Risk Factors for Pregnancy-Associated Stroke in Women With Preeclampsia
Background and Purpose—Preeclampsia affects 3% to 8% of pregnancies and increases risk of pregnancy-associated stroke (PAS). Data are limited on which women with preeclampsia are at highest risk for PAS.
Methods—Using billing data from the 2003 to 2012 New York State Department of Health inpatient database, we matched women with preeclampsia and PAS 1:3 to preeclamptic controls based on age and race/ethnicity. Pre-defined PAS risk factors included pregnancy complications, infection present on admission, vascular risk factors, prothrombotic states, and coagulopathies. We constructed multivariable conditional logistic regression models to calculate the odds ratios (ORs) and 95% confidence intervals (95% CIs) for independent risk factors for PAS.
Results—Among women aged 12 to 55 years admitted to New York State hospitals for any reason during the study period (n=3 373 114), 88 857 had preeclampsia, and 197 of whom (0.2%) had PAS. In multivariable analysis, women with preeclampsia and stroke were more likely than controls to have severe preeclampsia or eclampsia (OR, 7.2; 95% confidence interval [CI], 4.6–11.3), infections present on admission (OR, 3.0; 95% CI, 1.6–5.8), prothrombotic states (OR, 3.5; 95% CI, 1.3–9.2), coagulopathies (OR, 3.1; 95% CI, 1.3–7.1), or chronic hypertension (OR, 3.2; 95% CI, 1.8–5.5). Additional analyses matched and stratified by severity of preeclampsia confirmed these results.
Conclusions—Infections, chronic hypertension, coagulopathies, and underlying prothrombotic conditions increase PAS risk in women with preeclampsia. These women may warrant closer monitoring.
Preeclampsia is a multisystem hypertensive disorder unique to pregnancy, characterized by widespread endothelial dysfunction and immune dysregulation.1 Approximately 36% of women with pregnancy-associated strokes (PASs) have comorbid preeclampsia,2 and preeclampsia increases stroke risk during the puerperium up to 6-fold.3 Among women with PAS, women with preeclampsia are at increased risk of complications and death.4,5
Although preeclampsia affects 3% to 8% of all pregnancies,3,6 the overall occurrence of PAS remains rare (34.2 per 100 000 deliveries).7 Older age, black race, and lack of private insurance are associated with PAS in women with preeclampsia,4,8 and increased preeclampsia severity is associated with an increased risk of cardiovascular events.9 The rarity of PAS makes it difficult to predict which preeclampsia patients are at highest risk for cerebrovascular complications. The American Heart Association/American Stroke Association found insufficient evidence to make recommendations on prevention of stroke in pregnancy complicated by hypertensive disorders.10
We sought to identify modifiable risk factors that put women with preeclampsia at highest risk of PAS.
Study Design and Data Description
We performed a case–control study using billing data, coded according to the International Classification of Diseases, Ninth Revision, from the 2003 to 2012 New York State Department of Health Statewide Planning and Research Cooperative System inpatient database. We identified all women aged 12 to 55 years old admitted with preeclampsia, including mild preeclampsia (642.4x [where x is any fifth digit from 0–9], 642.7x), severe preeclampsia (642.5x), or eclampsia (642.6x) from January 1, 2003, through December 31, 2012. Women without preeclampsia were not included in the study, regardless of whether they had strokes during or after pregnancy. Cases were defined as women with preeclampsia and PAS, including diagnosis codes for transient ischemic attack or ischemic stroke, intracerebral hemorrhage, subarachnoid hemorrhage (SAH), cerebral venous thrombosis, and codes specific for nonspecified PAS. Pregnancy-specific stroke codes were validated using internal data sets (Table I in the online-only Data Supplement). In women with >1 stroke admission, we counted only the first admission. We characterized PAS in terms of timing and stroke subtype. Although stroke risk is increased up to 12 weeks after delivery,11 a 6-week post-partum time frame was chosen because of limitations of International Classification of Diseases, Ninth Revision, coding.12 Traumatic intracerebral hemorrhage and SAH and hospitalizations with a primary rehabilitation diagnosis (V57.89) were excluded. We matched each stroke case to 3 controls of the same age, race/ethnicity, and insurance status, selected randomly from the pool of women with preeclampsia without PAS.
