Hypertensive Disorders of Pregnancy and Incident Dementia

Key Points

Question

Are hypertensive disorders of pregnancy (HDPs) associated with long-term dementia risk?

Findings

In this nationwide cohort study of 1.37 million females, HDPs were associated with an increased risk of dementia over approximately 2 decades of follow-up after controlling for maternal, socioeconomic, and obstetric factors. This association was attenuated and was no longer statistically significant after additional adjustment for postdelivery vascular and metabolic conditions.

Meaning

These findings suggest that HDPs may serve as an early-life marker of long-term neurovascular vulnerability.

Abstract

Importance

Hypertensive disorders of pregnancy (HDPs) are established markers of later-life cardiovascular disease, but their association with dementia remains uncertain.

Objective

To investigate whether HDPs are associated with incident dementia in a nationwide population-based cohort with more than 2 decades of follow-up.

Design, Setting, and Participants

This retrospective cohort study used the Korean National Health Insurance Service database to identify females aged 16 years or older who delivered between January 1, 2002, and December 31, 2005, and followed up through December 31, 2024. Statistical analyses were conducted between October 2025 and May 2026.

Exposures

HDPs were defined as gestational hypertension, preeclampsia, HELLP (hemolysis, elevated liver enzymes, low platelet count) syndrome, or eclampsia.

Main Outcomes and Measures

The primary outcome was incident dementia, including Alzheimer disease and vascular dementia. Cox proportional hazards regression models estimated hazard ratios (HRs) and 95% CIs.

Results

Among 1 370 485 females (mean [SD] age, 29.2 [3.9] years), 28 747 (2.1%) experienced HDPs. The cumulative incidence of dementia was higher among females with HDPs than among those without HDPs (7.2 vs 2.8 per 10 000 females; P < .001), as was the incidence density of dementia (3.00 vs 1.05 per 100 000 person-years; P < .001). In model 1, HDPs were associated with an increased risk of incident dementia after adjustment for baseline maternal, socioeconomic, and obstetric factors (adjusted HR, 1.96; 95% CI, 1.19-3.23). In model 2, after additional adjustment for postdelivery hypertension, diabetes, myocardial infarction, and stroke, the association was attenuated and was no longer statistically significant (adjusted HR, 1.15; 95% CI, 0.70-1.91).

Conclusions and Relevance

In this cohort study, HDPs were associated with an increased risk of incident dementia after adjustment for baseline maternal, socioeconomic, and obstetric factors. However, this association was no longer statistically significant after additional adjustment for postdelivery vascular and metabolic conditions. Given the small number of dementia events among females with HDPs, these findings should be interpreted cautiously. These findings suggest that HDPs may serve as an early-life marker of long-term neurovascular vulnerability.

This cohort study of 1.4 million females with more than 2 decades of follow-up investigates whether hypertensive disorders of pregnancy are associated with incident dementia risk.

Introduction

Dementia encompasses a group of neurodegenerative disorders characterized by progressive cognitive decline, among which Alzheimer disease and vascular dementia are the most common. Women bear a disproportionate burden, experiencing greater reductions in life expectancy and quality of life compared with men.¹ Hypertension is a well-established and potentially modifiable risk factor for both Alzheimer disease and vascular dementia.^2,3,4,5,6 Hypertensive disorders of pregnancy (HDPs), including gestational hypertension; hemolysis, elevated liver enzymes, low platelet count (HELLP) syndrome; preeclampsia; and eclampsia, affect up to 15% of pregnancies and remain a major cause of maternal and perinatal morbidity worldwide, with increasing prevalence attributed to increasing maternal age and metabolic risk factors.^7,8,9

