Association of Alzheimer's disease blood-based biomarkers and visit-to-visit blood pressure variability: The HABS-HD study

Abstract

Background

The interplay between Alzheimer's disease (AD) and the cardiovascular system remains poorly understood. While vascular factors influencing AD have been studied, whether AD pathology conversely affects blood pressure regulation is unclear.

Objective

This study investigated whether baseline plasma AD biomarkers predict subsequent visit-to-visit blood pressure variability (BPV).

Methods

This prospective analysis included 470 community-based older adults from the Health and Aging Brain Study-Health Disparities (HABS-HD) cohort. Multiple linear regression models separately assessed associations between each baseline plasma AD biomarker (Aβ_42/40 ratio, total tau, p-tau181, NfL) and follow-up BPV, with progressive adjustments for demographics, clinical diagnosis, cardiovascular risk factors, and mean blood pressure. In fully adjusted models, all plasma AD biomarkers were included simultaneously to adjust for potential mutual confounding and identify independent associations. Subgroup and sensitivity analyses were conducted.

Results

In fully adjusted models, total tau remained independently associated with higher systolic blood pressure variability (β = 3.94, 95%CI: 1.00–6.87, p = 0.009). These associations were particularly prominent in males (β = 5.80, 95% CI: 1.66–9.95, p = 0.007), non-Hispanic White (β = 4.00, 95% CI: 0.12–7.88, p = 0.044), cognitively unimpaired individuals (β = 4.42, 95% CI: 1.14–7.70, p = 0.009), APOE ε4 non-carriers (β = 4.91, 95% CI: 1.42–7.70, p = 0.009), and individuals not using antihypertensive medication (β = 4.29, 95% CI: 0.71–7.87, p = 0.020). Sensitivity analyses confirmed robustness of findings.

Conclusions

Plasma total tau independently predicts BPV, especially during pre-clinical AD stages. This finding reveals an important connection between AD pathology and vascular dysregulation, supporting the paradigm of AD as a systemic disorder and providing new directions for early risk identification and intervention.

Introduction

Alzheimer's disease (AD) is a complex neurodegenerative disorder whose pathological mechanisms have long focused on two classical hallmarks: amyloid-β (Aβ) deposition and neurofibrillary tangles formed by hyperphosphorylated tau protein. However, as research advances, the pathological complexity of AD extends far beyond initial assumptions. Growing evidence suggests that AD may not be merely a brain-specific disease but rather a systemic disorder often coexisting with multiple comorbid pathologies. ¹ Among these comorbidities, cardiovascular and metabolic diseases—particularly hypertension—are considered to play a critical role in the onset and progression of AD.^2,3

Studies have revealed a complex age-dependent relationship between average blood pressure and dementia risk: midlife hypertension clearly increases the risk of dementia in later life, whereas in older age, this association weakens or even reverses, with higher blood pressure sometimes linked to a lower dementia risk. ⁴ This seemingly paradoxical phenomenon suggests that other, more sensitive blood pressure-related indicators may be involved. In recent years, blood pressure variability (BPV) has gained widespread attention as a novel vascular risk predictor independent of average blood pressure. ⁵ BPV refers to fluctuations in blood pressure over time and can be categorized into short-term variability (minutes to hours) and long-term variability (months to years). ⁶ Among these, visit-to-visit BPV has become a research focus due to its accessibility in clinical practice and significant prognostic value. ⁷ A growing body of evidence directly links BPV to cognitive decline and dementia risk. A systematic review and meta-analysis (20 studies, approximately 7.9 million participants) confirmed higher BPV links to greater dementia and cognitive impairment risk: 1-unit higher systolic BPV raises risk by 25%. ⁸ Notably, a Mendelian randomization study has revealed a significant causal effect of SBPV on AD. ⁹

