Interactions between blood pressure variability, cerebral autoregulation, and covert cerebral small vessel disease in relation to cognitive performance in community‐dwelling older adults

Abstract

INTRODUCTION

Both increased blood pressure variability (BPV) and impaired cerebral autoregulation (CA) have been linked to cerebral small vessel disease (CSVD) and cognitive performance, but whether they independently or interactively influence cognition remains unclear.

METHODS

We analyzed 300 community‐dwelling participants. Dynamic CA, beat‐to‐beat BPV measures, and MRI‐based CSVD markers were quantified. Multivariate linear regression and mediation analyses assessed the associations and interactions between CA, BPV, CSVD, and Montreal Cognitive Assessment (MoCA) scores.

RESULTS

Higher BPV was associated with greater CSVD burden. A significant interaction between CA and BPV indicated that the negative association between BPV and MoCA scores was evident only under impaired CA. Mediation analyses demonstrated that BPV influenced MoCA score directly rather than through CSVD.

DISCUSSION

Increased BPV and impaired CA interacted and are associated with poorer cognitive performance independently of detectable CSVD, suggesting a potential link between hemodynamic instability and cognitive vulnerability in community‐dwelling older adults.

Highlights

• Subclinical alterations in beat‐to‐beat blood pressure variability (BPV) are associated with cerebral small vessel disease (CSVD) burden and cognitive performance in community‐dwelling older adults.

• Cerebral autoregulation (CA) functions as a physiological buffer against beat‐to‐beat BPV, and cognitive vulnerability is observed when this buffering capacity is impaired, highlighting a hemodynamic interaction overlooked in prior CSVD research.

• Beat‐to‐beat BPV is directly associated with cognitive performance, with no mediation through CSVD.

1. BACKGROUND

Cerebral small vessel disease (CSVD) is a major contributor to cognitive impairment. CSVD encompasses white matter hyperintensities (WMHs), lacunes, cerebral microbleeds (CMBs), and perivascular spaces (PVSs). ¹ , ² A higher CSVD burden—particularly WMHs and lacunes—is associated with poorer cognitive performance in individuals with stroke and dementia. ³ , ⁴

CSVD has been attributed to genetic predisposition and traditional vascular risk factors—including hypertension, diabetes, and smoking—and reduced hemodynamic stability. Blood pressure variability (BPV) is a biomarker of hemodynamic stability, with emerging evidence indicating its association with CSVD. ⁵ Increased BPV reflects greater fluctuations in BP, has a detrimental effect on stable end‐organ perfusion, and has been linked to a higher risk of vascular diseases, dementia, and chronic kidney diseases. ⁶ Nevertheless, a physiological mechanism—cerebral autoregulation (CA)—may mitigate the detrimental effects of increased BPV. CA maintains stable and adequate cerebral blood flow (CBF) despite changes in cerebral perfusion pressure, ⁷ primarily governed by vascular myogenic reactivity and cerebral metabolic demand. ⁸ Therefore, intact CA may act as a protective mechanism against cerebrovascular damage arising from increased BPV, and these two mechanisms may interact in the pathogenesis of CSVD.

Together, these findings suggest a potential mechanism in which increased BPV and impaired CA may promote the development of CSVD, which in turn contributes to cognitive impairment.

Based on this conceptual model, prior studies have shown that both increased BPV and impaired CA are associated with greater CSVD burden, ⁹ , ¹⁰ , ¹¹ , ¹² cognitive impairment, ¹³ , ¹⁴ and unfavorable outcomes in cerebrovascular diseases. ¹⁵ , ¹⁶ However, most evidence has been derived from patients with stroke or dementia, making it unclear whether similar hemodynamic disturbances are present in normal older adults or whether they contribute to early, subclinical CSVD.

Among cognitively normal older adults, covert CSVD has been associated with reduced cognitive performance, ¹⁷ but whether subclinical BPV and CA alterations are associated with CSVD burden and cognitive performance—and whether CA and BPV affect cognition through the development of CSVD—remains uncertain.

In the present study, we hypothesized that BPV and CA interactively relate to cognitive performance, potentially through their association with CSVD in normal older adults. We investigated the individual roles and potential interactions of CA, BPV, and covert CSVD in relation to cognitive performance in community‐dwelling older adults.

2. METHODS

2.1. Study participants

This cross‐sectional observational study utilized data from the Longitudinal Aging Study of Taipei (LAST), an ongoing community‐based study in Taiwan. The LAST, initiated in 2016, focuses on age‐related risk factors and diseases such as sarcopenia, frailty, cognitive impairment, and vascular aging. The study was approved by the Institutional Review Board of National Yang Ming Chiao Tung University.

