 450 (73%) were confirmed through adjudication; the remainder were disconfirmed.

The primary analyses included all self-reported first onset strokes except those that were disconfirmed by neurologist adjudication, as well as fatal strokes identified through NDI, for a total of 1,555 incident cases in 42,924 women (Figure 1). A secondary analysis was conducted using only confirmed cases (N=576).

Covariates:

We adjusted for variables identified a priori as pre-pregnancy confounding factors: age at first birth (<20, 20–24, 25-29, 30-34, ≥35 years), education (<12, 12–15, 16, ≥17 years), body mass index (BMI) at age 18 years (weight in kg divided by height in meters squared; continuous), and parental history of stroke (yes, no). We assigned “No” to those women who did not check “yes” to parental history of stroke. BMI at age 18 years was used as a proxy for pre-pregnancy BMI because questionnaires had not specifically asked about BMI before the first pregnancy and 87% percent of first pregnancies in the BWHS occurred after age 18. Adult height was assessed in the baseline questionnaire and weight was assessed in every questionnaire cycle. Self-reported weight and height were validated among 115 BWHS participants, with Spearman correlations for weight and height being 0.93 and 0.97, respectively, for self-report versus technician measures¹⁸.

Participants were classified as having a history of hypertension at baseline if they reported physician-diagnosed hypertension together with use of an antihypertensive medications (beta-blocker, angiotensin-converting-enzyme inhibitor, calcium channel blocker, angiotensin receptors blocker, renin inhibitor) or diuretic, or reported use of an antihypertensive medication alone. In a prior validation study, 138 out of 139 (99%) self-reports of hypertension were confirmed by medical records¹⁹. Participants were classified as having a history of type 2 diabetes if they reported physician-diagnosed diabetes with onset at age 30 or older at baseline. In a validation study, 217 (94%) of 229 women who reported diabetes were confirmed by their physicians to have type 2 diabetes²⁰.

Statistical analyses:

Eligible women contributed person-years from enrollment in 1995 (or year of first birth if their first birth occurred after enrollment) until diagnosis of incident stroke, last questionnaire completed, death, or the end of follow up in December 2019, whichever came first. For women who were already parous by the time they enrolled in 1995, follow-up began when they enrolled in 1995, and exposure status was updated over time if later pregnancies occurred. For women who were nulliparous in 1995 but had a pregnancy after that time, follow-up began at the time of the onset of their first pregnancy and exposure status was updated over time if they had additional pregnancies. Analyses were censored on the occurrence of the birth of twins or triplets because such pregnancies are associated with increased risk of HDOP and also associated with higher cardiovascular disease risk, but not necessarily through the relation with HDOP. Cox proportional hazards models were used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs) for stroke risk in women with a history of preeclampsia or gestational hypertension relative to risk in women who had normotensive pregnancies only. In addition to the primary analysis, which included all cases, we conducted a secondary analysis restricted to confirmed cases only. We conducted a sensitivity analysis restricted to women who had only one term pregnancy in order to remove any potential confounding by number of births or birth order of the preeclamptic pregnancy, for which data were not available.

In multivariable models, we additionally adjusted for BMI at age 18 (continuous, kg/m²), age at first birth (<20, 20-24, 25-29, 30+ years), educational attainment (<12, 12-15, 16, ≥17 years), and parental history of stroke (yes, no). It would have been ideal to adjust for potential confounding by cardiometabolic status at the time of pregnancy, but that was not possible because most participants had already completed childbearing by the time of enrollment in 1995. Only 14% of parous women gave birth for the first time after enrollment in the study.

We conducted analyses stratified on age (<55, ≥55 years), BMI (<25, 25-29, ≥30 kg/m²), hypertension, and residence in the Stroke Belt (Alabama, Arkansas, Georgia, Indiana, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia²¹). We tested effect modification by creating interaction terms between HDOP and the stratified terms, and tested interaction using the Wald test. For missing data in covariates, the missing indicator method was used. Sensitivity analyses were carried out using the complete case method.

