Association of preeclampsia with long‐term risk of neurodegenerative disorders

Abstract

Objective

Preeclampsia (PE) is a common complication of pregnancy linked to increased lifetime risk of multiple comorbidities. However, the link between PE and development of adverse neurologic conditions is still poorly understood. Our objective was to investigate the risk of neurodegenerative disorders for women with PE compared to those without PE.

Study design

We utilized the TriNetX database including electronic health records from 66 healthcare organizations and identified women 18+ years old with their last delivery between January 2000 and December 2023. PE was defined using ICD‐10 codes that occurred before delivery up until December 2020. The comparator group included pregnant women in the same period without PE diagnosis. After 1:1 propensity score matching on age at index pregnancy, demographics, health utilization, comorbidities, depression, reproductive factors, and body mass index, hazard ratios (HR), and 95% confidence intervals (CIs) for cognitive disorders were estimated between the PE and comparator group using a Cox proportional hazard regression model. Cognitive outcomes were identified by ICD‐10 codes and censored until one of the following conditions was met: (1) development of an outcome, (2) July 26, 2024, or (3) loss to follow‐up in the system.

Results

The PE group had a 9% increased risk of symptomatic cognitive impairment (HR 1.09; 95% CI 1.03, 1.16) and a 9% increased risk of composite cognitive impairment outcomes, including clinical, symptomatic, mild cognitive impairment, dementia, or related medication use (HR 1.09; 95% CI 1.03, 1.15) compared to the non‐PE group. Similar trends were observed in the term birth subgroup for symptomatic cognitive impairment (HR 1.06; 95% CI 1.00, 1.12) and composite outcomes(HR 1.06; 95% CI 1.01, 1.13). No significant associations were found in the preterm subgroup due to limited events. Risk of Parkinson's disease, parkinsonism, or related outcomes was not statistically different in the main model but was higher in sensitivity analyses with natural language processing (NLP) and accounting for death as competing risk. Cognitive impairment, dementia/Alzheimer's, and vascular dementia showed no differences in the main or NLP sensitivity analyses but were elevated in the competing risk sensitivity analysis.

Conclusion

Our findings demonstrate an increased risk of cognitive impairment among individuals with a history of PE, with additional associations for Parkinson's disease and parkinsonism identified in sensitivity analyses. These results highlight the importance of obstetrical health as a predictor of lifelong neurocognitive health, underscoring the need for targeted surveillance and early intervention strategies for neurodegenerative diseases in this population.

1. INTRODUCTION

Preeclampsia (PE) is a life‐threatening condition that can cause severe neonatal and maternal complications if not treated promptly. It is considered a major cause of maternal and neonatal death worldwide, an estimated 4 million people are diagnosed with PE each year, leading to the deaths of more than 70,000 people and over 500,000 infants [1].

Neurodegenerative disorders, particularly dementia, often lead to significant disability, placing a substantial financial burden on society due to the high costs of treatment and long‐term care. A study examining a subpopulation from the Health and Retirement Study found that the financial burden of dementia is comparable to that of heart disease and cancer [2]. Given that neurodegenerative diseases contribute to a high number of disability‐adjusted life years, they are a major driver of global health burden. Twice as many women as men are affected by Alzheimer's disease, a disparity that may be explained by genetic risk factors, such as the presence of the APOE4 lipoprotein mutation, which can affect men and women equally; however, women who carry this mutation are more likely to develop mild cognitive impairment [3]. Women generally have higher levels of tau protein, a biomarker for Alzheimer's, in their brains compared to men. A study recently found that women with high β‐amyloid (Aβ) levels exhibited significantly faster tau accumulation than men with similar Aβ levels. Additionally, female carriers of the APOEε4 allele showed faster tau accumulation in the inferior temporal region compared to male carriers. These findings suggest that women may be at greater risk for Alzheimer's disease progression due to more aggressive tau pathology in the presence of Aβ [4]. Depression is another factor to consider, as it also contributes to an increased risk of Alzheimer's disease. Women are twice as likely to experience depression as men [3].

While prospective studies have identified an association between PE and dementia, as well as anatomical changes in the occipital region of the brain and cognitive impairment, retrospective studies have contradicted these findings, reporting no association between PE and Alzheimer's disease [5, 6]. Exploring the relationship between PE and neurodegenerative disorders is crucial for improving diagnostics, reducing costs, optimizing resource allocation, and alleviating public health burden. This report examines whether PE is associated with increased risk of dementia, Parkinson's disease, and cognitive impairment later in life with a robust sample size and follow‐up time. We hypothesize that PE increases the risk of these neurodegenerative diseases.