The pathophysiology of HDPs, particularly preeclampsia, involves abnormal remodeling of placental spiral arteries, leading to placental hypoperfusion and oxidative stress with inflammation.¹⁰ Emerging evidence suggests women with a history of HDPs have increased risks of chronic hypertension, stroke, myocardial infarction, and chronic kidney disease, supporting the concept that HDPs may reflect underlying systemic vascular vulnerability.^10,11,12,13 Although HDPs were traditionally considered transient conditions confined to pregnancy, accumulating epidemiologic evidence suggests HDP-related vascular injury may contribute to long-term cerebrovascular and neurological sequelae.^14,15 Several population-based studies and meta-analyses reported an increased risk of dementia among women with a history of HDPs,^8,16,17,18 whereas other studies have not demonstrated a clear association with subsequent cognitive decline.^11,16 To address these inconsistencies, we conducted a nationwide population-based cohort study using the Korean National Health Insurance Service (NHIS) database to evaluate the long-term association between HDPs and the risk of incident dementia in females.

Methods

Study Design and Data Sources

This retrospective cohort study used data from the Korean NHIS, which covers the entire Korean population and contains information on demographic characteristics, medical diagnoses, procedures, and prescription records. Each NHIS beneficiary is assigned a unique, encrypted personal identifier, which was used to link all delivery, exposure, covariate, and outcome (dementia) claims for the same individual across the study period and to remove duplicate records, ensuring that each female in the final cohort was represented only once. Diagnoses are recorded using the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10). The validity of ICD-10–based diagnoses in the NHIS database has been previously demonstrated in validation studies.^19,20,21 This study was approved by the institutional review board of Seoul National University Bundang Hospital. Because the study used anonymized administrative claims data, the requirement for informed consent was waived. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Study Population

This study included females aged 16 years or older who delivered between January 1, 2002 and December 31, 2005 (Figure 1). Deliveries were identified using procedure codes for vaginal or cesarean delivery together with corresponding ICD-10 diagnostic codes (eTable 1 in Supplement 1). Repeated deliveries with an interdelivery interval of 210 days or less were considered part of the same pregnancy episode and consolidated. Among 1 374 100 females who delivered during the study period, 3615 with chronic hypertension diagnosed 1 to 2 years before delivery were excluded. The final analytic cohort included 1 370 485 females, classified according to the presence or absence of HDPs and followed up for incident dementia through December 31, 2024 (Figure 1).

Figure 1. Flowchart of the Study Population Selection.

Deliveries recorded in the Korean National Health Insurance Service database between January 1, 2002, and December 31, 2005, were identified (n = 1 374 100). Females with a diagnosis of chronic hypertension during the prepregnancy period were excluded (n = 3615), yielding a final analytic cohort of 1 370 485 females. Participants were classified according to the presence or absence of hypertensive disorders of pregnancy (HDPs) and followed up for incident dementia through December 31, 2024. During follow-up, dementia occurred in 18 females with HDPs and 297 females without HDPs.

Definition of Outcome Variable

The primary outcome was incident dementia. Dementia cases were identified using ICD-10 diagnostic codes for Alzheimer disease (codes F00 and G30), vascular dementia (code F01), dementia in other diseases classified elsewhere (code F02), frontotemporal dementia (code G31.0), unspecified dementia (code F03), and Lewy body dementia (code G31.83), together with prescriptions for antidementia medications (eTable 2 in Supplement 1). To improve diagnostic specificity, dementia was operationally defined as the presence of a dementia diagnostic code accompanied by at least 1 prescription for antidementia medication, including donepezil (Anatomical Therapeutic Chemical code N06DA02), rivastigmine (N06DA03), galantamine (N06DA04), or memantine (N06DX01). Participants were followed up from the date of delivery until the first diagnosis of dementia, death, or the end of follow-up, whichever occurred first.

Definition of Exposure and Covariate Variables

HDPs were defined using ICD-10 diagnostic codes (eTable 3 in Supplement 1) and classified into 4 mutually nonexclusive subgroups based on clinical diagnosis codes recorded in the claims data: preeclampsia, HELLP syndrome, eclampsia, and gestational hypertension. For individuals with multiple eligible pregnancies during the study period, only 1 record per female was retained. In such cases, pregnancies complicated by HDPs were prioritized; otherwise, the first recorded delivery was selected.

