Associations between potential risk factors and blood-brain barrier water permeability in middle-aged and older adults

Abstract

Background: Blood-brain barrier (BBB) dysfunction is suggested to be a potential mediator between vascular risk factors and cognitive impairment, leading to vascular cognitive impairment. Objective: To investigate the relationships between age, sex, and vascular risk factors and BBB water permeability as well as their relationship with cognition. Methods: To measure BBB permeability, a novel arterial spin labelling MRI technique (ME-ASL) was applied to derive the time of exchange (Tex), arterial time transit (ATT), and cerebral blood flow (CBF). The association of potential risk factors, such as age, sex, body mass index (BMI), blood pressure (BP), and medical history, with these BBB parameters were assessed in 144 community-dwelling adults (median age 59 years, 57% females). The relationship between BBB permeability and cognitive performance measured by the Montreal Cognitive Assessment (MoCA) was also assessed. Results: We found that increased BMI was significantly associated with decreased CBF (β = −0.06). Systolic BP and diastolic BP showed significant associations with all ASL parameters; systolic BP was negatively correlated with Tex (β = −0.02) and CBF (β = −0.01) but positively with ATT (β = 0.02). Diastolic BP was negatively associated with Tex (β = −0.03) and CBF (β = −0.03) but positively with ATT (β = 0.03). MoCA scores had a borderline significant association with Tex (OR = 1.51) and a significant association with CBF (OR = 1.84), which became non-significant after adjusting for confounders. Conclusions: These outcomes underscore the potential of using ME-ASL, warranting further research to strengthen these findings.

Introduction

Vascular cognitive impairment (VCI) refers to a range of cognitive problems caused by damage to the blood vessels. VCI is often associated with various forms of vascular disease, including stroke and cerebral small vessel disease (cSVD). ¹ One of the key underlying mechanisms believed to contribute to the development of cSVD is dysfunction of the blood-brain barrier (BBB), ² where it was found that BBB dysfunction may promote the leakage of neurotoxins into the brain, leading to cognitive decline.^3,4

While aging has traditionally been linked to BBB dysfunction, ⁵ previous studies have also found associations between vascular risk factors and BBB dysfunction, such as hypertension, cardiovascular disease, hyperlipidemia, diabetes, and smoking.^6,7 These vascular risk factors can affect a myriad of mechanisms that degrade the BBB. For example, the inability to regulate arterial pressure can lead to altered BBB, contributing to the development of cSVD. ¹ Studies have also found that hyperglycemic conditions were associated with the thickening of the capillary and abnormal proliferation of endothelial cells affecting BBB integrity. In fact, diabetes, hypertension, and smoking all have in common an increase in oxidative stress in the bloodstream, which damages BBB integrity.^7,8 However, these mechanisms are hard to measure in a population setting and are often assessed in vitro and animal models.

In general, reducing the risk of cardiovascular diseases (myocardial infarction, stroke, atrial fibrillation, blocked coronary arteries, and heart failure) was also found to be protective against brain deterioration. ⁹ This protective effect aligns with what has been observed through the biological mechanisms mentioned above. Hence, it is also important to investigate BBB integrity in vivo in a population setting to better understand the relationship between vascular risk factors and BBB permeability.

Besides vascular risk factors, it has also been found that females tend to have better BBB integrity than males. These differences were proposed to be caused by the difference in the distribution of sex hormone receptors.^10–12 However, these sex differences may vary between cognitively healthy and cognitively impaired individuals, with no differences observed in BBB permeability among participants with mild cognitive impairment. ¹¹ Therefore, it is important to assess BBB integrity when the participant is still cognitively healthy.

