Association of cardiovascular disease with dementia: A longitudinal analysis using National Alzheimer's Coordinating Center data

Abstract

We analyzed how cardiovascular disease subtypes influence the prevalence and incidence of dementia among 30,582 individuals aged 50 and older in the National Alzheimer's Coordinating Center cohort. We calculated prevalence ratios (aPR) and hazard ratios (aHR), using models adjusted for age, sex, race, years of education, hypertension, diabetes, and dyslipidemia. Stroke (aPR: 1.26; 95% CI: 1.20–1.32) and history of arrhythmias (aPR: 1.17; 95% CI: 1.09–1.24) were associated with higher dementia prevalence. In survival analysis, stroke was associated with a 55% increased risk of incident dementia (aHR: 1.55; 95% CI: 1.36–1.77).

Introduction

Cardiovascular disease (CVD) and dementia are two of the most significant global health challenges, with their burden expected to grow as global populations age. ¹ CVD remains the leading cause of death worldwide, with ischemic heart disease accounting for 13%, and stroke for 10% of total global deaths. ² Meanwhile, dementia ranked as the seventh leading cause of death, ² affecting approximately 57 million people worldwide, ³ a figure projected to reach 152 million by 2050. ⁴

The relationship between CVD and dementia is complex and driven by shared risk factors such as hypertension, diabetes, and dyslipidemia.^1,5 The underlying mechanisms between CVD and dementia involve chronic cerebral hypoperfusion, ⁶ neurohormonal activation, and amyloid-β protein aggregation. ⁵ Among CVD subtypes, stroke, ischemic heart disease, heart failure, and atrial fibrillation are associated with an increased dementia risk.^7–9 While prior studies have explored this relationship, these often rely on administrative claims data or pooled risk estimates, which limit accuracy and therefore the ability to examine how specific CVD subtypes independently contribute to dementia outcomes. Furthermore, dementia diagnoses in these datasets are typically not informed by comprehensive clinical evaluation.

This study leverages data from the National Alzheimer's Coordinating Center (NACC) dataset, one of the most comprehensive dementia datasets in the United States. ¹⁰ NACC serves as a centralized data repository, aggregating longitudinal data from over 30 Alzheimer's Disease Research Centers (ADRCs) from across the country. Participants are recruited through clinician referrals, self-referrals, and active recruitment in community organizations. To ensure systematic data collection across ADRCs, in 2005 NACC established the Uniform Data Set (UDS), a standardized, longitudinal protocol that facilitates harmonized clinical, cognitive, and diagnostic data across centers.

This data set provides a valuable resource for studying the progression of cognitive decline and its relationship with CVD. By leveraging structured cognitive assessments and clinician-adjudicated diagnoses across a multi-center U.S. cohort, our study examines how individual CVD subtypes may contribute differently to dementia risk. These findings have important implications for risk stratification, preventive strategies, and the development of integrated care models that address the dual burden of CVD and dementia.

Methods

Study design

We conducted a cross-sectional and survival analysis using data from the NACC ¹⁰ dataset, from September 2005 to March 2023. We included patients who were aged 50 and older at the time of their baseline assessment. At each ADRC, participants were followed annually for as long as they remained able and willing to participate. During each visit, trained clinicians or clinical staff completed 18 standardized data-collection forms as part of the UDS. These forms capture demographic information, clinical history, neuropsychological test results, and other relevant health data. Based on all available information at the time of the assessment, a cognitive diagnosis was assigned at each visit by either a consensus panel or a single physician. ¹⁰

At each center, experienced clinicians made dementia diagnoses using a comprehensive clinical assessment based on the NACC UDS protocol. ¹¹ This protocol provides a structured framework for diagnostic decision-making, incorporating demographic and medical histories, physical and neurological examinations, and standardized scales to assess vascular and motor features. Informant-based instruments, such as the Neuropsychiatric Inventory Questionnaire and the Functional Assessment Questionnaire, are used to evaluate behavioral symptoms and functional decline. Cognitive performance is measured using a standardized neuropsychological battery (Form C1), which assesses memory, language, attention, and executive function. Clinicians integrate all available data, including Clinical Dementia Rating (CDR) scores and established diagnostic criteria by the National Institutes of Health, ¹² and the National Institute on Aging and Alzheimer's Association (NIA-AA), ¹³ to assign a final diagnosis (Form D1), specifying cognitive status, dementia presence, severity, and etiology.