Covariates were selected a priori based on previous claims-based obstetric cohort studies. A formal directed acyclic graph was not used. Variables not routinely captured in claims data, including smoking, obesity, and family history of dementia, could not be included. Obstetric and delivery–related variables included multifetal pregnancy, pregestational diabetes, gestational diabetes, preterm birth, mode of delivery, and postpartum hemorrhage. Postdelivery chronic diseases included hypertension, diabetes, cardiovascular disease, and cerebrovascular disease. Cardiovascular disease was operationalized as myocardial infarction and cerebrovascular disease as stroke, without further distinction between ischemic and hemorrhagic subtypes (eTable 3 in Supplement 1).

Maternal age was assessed at the time of delivery and categorized as younger than 35 years or 35 years or older, consistent with the standard clinical definition of advanced maternal age used in obstetric practice.²² Socioeconomic status was estimated using annual health insurance premiums at the time of pregnancy. The original 11-tier premium classification was recategorized into 5 groups (tiers 0-2, 3-4, 5-6, 7-8, and 9-10), with higher tier numbers indicating higher premiums.

Incidence Density and Cumulative Incidence of Dementia

The cumulative incidence of dementia was estimated using the Kaplan-Meier method and expressed per 10 000 females, separately for those with or without HDPs. The incidence density of dementia was calculated as the number of dementia cases per 100 000 person-years of observation, also calculated separately for females with or without HDP.

Statistical Analysis

Statistical analyses were conducted between October 2025 and May 2026. Baseline characteristics were summarized according to the presence or absence of HDPs. Continuous variables were presented as mean (SD), and categorical variables were expressed as frequencies and percentages. Differences between groups were evaluated using the t test or χ² test, as appropriate. The cumulative incidence rate of dementia was estimated using the Kaplan-Meier method and compared using the log-rank test. Incidence density was calculated as the number of dementia cases per 100 000 person-years.

Associations between HDPs and incident dementia were evaluated using Cox proportional hazards regression models and presented as hazard ratios (HRs) with 95% CIs. Univariable analyses were performed first, followed by 2 multivariable Cox proportional hazards regression models. Model 1 adjusted for maternal age at delivery, socioeconomic status, multifetal pregnancy, pregestational diabetes, gestational diabetes, preterm birth, cesarean delivery, and postpartum hemorrhage. Model 2 additionally adjusted for postdelivery hypertension, diabetes, myocardial infarction, and stroke, which occurred after the index pregnancy and were therefore considered potential intermediate variables rather than conventional baseline confounders. It is therefore presented as an exploratory, hypothesis-generating analysis rather than a conventional multivariable-adjusted model; a time-dependent Cox proportional hazards regression model was not fitted, as our study design retained only a single index pregnancy per female and could not support formal time-varying covariate modeling of subsequent comorbidities. The proportional hazards assumption was evaluated using Schoenfeld residuals and was satisfied for the primary exposure and all covariates included in the multivariable models.

Statistical analyses were conducted using SAS Enterprise Guide, version 8.2 (SAS Institute Inc). Figures were generated using RStudio with R, version 4.5.0 (R Project for Statistical Computing). A 2-sided P < .05 was considered statistically significant.

Results

General Characteristics of Study Population

Among 1 370 485 females (mean [SD] age, 29.2 [3.9] years) who delivered between 2002 and 2005, 28 747 (2.1%) had HDPs and 1 341 738 (97.9%) did not (Figure 1). Baseline characteristics of both groups are summarized in Table 1. Females with HDPs were older at delivery than those without HDPs (mean [SD] age, 30.0 [4.4] vs 29.2 [3.9] years; P < .001), and a higher proportion were aged 35 years or older (15.1% vs 8.6%; P < .001). Multifetal pregnancy (3.3% vs 0.7%; P < .001), pregestational diabetes (0.9% vs 0.2%; P < .001), gestational diabetes (10.5% vs 5.6%; P < .001), preterm birth (9.1% vs 1.3%; P < .001), cesarean delivery (55.9% vs 35.6%; P < .001), and postpartum hemorrhage (6.3% vs 3.8%; P < .001) were all significantly more frequent in females with HDPs. During follow-up, hypertension (50.0% vs 20.1%; P < .001), diabetes (28.5% vs 15.2%; P < .001), cerebrovascular disease (4.1% vs 1.7%; P < .001), and cardiovascular disease (2.1% vs 0.8%; P < .001) were also significantly more common in females with HDPs than in those without HDPs. The median follow-up duration was 21.0 years (IQR, 20.2-21.9 years) among females with HDPs and 21.3 years (IQR, 20.3-22.1 years) among females without HDPs.