Although the impact of vascular factors on AD has been extensively studied, understanding of the reverse relationship, i.e., whether AD pathology itself affects systemic physiological regulation, particularly blood pressure stability, remains limited. Most current studies are still based on a unidirectional causal model, emphasizing the influence of vascular factors on AD pathology while overlooking the potential feedback effect of AD pathology on the cardiovascular system.^10,11 Notably, several clues suggest that such a reverse influence may indeed exist. Patients with AD, even those at the mild cognitive impairment (MCI) stage, often exhibit clinical features of autonomic dysfunction, including orthostatic hypotension and altered heart rate variability, suggesting that the neurodegenerative process may have systemic physiological consequences.^12,13 Neuroimaging studies have confirmed that AD pathology involves brainstem and cortical regions that are central to cardiovascular autonomic regulation; damage to these areas may directly disrupt blood pressure stability.^7,11 AD-related neurofibrillary changes, primarily driven by tau protein, propagate through the brain in a hierarchical sequence (Braak staging), progressively invading not only cortical areas but also brainstem regions that are central to autonomic control.^14,15 Notably, the insular cortex is affected at a preclinical stage and has been implicated in early autonomic dysregulation and blood pressure variability. ¹⁶ Postmortem studies have identified significant tau pathology in key autonomic nuclei of the medulla oblongata, including the dorsal motor nucleus of the vagus and the nucleus of the solitary tract. ¹⁷ These specific brainstem nuclei are critical for baroreflex sensitivity and the maintenance of blood pressure homeostasis. If AD pathology can exacerbate vascular instability, it may create a potential vicious cycle, where AD pathology worsens vascular health, and conversely, vascular dysfunction accelerates the progression of AD pathology. Elucidating this feedback loop is essential for comprehensive risk stratification and the development of holistic intervention strategies.

However, direct empirical testing of this hypothesis remains scarce. There is a particular lack of prospective studies examining whether baseline AD pathology predicts long-term BPV. Recently, key plasma biomarkers based on the AT(N) (amyloid/tau/neurodegeneration) framework, including amyloid-β (Aβ), tau, and neurofilament light chain (NfL), have emerged as non-invasive tools for the detection and diagnosis of AD.^18,19 By measuring plasma amyloid-β40 (Aβ₄₀), amyloid-β42 (Aβ₄₂), total tau (t-tau), phosphorylated tau-181 (p-tau181), and NfL levels, we can non-invasively assess the burden of AD pathology at the preclinical stage and prospectively investigate its systemic impact on blood pressure regulation.

This study utilized data from the Health and Aging Brain Study-Health Disparities (HABS-HD) prospective cohort^20,21 to test the hypothesis that baseline plasma AD biomarker levels predict higher visit-to-visit BPV during follow-up, and that this association is independent of traditional vascular risk factors. Previous studies utilizing the HABS-HD cohort have explored the association of cardiovascular risk factors and blood biomarkers with cognition ²² ; however, no study to date has investigated the reverse relationship. The longitudinal follow-up design of the HABS-HD cohort provides a unique opportunity to examine the effect of baseline plasma AD biomarkers on subsequent visit-to-visit blood pressure variability. This research holds significance in directly exploring the potential impact of AD pathology on blood pressure stability, providing evidence for the new paradigm that “AD pathology can exacerbate systemic comorbidities.”

Methods

Participants and cognitive diagnosis

This study utilized data from the Health and Aging Brain Study—Health Disparities (HABS-HD) cohort, a large, ongoing, community-based prospective study investigating cognitive aging and health disparities among diverse populations.^21,23–25 The study methodology has been previously described in detail. ²⁶ Briefly, participants were recruited using a community-based participatory research approach and were eligible if they were 50 years of age or older, fluent in English or Spanish, and capable of undergoing the study protocol, which included blood draws and neuroimaging. The HABS-HD study is conducted under the University of North Texas Health Science Center (UNTHSC) Institutional Review Board (IRB)-approved protocols, with written informed consent obtained from all participants. The HABS-HD study follows a wave-based longitudinal design. At baseline (Visit 1), all participants undergo a comprehensive assessment. Follow-up evaluations are conducted in scheduled waves, with subsequent visits (Visit 2 and Visit 3) typically occurring at approximately 24–30 months and 48–60 months from baseline, respectively. Due to coronavirus disease 2019 (COVID-19), completion of ongoing Wave 1 and Wave 2 assessments was longer. ²⁷ For this specific analysis, we included HABS-HD participants who had completed baseline plasma AD biomarker measurements and had at least three separate follow-up visits with valid blood pressure measurements to ensure a reliable calculation of visit-to-visit BPV.

Cognitive diagnosis at baseline was determined through a consensus review process overseen by dementia experts associated with the study. A diagnostic algorithm was applied, following established diagnostic criteria for dementia. This algorithm incorporates three core elements: self- or other report of cognitive change, neuropsychological test performance, and the Clinical Dementia Rating (CDR) Sum of Boxes score. All cases meeting the diagnostic algorithm criteria for MCI and dementia, along with 10% of those classified as cognitively unimpaired (CU), underwent consensus review. This review process included individual chart review by dementia experts, a medical professional, and clinic staff associated with HABS-HD. Based on this comprehensive process, participants were classified as CU, MCI, or Dementia. CU was defined by a CDR Sum of Boxes score of 0, MCI by scores ranging from 0.5 to 2.0, and Dementia by scores greater than or equal to 2.5. ²³ For the purposes of this analysis and to enhance statistical power, we combined participants with MCI and Dementia into a single Cognitively Impaired (CI) group.