We calculated the population attributable risk fraction (PAF) and its 95% confidence interval using the formula $PAF=\frac{P_{pop}\times \left(RR-1\right)}{P_{pop}\times \left(RR-1\right)+1}$ (Ppop= the proportion of women with HDOP in the entire study population; RR=the multivariable hazard ratio for ever HDOP pregnancy relative to only normotensive pregnancies).

Results

Women who had had HDOP were similar to parous women whose pregnancies were normotensive with regard to age at first birth, parity, education, parental history of stroke, geographic region, and smoking status (Table 1). At the time of enrollment in 1995, 19% of women with HDOP had BMI at age 18 ≥ 25 kg/m², compared with 11% for normotensive women; 33% of women with HDOP had hypertension, compared with 15% for normotensive women.

Table 1.

Age-standardized baseline^† characteristics of parous Black Women’s Health Study (BWHS) participants by history of hypertensive disorder of pregnancy (HDOP)

Characteristics	No history of hypertensive disorder of pregnancy (n=37,986)	History of hypertensive disorder of pregnancy (n=4,938)	Standardized difference
Age at enrollment in 1995, mean (SD)	40.25 (10.88)	38.47 (10.14)	−0.168
Age at first birth, years (%)*			0.123
< 20	30	28
20 - 24	31	32
25 - 29	20	22
30 - 34	11	11
≥ 35	5	6
Body mass index (BMI) at age 18, kg/m2 (%)*			0.265
< 20	43	34
20 - 24	44	46
≥ 25	11	19
Education, years (%)*			0.065
< 12	3	3
12 – 15	56	57
16	21	20
≥ 17	20	20
Number of births (%)			0.085
1	43	39
2	31	31
≥ 3	26	30
Parental history of stroke (%)	17	20	0.059
Geographic region (%)
Northeast	26	24	0.156
South	30	32
Midwest	24	24
West	19	19
Stroke belt ††	25	28	0.067
Hypertension (%)	15	33	0.366

^†Baseline was year of enrollment in 1995 for 86% of participants and year of first birth for the 14% of participants who gave birth for the first time after enrollment

^*Proportion of participants with missing data: age at first birth, 1.4%; BMI at age 18, 1.8%; years of education, 0.2%.

^††Stroke belt was defined as Alabama, Arkansas, Georgia, Indiana, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, and Virginia

In the primary analyses, the age-adjusted HR for HDOP versus no HDOP was 1.72 (95% CI 1.52-1.95, Table 2); after additional control for age at first birth, BMI at age 18, educational attainment, and parental history of stroke, the HR was 1.66 (95% CI 1.46-1.89). The comparable multivariable HRs were 1.53 (95% CI 1.29-1.82) for preeclampsia and 1.81 (95% CI 1.53-2.13) for gestational hypertension without preeclampsia (Table 2). Twelve of the strokes (<0.01%) had occurred within one year of pregnancy; results were unchanged with exclusion of those women.

Table 2.

Hypertensive disorders of pregnancy in relation to incidence of stroke in the Black Women’s Health Study, 1995-2019

Hypertensive disorder of pregnancy	Stroke cases / person-years	Incidence rate	Age-adjusted HR (95% CI)	Multivariable* HR (95% CI)
No	1,245/692,670	1.80	1.00 (reference)	1.00 (reference)
Yes	310/109,495	2.83	1.72 (1.52, 1.95)	1.66 (1.46, 1.89)
Preeclampsia	148/60,012	2.47	1.56 (1.31, 1.85)	1.53 (1.29, 1.82)
Gestational hypertension only	162/49,483	3.27	1.90 (1.61, 2.24)	1.81 (1.53, 2.13)

^*Adjusted for age, age at first birth, body mass index at age 18, educational attainment, and parental history of stroke

HR, hazard ratio; CI, confidence interval. Incidence rate: per 1000 person-years.