Pre-defined exposures of interest, selected based on previously described associations with preeclampsia and PAS, included pregnancy-related variables, chronic vascular risk factors, comorbid prothrombotic states, coagulopathies/bleeding disorders, and infection of any type present on admission (POA), identified with a Statewide Planning and Research Cooperative System–specific indicator.13,14 Infections acquired during the stroke hospitalization were not included. Coding details are in Table II in the online-only Data Supplement.
Cerebral symptoms are sufficient to classify preeclampsia as severe.15,16 Because all our cases likely had neurological symptoms, this could create selection bias, confusing the relationship between severity of preeclampsia and stroke risk (ie, women with stroke and preeclampsia might by definition be coded as having severe preeclampsia). We therefore conducted a separate analysis matching each case to 3 controls of the same age, race/ethnicity, and severity of preeclampsia. Pre-specified subgroup analyses stratified the severity-matched cohort into 2 groups: those with mild preeclampsia and those with severe preeclampsia or eclampsia.
We conducted univariable analyses comparing risk factors among cases and controls using χ2 tests. Including only those PAS risk factors with P<0.2 from the univariable analysis,17 we calculated unadjusted and adjusted odds ratios (ORs) and 95% confidence intervals (95% CI) using multivariable conditional logistic regression. Pre-specified subgroup analyses included a sensitivity analysis excluding all cases and controls with eclampsia (642.6x) and a subgroup analysis of post-partum strokes. All analyses were completed with SAS version 9.3 (SAS Institute, Cary, NC). A P<0.05 was set as the significance level.
Approval was obtained from the Institutional Review Board of Columbia University Medical Center to conduct the analyses. Requirement for consent was waived because of the public, deidentified nature of the data.
Baseline Characteristics of Women With Preeclampsia With and Without PAS
Among women aged 12 to 55 years admitted for any reason during the study period (n=3 373 114), 88 857 had preeclampsia, including eclampsia. Of these, 197 (0.2%) had PAS and were identified as cases, giving a cumulative incidence of PAS in women with preeclampsia of 222/100 000 during the study period. In the unmatched sample, compared with the 88 660 women with preeclampsia who did not have PAS, women with preeclampsia and PAS were older, had higher proportion of black race, lower proportion of Hispanic ethnicity, and higher proportion of severe preeclampsia (42.1% versus 29%) or eclampsia (28.9% versus 2%; Figure 1). Among cases, median age was 32 years (interquartile range, 26–36); 57 women (28.9%) had mild preeclampsia, 83 (42.1%) had severe preeclampsia, and 57 (28.9%) had eclampsia. The 197 stroke cases were matched with 591 nonstroke controls of the same age, race/ethnicity, and insurance status (Figure 2).
Timing and Characteristics of Strokes in Women With Preeclampsia and PAS
Among the 197 women with strokes, 55 (27.9%) strokes occurred antepartum; 8 (4.1%) occurred during the delivery hospitalization without further characterization in timing; 131 (66.5%) occurred post-partum; and 3 (1.5%) occurred during an admission without delivery, with no further characterization of timing. Of 131 post-partum strokes, 82 (62.6%) occurred after discharge (Figure 2). Stroke types included 92 (46.7%) hemorrhagic strokes, 26 (13.1%) ischemic, 8 (4.1%) transient ischemic attack, 5 (2.5%) cerebral venous thromboses, and 70 (35.5%) nonspecific PAS; 9 women had multiple subtypes (Figure I in the online-only Data Supplement). In-hospital mortality was 13.2% among cases, compared with 0.2% among controls. Details of discharge disposition/mortality are in Table III in the online-only Data Supplement.