Plasma AD biomarkers

Fasting venous blood was collected from participants at baseline, in accordance with international guidelines for AD blood-based biomarker research. Plasma was separated, aliquoted, and stored at −80°C within a strict 2-h pre-analytical processing window to maintain biomarker stability. Electrochemiluminescence (ECL) assays from a validated AD blood screen were used for proteomic assessments. The assays were performed on the QuickPlex multiplex platform (Meso Scale Discovery) in line with previously established methods, and the preparatory steps for these assays were automated using the Hamilton Robotics StarPlus system. Concentrations of Aβ_42/40 ratio, total tau, p-tau181, and NfL were quantified using the highly sensitive single molecule array (Simoa) platform on a HD-X Automated Analyzer (Quanterix, Billerica, MA, USA). All assays were conducted in a single batch at the Institute for Translational Research Biomarker Core by technicians blinded to participant information. Rigorous quality control was maintained, with coefficients of variation for all assays below 5%. ²⁷ Due to skewed distributions, all biomarker values were natural log-transformed for statistical analyses. ²² To assess the potential influence of extreme values on our findings, we performed a sensitivity analysis excluding outliers. Outliers were defined as individuals with any plasma biomarker value (Aβ_42/40 ratio, total tau, p-tau181, NfL) exceeding ±5 standard deviations from the mean. We used data exclusively from the release containing the largest number of plasma biomarker measurements rather than merging or harmonizing data across different releases.

Blood pressure variability

Blood pressure was measured at each study visit using a sphygmomanometer following standard auscultatory protocol. Two measurements of systolic and diastolic blood pressure were taken at each visit, and the mean value was calculated for analysis.^28,29 These mean values from three scheduled follow-up visits were used to calculate visit-to-visit BPV. The primary metric for BPV was the coefficient of variation (CV), calculated as CV = (standard deviation / mean) × 100%. We also calculated BPV using the standard deviation (SD) and average real variability (ARV) to ensure robustness of findings. ARV was calculated as the average of the absolute differences between consecutive blood pressure measurements. All three metrics showed high intercorrelation (r range 0.89 to 0.98). Given the established stronger association of systolic BPV with cardiovascular and cerebrovascular outcomes, SBPV was designated as the primary analysis focus, with DBPV analyzed as a secondary outcome. ³⁰

Covariates

Covariates were selected based on their established relevance to blood pressure regulation or AD pathology. These included: (1) Demographic factors: age (in years), sex, self-reported race/ethnicity, and education (in years), which were obtained from self-reported data during the baseline assessment; (2) Vascular risk factors and comorbidities: body mass index (BMI, kg/m²), and histories of diabetes, dyslipidemia, hypertension, stroke, ever use of antihypertensive medication (AHM), smoking, all of which were ascertained through interview-administered questionnaires during the baseline assessment; (3) AD-related factors: APOE ε4 carrier status (determined by genotyping) and baseline cognitive diagnosis (cognitively normal versus cognitively impaired), both of which were acquired during the baseline assessment; (4) Blood pressure parameters: mean systolic and diastolic blood pressure. This parameter was calculated as the average of all blood pressure readings across follow-up visits, and was adjusted for in the models to isolate the effect of BPV from the absolute blood pressure level. (5) Estimated glomerular filtration rate (eGFR), which may affect plasma biomarker levels.