So far, the alterations of BBB permeability have been assessed using either imaging techniques or cerebrospinal biomarkers as a proxy to determine BBB permeability. Specifically for imaging techniques, dynamic susceptibility contrast (DSC) and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) using gadolinium chelates and positron emission tomography (PET) were employed. ¹³ Although such methods exist, they are considered invasive and are not suitable for all patients. PET exposes patients to radiation and is expensive, while DSC- and DCE-MRI rely on contrast agents that may not be suitable for all patients.^13,14

Several variations of arterial spin labelling (ASL) MRI have been introduced for research applications in recent years, a non-invasive method for measuring blood perfusion in the brain.^15–19 ASL uses magnetically labelled blood water as an endogenous tracer and can be extended to quantify water exchange dynamics across the BBB. Using the Multi-TE Hadamard pseudo-continuous (pCASL) sequence with 3D Gradient and Spin Echo (GRASE) readout technique (ME-ASL),^20,21 a three-compartment model was implemented to calculate three ASL parameters: cerebral blood flow (CBF), arterial transit time (ATT), and water exchange time (Tex). Tex is a time parameter that describes the residence time of labelled water at the capillary bed before transiting from intravascular to extravascular space.^18,20,21 While ASL has been used extensively, no study has employed ME-ASL to assess the relationship between vascular risk factors and BBB permeability.^22,23 The lack of information on ME-ASL warrants investigation to explore ME-ASL as a proxy marker in detecting or monitoring disease progression. In this paper, we investigated the relationships between age, sex, and vascular risk factors and BBB water permeability as well as their relationship with cognition.

Methods

Study design and participants

In this study, we included two studies with community-dwelling participants as one cohort. The differences in the inclusion criteria are described below. The NEUROlogical biomarkers of Blood, MRI, and Cognition (NEURO-BMC) study is an ongoing cohort at the National University of Singapore (NUS), comprising participants from a multi-ethnic study (MEC) as well as from the community aged 45–85 years. ²⁴ The NEURO-BMC cohort was recruited to examine early subclinical biomarkers for vascular and neurodegenerative changes. To be included in the study, participants with any physical disability, prevalent dementia, significant neurological disease, or contraindications to MRI were excluded. ²⁵ For this study, we included data collected from 14 February 2022 to 17 August 2023 (Figure 1). All participants underwent extensive questionnaire administration and neuropsychological assessment. Those with suspected cognitive decline based on a Clinical Dementia Rating scale (CDR) > 0.5, subjective cognitive complaints, or failure of any cognitive assessment domains or cognitive assessment scores below the education-adjusted threshold were invited for brain MRI (n = 144). Quality checks were conducted, as described elsewhere, to filter out images that were considered acceptable and good (Supplemental Figure 3). ²⁶ Eight out of 144 participants were excluded due to incomplete or missing ASL scans. A total of 110 participants were included in the study from NEURO-BMC, after excluding 25 participants for poor quality ASL images (ASL labelling errors, motion artefacts) and two outliers (CBF values above 3SD).

Figure 1.

Flowchart of excluded participants.

The Spatial Performance Assessment for Cognitive Evaluation (SPACE) study recruited participants from the community aged 55–80 years. To be eligible, participants with any physical disability, prevalent dementia, significant neurological disease, or contraindications to MRI were deemed ineligible to participate. A total of 40 out of 42 eligible participants from the community were invited for this study, with two rejecting the invitation. The participants in SPACE study were randomized into two arms, control and intervention, until recruitment numbers were met (n = 40). All participants in this cohort were invited for brain MRI for both baseline and follow-up. MRI scans were acquired for the 40 participants, and only baseline images were used for this study. Six participants were excluded for poor-quality ASL images (ASL labelling errors, motion artefacts).

Written informed consent was obtained in the participants’ preferred language for both NEURO-BMC and SPACE. Ethics approval for NEURO-BMC and SPACE were obtained from the National University of Singapore Institutional Review Board (NUS-IRB Reference Code: NUS-IRB-2021-531 and NUS-IRB-2022-466, respectively). In this study, both cohorts were combined as one cohort for the analysis, as both studies used identical MRI protocols. The final number of participants was 144.