Stroke Risk Factors in Cases and Controls Matched on Age/Race-Ethnicity/Insurance
In univariable analysis, women with PAS had a higher proportion than controls of severe preeclampsia or eclampsia, infections POA, chronic hypertension, prothrombotic states, coagulopathies, migraine, and heart disease (Table 1). The difference in infections between cases and controls was driven by genitourinary infections: 71% of infections in cases were genitourinary, compared with 39% of infections in controls, and genitourinary infections occurred in 10% of cases compared with 2% of controls (P<0.0001). After adjusting for other risk factors, significant risk factors for stroke were severe preeclampsia or eclampsia (OR, 7.2; 95% CI, 4.6–11.3), infections POA (OR, 3.0; 95% CI, 1.6–5.8), chronic hypertension (OR, 3.2; 95% CI, 1.8–5.5), prothrombotic states (OR, 3.5; 95% CI, 1.3–9.2), and coagulopathies (OR, 3.1; 95% CI, 1.3–7.1; Table 2).
Demographics and Risk Factors in Severity-Matched Cohort
When cases were matched to controls by preeclampsia severity, similar risk factors emerged: infections POA (OR, 2.6; 95% CI, 1.4–4.6), prothrombotic states (OR, 2.9; 95% CI, 1.3–6.4), and chronic hypertension (OR, 4.2; 95% CI, 2.4–7.4) conferred greater risk of stroke after adjusting for other risk factors (Table 3). In the severity-stratified subgroup analysis, after adjusting for other variables, chronic hypertension (OR, 3.0; 95% CI, 1.4–6.3) and infection POA (OR, 4.5; 95% CI, 1.6–12.6) were significant stroke risk factors in the mild subgroup (n=57). In the severe subgroup (n=140), chronic hypertension (OR, 8.4; 95% CI, 3.1–22.5) and prothrombotic states (OR, 6.6; 95% CI, 2.5–17.5) increased stroke risk; the association of infection POA did not reach statistical significance after adjusting for other risk factors (OR, 1.9; 95% CI, 0.9–4.0; Table IV in the online-only Data Supplement). Demographics of mild versus severe cases and controls are shown in Table V in the online-only Data Supplement.
In the subgroup analysis excluding women with eclampsia (International Classification of Diseases, Ninth Revision, 642.6x), results were similar: women with chronic hypertension (adjusted OR, 4.2; 95% CI, 2.2–7.8), infection POA (adjusted OR, 3.7; 95% CI, 1.8–7.7), and prothrombotic states (adjusted OR, 2.6; 95% CI, 1.1–6.3) had higher risk of stroke (Table VI in the online-only Data Supplement). In the subgroup analysis, including only post-partum stroke, fewer cases than controls delivered via cesarean section (52% versus 65%; P=0.01). However, after adjusting for other risk factors, there was no significant association between cesarean delivery and stroke risk (adjusted OR, 0.7; 95% CI, 0.4–1.1). Otherwise, similar risk factors were found in women with post-partum stroke (Tables VII and VIII in the online-only Data Supplement).
In this case–control study in a diverse population of women with preeclampsia, we found that risk factors for PAS included infections (predominantly genitourinary), chronic hypertension, prothrombotic conditions, and coagulopathies. The cumulative incidence of stroke in our preeclamptic population was 222/100 000, >6 times the incidence of stroke in the overall pregnant population, consistent with prior studies.3,4 Two thirds of cases of PAS (131 of 197) were diagnosed post-partum, consistent with prior research.7 More than 1 in 10 women with preeclampsia and PAS died during their admission for stroke, likely reflecting the large proportion of hemorrhagic strokes.18 In comparison, overall US maternal mortality in 2011 was 17.8 per 100 000 deliveries, or 0.02%,5,19–21 emphasizing the importance of identifying women at the highest risk of PAS and targeting them for closer monitoring.