The primary outcome of our study was dementia status, defined using the NACC variable NACCETPR or its equivalent in earlier UDS versions. This clinician-assigned variable captures the presence of all-cause dementia, regardless of subtype, as recorded at baseline and at each follow-up visit.

To examine the relationship between CVD and dementia, we used chronic medical conditions listed in the NACC dataset. These conditions are primarily participant- or informant-reported and are documented by the clinician in the UDS forms. CVD subtypes and interventions were categorized into nine groups: (1) stroke, (2) heart attack, (3) congestive heart failure (CHF), (4) transient ischemic attack (TIA), (5) atrial fibrillation, (6) other cardiovascular conditions (e.g., valvular heart disease, peripheral vascular disease, or cardiomyopathy), (7) history of arrhythmia requiring a pacemaker, (8) history of heart disease with angioplasty, endarterectomy, or stent placement, and (9) history of heart disease with cardiac bypass procedure. The “other cardiovascular conditions” category captures diagnoses not otherwise specified in the predefined subtypes and may reflect site-level variability in reporting less common CVD presentations.

Statistical analysis

To assess dementia prevalence, we calculated prevalence ratios (PR) using log-binomial regression, adjusting for sex, age, race, years of education, diabetes, hypertension, and dyslipidemia. To evaluate the risk of developing dementia, we conducted a survival (time-to-event) analysis using Cox proportional hazards models to calculate adjusted hazard ratios (aHR) for each CVD subtype using the same covariate adjustments. Hypertension, diabetes, and dyslipidemia are well-established risk factors for both CVD and dementia and were included to help account for confounding. These adjustments enabled us to examine whether the association between CVD and dementia persisted independently of these shared vascular risk factors. To establish temporality, we excluded participants with a dementia diagnosis at baseline; incident dementia was assessed during follow-up, ensuring that CVD conditions preceded dementia onset. Data were analyzed using RStudio. This study was deemed exempt by Emory University Institutional Review Board.

Results

We analyzed 30,582 participants aged 50 and older from the NACC dataset. The mean age was 73 years (SD = 9.3), with 57% female and 81% White. Educational attainment varied: 8.8% had less than high school education, 38% completed high school, 24.5% had a bachelor's degree, 19% held a master's degree, and 9.4% a doctorate degree. The prevalence of hypertension was 52%, dyslipidemia 51%, and diabetes 13.2%. Full demographic characteristics for the cross-sectional and survival analysis samples are presented in Table 1. The prevalence of each CVD subtype for both samples is detailed in Supplemental Table 1.

Table 1.

Baseline demographic and clinical characteristics of participants by cardiovascular disease (CVD) status.