Statistical analyses

All analyses were performed using R software (version 4.1.2). Descriptive statistics presented continuous variables as mean (standard deviation) and categorical variables as number (percentage). We employed multiple linear regression to examine the associations between log-transformed plasma biomarkers (independent variables) and BPV metrics (dependent variables). Analyses were conducted using four predefined models: crude model (unadjusted); Model 1 (adjusted for age and sex); Model 2 (additionally adjusted for diagnosis, race/ethnicity, education, BMI, APOE ε4 status, mean SBP/DBP, ever use of AHM, smoking history, eGFR, and comorbidities including diabetes, dyslipidemia, hypertension, and stroke); and Model 3 (further adjusted for all plasma AD biomarkers). To assess multicollinearity among predictors, we computed Spearman correlation matrices for continuous variables and calculated variance inflation factors (VIF) for all predictors in the fully adjusted models (Model 3). A VIF value greater than 5 was considered indicative of concerning collinearity. Additionally, to examine whether extreme blood pressure values might drive the results, we performed a sensitivity analysis excluding participants with systolic blood pressure <90 mmHg or >180 mmHg at any follow-up visit and re-ran the fully adjusted Model 3. To explore potential non-linear relationships, we conducted a sensitivity analysis by introducing a quadratic term for plasma biomarkers positively associated with BPV into the fully adjusted model (Model 3). Stratified analyses were performed for the biomarker that showed consistent significance in the main analyses by key subgroups: age (<65/≥65 years), sex, race/ethnicity, APOE ε4 status, diagnostic group, and antihypertensive medication use. To assess potential effect modification, we also performed interaction testing in addition to stratified analyses. These interaction analyses were conducted to complement the stratified analyses and to provide a more robust assessment of whether the association varied across subgroups. Sensitivity analyses assessed the robustness of our findings by calculating BPV using SD and ARV as alternative metrics. A two-sided p-value < 0.05 was considered statistically significant for all analyses. We applied the Bonferroni correction to account for the four main plasma biomarkers examined (Aβ_42/40 ratio, total tau, p-tau181, and NfL). Accordingly, the statistical significance threshold for Model 3 was set at p < 0.05/4 = 0.0125.

Results

Participant characteristics

The initial HABS-HD cohort comprised 4134 participants at first visit. For the present analysis, we included 507 participants who had complete baseline plasma biomarker data and at least three follow-up blood pressure measurements. There were no missing data for any of the variables included in the primary analysis for this selected sample. The baseline characteristics of the study population are summarized in Table 1. The mean (SD) age was 66 (8.18) years, 35.1% were male, and the sample was ethnically diverse (55.2% non-Hispanic White and 44.8% Hispanic). The mean (SD) years of education was 13.36 (4.39). The cohort exhibited a high burden of vascular risk factors, with a mean (SD) BMI of 30.34 (6.24) kg/m², and high prevalence of dyslipidemia (51.3%), hypertension (49.9%), and diabetes (23.1%). Among the participants, 34.3% had a history of smoking, and 38.9% had ever used AHM. The majority of participants were cognitively normal (86.2%) at baseline, and 23.3% were APOE ε4 carriers. During the follow-up period, the mean (SD) systolic and diastolic blood pressures were 133.39 (18.88) mm Hg and 80.48 (8.17) mm Hg, respectively. The mean levels of visit-to-visit systolic and diastolic BPV, quantified by the coefficient of variation, were 7.76 (5.15) and 7.03 (4.27), respectively.

Table 1.

Baseline characteristics of the study population.

Variable	N = 507
Age, mean (SD), years	66 (8.18)
Male, No. (%)	178 (35.1%)
Diagnosis, No. (%)
Cognitively normal	437 (86.2%)
Cognitively impaired	70 (13.8%)
Race/Ethnicity, No. (%)
non-Hispanic White	280 (55.2%)
Hispanic	227 (44.8%)
Education, mean (SD), years	13.36 (4.39)
BMI, mean (SD), kg/m2	30.34 (6.24)
APOE ε4 carrier, No. (%)	118 (23.3%)
Ever Smoking, No. (%)	174 (34.3%)
Ever use of AHM, No. (%)	197 (38.9%)
eGFR	80.86 (15.68)
Diabetes, No. (%)	117 (23.1%)
Dyslipidemia, No. (%)	260 (51.3%)
Hypertension, No. (%)	253 (49.9%)
Stroke, No. (%)	14 (2.8%)
Baseline SBP, mmHg, mean (SD)	135.09 (18.88)
Baseline DBP, mmHg, mean (SD)	81.48 (9.87)
Mean SBP, mean (SD), mm Hg	133.39 (14.94)
Mean DBP, mean (SD), mm Hg	80.48 (8.17)
SBPV, CV, mean (SD)	7.76 (5.15)
DBPV, CV, mean (SD)	7.03 (4.27)
Aβ40, median (IQR)	2.31 (2.24, 2.37)
Aβ42, median (IQR)	1.02 (0.95, 1.08)
Aβ42/40 ratio, median (IQR)	0.44 (0.42, 0.46)
Total tau, median (IQR)	0.31 (0.21, 0.41)
p-Tau181, median (IQR)	0.23 (0.14, 0.37)
NfL, median (IQR)	1.13 (0.98, 1.27)

Aβ: amyloid-β; SD: standard deviation; IQR: interquartile range; BMI: body mass index; AHM: antihypertensive medication; eGFR: estimated glomerular filtration rate; SBP: systolic blood pressure; DBP: diastolic blood pressure; SBPV: systolic blood pressure variability; DBPV: diastolic blood pressure variability; CV: coefficient of variation; NfL: neurofilament light chain; t-tau: total tau; p-tau181: phosphorylated tau 181.