Infection is increasingly recognized as a trigger for stroke, particularly in young people.14,22 Proposed mechanisms include increased levels of inflammatory cytokines leading to platelet aggregation and impaired endothelial function. Infections may also provoke cardiac arrhythmias or dehydration-induced thrombosis; in 1 study, infections tripled the odds of peripartum cerebral venous thrombosis.23 The pathophysiology of preeclampsia remains incompletely understood1 although imbalance of pro- and antiangiogenic factors and an increase in proinflammatory cytokines seem to play roles.24 Animal models of preeclampsia have demonstrated impaired cerebral autoregulation, increased blood-brain barrier permeability, and neuronal hyperexcitability, mediated by elevated levels of tumor necrosis factor-α.25–28 In the setting of preeclampsia-associated inflammation and puerperial hypercoagulability, superimposed infection could trigger stroke by exacerbating inflammation and coagulopathy.11
Although we included only infections POA in our analysis, it is important to acknowledge that we cannot be certain that the infection preceded the stroke without a prospective study. Nevertheless, a possible association between infection and preeclampsia-associated stroke has important clinical implications because it may represent a modifiable risk factor. The difference in infections between cases and controls in our cohort was driven entirely by genitourinary infections. Symptoms of urinary infections, such as dysuria or incontinence, may go unrecognized in post-partum women because similar symptoms may occur with normal post-partum recovery. Women with preeclampsia undergo post-partum blood pressure checks; screening for urinary tract infections at these visits and monitoring those with infections more closely for neurological symptoms may be warranted. The role for prophylactic antibiotics could be considered on a case by case basis and may be a target in clinical trials.
The role of chronic hypertension in preeclampsia-associated stroke is poorly characterized. Chronic hypertension shifts the cerebral autoregulatory curve29 but is not associated with loss of dynamic cerebral autoregulatory capacity because of rapid adaptation of the cerebral vasculature.30 However, acute malignant hypertension in chronically hypertensive patients severely impairs dynamic cerebral autoregulation.31 Preeclampsia may cause impairment of cerebrovascular autoreactivity,32 leading to hyperemia, hypertensive leukoencephalopathy, and the reversible cerebral vasocontriction syndrome,33 all of which can be associated with stroke.34 Reversible cerebral vasocontriction syndrome is highly associated with post-partum stroke.35 Unfortunately, we were unable to assess investigate the role of reversible cerebral vasocontriction syndrome and stroke in our population because reversible cerebral vasocontriction syndrome is not reliably captured by International Classification of Diseases, Ninth Revision, coding.
Prothrombotic Conditions and Coagulopathies
Women with an underlying propensity to thrombosis are at increased risk of stroke when pregnancy, itself a hypercoagulable state, is complicated by preeclampsia.4,36 Only 2.5% of strokes in our cohort were because of cerebral venous thrombosis, suggesting that prothrombotic states may put women at risk of arterial thrombosis and hemorrhagic stroke as well. However, cerebral venous thrombosis, which may also present with hemorrhage, may be underdiagnosed. Women with underlying thrombophilia may have been on antithrombotic treatment during or after their pregnancy, increasing their intracerebral hemorrhage/SAH risk. Patients with preeclampsia and a history of coagulopathy likely warrant increased vigilance in the peripartum period.
Other Vascular Risk Factors
Heart disease has been identified as a risk factor for PAS in women with preeclampsia.4 After adjusting for other risk factors, our results failed to show significant between-group differences in heart disease. Similarly, we found no significant differences in stroke risk comparing other traditional vascular risk factors, such as diabetes mellitus, obesity, chronic renal disease, smoking, and other substance abuse. In fact, proportions of these risk factors were low in both cases and controls. Our study may have been underpowered to detect between-group differences in some risk factors because other studies have found otherwise; alternatively, the pathophysiology of PAS in women with preeclampsia may bear little relationship to traditional vascular risk factors.
Preeclampsia is regarded as a disease along a continuum; women with gestational hypertension may go on to develop preeclampsia and then severe preeclampsia.37 Women with severe preeclampsia were over-represented in our cases compared with the overall population of women with preeclampsia; however, it is difficult to interpret this finding because any cerebral symptom in a woman with preeclampsia is considered a severe feature, creating a significant selection bias. Further complicating interpretation of our data is the updating of diagnostic criteria for severe preeclampsia in 2013, after our study period.16 The lack of data on the HELLP (hemolysis, elevated liver enzymes, and low platelets) syndrome, a variant of preeclampsia, is a limitation of our study because these women may be at particularly high stroke risk.38
Research in Context: Study Strengths and Limitations
Two prior population-based studies compared women with preeclampsia and stroke to women with preeclampsia without stroke.39,40 Neither study was designed to identify multiple independent risk factors for stroke in this population: one compared only mortality outcomes and the other compared methods of anesthesia during delivery. The ethnic and regional diversity of New York State increases the generalizability of our findings. Matching of cases and controls allowed for nuanced analysis of other risk factors. We conducted multiple subgroup analyses to explore possible biasing factors in our results.
