Women with Adverse Pregnancy Outcomes Have Higher Odds of Midlife Stroke: The Population Assessment of Tobacco and Health Study

Abstract

Background:

A history of adverse pregnancy outcomes (APOs) is associated with increased risk of future cardiovascular disease, including stroke. Few large U.S. population-based surveys included data on APOs.

Methods:

The Population Assessment of Tobacco and Health study is a nationally representative survey of 45,971 U.S. respondents. Female respondents ≥50 years old who reported pregnancy history at the 2013–2014 baseline interview were included in this cross-sectional analysis (n = 3,175; weighted n = 35,783,619). The primary exposure was a history of ≥1 APO, including preterm delivery, low birth weight, preeclampsia, placental abruption, and stillbirth. The primary outcomes were (1) stroke before age 60 and (2) any stroke. We used weighted logistic regression models to estimate odds ratios (OR) and 95% confidence intervals (95% CI) for the association between APO and stroke, adjusting for age, race/ethnicity, socioeconomic status, parity, and vascular risk factors.

Results:

Among stroke-free respondents, 15% reported ≥1 APO. Among women who reported a stroke before age 60, 39% reported ≥1 APO (p < 0.001); among women reporting stroke at any age, 25% reported ≥1 APO (p = 0.01). Controlling for covariates, women with APOs had increased odds of stroke before age 60 (adjusted OR 2.66, 95% CI 1.49, 4.75). The association of APOs with stroke at any age was not significant after controlling for covariates (adjusted OR 1.57, 95% CI 0.93, 2.64).

Conclusion:

In this analysis of U.S. nationally representative survey data, APOs were independently associated with midlife stroke. Women with APOs have higher odds of midlife stroke and warrant targeted prevention strategies.

Introduction

Stroke is a leading cause of death and disability in women in the United States. While the incidence of stroke in older age groups is decreasing, stroke and stroke risk factors in younger and middle-aged U.S. women are on the rise.^1,2 Adverse pregnancy outcomes (APOs) such as preterm delivery and preeclampsia have been associated with higher risk of future cerebrovascular disease in women, including stroke and vascular dementia.^3–5 Other APOs, such as gestational hypertension, fetal growth restriction, placental abruption, and stillbirth, have also been shown to be associated with incident hypertension and cardiovascular disease, including stroke.^6–8

However, estimates of the impact of APOs on future maternal cerebrovascular disease in the United States have been limited by the limited obstetric history included in nationally representative survey data (e.g., National Health and Nutrition Examination Survey [NHANES]).⁹ Several longitudinal U.S. studies that included data on APOs, such as the Nurses' Health Study II,^10,11 the California Teachers Study,¹² the Women's Health Initiative,¹³ or the CARDIA (Coronary Risk Development in Young Adults) study,¹⁴ have had limited racial, ethnic, geographic, and/or socioeconomic representation.

Major racial, socioeconomic, and geographic disparities exist for both APOs and stroke; therefore, nationally representative data are needed for estimates of the impact of APOs on future cerebrovascular health, including the role of race/ethnicity and social determinants of health as moderators.^15–21 In addition, U.S. survey data suggest that before age 60, stroke prevalence in women may be comparable or even higher than in men, a trend that reverses in the 60–80-year age group, suggesting a possible impact of sex-specific risk factors in the younger age group.²²

The Population Assessment of Tobacco and Health (PATH) study is an ongoing, publicly funded, longitudinal cohort study launched in 2011, designed as a nationally representative U.S. sample.²³ We hypothesized that U.S. women with self-reported history of APOs would have increased odds of (1) stroke before age 60 and (2) any stroke. We also hypothesized that the association of APOs with stroke would be partially mediated by hypertension. We further explored whether these associations varied by race/ethnicity, income, and education.

Methods

The PATH study is a U.S. national longitudinal study of 45,971 adults and youth aged 12 and older at the time of Wave 1 data collection in 2013–2014,²³ funded jointly by the National Institutes of Health, the National Institute on Drug Abuse, and the U.S. Food and Drug Administration. The study sample is drawn from all nine U.S. regions (Northeast, Mid-Atlantic, Northeast Central, Northwest Central, South Atlantic, Southeast Central, Southwest Central, Mountain, and West Coast), and is representative of the U.S. civilian, noninstitutionalized population aged 12 and older in 2013. A detailed description of the survey design is available in the User's Guide to Restricted Use Files (http://doi.org/10.3886/ICPSR36231).