Associations of plasma biomarkers with blood pressure variability

The associations between plasma AD biomarkers and visit-to-visit BPV are presented in Table 2. Among the four biomarkers investigated, only total tau demonstrated a consistent and independent association with systolic BPV across progressively adjusted models. The Spearman correlation matrix revealed no highly correlated predictor pairs. VIFs for all variables in the final models were below 5, indicating no significant multicollinearity (see Supplemental Figures 1 and 2).

Table 2.

Association between plasma biomarkers and visit-to-visit blood pressure variability.

Biomarker	SBPV		DBPV
	β (95% CI)	p	β (95% CI)	p
Aβ42/40 ratio
Crude	8.10 (−3.11 to 19.32)	0.157	4.86 (−4.44 to 14.16)	0.306
Model 1 a	7.27 (−3.91 to 18.46)	0.203	3.14 (−6.13 to 12.40)	0.507
Model 2 b	7.03 (−4.39 to 18.45)	0.228	3.82 (−5.75 to 13.40)	0.434
Model 3 c	5.19 (−6.27 to 16.65)	0.375	3.64 (−6.15 to 13.43)	0.467
t-tau
Crude	4.93 (2.24 to 7.61)	<0.001	1.12 (−1.13 to 3.37)	0.330
Model 1 a	4.54 (1.78 to 7.30)	0.001	1.37 (−0.93 to 3.67)	0.244
Model 2 b	3.72 (0.82 to 6.61)	0.012	0.64 (−1.80 to 3.09)	0.605
Model 3 c	3.94 (1.00 to 6.87)	0.009	0.55 (−1.96 to 3.05)	0.668
p-tau181
Crude	0.62 (−1.77 to 3.01)	0.612	2.01 (0.04 to 3.98)	0.047
Model 1 a	−0.38 (−2.98 to 2.22)	0.774	0.70 (−1.46 to 2.85)	0.526
Model 2 b	−1.36 (−4.03 to 1.31)	0.320	0.33 (−1.91 to 2.57)	0.774
Model 3 c	−2.35 (−5.15 to 0.45)	0.101	0.12 (−2.28 to 2.51)	0.925
NfL
Crude	2.90 (0.94 to 4.86)	0.004	2.40 (0.78 to 4.03)	0.004
Model 1 a	2.12 (−0.22 to 4.46)	0.076	1.12 (−0.82 to 3.06)	0.256
Model 2 b	1.37 (−1.12 to 3.87)	0.280	0.56 (−1.53 to 2.65)	0.598
Model 3 c	1.58 (−1.02 to 4.18)	0.233	0.58 (−1.63 to 2.80)	0.607

SBPV: systolic blood pressure variability; DBPV: diastolic blood pressure variability; NfL: neurofilament light chain; Aβ: amyloid-β; t-tau: total tau; p-tau181: phosphorylated tau 181.

^a Model 1 adjusted for age and gender.

^b Model 2 additionally adjusted for diagnosis, ethnicity, education, body mass index, APOE ε4 status, mean SBP or DBP, smoking status, use of antihypertensive medication, estimated glomerular filtration rate, and comorbidities including diabetes, dyslipidemia, hypertension, and stroke.

^c Model 3 additionally adjusted for plasma biomarkers including Aβ₄₂, Aβ₄₀, Aβ_42/40 ratio, total Tau, pTau181, and NfL.

In the unadjusted analysis, significant positive associations were observed between SBPV and total tau (β = 4.93, 95% CI: 2.24 to 7.61, p < 0.001), and NfL (β = 2.90, 95% CI: 0.94 to 4.86, p = 0.004). For DBPV, significant crude associations were found with NfL (β = 2.40, 95% CI: 0.78 to 4.03, p = 0.004). However, after adjustment for age and sex (Model 1), the associations for NfL became non-significant (all p > 0.05). In the fully adjusted clinical model (Model 2), total tau remained significantly associated with higher SBPV (β = 3.72, 95% CI: 0.82 to 6.61, p = 0.012).