Table 1. Characteristics of the Study Population .

Characteristic	No. (%)		P value
	Without HDPs (n = 1 341 738)	HDPs (n = 28 747)
Age, mean (SD), y	29.2 (3.9)	30.0 (4.4)	<.001
Maternal age ≥35 y	115 050 (8.6)	4337 (15.1)	<.001
Health insurance premium tiera
0-2 (lowest tier; Medical Aid enrollees)	219 737 (17.1)	5241 (19.1)	<.001
3-4	288 352 (22.5)	6251 (22.8)
5-6	296 767 (23.1)	6044 (22.0)
7-8	265 178 (20.7)	5574 (20.3)
9-10 (highest tier)	213 440 (16.6)	4364 (15.9)
Multifetal pregnancy	9340 (0.7)	955 (3.3)	<.001
Pregestational diabetes	3211 (0.2)	252 (0.9)	<.001
Gestational diabetes	75 615 (5.6)	3015 (10.5)	<.001
Preterm birth (<37 wk)	16 784 (1.3)	2618 (9.1)	<.001
Cesarean delivery	478 182 (35.6)	16 061 (55.9)	<.001
Postpartum hemorrhageb	50 371 (3.8)	1814 (6.3)	<.001
Postpartum period morbidity
Hypertension	270 178 (20.1)	14 359 (50.0)	<.001
Diabetes	203 757 (15.2)	8193 (28.5)	<.001
Cerebrovascular disease	23 406 (1.7)	1192 (4.1)	<.001
Cardiovascular disease	11 121 (0.8)	601 (2.1)	<.001
Follow-up duration, median (IQR), y	21.3 (20.3-22.1)	21.0 (20.2-21.9)	NAc

Abbreviations: HDPs, hypertensive disorders of pregnancy; NA, not applicable.

^a Health insurance premium tiers are based on annual premiums assessed at the time of pregnancy. Higher tier numbers indicate higher premiums and, by proxy, higher socioeconomic status. Tier 0 represents enrollees in the Medical Aid program for low-income households. Percentages for health insurance premium tier are calculated among females with nonmissing data. Tier information was missing for 58 264 females without HDPs (4.3%) and 1273 females with HDPs (4.4%).

^b Requiring transfusion.

^c Follow-up duration was reported descriptively and was not subjected to between-group hypothesis testing.

Incidence Density and Cumulative Incidence of Dementia

During approximately 2 decades of follow-up after delivery, dementia occurred in 18 females with HDPs and 297 females without HDPs (Table 2). The incidence density of dementia was 3.00 per 100 000 person-years in females with HDPs and 1.05 per 100 000 person-years in females without HDPs, and was significantly higher among females with HDPs across all dementia subtypes. Kaplan-Meier analysis demonstrated that the cumulative incidence of dementia at the end of follow-up was 7.2 per 10 000 females with HDPs and 2.8 per 10 000 females without HDPs (log-rank P < .001) (Figure 2). The cumulative incidence curves remained consistently higher among females with HDPs throughout the follow-up.

Table 2. Incidence and Incidence Density of Dementia According to HDP Subtype.