In a secondary analysis restricted to 576 confirmed strokes (Table 3), the multivariable HR for any HDOP relative to none was 1.73 (95% CI 1.41-2.13) and the comparable HRs for preeclampsia and gestational hypertension were 1.63 (95% CI 1.23-2.16) and 1.85 (95% CI 1.41-2.43) respectively.

Table 3.

Association of hypertensive disorder of pregnancy with incidence of stroke in the Black Women’s Health Study, 1995-2019, confirmed cases only

Hypertensive disorder of pregnancy	Stroke cases / person-years	Incidence rate	Age-adjusted HR (95% CI)	Multivariable* HR (95% CI)
No	460/705,263	0.65	1.00 (reference)	1.00 (reference)
Yes	116/112,803	1.03	1.78 (1.45, 2.18)	1.73 (1.41, 2.13)
Preeclampsia	56/61,378	0.91	1.67 (1.26, 2.21)	1.63 (1.23, 2.16)
Gestational hypertension only	60/51,425	1.17	1.89 (1.44, 2.48)	1.85 (1.41, 2.43)

^*Adjusted for age, age at first birth, body mass index at age 18, educational attainment, and parental history of stroke

HR, hazard ratio; CI, confidence interval. Incidence rate: per 1000 person-years.

There were 393 incident stroke cases among uniparous women (Table 4) and the HR for stroke risk for any HDOP versus no HDOP was 1.80 (95% CI 1.39-2.34). The comparable HRs for preeclampsia and gestational hypertension were 1.83 (95% CI 1.29-2.58) and 1.77 (95% CI 1.24-2.53).

Table 4.

Association of hypertensive disorders of pregnancy with stroke incidence among participants who had only one birth

Hypertensive disorder of pregnancy	Stroke cases / person-years	Incidence rate	Age-adjusted HR (95% CI)	Multivariable* HR (95% CI)
No	320/247,939	1.29	1.00 (reference)	1.00 (reference)
Yes	73/33,464	2.18	1.90 (1.47, 2.46)	1.80 (1.39, 2.34)
Preeclampsia	38/18,721	2.03	1.88 (1.34, 2.64)	1.83 (1.29, 2.58)
Gestational hypertension only	35/14,743	2.37	1.92 (1.35, 2.73)	1.77 (1.24, 2.53)

^*Adjusted for age, age at first birth, body mass index at age 18, educational attainment, and parental history of stroke

HR, hazard ratio; CI, confidence interval. Incidence rate: per 1000 person-years.

In age-stratified analyses, HRs for HDOP and stroke risk were 1.88 (95% CI 1.55-2.26, Table S1) among women under 55 and 1.52 (95% CI 1.28-1.80) among women age 55 and older. Associations were consistent across strata of BMI (Table S2) and residence in the Stroke Belt (Table S3). In analyses stratified by hypertension in 1995 (Table S4), an association of HDOP with increased stroke risk was present both among women with hypertension and those without. Our results and conclusions remained the same after sensitivity analyses using the complete case method for missing data in covariates.

We estimated that 7% (95% CI 5%−9%) of stroke cases in this population were attributable to HDOP. Population attributable fractions were 6% (95% CI 3%−9%) for preeclampsia, and 9% (95% CI 6%−12%) for gestational hypertension.

Discussion

In this prospective cohort study of 42,924 Black women, women who reported a history of HDOP were estimated to have 1.66 times the risk of stroke relative to parous women who had not had those pregnancy complications; for history of preeclampsia in particular, the estimated HR was 1.53. The association was present across strata of age, BMI, and residence in the Stroke Belt. Approximately 7% of stroke cases were estimated to be attributable to HDOP.