Cross-Sectional Sample				Survival Analysis Sample
	Yes	No	Overall		Yes	No	Overall
CVD	(N = 9625)	(N = 20,957)	(N = 30,582)	CVD	(N = 5853)	(N = 13,371)	(N = 19,224)
Sex n (%)				Sex n (%)
Male	5090 (52.9)	8081 (38.6)	13,171 (43.1)	Male	2939 (50.2)	4825 (36.1)	7764 (40.4)
Female	4535 (47.1)	12,876 (61.4)	17,411 (56.9)	Female	2914 (49.8)	8546 (63.9)	11,460 (59.6)
Age at baseline				Age at baseline
Mean in years (SD)	76.1 (8.94)	71.4 (9.44)	72.9 (9.53)	Mean in years (SD)	75.5 (9.01)	71.1 (9.13)	72.4 (9.32)
Race n (%)				Race n (%)
White	7936 (82.5)	16,836 (80.3)	24,772 (81.0)	White	4796 (81.9)	10,525 (78.7)	15,321 (79.7)
Black -AA	1283 (13.3)	3016 (14.4)	4299 (14.1)	Black -AA	836 (14.3)	2174 (16.3)	3010 (15.7)
Other	406 (4.2)	1105 (5.3)	1511 (4.9)	Other	221 (3.8)	672 (5.0)	893 (4.6)
Education n (%)				Education n (%)
Less than high school	933 (9.7)	1756 (8.4)	2689 (8.8)	Less than high school	400 (6.8)	810 (6.1)	1210 (6.3)
GED/High school	3753 (39.0)	7908 (37.7)	11,661 (38.1)	GED/High school	2193 (37.5)	4700 (35.2)	6893 (35.9)
Bachelor	2312 (24.0)	5194 (24.8)	7506 (24.5)	Bachelor	1468 (25.1)	3475 (26.0)	4943 (25.7)
Master	1679 (17.4)	4161 (19.9)	5840 (19.1)	Master	1161 (19.8)	3051 (22.8)	4212 (21.9)
Doctorate	948 (9.8)	1938 (9.2)	2886 (9.4)	Doctorate	631 (10.8)	1335 (10.0)	1966 (10.2)
Diabetes n (%)	1757 (18.3)	2282 (10.9)	4039 (13.2)	Diabetes n (%)	1113 (19.0)	1496 (11.2)	2609 (13.6)
Hypertension n (%)	6333 (65.8)	9664 (46.1)	15,997 (52.3)	Hypertension n (%)	3862 (66.0)	6268 (46.9)	10,130 (52.7)
Dyslipidemia n (%)	6166 (64.1)	9455 (45.1)	15,621 (51.1)	Dyslipidemia n (%)	3791 (64.8)	6134 (45.9)	9925 (51.6)
Dementia n (%)	3772 (39.2)	7586 (36.2)	11,358 (37.1)

AA: African American; CVD: cardiovascular disease; GED: General Education Development; SD: standard deviation.

Dementia is reported only in the cross-sectional sample; the survival sample excludes baseline dementia to assess incident cases.

Prevalence

At baseline, 11,358 individuals had dementia, representing an overall prevalence of 37%. Among all CVD subtypes, stroke was associated with the highest prevalence of dementia (aPR: 1.26; 95% CI: 1.20–1.32), followed by history of arrhythmias requiring pacemaker use (aPR: 1.17; 95% CI: 1.09–1.24), and TIA (aPR: 1.08; 95% CI: 1.02–1.14). Figure 1 illustrates these findings.

Figure 1.

Adjusted prevalence ratios for dementia by cardiovascular subtype. Estimates are adjusted for sex, age, race, years of education, diabetes, hypertension, and dyslipidemia. CVD: cardiovascular disease.

Incidence

Participants were followed for a mean duration of 1091 days (SD = 1004; range: 0–3458), with an average of 3.58 visits (SD = 2.45; range: 1–10) per individual (Supplemental Table 2). Among 19,224 participants without a dementia diagnosis at baseline, 5853 developed dementia during the study period (2005–2023), corresponding to an incidence rate of 30%. After adjusting for age, race, sex, years of education, hypertension, diabetes and dyslipidemia, stroke was the only condition associated with increased risk of dementia development, elevating the risk by 55% (aHR: 1.55; 95% CI: 1.36–1.77) (Figure 2).

Figure 2.

Adjusted hazard ratios of incident dementia by cardiovascular subtype. Estimates are adjusted for sex, age, race, years of education, diabetes, hypertension, and dyslipidemia. CVD: cardiovascular disease.

Discussion

Our results align with prior studies demonstrating that stroke significantly increases the risk of dementia.^14,15 These findings support existing hypotheses linking cerebrovascular pathology to cognitive decline and dementia.^1,16–20 Several mechanisms may explain this association. Small vessel disease contributes to microinfarcts, white matter changes, and hippocampal atrophy, all of which impair cognitive function. ²¹ Blood-brain barrier disruption leads to vascular leakage and inflammation, increasing the risk of microhemorrhages. ²² Hemodynamic changes from atherosclerosis and arterial stenosis further reduce brain perfusion, exacerbating ischemic damage. ²³ Cerebral amyloid angiopathy promotes amyloid-β accumulation in blood vessels, promoting cortical microinfarcts and increased risk of hemorrhage. ²⁴

In addition, hypertension, dyslipidemia and diabetes are well-established and modifiable risk factors for both stroke and dementia.^25–33 Our findings highlight the development of stroke superimposed on preexisting hypertension, diabetes, or dyslipidemia significantly amplifies dementia risk. Specifically, individuals who experienced a stroke had a 55% higher risk of developing dementia compared to those without stroke, independent of other risk factors.