Dementia subtypea	Without HDPs (n = 1 341 738)	HDPs (n = 28 747)b				P valuec
		Gestational HT (n = 2030)	Preeclampsia (n = 23 357)	Eclampsia (n = 3687)	Total HDPs (n = 28 747)
Overall dementia
No. of cases	297	3	13	2	18	<.001
Incidence density, per 100 000 person-years	1.05	7.06	2.66	2.64	3.00
Alzheimer disease
No. of cases	222	3	7	2	12	<.001
Incidence density, per 100 000 person-years	0.78	7.06	1.43	2.64	2.00
Vascular dementia
No. of cases	80	1	5	0	6	<.001
Incidence density, per 100 000 person-years	0.28	2.35	1.02	0	1.00
Other dementia
No. of cases	16	0	1	0	1	NA
Incidence density, per 100 000 person-years	0.06	0	0.20	0	0.17
Unclassified dementia
No. of cases	22	0	2	0	2	NAd
Incidence density, per 100 000 person-years	0.08	0	0.41	0	0.33
Lewy body dementia e
No. of cases	0	0	0	0	0	NAd
Incidence density, per 100 000 person-years	0	0	0	0	0

Abbreviations: HDPs, hypertensive disorders of pregnancy; HT, hypertension.

^a Dementia subtype categories were not mutually exclusive.

^b HELLP (hemolysis, elevated liver enzymes, low platelet count) syndrome was identified in 3 females, none of whom developed dementia during follow-up, and is therefore not presented as a separate subgroup.

^c P values reflect comparison of incidence density between the total HDP group and the non-HDP group, calculated using person-years of observation. Other dementia and unclassified dementia had insufficient case numbers for meaningful statistical comparison; P values are not reported for these subtypes.

^d P values were not calculated for these subtypes because of insufficient event counts for meaningful statistical comparison.

^e No cases of Lewy body dementia occurred in either group during follow-up; this subtype was retained in the a priori case definition for completeness but no association between HDPs and Lewy body dementia could be examined in this cohort.

Figure 2. Kaplan-Meier Curves for the Cumulative Incidence of Dementia According to Hypertensive Disorders of Pregnancy (HDPs).

Kaplan-Meier estimates of the cumulative incidence of dementia (per 10 000 females) among females with and without HDPs over approximately 2 decades of follow-up. The estimated cumulative incidence at the end of follow-up duration was 7.2 per 10 000 females in the HDP group and 2.8 per 10 000 females in the non-HDP group. Differences between groups were assessed using the log-rank test (P < .001). The 19-year time point corresponds to the minimum follow-up duration available for all participants regardless of enrollment year.

Risk Factors for Dementia

In univariable Cox proportional hazards regression analysis, HDPs were associated with an increased risk of dementia (HR 2.73, 95% CI 1.68-4.45) (Table 3). In model 1, the association remained significant after adjustment for baseline maternal, socioeconomic, and obstetric factors (adjusted HR [AHR], 1.96; 95% CI, 1.19-3.23). Additional adjustment for postdelivery hypertension, diabetes, myocardial infarction, and stroke in model 2 attenuated the association, which was no longer statistically significant (AHR, 1.15; 95% CI, 0.70-1.91). Among the model 2 covariates, stroke showed the largest risk of dementia (AHR, 22.83; 95% CI, 17.46-29.05), higher than that of hypertension (AHR, 1.68; 95% CI, 1.32-2.15), diabetes (AHR, 1.90; 95% CI, 1.49-2.42), or myocardial infarction (AHR, 1.01; 95% CI, 0.55-1.85).

Table 3. Cox Proportional Hazards Analysis of Risk Factors for Incident Dementia During Follow-up.