Inclusion criteria were as follows: (1) aged ≥ 50 years, (2) community‐dwelling individuals residing in Taipei city or New Taipei city, and (3) willingness to adhere to the study protocol and follow‐up. Exclusion criteria were: (1) inability to perform independent daily activities; (2) speech or language impairment hindering communication; (3) medical conditions with an expected life expectancy of less than 6 months; (4) history of atrial fibrillation, heart failure, or severe valvular heart disease; (5) history of stroke, dementia, or other diseases that may affect cognitive function; and (6) intracranial arterial stenosis > 50% stenosis in any major intracranial artery on time‐of‐flight magnetic resonance angiography (TOF‐MRA).

Written informed consent was obtained from all participants. Demographic information, including age, sex, education level, history of hypertension, diabetes, hyperlipidemia, smoking status, body weight, height, and brachial blood pressure measured at the interview, was collected.

2.2. Neuropsychological assessment

Neuropsychological assessment involved the use of two tests: (1) Chinese version of the Montreal Cognitive Assessment (MoCA) to evaluate general cognitive performance, and (2) Geriatric Depressive Scale (GDS‐15) to evaluate mood status. Trained interviewers, blinded to the results of CA, BPV, and neuroimaging, conducted cognitive function and mood assessments.

2.3. CA and BPV analyses

CA was assessed using the indices of dynamic CA (dCA) derived from a 10‐minute recording of spontaneous beat‐to‐beat fluctuations in blood pressure (BP) and cerebral blood flow velocity (CBFV). BP and CBFV were used as surrogates for cerebral perfusion pressure and CBF, respectively.

Participants were tested in a supine position with the head elevated at 30°. Bilateral CBFV of the extracranial internal carotid artery was measured using a Doppler ultrasonography system (DWL Doppler‐Box X, Compumedics DWL, Singen, Germany) equipped with two 2‐MHz probes fixed on a custom‐made head frame with an insonation depth 40–50 mm. ¹⁸ BP was measured by finger plethysmography (CNAP monitor, CNSystems, Graz, Austria). CBFV and BP signals were simultaneously recorded at a sampling rate of 100 Hz.

Dynamic CA was quantified using transfer function analysis (TFA) algorithm according to the recommendations of the Cerebral Autoregulation Research Network (https://www.car‐net.org/tools). ¹⁹ TFA estimated the phase shift (degrees) and gain (%/%) between BP and CBFV within the very‐low‐frequency (VLF, 0.02–0.07 Hz) and low‐frequency (LF, 0.07–0.20 Hz) ranges. A larger phase shift and lower gain indicate that CBFV changes are smaller and recover faster relative to BP fluctuations, reflecting more effective dCA. Each dCA index was calculated as the average of the left and right hemispheres for subsequent analyses.

The same 10‐minute beat‐to‐beat BP recording obtained for dCA assessment was used for BPV analysis. ²⁰ Three BPV indices that is, standard deviation (SD), coefficient of variation (CV), and average real variability (ARV), were calculated for systolic BP (SBP) and diastolic BP (DBP) respectively using custom Python scripts. The formulas of BPV indices were provided in the Supporting Information. Prior to analysis, all raw BP and CBFV data were inspected by an experienced neurologist to exclude data of poor quality, including motion artifacts and > 10% incidence of ectopic heart beats.

2.4. Brain MRI characteristics quantification

MRI scans were performed on a Siemens Tim Trio 3‐Tesla MRI scanner (Siemens Medical Solutions, Erlangen, Germany) at National Yang Ming Chiao Tung University. High‐resolution T1‐weighted magnetization‐prepared rapid‐acquisition gradient echo (3D T1w‐MPRAGE), T2‐weighted fluid‐attenuated inversion recovery (FLAIR), susceptibility‐weighted image (SWI), and TOF‐MRA were acquired for each participant. Prior to analysis, all images were inspected by an experienced neurologist to exclude images affected by motion artifacts or structural abnormalities, such as tumors, hemorrhages, or non‐lacunar infarction. Detailed brain MRI acquisition parameters are provided in the Supporting Information.

RESEARCH IN CONTEXT

Systematic review: Prior studies have shown that both increased blood pressure variability (BPV) and impaired cerebral autoregulation (CA) are associated with cerebral small vessel disease (CSVD) and cognitive decline. However, few investigations have simultaneously examined beat‐to‐beat BPV, dynamic CA, CSVD, and cognition, and the potential interaction between BPV and CA has been largely overlooked.

Interpretation: In this community‐based cohort, higher beat‐to‐beat BPV was linked to greater CSVD burden and poorer cognitive performance. Crucially, the adverse association between BPV and cognition were observed only when CA was impaired, indicating a hemodynamic interaction. Mediation analysis showed that BPV was directly associated with cognition rather than through CSVD.

Future directions: Future longitudinal and interventional studies should determine whether stabilizing BPV or enhancing CA buffering capacity can mitigate cognitive vulnerability and whether the BPV–CA interaction predicts subsequent CSVD progression or cognitive decline.