1.1. Study design

We utilized the TriNetX Research cohort, which includes electronic health record data from 104 healthcare organizations (Supporting Information) to identify female patients who are over 18 years old at the time of delivery between January 2000 and December 2020. PE was defined using ICD‐10 diagnostic codes O14 and O15 that occurred before delivery. The comparator group included female patients identified by delivery in the same period without PE diagnosis. After 1:1 propensity score matching (PSM) on age at index pregnancy, demographics, substance use (nicotine dependence, alcohol, cannabis, opioid, and other stimulant related disorders), health utilization (history of inpatient, outpatient, emergency services, preventive medicine, and consultations), comorbidities (hyperlipidemia, type 2 diabetes, hypertension, cerebrovascular, and ischemic heart diseases), depression, body mass index, hazard ratios (HR), and 95% confidence intervals (CIs) for cognitive disorders were estimated between the PE group and the control group using a Cox proportional hazard regression model. Covariate balance was pursued if standard mean difference was below the 0.1 threshold. The cohort was followed up from 1 day after delivery to anytime until one of the following conditions was met: (1) development of any neurodegenerative outcomes including Parkinson's disease, parkinsonism, cognitive impairment, dementia/Alzheimer's disease, and composite outcomes with related medication of the conditions, which were identified by ICD‐10 and RxNorm codes (Table S3); (2) November 7, 2024 (date of study query); or (3) loss to follow‐up in the system. Three sensitivity analyses were performed: (1) combining death with neurodegenerative outcomes as composite outcomes to account for competing risk and survivorship bias; (2) utilizing the TriNetX Research NLP cohort to include natural language processing (NLP) derived facts to improve capture of cases, controls, and outcomes; (3) restricting the PE cohort to those who developed it during January 2000 to December 2017 to allow at least 8 years of follow‐up, but data were queried up to 2024, 24 years being the maximum possible follow‐up duration. We conducted two subgroup analyses stratified by: (1) term or preterm delivery; (2) PE with and without severe features. The Chung‐Shan Medical University Hospital Institutional Review Board approved using the TriNetX database for this investigation (approval CS2‐21176), waiving informed consent requirement, considering only de‐identified aggregated counts and statistical summaries compliant with HIPAA privacy rule standards were analyzed. This investigation conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement.

2. RESULTS

The cohort included 128,839 patients with PE, and 1:1 PSM matched to 128,839 patients without PE (Table 1). The mean follow‐up time was 2066.57 ± 1627.7 days for the PE group and 2142.4 ± 1649.8 days for the non‐PE group. The mean age at delivery was 28.7 ± 6.4 for the PE group and 29.0 ± 6.2 years for the non‐PE group. Most of the cohort was White (55.4%) and about 10% of the cohort had hypertensive disorders other than PE before pregnancy. Compared to their non‐PE counterparts, PE patients had a 9% higher risk of developing symptomatic cognitive impairment outcomes (HR 1.09; 95% CI 1.03, 1.16) and a 9% higher risk (HR 1.09; 95% CI 1.03, 1.15) of developing composite cognitive impairment outcomes that includes clinical, symptomatic, mild cognitive impairment, dementia, or having a related medication prescription (donepezil, memantine, rivastigmine, galantamine; Figure 1 and Table S1). Comparable trends were observed in the term birth subgroup for both symptomatic cognitive impairment (HR 1.06; 95% CI 1.00, 1.12) and composite outcomes (HR 1.06; 95% CI 1.01, 1.13), but not in the preterm subgroup, likely due to the limited event count. In addition to the significantly increase risk of symptomatic (HR 1.13; 95% CI 1.04, 1.23) and composite cognitive impairment outcomes (HR 1.16; 95% CI 1.07, 1.25) as seen in the main model, the PE with severe features group also demonstrated significantly increased risk of mild cognitive impairment (HR 1.61; 95% CI 1.06, 2.44) and vascular dementia (HR 6.25; 95% CI 1.40, 27.94), which non were statistically increased in the PE without severe features group (Table S2). No statistical differences in Parkinson's disease, parkinsonism, or the composite outcome (including relevant medication) were observed between the PE and non‐PE groups in the main model, but all were significantly higher in the sensitivity analyses that accounted for death as competing risk. For clinical cognitive impairment, mild cognitive impairment, dementia/Alzheimer's disease, and vascular dementia, no differences were found in the main model or the NLP sensitivity analysis. All outcomes were significantly higher in the sensitivity analysis accounting for death as a competing risk whether it is the 2000–2020 cohort or 2000–2017 cohort that allowed longer follow‐up (Table S2).