Wave 1 baseline data were collected via in-person and computer-assisted interviews using validated questionnaire tools,²³ with participants reporting medical history, including a history of stroke and vascular risk factors, as well as detailed pregnancy history, including a history of APOs. For this cross-sectional analysis, only data collected from the Wave 1 interview were used.

We included all adult women in our analysis who were aged 50 or older at the time of the baseline (Wave 1) interview, and who answered the Wave 1 questions: “Ever had pregnancy outcome of stillbirth” and “Ever had outcome for live birth” for the following pregnancy complications: preterm delivery, preeclampsia, low birth weight, and placental abruption. All pregnancy complications were ascertained solely from self-report based on response to this question.

We excluded women who reported at least one pregnancy and no live births, but did not report whether or not they experienced stillbirths, and men, who were not at risk of APO. We also excluded parous women in the study sample who did not respond to the question regarding pregnancy complications. Women who reported ≥1 pregnancy ending in stillbirth, but no live births, were included in our sample and considered nulliparous. We compared characteristics of women with and without missing pregnancy history data or covariates to estimate the extent to which missing data might bias our sample.

The primary exposure of interest was self-reported history of any APO, defined as one or more of the reported pregnancy complications, including stillbirth, preterm delivery, preeclampsia, low birth weight, and placental abruption. Due to PATH study policy regarding small cell sizes, in analyses we combine placental abruption, preeclampsia, and stillbirth into one category of “other complications.” The primary outcomes of interest were (1) stroke before the age of 60 and (2) any self-reported stroke. Stroke types (hemorrhagic vs. ischemic) were not available for analysis.

To help ensure that the outcome (stroke) occurred after the primary exposure of interest (APO), we considered women to have met the outcome only if they reported a stroke either after the year of the most recent reported pregnancy, or after age 50, if age at last pregnancy was not reported. Prespecified covariates of interest included self-reported hypertension with onset either after the last pregnancy or after age 50 as a potential mediator, as well as age, race/ethnicity, U.S. geographic region, parity, income, level of education, and self-reported vascular risk factors, including obesity, diabetes, cigarette smoking, and high cholesterol, reported at the time of baseline interview.

Due to the small cell sizes of Hispanic women, we combined Hispanic women with other race women for analysis. We compared demographics and clinical characteristics between women who did and did not report a stroke, using chi-square tests for categorical variables and Student's t-test for continuous variables.

We created multivariable logistic regression models to estimate the odds ratios (OR) and 95% confidence intervals (95% CI) for the association of APOs with subsequent stroke, including an unadjusted model (Model 1) followed by nested models adjusting for the addition of age and race/ethnicity (Model 2), income and education level (Model 3), and parity and vascular risk factors that were significant in the univariable analysis (Model 4). To assess whether hypertension might mediate the association of APOs with stroke, we compared effect sizes in models with hypertension (Model 5) and without (Model 4).

To assess whether the relationships between any APO and stroke outcomes varied by race/ethnicity, income, and education, we ran single models with interaction terms as well as stratified models for black and white women (the group of Hispanic and other race women was too small to produce statistically stable results), women in the lowest quartile of income versus women in all other income quartiles, and women who completed through high school versus women who completed beyond high school. We performed a number of exploratory post-hoc sensitivity analyses, including adding parity as a potential confounder or mediator and using the outcome of stroke age <65 years. All results were weighted to account for the complex survey design.²³ A p-value <0.05 for main analyses and a p-value <0.20 for interaction analyses were considered statistically significant.

Standard protocol approvals, registrations, and patient consents

Approval was obtained from the Westat Institutional Review Board before initiation of the PATH study, and all participants provided informed consent before enrollment. For this analysis, a Restricted Data Use Agreement was executed between the Inter-university Consortium for Political and Social Research (ICPSR) and Columbia University, and approval was obtained from Columbia University Irving Medical Center's Institutional Review Board before initiation of data analysis.