Notably, the association between SBPV and total tau persisted even after further simultaneous adjustment for all other plasma biomarkers (Aβ_42/40 ratio, p-tau181, NfL) in Model 3 for SBPV (β = 3.94, 95% CI: 1.00 to 6.87, p = 0.009). In the fully adjusted Model 3, after applying Bonferroni correction (significance threshold set at p < 0.0125), the association between total tau and systolic BPV remained statistically significant. This indicates that the link between plasma total tau and BPV is independent of co-existing amyloid pathology and neuroaxonal injury.

Subgroup and sensitivity analyses

To explore the heterogeneity of the robust association between total tau and BPV, we performed stratified analyses. Comparisons between each group are shown in Supplemental Tables 2–8. The results revealed that this association was not uniform across all subgroups, as illustrated in Figure 1.

Figure 1.

Subgroup analysis of the association between plasma total tau and systolic blood pressure variability (SBPV). Forest plot displaying the adjusted β coefficients and 95% confidence intervals (CIs) for the association between log-transformed total tau and SBPV (coefficient of variation) across prespecified subgroups. Estimates were derived from linear regression models fully adjusted for demographics, estimated glomerular filtration rate, cardiovascular risk factors, mean blood pressure, APOE ε4 status, ever use of AHM, and other plasma biomarkers (Aβ_42/40 ratio, p-tau181, and NfL; Model 3). AHM: antihypertensive medication.

For SBPV, the positive association with total tau was primarily evident in cognitively normal individuals (β = 4.42, 95% CI: 1.14–7.70, p = 0.009), APOE ε4 non-carriers (β = 4.91, 95% CI: 1.42–7.70, p = 0.009), males (β = 5.80, 95% CI: 1.66–9.95, p = 0.007), non-Hispanic White participants (β = 4.00, 95% CI: 0.12–7.88, p = 0.044), and individuals not using antihypertensive medication (β = 4.29, 95% CI: 0.71–7.87, p = 0.020). Interaction testing was performed to evaluate whether the association between total tau and systolic BPV differed across subgroups. Interaction testing revealed no significant effect modification by any of the examined factors (all p for interaction > 0.05; Supplemental Table 1).

In the outlier sensitivity analysis, all plasma biomarker values fell within ±5 standard deviations of the mean; consequently, no participants were excluded. In a sensitivity analysis excluding participants with extreme blood pressure values (systolic BP <90 mmHg or >180 mmHg at any follow-up visit, N = 22), the association between total tau and SBPV remained significant in the fully adjusted Model 3 (β = 4.17, 95% CI: 1.54–6.81, p = 0.003), suggesting that our primary findings were not driven by extreme cases. In sensitivity analyses, the core finding of an association between total tau and BPV proved robust when using alternative metrics for BPV calculation. When BPV was quantified using the standard deviation in the fully adjusted Model 3, higher total tau levels remained significantly associated with greater SBPV-SD (β = 5.33, 95% CI: 1.28 to 9.39, p = 0.010). Similarly, using the ARV metric, significant positive associations were maintained for both SBPV-ARV (β = 7.86, 95% CI: 2.90 to 12.81, p = 0.002). The consistency of these associations across three distinct BPV calculation methods underscores the robustness of the relationship between plasma total tau and visit-to-visit BPV, independent of the specific metric employed. In sensitivity analyses examining non-linearity, the quadratic term of total tau was not statistically significant (β=-7.35, 95% CI: −16.35 to 1.65, p = 0.110) in the fully adjusted model. Furthermore, inclusion of the quadratic term did not meaningfully improve the model fit (ΔR² < 0.01). These findings indicate that a linear model adequately captures the relationship between total tau and BPV.

Discussion

In this prospective analysis of the multi-ethnic HABS-HD cohort, we investigated the association between baseline plasma Alzheimer's disease biomarkers and subsequent visit-to-visit BPV. Our findings revealed a distinctive pattern: among all biomarkers analyzed, total tau demonstrated the most robust and consistent association with systolic BPV that persisted across progressively adjusted models. This association remained significant even after comprehensive adjustment for demographic factors, clinical diagnoses, ever use of AHM, eGFR, and cardiovascular risk factors. The stratified analyses provided additional nuance, revealing that this relationship was primarily evident in males, non-Hispanic White, cognitively unimpaired individuals and APOE ε4 non-carriers. Sensitivity analyses further strengthened our confidence in these findings by demonstrating consistent results across different BPV calculation methods. Collectively, these results advance the emerging paradigm of AD as a systemic disorder by suggesting that underlying AD-related neuronal injury, reflected by plasma total tau, may contribute to impaired systemic blood pressure regulation, highlighting a potential pathway through which AD pathology could exacerbate vascular comorbidities.