Our study has limitations. Data were drawn from a large administrative database, and there may be coding errors. Some preeclampsia cases may not have been formally diagnosed and thus not included in the study. Pregnancy-specific stroke codes do not distinguish between stroke subtypes, limiting the granularity of the data, although the majority of patients had additional codes to identify stroke subtypes with more precision (Figure I in the online-only Data Supplement). Although the proportion of each of the prothrombotic conditions individually (hypercoagulable states, history of thromboembolic events, systemic lupus erythematosis, and sickle cell disease) was higher in cases than controls, grouping them together may overestimate their effects. Diagnosis of prothrombotic conditions or coagulopathies may have occurred only after the stroke, leading to ascertainment bias. Although we included only infections POA for stroke, exact timing of infections in relation to stroke onset cannot be confirmed without a prospective study. Causality cannot be inferred from this observational study, and results should be interpreted cautiously.
In summary, urinary tract infections, chronic hypertension, prothrombotic conditions, and coagulopathies increased stroke risk in women with preeclampsia. Infections may be an important treatable risk factor in this population; similarly, screening for coagulopathies and prothrombotic conditions may be warranted in women with preeclampsia. Prospective studies are needed to confirm these findings and develop interventions aimed at preventing strokes in this uniquely vulnerable group.
Sources of Funding
Dr Miller receives support from a National Institutes of Health (NIH) National Institute of Neurological Disorders and Stroke (NINDS) StrokeNet Training Fellowship. Dr Marshall receives support from NIH NINDS 1U10 NS086728.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.017374/-/DC1.
- Received March 17, 2017.
- Revision received April 14, 2017.
- Accepted April 27, 2017.
- © 2017 American Heart Association, Inc.
- Ahmed A,
- Rezai H,
- Broadway-Stringer S
- Crovetto F,
- Somigliana E,
- Peguero A,
- Figueras F
- Giannubilo SR,
- Landi B,
- Ciavattini A
- Hovsepian DA,
- Sriram N,
- Kamel H,
- Fink ME,
- Navi BB
- Bushnell C,
- McCullough LD,
- Awad IA,
- Chireau MV,
- Fedder WN,
- Furie KL,
- et al
- 12.↵National Center for Health Statistics. ICD-9-CM Official Guidelines for Coding and Reporting. Washington, DC: US GPO; 2011.
- Andersen KK,
- Olsen TS,
- Dehlendorff C,
- Kammersgaard LP
- Leffert LR,
- Clancy CR,
- Bateman BT,
- Cox M,
- Schulte PJ,
- Smith EE,
- et al
- 21.↵Pregnancy Mortality Surveillance System. Centers for Disease Control and Prevention: Reproductive Health; 2016. http://www.cdc.gov/reproductivehealth/maternalinfanthealth/pmss.html. Accessed April 12, 2017.
- Elkind MS,
- Hills NK,
- Glaser CA,
- Lo WD,
- Amlie-Lefond C,
- Dlamini N,
- et al
- Lanska DJ,
- Kryscio RJ
- Warrington JP,
- Drummond HA,
- Granger JP,
- Ryan MJ
- Porcello Marrone LC,
- Gadonski G,
- de Oliveira Laguna G,
- Poli-de-Figueiredo CE,
- Pinheiro da Costa BE,
- Lopes MF,
- et al
- Strandgaard S
- Immink RV,
- van den Born BJ,
- van Montfrans GA,
- Koopmans RP,
- Karemaker JM,
- van Lieshout JJ
- Hasegawa J,
- Ikeda T,
- Sekizawa A,
- Tanaka H,
- Nakata M,
- Murakoshi T,
- et al
- Gastrich MD,
- Gandhi SK,
- Pantazopoulos J,
- Zang EA,
- Cosgrove NM,
- Cabrera J,
- et al
- Huang CJ,
- Fan YC,
- Tsai PS