Data availability statement

For this analysis, we made use of the restricted-use data set of the PATH study, which is available to qualified researchers upon request to ICPSR and approval of an appropriate Restricted Data Use Agreement. The restricted-use data set differs from the unrestricted, publicly available data set in that additional details of medical and pregnancy history are included, such as participant age and age at which significant events occurred. Data were provided via ICPSR's Virtual Data Enclave (VDE), and all analyses were conducted within the VDE using Stata/MP version 15.1 (College Station, TX, USA). Statistical code used for this analysis will be made available within the VDE to qualified researchers using the restricted PATH data set.

Results

There were 3,175 eligible women age ≥50 years with complete baseline data in the study sample, representing a weighted population of 37 million women from all U.S. regions (mean age 63 years [standard deviation (SD) 8], 74% non-Hispanic white, 11% non-Hispanic black, and 15% Hispanic or other race). A flow diagram of participants included in the analysis is shown in Figure 1. Demographics and clinical characteristics of the weighted study population are shown in Table 1. An additional 10,650,554 women (weighted population) were missing analysis covariates. Comparing the combined group of respondents included in the analysis sample and respondents missing analysis covariates with respondents included in the analysis sample, women with missing data were older and of lower income (Table 2).

FIG. 1.

Weighted sample size selection. Flow diagram of analysis sample selection. All n's are weighted.

Table 1.

Characteristics of Women in the Population Assessment of Tobacco and Health Study Aged ≥50 at Time of Wave 1 Interview (2013–2014), With and Without Postpregnancy Stroke

Characteristica	Full analysis sample, n%	No stroke, n%	Any stroke, n%	p	No stroke <60 years, n%	Stroke <60 years, n%	p
Population distribution	100	97	3		98	2
Pregnancy complication
Any	15	15	25	0.01	15	39	<0.001
Preterm birthb	7	7	14	0.03	7	20	0.002
Low birth weightb	8	8	10	0.61	8	18	0.03
Otherc	5	5	8	0.09	5	15	0.007
Age, mean	63 (8)	63 (8)	68 (9)	< 0.001	64 (8)	62 (7)	0.14
Race/ethnicity				0.15			0.03
White, non-Hispanic	74	74	71		74	60
Black, non-Hispanic	11	10	17		11	23
Hispanic/other race	15	15	12		15	17
Education				0.02			0.04
< High school	17	17	26		17	22
High school	29	28	33		29	25
Some college	31	31	30		30	44
College or more	24	24	11		24	9
Income				< 0.001			<0.001
≤$14,999	21	20	43		20	61
$15,000–$34,999	24	24	33		25	18
$35,000–$74,999	28	29	11		28	12
≥$75,000	27	27	13		27	9
Parityd				0.02			0.50
0–1	18	18	14		18	16
2	38	38	26		38	29
3	24	24	27		24	28
≥4	20	20	34		20	27
Region				0.85			0.94
North East	7	7	6		6	8
Mid-Atlantic	12	12	11		12	11
East North Central	15	15	16		15	13
West North Central	7	7	5		7	3
South Atlantic	20	20	24		21	19
South East Central	7	7	10		7	10
West South Central	10	10	12		10	11
Mountain	9	9	6		9	8
West Coast	13	13	10		13	16
Cigarette use				0.01			0.001
Never	31	31	25		31	21
Former established	27	26	27		26	45
Current established	13	13	13		13	20
Former experimental	28	28	27		28	9
Current experimental	2	2	2		2	5
Baseline conditions
Hypertension				<0.001			<0.001
None	55	57	16		56	16
Onset before last pregnancy	3	3	9		3	9
Onset after last pregnancy	42	41	75		41	75
Diabetes	23	22	45		22	60	<0.001
High cholesterol	39	38	59	0.002	39	66	<0.001
Baseline body mass index				0.73			0.70
<25.0	35	35	33		35	29
25.0–29.9	32	32	36		32	35
≥30.0	33	33	32		33	36

^a Results are weighted for complex survey design, N = 35,783,619.