Image acquisition and processing

MR imaging was performed at the Singapore Centre for Translational MR Research using a 3T Prisma-Fit scanner (Siemens Healthineers, Erlangen, Germany). The standardized neuroimaging protocol used in this study included 3D T1-weighted (TR = 2300 ms, TE = 1.96 ms, TI = 900 ms, flip angle = 9॰, voxel size = 1 × 1 × 1 mm³), T2-weighted FLAIR (TR = 9000 ms, TE = 82 ms, TI = 2500 ms, flip angle = 180॰, voxel size = 1 × 1 × 3 mm³), and ASL sequences. ASL was acquired according to the DEBBIE consortium protocol. ¹⁵ Briefly, two Walsh-ordered Hadamard time-encoded pseudo-continuous arterial spin-labelling acquisitions with a segmented 3D-GRASE readout were acquired: One Hadamard 8 acquisition with an 8 × 8 matrix with two repetitions (post labelling delay (PLD) = 0.6 and 0.8 ms, repetition time (TR) = 4000 ms) and sub-bolus duration 400 ms, resulting in 14 PLDs [600,800,1000,1200,1400,1600,1800,2000,2200, 2400,2600,2800,3000,3200] ms, and a single-echo (echo time (TE) = 11.6 ms) readout optimized for ATT and CBF quantification.^21,27–30 One with Hadamard 4 acquisition with a 4 × 4 matrix, sub-bolus duration 1000 ms, three PLDs [500,1500,2500] ms, one repetition and a multi-echo (TE = [11.6:34.9:174.6] ms) readout optimized for BBB Tex quantification. The spatial resolution was 2.5 × 2.5 × 5 mm³ with 32 slices and two background suppression pulses timed to suppress T1 values of 700 and 1400 ms. A separate M0 image was acquired with a TI 1700 ms and TR 5000 ms. Constant T2 values were assumed to be 165 ms for blood and 85 ms for tissue. Tex, CBF, and ATT quantification was performed using the default settings of ExploreASL version v1.11 (https://exploreasl.github.io/Documentation/1.11.0/), including SPM12 version 7219 (https://www.fil.ion.ucl.ac.uk/spm/software/spm12/). The image processing by ExploreASL, ³¹ except for the recent expansion with ME-ASL analysis, ²¹ was detailed previously. In short, ExploreASL uses several toolboxes to process ASL images: (1) Structural segmentation was performed using CAT12 version r1615 (https://neuro-jena.github.io), (2) white matter hyperintensity segmentation was performed using LST version 2.0.15 (www.statisticalmodelling.de/lst.html), and (3) ASL quantification was performed with FSL version 6.0.7.11 (https://fsl.fmrib.ox.ac.uk/), which has a dedicated ME-ASL quantification module. ²¹ A three-compartment model was used with an arterial compartment, capillary intravascular compartment, and extravascular compartment. Four parameters were fitted within this model: ATT, CBF, time parameter Tex describing the exchange across the BBB, and intra-voxel transit time (ITT) parameter describing the time delay the labelled blood needs to reach the capillary compartment from the arterial compartment. Hadamard 4 and Hadamard 8 data were concatenated, and all four parameters were fitted jointly on a voxel-wise basis. ²¹ The grey matter (GM) region was created by thresholding the GM map at 70%.

Potential risk factors measurements

All participants underwent the same protocol for each of the risk factors measured. After the participant was rested in a seating position for at least 15 min, blood pressure (BP) on their upper arm was measured three times during the visit using an OMRON HEM-7121. The mean was calculated for all three systolic and diastolic readings. Body mass index (BMI) was calculated from measured height and weight. Smoking history was self-reported by using an interviewer-administered questionnaire. Anyone who reported having smoked more than 100 cigarettes is categorized in the “Ever Smoked” group. History of hypertension, hyperlipidemia, diabetes, cardiovascular disease, and medication status were recorded based on medical records provided by the participants. Cardiovascular disease was defined as having a history of myocardial infarction, stroke, atrial fibrillation, coronary angioplasty, coronary bypass surgery, or heart failure. The Montreal Cognitive Assessment (MoCA) was administered in both cohorts to assess cognition. ³²