MATLAB R2023b (MathWorks, Natick, MA) and the FMRIB Software Library (FSL v6.0.6, https://fsl.fmrib.ox.ac.uk/fsl/docs/) were used for imaging processing and feature quantification. Total intracranial volume (TIV) and total gray matter volume were measured from 3D T1w‐MPRAGE images using the Computational Anatomy Toolbox (CAT12, r2560, https://www.neuro.uni‐jena.de/cat/) implemented in Statistical Parametric Mapping (SPM12, v7771, https://www.fil.ion.ucl.ac.uk/spm/). ²¹ Native 3D T1w‐MPRAGE images were normalized to the standard Montreal Neurological Institute (MNI) template, and individual deformation fields were obtained using FMRIB's Linear Image Registration Tool (FLIRT) and Nonlinear Image Registration Tool (FNIRT).

WMH volumes were analyzed using the Lesion Segmentation Tool (LST v3.0.0, https://www.applied‐statistics.de/lst.html) implemented in SPM12, ²² which coregistered FLAIR images to the corresponding 3D T1w‐MPRAGE images and then normalized them to MNI space to generate WMH probability maps. CMBs and PVSs were analyzed using the SHIVA‐CMB (https://github.com/pboutinaud/SHIVA_CMB) and SHIVA‐PVS (https://github.com/pboutinaud/SHIVA_PVS) tools, ²³ , ²⁴ which detected CMBs and PVSs in SWI and 3D T1w‐MPRAGE images, respectively, and then normalized them to MNI space to generate probability maps. WMHs, PVSs, and CMBs were quantified using custom Python scripts. WMH probability maps were segmented and classified into periventricular and deep types based on the 10‐mm distance from the lateral ventricles. PVS and CMB probability maps were binarized (threshold > 0.5), identified as 3D connected components, and counted within each brain region defined by the atlas in CAT12. CMBs were classified into lobar, deep (including the basal ganglia, thalamus, and brainstem), and total (lobar + deep). PVSs were classified into the basal ganglia, corona radiata, and total (basal ganglia + corona radiata). Lacunes and intracranial arterial stenosis were visually assessed by a neurologist using FLAIR and TOF‐MRA images.

CSVD scoring incorporated the WMH Fazekas score, the presence of lacunes, the presence of CMBs, and the presence of more than 10 PVSs in the basal ganglia. ²⁵ , ²⁶ In the current study, a normalized total WMH volume (total WMH volume / TIV) > 1.15% was used to indicate a WMH Fazekas score of ≥2, ²⁷ instead of visual rating, contributing one point to the total CSVD score.

2.5. Statistical analyses

Continuous variables are presented as median (interquartile range, IQR). MoCA scores (range 0–30) were analyzed as a continuous outcome using linear regression. Multivariate linear regression analyses explaining MoCA scores—including CSVD scores, dCA and BPV indices, and an interaction term (dCA × BPV)—were conducted while adjusting for potential covariates: age, sex, education level, hypertension, diabetes, hyperlipidemia, smoking, body mass index (BMI), normalized gray matter volume (total gray matter volume / TIV), and GDS‐15 score.

Mediation analysis was performed to evaluate whether the association between an exposure (BPV) and an outcome (MoCA score) was transmitted through a potential mediator (CSVD score). Separate models were constructed for each exposure–mediator–outcome combination, and all models were adjusted for the same covariates used in the multiple regression analyses. The indirect effect (average causal mediation effect, ACME) and direct effect (average direct effect, ADE) were estimated using nonparametric bootstrapping with 5000 resamples, and 95% confidence intervals were obtained. A two‐tailed P value < 0.05 was considered statistically significant. Statistical analyses were performed using MedCalc Statistical Software version 23.0.8 (MedCalc Software Ltd, Ostend, Belgium) and R version 4.5.1 (R Foundation for Statistical Computing, Vienna, Austria).

3. RESULTS

A total of 331 participants were screened for eligibility between September 2020 and March 2025. Thirty‐one were excluded for the following reasons: incomplete cognitive testing (n = 2), incomplete acquisition or motion artifacts on MRI (n = 3), > 50% stenosis of any intracranial artery on MRA (n = 7), brain tumor or non‐lacunar infarction on MRI (n = 2), > 10% incidence of ectopic heartbeats detected during the test (n = 12), and poor CBFV or BP signal quality due to motion artifacts (n = 5). Consequently, 300 participants were included in the final analysis. The characteristics of the participants are summarized in Table 1. Overall, 90% of the participants had at least a senior high school level of education (> 12 years). Most participants (73.3%) had a low CSVD burden (CSVD score of 0 or 1), and CSVD burden was mainly attributable to the presence of CMBs or PVSs. Most participants (76.5%) had a MoCA score ≥26, and only five participants (1.7%) had a GDS‐15 score > 4.

TABLE 1.

Characteristics of the participants (n = 300).