^b Weighted sample N for parous respondents = 34,488,162.

^c “Other” includes preeclampsia, placental abruption, and stillbirth.

^d Women reporting ≥1 pregnancy ending in stillbirth, and no live births, were included with parity 0. Parity 0 and 1 were combined for outcomes of any stroke and stroke <60 years due to small cell sizes.

Table 2.

Characteristics of Women in the Population Assessment of Tobacco and Health Study With and Without Baseline Missing Data on Pregnancy History, Analysis Covariates

Characteristica	In analysis sample, n%	Not in analysis sample, with missing covariates, n%	Weighted sample in analysis or not in analysis with missing covariates	p
Pregnancy complication
Any	15	16	46,434,173	0.69
Preterm birthb	7	6	44,832,984	0.25
Low birth weightb	8	9	44,832,984	0.81
Otherc	5	5	46,434,173	0.84
Age, mean	63 (8)	66 (8)	46,434,173	<0.001
Race/ethnicity			45,115,796	0.53
White, non-Hispanic	74	72
Black, non-Hispanic	11	10
Hispanic/other race	15	17
Education			46,434,173	0.02
< High school	17	21
High school	29	32
Some college	31	27
College or more	24	19
Income			39,370,382	0.01
<$15,000	21	29
$15,000–$34,999	24	30
$35,000–$74,999	28	22
≥$75,000	27	19
Parity			46,434,173	0.67
0	4	3
1	14	15
2	38	36
3	24	24
≥4	20	22
Region			46,434,173	0.44
North East	7	7
Mid-Atlantic	12	12
East North Central	15	16
West North Central	7	8
South Atlantic	20	19
South East Central	7	8
West South Central	10	9
Mountain	9	6
West Coast	13	14
Cigarette use			46,119,062	0.01
Never	31	38
Former established	27	23
Current established	13	10
Former experimental	28	26
Current experimental	2	2
Baseline conditions
Hypertension			46,434,173	0.95
None	55	55
Onset before last pregnancy	3	3
Onset after last pregnancy	42	42
Diabetes	23	24	464,424,216	0.72
High cholesterol	39	40	46,434,173	0.83
Baseline body mass index			44,830,497	0.14
<25.0	35	37
25.0–29.9	32	35
≥30.0	33	29

^a Results are weighted for complex survey design.

^b Weighted sample N for parous respondents.

^c “Other” includes preeclampsia,, placental abruption, and stillbirth.

Of women in the weighted study sample, 15% reported a history of APOs. A total of 7% of women reported preterm delivery, 8% reported low birth weight, and 5% reported placental abruption, preeclampsia, or stillbirth. A total of 3% of the sample population reported a stroke after the last pregnancy, with a mean age at stroke of 58 years (SD 11).

Stroke before age 60 was reported by 5% of women with APOs, versus 1% of women without APOs (p < 0.01), with a mean age at stroke of 49 years (SD 8). Stroke at any age was reported by 6% of women with APOs and 3% of women without APOs (p = 0.01). Of women who reported stroke before age 60, 39% reported a history of APO; among women who reported any stroke, 15% reported a history of APO (p < 0.001). No one reported stroke within 1 year postpartum. Of those without a history of APOs, mean age at stroke was 61 years (SD 11) compared with mean age at stroke of 52 years (SD 9) for those with a history of APOs (p = 0.01).

Compared with women who did not report a history of stroke, women who reported a history of any stroke were older, had higher parity, and lower income levels (Table 1). Women who reported a history of any stroke also reported higher proportions of vascular risk factors, including hypertension onset after last pregnancy (75% vs. 41%, p < 0.001), diabetes (45% vs. 22%, p < 0.001), and high cholesterol (59% vs. 38%, p = 0.001). These between-group differences were seen in both comparison groups (“any stroke” and “stroke before age 60”), although proportions differed, and difference in parity was not statistically significant in the group with stroke before age 60, compared with those without stroke before age 60.