Pregnancy is a unique time period for women, and the risk of stroke is elevated during the third trimester and postpartum period due to cardiometabolic physiological changes^22,23. Preeclampsia triggers a cascade of downstream vascular, endothelial, and coagulation dysfunction events that may lead to long-term increased stroke risk²⁴. In studies based primarily on White women^25–32, preeclampsia has been estimated to be associated with 1.5 to 2 times the risk of all cardiovascular disease and of stroke specifically. In a single study of Black women (42 incident strokes) HDOP was associated with 1.7 times the risk of stroke within 5 years of the delivery³³. Although previous findings from studies of White women may well be generalizable to Black women, this has not been demonstrated until now. In our study of 42,924 parous Black women with a median of 22 years of follow up, we found a 1.7-fold increased stroke risk associated with HDOP, 1.5-fold increased stroke risk for preeclampsia and 1.8-fold increased stroke risk for gestational hypertension. The slightly stronger association for gestational hypertension than for preeclampsia in our study is consistent with findings from prior studies in White women^32,34. We speculate that HDOP subtypes may be more complicated than representing simply a spectrum of severity and may represent different disease phenotypes.

Strengths of the study include the large sample size of Black women, large number of incident strokes, with stroke outcomes adjudicated by neurologist review of medical records. Other strengths include the prospective cohort study design, length of follow up, as well as the ability to examine associations across different subgroups such as geographic regions (including within the Stroke Belt).

The study has limitations. It was not possible to obtain medical records for all participants who reported stroke, either because individuals would not sign a medical record release or because appropriate records could not be obtained from the hospitals. While it was possible to confirm stroke events for participants who self-reported stroke, there may have been women had stroke events but did not report them and would have been incorrectly classified as non-cases. We think that such misclassification would have been nondifferential with regard to HDOP, and, regardless, would have had very little effect, if any, on analysis results because there were over 37,000 non-cases, among whom only a very few would represent true stroke cases. Hazard ratios were very similar in analyses from the overall sample of self-reported cases (N= 1,555) and the sample of confirmed cases (N= 576), with HRs of 1.66 (1.46, 1.89) and 1.73 (1.41, 2.13), respectively. Preeclampsia and gestational hypertension were self-reported, allowing the possibility of exposure misclassification and underreporting of exposure. The literature suggests modest accuracy of self-reported, preeclampsia^35–39 , history of gestational hypertension⁴⁰, birth⁴¹ , and pregnancy characteristics^42–44. ⁴⁵. It was not possible to obtain pregnancy medical records for all BWHS participants to validate self-reports. However, misclassification of exposure due to underreporting of these conditions would have biased results in the direction of no association rather than towards a positive association. It is possible that some women who reported gestational hypertension may actually have had undiagnosed chronic hypertension; such misclassification may have led to inflated estimates of the relative risk. However, given the young age at first pregnancy in this cohort (22 years), the number of women affected may be minimal. In the BWHS, we do not have detailed data on each pregnancy. However, in our sensitivity analyses restricting to women with only one birth, the magnitude of association was similar to that in the larger group of women studied. A limitation of the study is lack of data on the exact timing of each pregnancy and the pregnancy in which preeclampsia occurred. Therefore, it was not possible to do a time-varying analysis for pregnancies that occurred before entry into the study. Instead we began follow-up in 1995 (year of enrollment for all participants) for all women who had had at least one pregnancy before enrolling in the study and at the time of the first pregnancy for women whose first birth occurred after enrollment. Because women who experienced a stroke before enrollment in the study were excluded from the analyses, there may have been some “depletion of susceptibles”. If this occurred, it would lead to underestimation of the true association between HDOP and stroke risk.

Women who developed HDOP may already have had a worse underlying cardiometabolic risk profile before pregnancy. Despite our efforts to control for most potential confounders, we did not have detailed information on pre-pregnancy cardiometabolic factors for most of the women, and therefore were not able to adjust for these factors.

Black women are disproportionately affected by both stroke and pregnancy complications. Our study provides evidence that pregnancy history may be an important factor for risk assessment and prevention of long-term stroke. Cardiovascular screening recommendations for Black women in particular should take history of HDOP into account.

Conclusion

In this large prospective study of Black women, a vulnerable population with a high prevalence of preeclampsia and at high risk for stroke relative to other U.S. racial groups, women who had had a preeclamptic pregnancy were estimated to have a 66% increased risk of developing stroke many years later. Our results may explain, in part, the disproportionately high incidence of stroke in Black women relative to other populations.