Demographic variables
Age, years	71 (69–74)
Sex, female/male	201/99
Education level, years	14 (12–16)
Hypertension, %	35.7
SBP, mmHg	126 (114–136)
DBP, mmHg	73 (67‐80)
Diabetes, %	16.3
Hyperlipidemia, %	39.0
Smoking (no/quit/current), %	84.7/12.7/2.7
Body mass index, kg/m2	23.6 (21.8–25.2)
Total intracranial volume, mL	1385 (1301–1473)
Gray matter volume, mL	583 (557–612)
normalized, %	42.2 (40.5–43.8)
Dynamic cerebral autoregulation
Phase shift‐VLF, degree	52 (36–67)
Phase shift‐LF, degree	24 (15–38)
Gain‐VLF, %/%	1.32 (1.05–1.65)
Gain‐LF, %/%	1.33 (1.07–1.66)
Blood pressure variability
SBP‐SD, mmHg	3.57 (2.65–4.80)
SBP‐CV, %	3.04 (2.31–4.22)
SBP‐ARV, mmHg	1.52 (1.17–1.96)
DBP‐SD, mmHg	2.65 (20.8–3.36)
DBP‐CV, %	3.73 (2.94–5.14)
DBP‐ARV, mmHg	1.15 (0.92–1.56)
CSVD score
Total score	1 (0–2)
0, %	29.0
1, %	44.3
2, %	25.0
3, %	1.7
4, %	0.0
WMH+, %	1.7
CMB+, %	35.3
PVS+, %	59.0
Lacune+, %	3.3
CSVD characteristics
WMH‐total volume, mL	2.4 (1.3–4.6)
normalized, %	0.17 (0.09–0.32)
WMH‐periventricular volume, mL	2.2 (1.2–4.2)
normalized, %	0.16 (0.08–0.29)
WMH‐deep volume, mL	0.1 (0.0–0.2)
normalized, %	0.01 (0.00–0.02)
CMB‐total, counts	0 (0–1)
CMB‐lobar, counts	0 (0–1)
CMB‐deep, counts	0 (0–0)
PVS‐total, counts	15 (10–22)
PVS‐basal ganglia, counts	12 (9–16)
PVS‐corona radiata, counts	2 (0–6)
Lacune, counts	0 (0–0)
Neuropsychological test
MoCA score	28 (26–29)
<26, %	23.5
<24, %	8.0
GDS‐15 score	1 (0–1)
>4, %	1.7

Abbreviations: ARV, average real variability; CMB, cerebral microbleed; CSVD, cerebral small vessel disease; CV, coefficient of variation; DBP, diastolic blood pressure; GDS: Geriatric Depression Scale; LDL‐C, low density lipoprotein cholesterol; LF, low frequency; MoCA, Montreal Cognitive Assessment; PVS, perivascular space; SBP, systolic blood pressure; SD, standard deviation; VLF, very low frequency; WMH, white matter hyperintensity.

The results of the linear regression analysis examining the association between hemodynamic parameters and CSVD features are summarized in Table 2. Both SBP‐ARV and DBP‐ARV were positively associated with CSVD score, and SBP‐ARV was positively associated with normalized total WMH volume and total PVSs. The other BPV indices and all dCA indices were not associated with CSVD features including CSVD score, normalized total WMH volume, total CMBs, total PVSs, and total lacunes. SBP‐ARV was consistently associated with most CSVD features, and was selected as the representative BPV index for subsequent analyses.

TABLE 2.

Linear regression analysis between hemodynamic parameters and CSVD features (n = 300)

Parameter	CSVD score
	β	95% CI	p‐Value
Dynamic cerebral autoregulation
Phase shift‐VLF, z‐score	0.034	−0.055 to 0.123	0.453
Phase shift‐LF, z‐score	0.042	−0.208 to 0.292	0.741
Gain‐VLF, z‐score	−0.106	−0.357 to ‐0.834	0.405
Gain‐LF, z‐score	0.022	−0.232 to 0.276	0.865
Blood pressure variability
SBP‐SD, mmHg	0.040	−0.008 to 0.087	0.102
SBP‐CV, %	0.022	−0.029 to 0.072	0.398
SBP‐ARV, mmHg	0.155	0.058 to 0.0252	0.002 *
DBP‐SD, mmHg	0.050	−0.020 to 0.119	0.162
DBP‐CV, %	0.026	−0.014 to 0.066	0.204
DBP‐ARV, mmHg	0.174	0.044 to 0.030	0.009 *