TABLE 1.

Baseline characteristics in women with or without preeclampsia.

	Before propensity score matching			After propensity score matching
Characteristic	Preeclampsia (n = 141,035)	Non‐preeclampsia (n = 1,924,629)	SMD	Preeclampsia (n = 128,839)	Non‐preeclampsia (n = 128,839)	SMD
Age at index, mean (SD), year	28.7 ± 6.4	29.0 ± 6.2	0.046	28.7 ± 6.4	28.7 ± 6.4	0.003
Race, n (%)
White	71,375 (55.4)	939,543 (52.7)	0.055	71,375 (55.4)	71,558 (55.5)	0.003
Black or African American	29,307 (22.7)	290,297 (16.3)	0.164	29,307 (22.7)	29,935 (23.2)	0.012
Asian	5998 (4.7)	143,676 (8.1)	0.140	5998 (4.7)	5770 (4.5)	0.008
Other a	7008 (5.4)	95,695 (5.4)	0.003	7008 (5.4)	6795 (5.3)	0.007
Ethnicity, n (%)
Hispanic	28,816 (22.4)	344,808 (19.3)	0.075	28,816 (22.4)	29,015 (22.5)	0.004
Non‐Hispanic	77,368 (60.1)	1,016,245 (57.0)	0.063	77,368 (60.1)	77,600 (60.2)	0.004
Unknown	22,655 (17.6)	422,911 (23.7)	0.152	22,655 (17.6)	22,224 (17.2)	0.009
Body mass index, mean (SD), kg/m2	34.1 ± 8.1	30.9 ± 7.1	0.419	29.6 ± 7.8	29.6 ± 7.9	0.002
Substance use, n (%)
Nicotine dependence (smoking)	6870 (5.3)	70,012 (3.9)	0.067	6870 (5.3)	7644 (5.9)	0.026
Alcohol‐related disorders	1215 (0.9)	12,263 (0.7)	0.028	1215 (0.9)	1374 (1.1)	0.012
Cannabis‐related disorders	2876 (2.2)	25,379 (1.4)	0.060	2876 (2.2)	2876 (2.2)	<0.001
Opioid‐related disorders	1578 (1.2)	16,766 (0.9)	0.028	1578 (1.2)	1856 (1.4)	0.019
Other stimulant‐related disorders	629 (0.5)	4711 (0.3)	0.037	629 (0.5)	567 (0.4)	0.007
Potential socioeconomic and psychosocial health hazards, n (%)	4407 (3.4)	43,551 (2.4)	0.058	4407 (3.4)	4277 (3.3)	0.006
Comorbidities, n (%)
Hyperlipidemia	3356 (2.6)	23,847 (1.3)	0.091	3356 (2.6)	3099 (2.4)	0.013
Type 2 diabetes mellitus	6578 (5.1)	26,816 (1.5)	0.203	6578 (5.1)	6332 (4.9)	0.009
Hypertensive diseases	13,709 (10.6)	44,179 (2.5)	0.334	13,709 (10.6)	13,186 (10.2)	0.013
Cerebrovascular diseases	715 (0.6)	4412 (0.2)	0.049	715 (0.6)	570 (0.4)	0.016
Ischemic heart diseases	626 (0.5)	4212 (0.2)	0.042	626 (0.5)	565 (0.4)	0.007
Depression	12,216 (9.5)	103,720 (5.8)	0.138	12,216 (9.5)	12,449 (9.7)	0.006
Healthcare utilization, n (%)
Outpatient services	64,681 (50.2)	664,972 (37.3)	0.263	64,681 (50.2)	65,375 (50.7)	0.011
Inpatient services	22,767 (17.7)	101,102 (5.7)	0.381	22,767 (17.7)	22,631 (17.6)	0.003
Emergency services	35,292 (27.4)	349,063 (16.2)	0.185	35,292 (27.4)	35,967 (27.9)	0.012
Preventive medicine	18,247 (14.2)	210,286 (11.8)	0.071	18,247 (14.2)	18,370 (14.3)	0.003
Consultations	18,381 (14.3)	150,460 (8.4)	0.185	18,381 (14.3)	18,273 (14.2)	0.002

Note: If the patient is less or equal to 10, results show the count as 10.

Abbreviations: SD, standard deviation; SMD, standardized mean difference.

^a Includes American Indian or Alaska Native, Native Hawaiian or Other Pacific Islander, other race, and those with unknown race.

FIGURE 1.