A higher proportion of women with stroke before age 60 were black, compared with women without stroke before age 60. There were no significant regional differences between the stroke and nonstroke groups, and no significant differences in body mass index. The combined analysis sample and sample missing covariates had a similar prevalence of stroke at any age compared with the analysis sample (4.2% vs. 3.4%, respectively; p value = 0.42) and stroke age <60 years (1.3% vs. 1.9%, respectively; p value = 0.33).

Unadjusted and adjusted ORs with 95% CIs for the association of APOs with future stroke are shown in Table 3, with nested models adjusting for age and race/ethnicity, education and income, and parity and vascular risk factors. In Model 4, adjusting for demographic and vascular risk factors, APOs were associated with a 2.66-fold higher odds of stroke before age 60 (95% CI 1.49, 4.75), and a more modest and nonsignificant 1.57-fold higher odds of any future stroke (95% CI 0.93, 2.53). These relationships remained similar when accounting for hypertension (Model 5).

Table 3.

Association of Self-Reported History of Adverse Pregnancy Outcomes with Odds of Subsequent Stroke in U.S. Women Aged ≥50 Years

Outcome: Stroke <60 years	Model 1 OR (95% CI)	Model 2 OR (95% CI)	Model 3 OR (95% CI)	Model 4 OR (95% CI)	Model 5 OR (95% CI)
Any APOa	3.67 (1.91, 7.07)	3.50 (1.81, 6.76)	3.28 (1.74, 6.21)	2.66 (1.49, 4.75)	2.62 (1.49, 4.63)
Pretermb	3.39 (1.43, 8.01)	3.35 (1.43, 7.84)	3.63 (1.57, 8.41)	2.62 (1.18, 5.81)	2.41 (1.13, 5.17)
Low birth weightb	2.42 (0.99, 5.87)	2.30 (0.95, 5.59)	2.21 (0.95, 5.11)	1.86 (0.82, 4.21)	1.80 (0.82, 3.98)
Other complicationsa	3.46 (1.58, 7.57)	3.41 (1.56, 7.46)	2.68 (1.16, 6.19)	2.26 (0.93, 5.55)	2.55 (1.04, 6.26)
Outcome: stroke all ages
Any APOa	1.92 (1.12, 3.32)	1.93 (1.11, 3.36)	1.83 (1.07, 3.12)	1.57 (0.93, 2.64)	1.46 (0.85, 2.53)
Pretermb	2.24 (1.03, 4.91)	2.30 (1.05, 5.07)	2.41 (1.12, 5.17)	1.96 (0.92, 4.22)	1.83 (0.87, 3.85)
Low birth weightb	1.22 (0.54, 2.73)	1.17 (0.51, 2.64)	1.10 (0.50, 2.43)	1.00 (0.47, 2.15)	0.91 (0.42, 1.99)
Other complicationsa	1.78 (0.88, 3.61)	1.76 (0.85, 3.70)	1.54 (0.74, 3.21)	1.38 (0.64, 2.96)	1.40 (0.64, 3.04)
Outcome: stroke <65 years
Any APOa	2.35 (1.76, 6.00)	3.14 (1.70. 5.80)	2.95 (1.63, 5.35)	2.38 (1.36, 4.18)	2.31 (1.31, 4.08)
Pretermb	3.33 (1.48, 7.47)	3.32 (1.49, 7.41)	3.57 (1.60, 7.97)	2.61 (1.20, 5.67)	2.44 (1.14, 5.23)
Low birth weightb	1.98 (0.82, 4.82)	1.90 (0.78, 4.59)	1.79 (0.78, 4.14)	1.51 (0.67, 3.41)	1.46 (0.64, 3.31)
Other complicationsa	2.94 (1.39, 6.22)	2.90 (1.38, 6.09)	2.31 (1.06, 5.04)	1.93 (0.83, 4.46)	2.04 (0.88, 4.73)

Model 1 is an unadjusted model.

Model 2 adjusts for age and race/ethnicity.

Model 3 additionally adjusts for education and income.

Model 4 additionally adjusts for baseline diabetes and cholesterol, parity, and smoking status.

Model 5 additionally adjusts for post-pregnancy onset hypertension.

^a Weighted N = 35,783,619.

^b Weighted N = 35,895,082 for parous respondents.