	WMH‐total volume normalized, %			CMB‐total, counts
	β	95% CI	p‐Value	β	95% CI	p‐Value
Dynamic cerebral autoregulation
Phase shift‐VLF, z‐score	0.001	−0.032 to 0.034	0.968	0.069	−0.127 to 0.264	0.490
Phase shift‐LF, z‐score	−0.003	−0.036 to 0.029	0.840	0.075	−0.122 to 0.274	0.455
Gain‐VLF, z‐score	0.029	−0.004 to 0.062	0.084	−0.050	−0.248 to 0.147	0.617
Gain‐LF, z‐score	−0.015	−0.048 to 0.019	0.389	0.106	−0.096 to 0.307	0.304
Blood pressure variability
SBP‐SD, mmHg	−0.002	−0.020 to 0.016	0.836	0.080	−0.024 to 0.185	0.132
SBP‐CV, %	−0.006	−0.025 to 0.013	0.530	0.042	−0.070 to 0.153	0.462
SBP‐ARV, mmHg	0.056	0.019 to 0.092	0.003 *	0.135	−0.082 to 0.352	0.222
DBP‐SD, mmHg	−0.001	−0.027 to 0.025	0.926	0.019	−0.135 to 0.173	0.805
DBP‐CV, %	0.001	−0.014 to 0.016	0.880	0.002	−0.086 to 0.0.91	0.957
DBP‐ARV, mmHg	0.020	−0.029 to 0.068	0.433	0.043	−0.248 to 0.334	0.770

	PVS‐total, counts			Lacune, counts
	β	95% CI	p‐Value	β	95% CI	p‐Value
Dynamic cerebral autoregulation
Phase shift‐VLF, z‐score	0.616	−0.952 to 2.185	0.440	−0.004	−0.034 to 0.026	0.805
Phase shift‐LF, z‐score	0.470	−1.075 to 2.015	0.550	−0.013	−0.043 to 0.017	0.399
Gain‐VLF, z‐score	1.210	−0.367 to 2.788	0.132	0.009	−0.021 to 0.039	0.540
Gain‐LF, z‐score	−0.059	−1.660 to 1.542	0.942	0.002	−0.028 to 0.033	0.876
Blood pressure variability
SBP‐SD, mmHg	0.043	−0.798 to 0.884	0.921	0.002	−0.014 to 0.018	0.081
SBP‐CV, %	−0.369	−1.261 to 0.524	0.417	−0.001	−0.018 to 0.016	0.906
SBP‐ARV, mmHg	3.592	1.898 to 5.286	<0.001 *	0.031	−0.002 to 0.064	0.065
DBP‐SD, mmHg	0.300	−0.934 to 1.534	0.633	0.000	−0.023 to 0.024	0.974
DBP‐CV, %	0.200	−0.508 to 0.908	0.579	−0.001	−0.015 to 0.012	0.863
DBP‐ARV, mmHg	2.276	−0.042 to 4.593	0.054	0.004	−0.041 to 0.048	0.869

Abbreviations: ARV, average real variability; CI, confidence interval; CMB, cerebral microbleed; CSVD, cerebral small vessel disease; CV, coefficient of variation; DBP, diastolic blood pressure; LF, low frequency; PVS, perivascular space; SBP, systolic blood pressure; SD, standard deviation; VLF, very low frequency; WMH, white matter hyperintensity.

^* p < 0.05.

The results of the univariate linear regression analyses of factors associated with MoCA score are summarized in Table 3. In the univariate analyses, age, diabetes, hyperlipidemia, CSVD score, normalized WMH volume (total, periventricular, and deep), PVS counts (total and corona radiata), and SBP‐ARV were significantly and negatively associated with MoCA score, whereas education level was significantly and positively associated. None of the dCA indices were associated with MoCA score. Although none of the dCA indices were associated with CSVD features or MoCA score, we retained phase shift‐VLF as the representative dCA index in subsequent analyses. Figure 1 and Figure 2 display scatter plots with regression lines for CSVD score plotted against phase shift–VLF and SBP‐ARV, and MoCA score plotted against CSVD score, phase shift‐VLF, and SBP‐ARV, respectively.

TABLE 3.

Multiple linear regression analysis of factors associated with MoCA score (n = 300)