Event number, incidence, and hazard ratio of neurodegenerative outcomes among women with preeclampsia compared to women without preeclampsia diagnosis after propensity score matching. CI, confidence interval; HR, hazard ratio. *Starred outcomes are composite outcomes.

3. DISCUSSION

Evidence regarding the association between PE and Alzheimer's disease or other forms of dementia remains inconclusive. Andolf et al. found a weak association between PE and Alzheimer's disease (HR 1.05; 95% CI 0.81, 1.26), but a stronger association with vascular dementia (HR 2.43; 95% CI 1.84, 3.21), which is a finding that was not validated in our main model nor the PE without severe features group [7]. While our study did not find a statistically significant association between PE and Alzheimer's disease per se (HR 0.85; 95% CI 0.54, 1.35), a sensitivity analysis incorporating death as a competing risk showed significant associations across all outcomes after PSM (Table S2). Fields et al. followed up participants for 30 years postdelivery and found no PE‐related cognitive decline or memory loss though limited to 40 participants [8]. Conversely, a French 8‐year study first showed increased early‐onset dementia risk with PE or hypertensive disorders versus uncomplicated pregnancies. Our main model detected no significant PE‐dementia association possibly due to short mean follow‐up (∼5.7 years) and young cohort (∼28 years), insufficient to capture late‐onset dementia seen on longer follow‐up studies.

Our analysis is the first to examine the association between PE and Parkinson's disease. Aarsland et al. report that over 75% of PD patients surviving beyond 10 years eventually develop dementia, with early mild cognitive impairment accelerating this progression. Although our main model did not show a significant link between PE and PD, sensitivity analyses that accounted for death (HR 1.23; 95% CI 1.12, 1.34) and NLP‐identified events (HR 1.39; 95% CI 1.04, 1.86) suggest that PE may indirectly increase PD risk [9]. Choi et al. found that age‐related macular degeneration, which reflects vascular and inflammatory changes, raises PD risk even among individuals with healthy lifestyles [10]. This supports our hypothesis that systemic inflammatory disease with vascular etiology like PE could similarly result in PD. The lack of significance in our primary model might be due to insufficient follow‐up for capturing late‐onset PD (with mean PD‐dementia onset around 10 years) or the impact of competing mortality risks [11].

We combined dementia and AD into a composite outcome because both share similar downstream mechanisms, such as neuroinflammation, oxidative stress, mitochondrial dysfunction, and vascular contributions. PE may also trigger vascular and inflammatory processes that predispose individuals to a range of neurodegenerative outcomes. This composite approach reflects a broader neurodegenerative burden and also enhances statistical power, as shown by our main and sensitivity analyses with significant associations even when individual dementia or AD outcomes were borderline statistically insignificant.

3.1. Strengths and limitations

One of the strengths of our study is the robust sample size from a large nationwide EHR database with a 24‐year follow‐up period and well‐balanced groups using propensity scores to adjust for multiple confounders. However, our study has several limitations. First, we could not adjust for significant variables that were not captured within the TriNetX database, such as insurance status, lifestyle factors, and environmental influences. Additionally, our study is based on ICD codes, which could cause a misclassification bias depending on the assessment of medical professionals and coding errors. We could not determine the exact age at the individual level in which participants were diagnosed with neurodegenerative disorders or related symptoms, which limits our ability to assess an association with early‐onset symptoms. We cannot accurately ascertain the number of individuals lost to follow‐up within the network as the TriNetX database does not explicitly track loss to follow‐up. However, the follow‐up duration was similar between groups suggesting minimal differential bias of incomplete records due to patients receiving care outside the network between groups. Another limitation is that this study is not designed to determine whether the symptoms of neurodegenerative disorders are directly attributable to PE itself or if they come from underlying vascular pathology or other predisposing factors that may have independently contributed to both PE and an elevated lifetime risk of neurodegenerative disorders.

4. CONCLUSION

It is important to note that this study aims to explore the association rather than the causal effect between PE and the risk of developing neurodegenerative disorders later in life. Our results show that PE may be associated with an increased risk of developing neurodegenerative disorders later in life. However, further research is needed to explore other symptoms associated with these neurodegenerative disorders, such as memory loss and other cognitive decline markers. Prospective research should be conducted on this topic to allow for better long‐term follow‐up of patients to capture symptoms during the age range when neurodegenerative disorders are most prevalent, providing more accurate data on the onset and progression of these conditions. In addition, the continued exploration of biomarkers is important for enabling early diagnosis. Early detection would allow for targeted screening and treatment of individuals exposed to hypertensive disorders, with PE being one of the most common, ultimately improving outcomes and reducing the long‐term impact on both individuals and society.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Supporting Information