APO, adverse pregnancy outcome; CI, confidence intervals; OR, odds ratio.

Stratified models suggested that the association between APO and stroke at any age was stronger among women of lowest income (OR 2.84, 95% CI 1.18, 6.80) compared with women of other income levels (OR 0.79, 95% CI 0.29, 2.14, p-value for interaction from single model = 0.07). The relationships between APO and both stroke outcomes did not vary by race or educational attainment Table 4. In sensitivity analyses adding parity to the saturated models for the outcomes of stroke at any age, stroke < age 60, and stroke < age 65, results were consistent (Supplementary Table S1).

Table 4.

Interactions Between Self-Reported Adverse Pregnancy Outcome and Race, Income, and Education

	Any stroke OR (95% CI)	Interaction p-value	Stroke <60 years OR (95% CI)	Interaction p-value	Stroke <65 years OR (95% CI)	Interaction p-value
Racea		0.44		0.44		0.37
Black women	2.08 (0.65, 6.67)		3.16 (.90, 11.12)		3.16 (0.90, 11.12)
White women	1.10 (0.54, 2.23)		1.65 (0.72, 3.80)		1.48 (0.67, 3.27)
Incomeb		0.07		0.31		0.28
Lowest quartile (Q1)	2.84 (1.18, 6.80)		3.52 (1.34, 9.25)		3.43 (1.32, 8.94)
All other quartiles (Q2–Q4)	0.79 (0.29, 2.14)		1.58 (0.58, 4.29)		1.53 (0.57, 4.13)
Educationb		0.99		0.57		0.63
Through high school	1.64 (0.76, 3.57)		3.08 (1.11, 8.57)		2.73 (1.04, 7.14)
Greater than high school	1.39 (0.59, 3.25)		2.00 (0.84, 4.78)		1.75 (0.74, 4.14)

All models adjust for age, race/ethnicity, education, income, and baseline diabetes, cholesterol, and smoking status.

^a Models do not include Hispanic and other race women due to convergence problems related to small number of outcomes. Weighted N = 35,783,619.

^b Models include Hispanic and other race women. Weighted N = 35,783,619.

Discussion

In this analysis of a nationally representative sample of U.S. women, we demonstrate that women who reported a history of APOs had more than twice the odds of stroke before the age of 60, even after adjusting for other risk factors. Furthermore, hypertension did not appear to be a major mediating factor in this analysis. The relationship between history of APOs and stroke at any age was stronger among women of lower income.

Common APOs such as preterm delivery and preeclampsia are sex-specific risk factors that have been consistently associated with future maternal cardiovascular disease, particularly hypertension.^3,4 However, the relationship between these conditions and stroke, specifically, is less well-described. Furthermore, other APOs, such as low birth weight and stillbirth, have not previously been specifically associated with future maternal stroke risk.

Stroke is often included in a composite outcome of “cardiovascular disease,” and few studies have specifically looked at stroke as an outcome, let alone evaluated differences between subgroups, age at stroke onset, or stroke subtype (ischemic vs. hemorrhagic). Those that have, have primarily been in non-U.S. populations^24,25 or subpopulations in the United States.^3,4 However, prior studies have consistently demonstrated an association between APOs and future stroke. The association is most well characterized for preeclampsia, where some studies have found up to triple the risk of stroke in women with a preeclampsia history.⁷

In terms of other APOs, women with small-for-gestational-age babies had 30% higher stroke risk²⁴; gestational hypertension, up to 80% higher risk⁷; and preterm delivery, up to 65% higher risk.⁴ Recently, data from the Women's Health Initiative, a large, diverse U.S. cohort, were published, again demonstrating a consistent association between history of APOs and future atherosclerotic cardiovascular disease (defined as myocardial infarction, stroke, peripheral artery disease, or coronary revascularization).²⁶ Stroke was not considered separately. The paucity of granular data and the heterogeneity of study designs, outcomes, and effect estimates underscore the urgent need for rigorous prospective studies to characterize the relationship between APOs and future stroke.