Parameter	Unadjusted β	95% CI	p‐Value	Adjusted β	95% CI	p‐Value
Demographic variables
Age, years	−0.073	−0.122 to ‐0.023	0.004 *	−0.056	−0.109 to ‐0.004	0.037 *
Sex, male	−0.414	−0.939 to 0.112	0.122	−1.065	−1.728 to ‐0.401	0.002 *
Education level, years	0.228	0.153 to 0.302	<0.001 *	0.201	0.126 to 0.276	<0.001 *
Hypertension, yes/no	−0.497	−1.011 to 0.018	0.059	0.038	−0.491 to 0.567	0.887
Diabetes, yes/no	−1.186	−1.843 to ‐0.529	<0.001 *	−1.018	−1.694 to ‐0.343	0.003 *
Hyperlipidemia, yes/no	−0.607	−1.111 to ‐0.104	0.018 *	−0.146	−0.667 to 0.375	0.581
Smoking, no/quit/current	0.171	−0.382 to 0.724	0.544	0.425	−0.184 to 1.034	0.171
Body mass index, kg/m2	−0.064	−0.144 to 0.016	0.115	0.004	−0.073 to 0.081	0.923
Gray matter volume normalized, %	0.033	−0.071 to 0.138	0.533	−0.116	−0.237 to 0.005	0.061
GDS‐15 score	−0.212	−0.433 to 0.010	0.061	−0.124	−0.334 to 0.091	0.258
CSVD score	−0.338	−0.655 to ‐0.022	0.036 *	−0.134	−0.434 to 0.166	0.382
CSVD characteristics
WMH‐total volume normalized, %	−3.475	−5.372 to ‐1.578	<0.001 *
WMH‐periventricular volume normalized, %	−1.453	−2.347 to ‐0.559	0.002 *
WMH‐deep volume normalized, %	−8.517	−15.510 to ‐1.523	0.017 *
CMB‐total, counts	−0.126	−0.270 to 0.018	0.086
CMB‐lobar, counts	−0.098	−0.256 to 0.060	0.224
CMB‐deep, counts	−0.432	−0.879 to 0.016	0.059
PVS‐total, counts	−0.023	−0.041 to ‐0.006	0.011 *
PVS‐basal ganglia, counts	−0.031	−0.065 to 0.002	0.068
PVS‐corona radiata, counts	−0.034	−0.061 to ‐0.006	0.016 *
Lacune, counts	0.547	−0.401 to 1.496	0.257
Dynamic cerebral autoregulation
Phase shift‐VLF, z‐score	0.053	−0.195 to 0.302	0.673	0.115	−0.116 to 0.346	0.329
Phase shift‐LF, z‐score	0.042	−0.208 to 0.292	0.741
Gain‐VLF, z‐score	−0.106	−0.357 to ‐0.834	0.405
Gain‐LF, z‐score	0.022	−0.232 to 0.276	0.865
Blood pressure variability
SBP‐SD, mmHg	−0.081	−0.214 to 0.051	0.228
SBP‐CV, %	−0.088	−0.229 to 0.053	0.220
SBP‐ARV, mmHg	−0.554	−0.823 to ‐0.286	<0.001 *	−0.284	−0.548 to ‐0.019	0.036 *
DBP‐SD, mmHg	−0.121	−0.315 to 0.074	0.225
DBP‐CV, %	−0.11	−0.222 to 0.001	0.052
DBP‐ARV, mmHg	−0.339	−0.706 to 0.028	0.071
Interaction between dynamic cerebral autoregulation and blood pressure variability
Phases shift‐VLF × SBP‐ARV (z‐score)				0.238	0.044 to 0.433	0.016 *

Abbreviations: ARV, average real variability; CI, confidence interval; CMB, cerebral microbleed; CSVD, cerebral small vessel disease; CV, coefficient of variation; DBP, diastolic blood pressure; GDS: Geriatric Depression Scale; LF, low frequency; MoCA, Montreal Cognitive Assessment; PVS: perivascular space; SBP, systolic blood pressure; SD, standard deviation; VLF, very low frequency; WMH, white matter hyperintensity.

^* p < 0.05.

FIGURE 1.

Scatter plots with regression lines showing the associations of CSVD burden (CSVD score) with (A) dynamic cerebral autoregulation (phase shift‐VLF) and (B) blood pressure variability (SBP‐ARV). ARV, average real variability; CSVD, cerebral small vessel disease; SBP, systolic blood pressure; VLF, very low frequency.

FIGURE 2.

Scatter plots with regression lines showing the associations of cognitive performance (MoCA score) with (A) CSVD burden (CSVD score), (B) dynamic cerebral autoregulation (phase shift‐VLF), and (C) blood pressure variability (SBP‐ARV). ARV, average real variability; CSVD, cerebral small vessel disease; MoCA, Montreal Cognitive Assessment; SBP, systolic blood pressure; VLF, very low frequency.

The results of the multivariate linear regression analyses of factors associated with MoCA score are summarized in Table 3. Demographic variables, GDS‐15 score, CSVD score, phase shift‐VLF, SBP‐ARV, and the interaction term between phase shift‐VLF and SBP‐ARV were included in the multivariate model. Age, education level, diabetes, and SBP‐ARV remained significantly associated with MoCA score, whereas male sex became significantly negatively associated. CSVD score was no longer associated with MoCA score. In addition, the interaction between phase shift‐VLF and SBP‐ARV was significantly and positively associated with MoCA score, indicating that the association between SBP‐ARV and MoCA score varied across different levels of phase shift‐VLF. We further tested normalized total WMH volume and total PVSs as alternative CSVD features in sensitivity analyses, and the results consistently showed that WMH volume and total PVSs were no longer associated with MoCA score, whereas the interaction between phase shift‐VLF and SBP‐ARV remained significantly and positively associated with MoCA score in multivariate models (Table S1).

To further clarify the interaction between phase shift‐VLF and SBP‐ARV, we produced a Johnson–Neyman plot (Figure 3). The plot showed that the adjusted β coefficient of SBP‐ARV was negative and significantly different from 0 when phase shift‐VLF was below 54.068°, indicating that BPV affected MoCA score when dCA was impaired.