The specificity of the total tau-BPV association across our analytical models is noteworthy. Even when simultaneously adjusting for other AD biomarkers including Aβ_42/40 ratio, p-tau181, and NfL in Model 3, the association between total tau and BPV remained statistically significant. This observation indicates that total tau, rather than co-existing amyloid pathology or neuroaxonal injury, is independently associated with blood pressure stability. Plasma tau is not only as a well-established biomarker of cerebral neurodegeneration but also as a potential indicator of dysregulated systemic physiological control. Prior research supports a unique role of tau protein, rather than Aβ protein deposition, in mediating autonomic nervous system impairment. A postmortem study found marked immunoreactivity for tau markers on both axons and somata within distinct regions of the medulla oblongata in AD patients. ¹⁷ This immunoreactivity was particularly pronounced in the dorsal motor nucleus of the vagus, nucleus of the solitary tract, and reticular nuclei, whereas no accompanying amyloid pathology was observed. The robustness of this association after accounting for mean blood pressure levels further underscores that the relationship specifically involves blood pressure fluctuation rather than absolute blood pressure elevation.

The potential biological mechanisms underlying the observed tau-BPV relationship are likely multifaceted and may involve the strategic anatomical progression of AD pathology. It is noteworthy that AD-related neurofibrillary changes propagate through the brain in a hierarchical sequence, beginning in the medial temporal lobe and progressively invading frontal, temporal, and parietal association cortices.^31,32 Critically, the insular cortex, a key hub within the cortical network for central autonomic control, may be affected in this process.^16,33,34 One study found that tau uptake was associated with cortical thinning in multiple brain regions, particularly in the left insula. ³³ The early involvement of the insular cortex by neurofibrillary tangles could directly lead to impaired central regulation of cardiac function. ¹¹ Importantly, while brainstem nuclei such as the dorsal motor nucleus of the vagus and the nucleus of the solitary tract are also involved in autonomic regulation, they may be affected at much later Braak stages, which correspond to advanced dementia. In contrast, the insular cortex is already affected at Braak stage III, a preclinical stage. ¹⁶ This finding supports the positive association between tau pathology and BPV observed in our community-based population, whose pathological changes likely reflect a preclinical stage. This provides a compelling neuroanatomical basis for our findings, suggesting that even at the subclinical stage, blood pressure instability may partially reflect and represent an epiphenomenon and marker of advancing AD pathology. This perspective, however, also introduces the possibility of a more complex, bidirectional relationship between BPV and AD. On one hand, our study may support the trajectory wherein tau pathology leads to BP instability via autonomic dysfunction. On the other hand, we cannot rule out that higher BPV might also act as an upstream risk factor for AD pathogenesis and progression. The pathological process itself could, in turn, amplify and mediate this relationship by exacerbating BPV and its adverse effects on the brain, potentially through mechanisms such as cerebral hypoperfusion and blood-brain barrier disruption.^35–37 Therefore, the association we observed may represent a vicious cycle wherein initial AD pathology disrupts central cardiovascular regulation, and the resulting BPV further accelerates neurodegenerative processes.

The pronounced association in cognitively unimpaired individuals at baseline suggests that the interplay between tau pathology and vascular dysregulation may represent an early process that occurs before overt cognitive symptoms emerge. This temporal relationship raises the hypothesis that blood pressure instability could serve as an early indicator of underlying neurodegeneration. If unexplained elevated BPV is observed, the potential involvement of an underlying neurodegenerative process should be considered, thereby prompting earlier neurocognitive assessment. In addition, we further observed that the association between total tau protein and BPV is particularly significant in the population of APOE ε4 non-carriers. Interestingly, the results of an analysis based on the Multi-Ethnic Study of Atherosclerosis (MESA) cohort support our findings—this study showed a stronger association between BPV and global cognitive function in APOE ε4 non-carriers. ³⁸ The effect modification by APOE ε4 status further underscores the pathological heterogeneity of AD. The robust association observed in ε4 non-carriers indicates that in the absence of a strong genetically-driven Aβ pathology, the progression of AD pathology in APOE ε4 non-carriers relies more heavily on “environmental” or “systemic” risk factors, particularly vascular health. ³⁹ In contrast, among ε4 carriers, the powerful genetic signal dominates, partially masking the independent contribution of vascular factors. It is important to note that, although numerical differences were observed across certain subgroups in the stratified analyses, there is no statistical evidence of effect modification by these factors. Subgroup analyses should be interpreted as exploratory and hypothesis-generating only, and the observed differences across subgroups should be viewed with caution.