Remarkably, nearly 4 in 10 women in our sample who suffered a stroke before the age of 60 had a history of APOs. Our results suggest that common APOs may offer an early glimpse into women's cerebrovascular health, and a unique opportunity to implement primary preventive strategies in a population at high risk of future stroke.

Primary prevention of stroke in women is of paramount importance. Stroke affects more women than men,²⁷ and stroke in midlife can have a devastating impact on a woman's health and well-being. The mean age at stroke was 49 among women in our sample who reported a stroke before age 60. Women in their 40s and 50s are in their prime working years, and often juggle multiple responsibilities, including childcare and eldercare.²⁸ Women are more likely to be single parents with sole caregiver responsibilities²⁹; a disabling stroke in midlife may therefore have ripple effects with ramifications extending to multiple generations. In addition, women have poorer functional recovery and worse quality of life after stroke, compared with men.³⁰

Women with hypertension have a higher risk of first stroke, compared with men with hypertension.³¹ Recently published evidence demonstrated that APOs are independently associated with earlier onset of hypertension.⁶ Since hypertension is a major risk factor for both ischemic and hemorrhagic stroke,³² we hypothesized that hypertension might mediate the relationship between APOs and stroke risk. However, our analysis did not show significant mediation of hypertension in the relationship between APOs and future stroke.

It is possible that other mechanisms, such as inflammatory pathways, shared underlying vascular risk factors, or social determinants of health, play roles in the association between APOs and future stroke. In particular our results suggest the need to further understand the role of social determinants of health as risk moderators. While evidence on the importance of social determinants on maternal mortality and severe morbidity in the perinatal period is mounting,³³ the role of social determinants on APOs and longer term maternal health is unclear.

Future work should also investigate the role of structural determinants of health in the relationship between APOs and longer term maternal health.³⁴ Measurement error for hypertension status may also explain the lack of evidence of mediation by hypertension.

The pathophysiological connection between APOs and future stroke in women remains unclear. One possibility is that APOs may occur more frequently in women with an underlying high-risk vascular phenotype, who are subsequently at greater risk for cerebrovascular disease.^35–37 Fetal growth restriction, preterm delivery, placental abruption, preeclampsia, and stillbirth can all result from aberrant placental implantation, which may occur more frequently in women with an unhealthy vasculature.³⁸

Another possibility is that APOs may be associated with immunological intolerance and a heightened inflammatory state that portend adverse long-term consequences for the maternal vasculature.³⁹ Of note, low birth weight has also been associated with an increased risk of future stroke in offspring, suggesting that placental cross talk may play a role in future cerebrovascular risk for both mother and child.⁴⁰

Our study has limitations. History of APOs, other risk factors, and stroke were based on self-report, with no ability to verify APO and stroke diagnoses via medical record review, or clarify if strokes were hemorrhagic or ischemic. In addition, without knowledge of stroke mechanisms, our results cannot help to elucidate the mechanism of the association (e.g., coagulopathic, inflammatory, embolic, or related to small- or large-vessel cerebrovascular disease).

Strokes may have been under- or overreported; self-reported stroke has been shown to have moderate concordance with hospital-recorded diagnoses, with better concordance in younger populations.^41–43 Pregnancy complications, too, may have been under- or overreported, although the prevalence of 16% is similar to that reported in prospective studies.⁶ Validation studies have shown high agreement between self-reported APOs and medical record-confirmed diagnoses^44,45; however, these studies were for specific APOs and may not accurately reflect differences in validity across the outcomes included in our analyses.

The smaller number of women reporting preeclampsia limited our power to assess the impact of this known risk factor on midlife stroke, and precluded our reporting preeclampsia separately; in addition, changing definitions of preeclampsia over the last 50 years could have an impact on the prevalence of the diagnosis. Furthermore, a history of gestational hypertension and gestational diabetes, both important cardiovascular risk factors, was not ascertained in the baseline interview. The lack of significant association for the outcome of “any stroke” may reflect poorer recall of APOs in older women.