FIGURE 3.

Johnson–Neyman plot: The influence of dynamic cerebral autoregulation (phase shift‐VLF) to the adjusted β coefficient of blood pressure variability (SBP‐ARV). ARV, average real variability; SBP, systolic blood pressure; VLF, very low frequency.

We conducted a mediation analysis to examine whether SBP‐ARV influenced MoCA score through CSVD score (Figure 4). The results indicated that SBP‐ARV was directly associated with MoCA score, with no significant mediation effect through CSVD score. We further tested normalized total WMH volume and total PVSs as alternative mediators in sensitivity analyses (Figure S1), and the results consistently showed that SBP‐ARV was directly associated with MoCA score rather than through these CSVD features.

FIGURE 4.

The mediation role of CSVD burden (CSVD score) between blood pressure variability (SBP‐ARV) and cognitive performance (MoCA score). ARV, average real variability; CSVD, cerebral small vessel disease; MoCA, Montreal Cognitive Assessment; SBP, systolic blood pressure.

4. DISCUSSION

In the present study, we found that functional hemodynamic alterations may occur in the absence of detectable structural microvascular injury, and in a community‐dwelling cohort of older adults, increased BPV—particularly higher beat‐to‐beat SBP‐ARV—was associated with poorer cognitive performance in individuals with poorer CA. In addition, although elevated SBP‐ARV was associated with increased CSVD burden, the negative impact of elevated SBP‐ARV on cognitive performance appeared to be a direct effect rather than one mediated through CSVD, which was in contrast to our initial expectation. Notably, the observed ranges of dCA indices and BPV in this cohort were comparable to previously published normative data, ¹⁹ , ²⁰ , ²⁸ suggesting that our participants represent a relatively healthy, community‐dwelling older population. Despite this generally preserved hemodynamic profile, variations in SBP‐ARV and phase shift–VLF were still significantly associated with cognitive performance. These findings collectively suggest that early hemodynamic instability may be associated with increased cognitive vulnerability, and the subclinical BP and CA alterations might represent potential targets for intervention.

In patients with higher phase shift‐VLF, the association between SBP‐ARV and cognitive performance was not significant. This interaction supports the idea that CA may serve as a physiological buffer against beat‐to‐beat BP fluctuations; when this buffering capacity is impaired, greater cognitive vulnerability may be observed. Therefore, either impaired CA or increased BPV alone may not be sufficient to contribute to cognitive impairment. In the current study, most participants had normal BP and adequate BPV; therefore, impaired CA may not be associated with inadequate CBF in this population. However, in participants with both impaired CA and increased BPV, the likelihood of inadequate CBF may be increased, which may be associated with poorer cognitive performance. In patients with mild ischemic stroke, decreased phase shift‐VLF is associated with post‐stroke cognitive impairment and subsequent cognitive decline. ¹⁴ However, in patients with Alzheimer's disease (AD) or mild cognitive impairment, phase shift‐VLF or other dCA indices do not differ from those in healthy controls. ²⁹ Therefore, impaired CA may not be a consequence of cerebral damage, but rather a pre‐existing vulnerability factor that requires other co‐occurring factors, such as increased BPV, to result in unfavorable neurological outcomes.

Numerous studies have investigated the influences of short term (hours‐to‐days) BPV and long term (visit‐to‐visit) BPV on cardiovascular and cerebrovascular diseases, dementia, and chronic kidney diseases. ⁶ Although studies regarding beat‐to‐beat BPV are relatively limited, evidence linking beat‐to‐beat BPV with diseases is increasingly recognized. ¹⁶ , ³⁰ , ³¹ Increased beat‐to‐beat BPV is associated with impaired baroreflex and sympathovagal imbalance, ³² , ³³ which is not the primary pathophysiology underlying increased hours‐to‐day or visit‐to‐visit BPV. Impaired baroreflex has been reported in numerous cardiovascular diseases. ³⁴ Therefore, increased beat‐to‐beat BPV has been associated with increased risks of cardiovascular and cerebrovascular diseases. The reason that SBP‐ARV was the most sensitive marker of BPV, compared to SBP‐SD and SBP‐CV, in explaining CSVD and cognitive performance in this study may be that the test‐retest reliability of ARV is better than SD and CV in beat‐to‐beat BPV. ²⁰ In addition, SD and CV reflect only global fluctuations around the mean level, whereas ARV accounts for the temporal sequence of changes. ³² Therefore, ARV may be less vulnerable to outliner values caused by measurement noise and may have higher disease specificity than SD and CV in BPV measurement.