Characteristic	Univariate, HR (95% CI)	Multivariate, AHR (95% CI)a
		Model 1	Model 2
Primary exposure
HDPs (reference: no HDPs)	2.73 (1.68-4.45)	1.96 (1.19-3.23)	1.15 (0.70-1.91)
Maternal age, y (reference <35 y)
≥35	6.29 (5.01-7.92)	5.92 (4.68-7.49)	3.90 (3.06-4.96)
Socioeconomic status (health insurance premium tier) (reference: 9-10)b
0-2	2.82 (1.88-4.25)	2.78 (1.85-4.18)	2.57 (1.71-3.87)
3-4	1.78 (1.17-2.71)	1.95 (1.28-2.97)	1.84 (1.21-2.81)
5-6	1.49 (0.97-2.28)	1.67 (1.09-2.57)	1.59 (1.04-2.45)
7-8	1.19 (0.76-1.88)	1.30 (0.83-2.56)	1.27 (0.81-2.01)
Obstetric and delivery factors
Multifetal pregnancy (reference: singleton pregnancy)	0.41 (0.06-2.91)	0.33 (0.05-2.33)	0.38 (0.05-2.69)
Pregestational diabetes (reference: no pregestational diabetes)	4.28 (1.37-13.34)	2.70 (0.86-8.47)	1.75 (0.55-2.69)
Gestational diabetes (reference: no gestational diabetes)	1.46 (0.97-2.20)	1.23 (0.82-1.86)	1.08 (0.71-1.63)
Preterm birth (<37 wk) (reference: term birth [≥37 wk])	1.82 (0.93-3.67)	1.34 (0.66-2.73)	1.08 (0.53-2.21)
Cesarean delivery (reference: vaginal delivery)	1.76 (1.41-2.20)	1.40 (1.13-1.76)	1.17 (0.93-1.47)
Postpartum hemorrhage (reference: no postpartum hemorrhage)	1.14 (0.65-1.98)	1.08 (0.62-1.89)	1.02 (0.59-1.78)
Postpartum vascular and metabolic diseasesc
Hypertension (reference: no hypertension)	3.94 (3.15-4.91)	NA	1.68 (1.32-2.15)
Diabetes (reference: no diabetes)	3.65 (2.92-4.58)	NA	1.90 (1.49-2.42)
Myocardial infarction (reference: no myocardial infarction)	4.16 (2.28-7.59)	NA	1.01 (0.55-1.85)
Stroke (reference: no stroke)	37.12 (29.62-46.52)	NA	22.83 (17.46-29.05)

Abbreviations: AHR, adjusted hazard ratio; HDPs, hypertensive disorders of pregnancy; HR, hazard ratio; NA, not applicable.

^a Model 1 was adjusted for maternal age, socioeconomic status, multifetal pregnancy, pregestational diabetes, gestational diabetes, preterm birth, cesarean delivery, and postpartum hemorrhage. Model 2 additionally adjusted for postdelivery hypertension, diabetes, myocardial infarction, and stroke.

^b Socioeconomic status was estimated using annual health insurance premium tiers at the time of pregnancy; higher tier numbers indicate higher premiums. Tier 0 represents enrollees in the Medical Aid program.

^c Postpartum vascular and metabolic diseases were included only in model 2. Because these conditions developed after the index pregnancy, they are presented as exploratory covariates reflecting potential intermediate pathways rather than conventional confounders. This model does not account for the timing of covariate onset relative to dementia diagnosis and may be subject to immortal time bias; results should be interpreted as hypothesis-generating rather than as a definitive adjusted estimate. These variables were not included in model 1.

Discussion

In this nationwide population-based cohort study including more than 1.3 million females, HDPs were associated with a higher incidence of dementia over approximately 2 decades of follow-up. After adjustment for maternal age, socioeconomic status, and obstetric factors, HDPs remained associated with an increased risk of dementia. This association was attenuated after additional adjustment for subsequent hypertension, diabetes, myocardial infarction, and stroke, consistent with the hypothesis that postpregnancy vascular and metabolic morbidity may contribute to the association between HDPs and later-life dementia.