Our results suggest that elevated plasma total tau, a marker of neurodegeneration, could be considered a novel risk indicator for blood pressure instability, independent of traditional cardiovascular risk factors. This implies that in older adults with labile blood pressure, especially those without a clear cardiovascular cause, a neurodegenerative contribution might be considered. While our findings do not immediately support the integration of AD biomarker screening into cardiovascular practice, they illuminate several avenues for future research. First, they suggest that plasma total tau could be investigated as a novel risk marker for blood pressure dysregulation, potentially improving risk stratification beyond traditional vascular factors. In older adults with unexplained high BPV, a neurodegenerative contribution might be considered. Second, the pronounced association in cognitively normal individuals raises the hypothesis that increased BPV could serve as an early, detectable sign of neurodegeneration. This suggests that longitudinal studies are warranted to explore whether managing BPV in individuals with elevated tau could modify the trajectory of cognitive decline. Finally, by revealing a link between tau pathology and a systemic physiological outcome, our study broadens the potential impact of future anti-tau therapies. It prompts the question of whether such treatments, if proven effective, might confer not only cognitive but also cardiovascular benefits by stabilizing autonomic function. These potential implications, however, remain speculative and require confirmation in interventional studies.

Our study has several strengths. The prospective design within a well-characterized, multi-ethnic cohort minimizes concerns about reverse causation and enhances generalizability. The comprehensive assessment of covariates allowed for rigorous adjustment for potential confounders, while the simultaneous evaluation of multiple plasma biomarkers enabled direct comparison of their associations with BPV. Furthermore, the consistency of our findings across different BPV metrics strengthens the robustness of our conclusions.

Several limitations must be considered when interpreting our results. First, despite our prospective design, the observational nature of our study precludes definitive causal inference. The relationship between tau pathology and BPV is likely complex and potentially bidirectional. Future research using Mendelian randomization approaches or leveraging natural experiments involving AD-targeted therapies in real-world settings could provide valuable evidence to help establish causal directionality. Second, despite our extensive adjustment for known potential confounders, we cannot completely rule out residual confounding by unmeasured or imperfectly measured factors. The lack of detailed data on antihypertensive medication regimens represents a particular limitation, as medication type, dosage, and adherence patterns can significantly influence BPV. Third, our study relied on single baseline measurements of plasma biomarkers, which may not fully capture the dynamic progression of AD pathology. As both biomarker levels and BPV can fluctuate over time, future studies with repeated serial assessments would help elucidate the temporal relationship between tau accumulation and blood pressure dysregulation. Fourth, while we employed standard and clinically relevant BPV metrics (CV, SD, ARV), we acknowledge that more complex approaches, such as blood pressure area under the curve (AUC) calculated via the trapezoid rule or time in hypertensive range, could provide complementary insights. These methods capture cumulative blood pressure load and may better reflect long-term vascular burden. Fifth, due to the requirement for at least three follow-up visits with complete blood pressure measurements, our analytical sample had limited representation of Non-Hispanic Black participants, and thus our findings may not be fully generalizable to this population. Future studies with longer follow-up and greater diversity are needed to validate these findings across all racial and ethnic groups. Sixth, our plasma biomarkers were not validated against gold-standard PET imaging, which limits the interpretation of their pathological specificity. Finally, despite the prospective design, the observational nature of our study precludes causal conclusions, and we cannot rule out residual confounding from unmeasured factors or shared underlying pathophysiology. Therefore, our findings are hypothesis-generating and warrant future Mendelian randomization or interventional studies to establish causality.

Conclusions

In conclusion, this prospective study establishes plasma total tau as an independent predictor of long-term BPV, particularly during the pre-clinical stage. These findings reveal a tau-driven pathway to vascular dysregulation and advance our understanding of AD as a systemic illness. Future research should focus on elucidating the underlying mechanisms and exploring whether targeting tau pathology may improve cardiovascular outcomes in at-risk populations.