Although SBP‐ARV was associated with CSVD, SBP‐ARV exerted a direct effect on MoCA score rather than an indirect effect mediated through CSVD in the present study. One possible explanation is that elevated BPV may be linked to cerebral functional disturbances before overt structural injury develops. Another possibility is that the relatively low and restricted range of CSVD scores in this community‐dwelling cohort (median CSVD score 1, IQR 0–2) may have limited our ability to detect a significant mediating effect of CSVD on the association between BPV and cognitive performance. However, in sensitivity analyses using normalized WMH volume and PVS counts as alternative mediators, the results similarly indicated a direct effect of SBP‐ARV on MoCA score rather than a mediated effect. In addition, CSVD score did not show an independent association with MoCA score in the multivariate linear regression model, suggesting that covert CSVD may have a relatively modest impact and may not be sufficient to account for the observed hemodynamic‐cognitive associations in this relatively healthy cohort.

Another consideration is the potential contribution of AD–related pathophysiology to the observed associations between BPV and cognitive performance. Emerging evidence suggests that increased BPV may be related to AD pathophysiological processes, although findings remain inconsistent. For example, prior studies have reported associations between increased visit‐to‐visit and beat‐to‐beat BPV and decreased amyloid‐β as well as increased phosphorylated tau levels in cerebrospinal fluid, ³⁵ , ³⁶ whereas another study has reported no association between 24‐hour ambulatory BPV and blood amyloid‐β and phosphorylated tau levels. ³⁷ It is therefore possible that the observed direct association between SBP‐ARV and cognitive performance may partly reflect underlying AD‐related mechanisms that were not assessed in this study. We did not assess AD biomarkers in the present study; therefore, we cannot exclude the possibility that some participants were in a prodromal stage of AD. Future investigations incorporating AD biomarker characterization are needed to clarify the interplay among BPV, cerebral hemodynamics, AD pathology, and cognitive decline.

BPV calculated from different time scales and using different formulas reflects distinct physiological mechanisms. Beat‐to‐beat BPV reflects the combined influence of sympathetic activation, baroreflex function, arterial compliance, and humoral status, whereas visit‐to‐visit BPV depends mainly on behavior and environmental factors. ³⁸ Therefore, the pathophysiological mechanisms underlying beat‐to‐beat BPV may differ from those of visit‐to‐visit BPV, potentially leading to different treatment strategies. Evidence supporting medical treatment for increased beat‐to‐beat BPV is limited. Aerobic exercise training can enhance baroreflex function, which may help reduce beat‐to‐beat BPV. ³⁹ , ⁴⁰ In addition, animal studies have shown that angiotensin‐converting enzyme inhibitors improve baroreflex sensitivity and lower BP by attenuating sympathetic tone, which may also reduce beat‐to‐beat BPV. ⁴¹ , ⁴² The Anglo‐Scandinavian Cardiac Outcomes Trial‐Blood Pressure‐Lowering Arm (ASCOT‐BPLA) revealed that use of a dihydropyridine calcium channel blocker (amlodipine) was associated with decreased visit‐to‐visit BPV and fewer cardiovascular events compared with a beta‐blocker (atenolol). ⁴³ In the Systolic Blood Pressure Intervention Trial (SPRINT), the use of dihydropyridine calcium channel blockers was also associated with reduced visit‐to‐visit BPV. ⁴⁴ However, whether BPV reduction can mitigate the risk of cognitive impairment—and whether beat‐to‐beat BPV can be reduced by using dihydropyridine calcium channel blockers— remains to be determined.

There are limitations to the current study. First, the participants were from urban areas and had relatively high education levels; thus, they might have better health literacy and access to healthcare resources than individuals from rural areas, and their CSVD burden was relatively low, potentially limiting the generalizability of the findings. Second, this study has a cross‐sectional observational design; therefore, causal relationships between variables cannot be established. For example, CSVD may result from alterations in CA and BPV, whereas CSVD itself may also induce reciprocal changes in CA and BPV. Moreover, it remains unclear whether alterations in BPV and CA precede, accompany, or follow early cognitive decline. Third, CA and BPV were assessed using Doppler ultrasonography and finger plethysmography, which are indirect measures of CBF and BP. Nevertheless, these techniques are well established and have been widely applied in relevant studies. ¹⁹ Fourth, we did not assess Alzheimer's disease biomarkers, which limits our ability to determine whether underlying AD pathophysiology contributed to the observed associations between BPV and cognitive performance.

In conclusion, in a community‐dwelling cohort of older adults, increased beat‐to‐beat BPV and poorer CA interacted and were associated with poorer cognitive performance. The negative impact of their interaction on cognitive performance appeared to show a direct association rather than one mediated through CSVD, suggesting that BPV and CA may be associated with cerebral function independently of detectable CSVD. Longitudinal investigations are necessary to clarify their temporal relationships and underlying mechanisms.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Any author disclosures are available in the Supporting Information.

CONSENT STATEMENT

The study was approved by the Institutional Review Board of National Yang Ming Chiao Tung University (YM109157E). Participants gave informed consent to participate in the study before taking part.

Supporting information

Supporting Information

Supporting Information