Several population-based studies and meta-analyses have reported that women with a history of preeclampsia have an increased risk of both Alzheimer disease and vascular dementia.^15,23,24 Consistent with more recent evidence,^25,26,27 our findings similarly support this association, although our model 2 results are consistent with the hypothesis that postpregnancy vascular and metabolic morbidity may contribute to this association. A recent nationwide cohort study further reported an association between preeclampsia and young-onset dementia, particularly in women with early-onset or superimposed preeclampsia.²⁸

The present study extends prior work in several important respects. First, whereas most previous studies focused primarily on preeclampsia, this analysis examined the full clinical spectrum of HDPs, including gestational hypertension, preeclampsia, HELLP syndrome, and eclampsia. However, because dementia events within each HDP subtype were few, subtype-specific comparisons should be regarded as exploratory rather than a primary strength of this analysis; the overall HDP category remains the most statistically robust exposure definition in this cohort. Second, with more than 1.3 million females and approximately 2 decades of follow-up, this is among the largest and longest cohort studies to address this issue, providing evidence that an association between HDP and dementia may be detectable as early as midlife. Third, this study was conducted in an East Asian population using a nationwide database with near-universal coverage, addressing a gap in a literature dominated by Western cohorts. Fourth, the 2-model analytic approach offered additional insight into the potential vascular and metabolic pathways through which HDPs may confer long-term neurologic risk.

Consistent with prior literature, females with HDPs in our cohort also experienced higher rates of hypertension, diabetes, cardiovascular disease, and cerebrovascular disease during follow-up, supporting the concept that pregnancy complications may reflect underlying systemic vascular susceptibility that persists across the life course.^29,30,31,32 The attenuation of the association between HDP and dementia after adjustment for these postdelivery conditions in model 2 further suggests that vascular and metabolic sequelae of HDPs may contribute to the association between pregnancy-related hypertensive disorders and later-life neurodegeneration.

Among the model 2 covariates, stroke showed the largest risk of dementia (AHR, 22.83; 95% CI, 17.46-29.05), higher than that of hypertension (AHR, 1.68; 95% CI, 1.32-2.15), diabetes (AHR, 1.90; 95% CI, 1.49-2.42), or myocardial infarction (AHR, 1.01; 95% CI, 0.55-1.85) (Table 3). This pattern suggests that stroke, rather than metabolic disease more broadly, may play a particularly important role in the association between HDPs and later-life dementia, consistent with the well-established association between HDPs and subsequent stroke risk and with the particular relevance of cerebrovascular injury, including hemorrhagic stroke, to vascular dementia.³³ This interpretation is consistent with the marked attenuation of the association between HDP and dementia observed in model 2 once stroke and related vascular and metabolic conditions were accounted for.

The baseline characteristics of our cohort were consistent with national obstetric trends during the study period. Mean maternal age (29.2-30.0 years) was similar to the national mean, which first exceeded 30 years in 2005,³⁴ and the cesarean delivery rate in the non-HDP group (35.6%) was comparable with the national rate (33.6%-36.9%).³⁵ The prevalence of HDPs (2.1%) was lower than contemporary international estimates, likely reflecting the demographic profile of Korea in 2002 to 2005, when advanced maternal age and multifetal pregnancies were uncommon. National data show that the multiple birth rate increased from 20.0 per 1000 births in 2003 to 27.5 per 1000 births in 2008, coinciding with increasing maternal age and use of assisted reproductive technology, both established risk factors for HDPs.³⁶

Several biological pathways may plausibly link HDPs, a placental-origin disorder of early adulthood, with dementia risk decades later. HDPs could serve as early vascular insults priming the cerebrovascular system for future injury, consistent with the stroke-associated attenuation in model 2.^32,33,37,38 Molecular overlap with neurodegeneration has also been reported, including elevated circulating phosphorylated tau (p-tau181)^39,40 and neuronal injury biomarkers such as neurofilament light chain, glial fibrillary acidic protein, and S100B,^41,42,43 with experimental evidence that circulating factors from preeclamptic pregnancies may impair blood-brain barrier integrity.⁴³ This early-life vascular hypothesis motivated the present study and is summarized schematically in the eFigure in Supplement 1.