Our results could also be biased by poorer recall in participants with a history of stroke, compared with stroke-free participants, since stroke patients have a high prevalence of poststroke cognitive impairment.⁴⁶ A cross-sectional study of 1,000 older U.S. women with disabilities, including cognitive impairment, found high levels of agreement between self-reported diagnoses and medical record review, but levels of agreement decreased with increasing age and cognitive impairment.⁴⁷ However, this would likely bias our results toward the null rather than otherwise. In addition, while we limited our outcomes to strokes that were reported after age 50 or after the most recent pregnancy, women may have had inaccurate recall of dates, and thus, we could not confirm conclusively that strokes occurred after (and not concurrent with) the APO.

Similarly, we could not confirm for all respondents at what age comorbidities such as obesity, hypertension, and diabetes were first diagnosed. Lastly, the competing risk of death in this analysis may have led to selection bias and underestimation of the association (i.e., women may have died before enrolling in the study). As PATH is an ongoing longitudinal study, extensions to this analysis should consider longitudinal relationships between APOs and stroke risk. Given these important limitations, our results should be considered hypothesis generating.

A strength of our study, in comparison with prior studies, is the nationally representative nature of the sample, making the results highly generalizable to the regionally, socioeconomically, and ethnically diverse U.S. population of middle-aged and older women. One caveat is that our analysis sample included only women aged ≥50 years. Due to the young age distribution of Hispanics in the United States, this resulted in a lower proportion of Hispanic women in our sample, limiting our ability to consider Hispanic women as a specific subpopulation.

The rapidly changing demographics in the United States may render these 2013–2014 data out of date and underrepresentative of the current U.S. Hispanic population. However, other more current U.S. nationally representative samples such as NHANES have limited information on APOs. Furthermore, our study included information on additional APOs, including low birth weight and stillbirth, which have not previously been associated with future stroke.⁷

Conclusion

In this cross-sectional analysis of women in a nationally representative longitudinal study of U.S. respondents, a self-reported history of APOs was independently associated with a greater than twofold increased odds of stroke in midlife. Our results suggest that clinicians caring for obstetric patients can play a critical role in early identification of young women who should be targeted for intensive interventions aimed at primary stroke prevention. The role of multidisciplinary postpartum clinics and “cardio-obstetrics” has emerged as a promising strategy for optimizing heart health in women.

Our results demonstrate the importance of including cerebrovascular health and “neuro-obstetrics” in this discussion. Furthermore, neurologists often see young and middle-aged women for common conditions such as migraine, epilepsy, back pain, and transient neurological symptoms. A careful reproductive history should be included in the assessment of future stroke risk in these patients. Future studies should consider APOs and stroke in the broader context of life course reproductive health, including age at menarche and menopause, use of hormonal contraception and hormone replacement, and breastfeeding.

Authors' Contributions

All authors confirm that they fulfill all four criteria for authorship as defined by the International Committee of Medical Journal Editors, including making substantial contributions to the conception or design of the work; or the acquisition, analysis, or interpretation of data for the work; drafting the work or revising it critically for important intellectual content; final approval of the version to be published; and agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Author Disclosure Statement

E.C.M. received personal compensation from medicolegal consulting related to maternal stroke and from Elsevier for editing two volumes of Handbook of Clinical Neurology on Neurology and Pregnancy, and is on the editorial board of Stroke (no personal compensation). N.A.B. has nothing to disclose. R.D. has nothing to disclose. A.M.F. has nothing to disclose. M.S.V.E. receives personal compensation from Merck/Organon for expert witness testimony related to hormonal contraception and stroke, and from UpToDate for chapters related to stroke. R.W. has nothing to disclose. S.E.T. has nothing to disclose.

Funding Information

E.C.M. is supported by a National Institutes of Health (NIH)/National Institute of Neurological Disorders and Stroke Career Development Award (K23NS107645) and the Gerstner Family Foundation (Gerstner Scholars Award). N.A.B. is supported by the NIH/National Heart Lung and Blood Institute (K23 HL136853) and the Katz Foundation. S.E.T. is supported by the NIH/National Institute on Aging Career Development Award (K01AG050723). Funding for the overall PATH study is provided by the U.S. Department of Health and Human Services, the NIH, the National Institute on Drug Abuse, and by the Food and Drug Administration, Center for Tobacco Products.

Supplementary Material

Supplementary Table S1