These results underscore the critical role of comprehensive cardiovascular risk management in dementia prevention strategies. Although this study did not directly assess treatment effects, prior research suggests that effective control of hypertension, ³⁴ diabetes,^35,36 and dyslipidemia ³⁷ may reduce stroke incidence and delay or prevent the onset of dementia.

In contrast with previous studies, including meta-analyses reporting a positive association between atrial fibrillation and dementia,^7,38,39 we did not observe a significant association in our adjusted models. This discrepancy may reflect selection bias within the NACC cohort, underreporting of asymptomatic or paroxysmal atrial fibrillation, differences in comorbidity profiles, variation in covariate adjustment across studies, and the absence of detailed treatment data, such as anticoagulation use, in our dataset.

Several additional factors may explain differences between our findings and those of prior studies, particularly in relation to study populations, diagnostic approaches, and disease classification. For example, a territory-wide study from the CDARS database in Hong Kong found that 11.0% of patients with heart failure developed dementia over a median follow-up of 4.1 years, with atrial fibrillation independently associated with increased dementia risk. ⁴⁰ In contrast, our models did not identify statistically significant associations for heart failure or atrial fibrillation. Differences in cohort composition (e.g., the CDARS cohort was Asian), diagnostic methodology, and health system–level treatment patterns may have contributed to these discrepancies.

Similarly, we did not find a significant association between heart attack and dementia. However, prior work has shown that earlier onset of coronary heart disease (CHD) is associated with greater dementia risk. ⁴¹ Our dataset did not include information on the age of CVD onset, which may explain the null association observed. Future studies should incorporate timing of cardiovascular events to clarify their relationship with dementia risk.

Meta-analytic studies reinforce the link between CVD and dementia, showing consistent associations for CHD, heart failure, and atrial fibrillation with dementia.^42,43 Notably, Sun et al. found that CHD was associated with increased risk of Alzheimer's disease (pooled OR/RR: 1.22; 95% CI: 1.00–1.48), while heart attack alone was not (pooled RR: 1.09; 95% CI: 0.91–1.30). ⁴³ These findings emphasize the importance of how cardiovascular conditions are defined and analyzed. In our study, heart attack was evaluated as a discrete category, consistent with the NACC Uniform Data Set (UDS), where it is classified separately from broader CHD. Other related diagnoses—such as valvular heart disease, peripheral artery disease, and cardiomyopathies—are grouped under “other cardiovascular conditions.” The lack of association observed in our models may reflect these classification structures, along with unmeasured variation in disease severity, treatment, and onset.

Although our models adjusted for race, the limited racial representation in the cohort may constrain the generalizability of our findings; particularly given evidence that African American individuals face higher dementia risk and differ from White individuals in both, risk factors and disease presentation. ⁴⁴

One key strength of our study is the large, national sample, with longitudinal, standardized clinical data. The NACC data, however, has limitations, including being a referral-based sample and not statistically representative of the U.S. population. In addition, attrition bias may be present if participants lost to follow-up differ from those retained; particularly if dropout is related to both CVD and dementia risk. Future studies should aim to validate these findings in more diverse populations and consider integrating biomarkers and pathological data to improve diagnostic accuracy and elucidate underlying mechanisms. Ongoing programs and targeted interventions, including lifestyle modifications, early cognitive screening, and integrated care models, may help reduce dementia risk and improve long-term outcomes in patients with CVD. ³⁹

In conclusion, our findings highlight a significant association between stroke­ and increased risk of dementia. These results underscore the importance of integrating targeted CVD prevention and management interventions into primary care to reduce dementia risk.