An important consideration is that females in this cohort remained relatively young at the end of follow-up. Because the NHIS database in the Republic of Korea was established in the early 2000s, the maximum follow-up period was approximately 2 decades. As a result, many females with HDPs were in their 50s or early 60s at the end of follow-up. Despite this, an association between HDPs and incident dementia was observed, suggesting that pregnancy-related vascular complications may be associated with earlier manifestations of cognitive risk. However, the attenuation of the association to nonsignificance after additional adjustment should not be interpreted as evidence of mediation. Formal mediation analysis was not performed because only 18 dementia events occurred among females with HDPs, limiting the statistical precision required to estimate direct and indirect effects. In addition, postdelivery hypertension, diabetes, myocardial infarction, and stroke occurred at varying times after the index pregnancy and may represent multiple, interrelated, time-varying mediators. Therefore, model 2 should be interpreted as an exploratory model evaluating the attenuation of the association after accounting for postdelivery vascular and metabolic conditions rather than as a formal mediation analysis.

These findings are consistent with prior studies demonstrating that vascular risk factors across the life course are associated with an increased risk of dementia, supporting the concept that early or midlife vascular insults may contribute to later neurodegenerative processes.⁴⁴ In addition, recent population-based evidence has reported an association between preeclampsia and early-onset dementia, further reinforcing the potential link between HDPs and long-term cognitive outcomes.²⁸ From a clinical perspective, these results support consideration of HDPs as a marker of long-term neurologic risk. Women with a history of HDPs may benefit from closer cardiovascular risk assessment and long-term monitoring of cognitive health.

Limitations

Several limitations should be acknowledged. First, dementia ascertainment based on administrative claims data may be subject to misclassification because the Korean NHIS database was developed for reimbursement rather than research purposes. To improve specificity, we prespecified dementia as the presence of both an ICD-10 diagnosis code and an antidementia medication prescription. However, this approach may have missed untreated or early-stage dementia cases, resulting in underascertainment of dementia incidence in both groups. Future studies using newly constructed NHIS datasets and alternative operational definitions, including ICD code–only definitions, may help assess the robustness of our findings. Similarly, HDP exposure was identified using ICD-10 claims codes and may also have been subject to misclassification. The relatively low prevalence of HDPs in this cohort (2.1%) compared with contemporary reports may reflect differences in population characteristics in early-2000s Korea, as well as possible underascertainment of milder cases.

Second, postdelivery hypertension, diabetes, myocardial infarction, and stroke were included in model 2 without time-dependent modeling. As these conditions may lie on the causal pathway between HDPs and dementia, and onset dates were not incorporated, model 2 may display immortal time bias. Consequently, its estimate (AHR 1.15) should be interpreted with caution as exploratory.

Third, residual confounding from unmeasured factors including educational level, smoking, obesity, family history, and APOE genotype cannot be excluded, particularly given obesity’s links with both HDPs and dementia. NHIS claims lacked complete reproductive history, restricted HDP exposure to the index pregnancy, and could not differentiate stroke subtypes (ischemic vs hemorrhagic). Furthermore, low dementia event counts in the HDP group limited formal interaction analyses. Future studies combining clinical and registry data should address these limitations.

Fourth, the relatively young age of the cohort at the end of follow-up likely resulted in underestimation of lifetime dementia risk. In addition, the small number of dementia events among females with HDPs limited statistical precision and precluded reliable multivariable analyses by dementia subtype. Therefore, the findings should be interpreted cautiously, and longer follow-up is needed to obtain more stable estimates.

Conclusions

In this cohort study of 1 370 485 females, HDPs were associated with an increased risk of incident dementia after adjustment for baseline maternal, socioeconomic, and obstetric factors. However, this association was attenuated and was no longer statistically significant after additional adjustment for postdelivery vascular and metabolic conditions. Given the small number of dementia events among females with HDPs, these findings should be interpreted cautiously. The present study suggests HDPs may represent an early-life marker of long-term neurovascular vulnerability.

Supplement 1.

eTable 1. ICD-10 Code for Delivery-Related Procedure

eTable 2. ICD-10 Codes for Dementia

eTable 3. ICD-10 Codes for Hypertensive Disorder During Pregnancy and Other Diseases

eFigure. Schematic Diagram of Proposed Mechanisms Linking Hypertensive Disorders of Pregnancy to later-Life Dementia

Supplement 2.

Data Sharing Statement