Statistical analysis

Table 1 presents the demographic characteristics between NEURO-BMC and SPACE. Further comparisons for demographic characteristics between included and excluded datasets, and between males and females can be found in Supplemental Tables 1 and 2, respectively. All continuous variables were tested for normality using the Shapiro-Wilk normality test. Continuous variables were compared using Wilcoxon rank sum test, while categorical variables were compared using Fisher's exact and Pearson's Chi-square test. All the dependent variables analyzed in linear regression models were standardized for easier comparison and non-standardized results were presented in the Supplemental Material. Associations between vascular risk factors and MoCA were determined using univariate logistic regression models. Associations between sex, age, existing medical conditions, blood pressure, hypertension status, and ME-ASL parameters were determined using univariate linear regression models. In each univariate model, a single dependent variable was regressed on a single independent variable without adjusting for covariates. The multivariate models included variables that were found to have a significant association from univariate analysis with all ASL parameters. For multivariate models with Tex as dependent variables, CBF and ATT were adjusted as covariates. To assess for multicollinearity in the multivariate regression models, variance inflation factor (VIF) was calculated for each independent variable, with a threshold above 5 indicating high correlation. The MoCA score was binarized into two groups (<27 and ≥27) based on previous validation. ³³ Logistic regression was used to assess the relationship between ASL parameters and MoCA, with the score <27 grouped as the reference for the dependent variable. MoCA was the dependent variable in these models, with ASL parameters included as the main exposure. Multivariate models were also adjusted for covariates, which were associated to MoCA from the previous univariate analysis, to further examine the relationship between ASL parameters and MoCA. Due to the lack of validity for WM ASL signal, multivariate analyses WM were conducted but not shown in our main results, analyses can be found in the Supplemental Material. Statistical analysis was performed using R version 4.2.2 (2022-10-31) on RStudio.

Table 1.

Characteristics of participants in NEURO-BMC and SPACE studies.

Characteristic	Overall n = 144	NEURO-BMC n = 111	SPACE n = 34	p
Age, median (IQR)	59 (54,65)	57 (52,63)	66 (63,70)	<0.001
Sex, no. (%)
Male	62 (43%)	28 (25%)	34 (100%)
Race, no. (%)				0.2
Chinese	137 (95%)	103 (94%)	34 (100%)
Ever Smoked, no. (%)	14 (9.7%)	8 (7.3%)	6 (18%)	0.1
BMI (kg/m2), mean (SD)	23.7 (3.7)	23.1 (3.3)	25.5 (4.1)	0.003
Hypertension, no. (%)	33 (23%)	18 (16%)	15 (44%)	<0.001
Diabetes, no. (%)	14 (9.7%)	5 (4.5%)	9 (26%)	<0.001
Hyperlipidemia, no. (%)	45 (31%)	32 (29%)	13 (38%)	0.3
Cardiovascular disease, no. (%)	12 (8.3%)	6 (5.5%)	6 (18%)	0.035
Anti-hypertensives, no. (%)	14 (9.7%)	12 (11%)	2 (5.9%)	0.5
Systolic Blood Pressure, mean (SD)	123 (19)	120 (18)	135 (17)	<0.001
Diastolic Blood Pressure, mean (SD)	77 (10)	76 (10)	81 (10)	0.014
GM CBF (ml/100 g/min), mean (SD)	72 (15)	75 (14)	62 (12)	<0.001
GM ATT (s), mean (SD)	1.17 (0.13)	1.15 (0.13)	1.22 (0.11)	0.014
GM Tex (ms), mean (SD)	186 (32)	190 (32)	175 (28)	0.023
MoCA, median (IQR)	28 (27,29)	28 (27,29)	27 (25,29)	0.034
MoCA, no. (%)
Above 27	110 (76)	88 (80)	22 (65)	0.066

GM: grey matter; CBF: cerebral blood flow; ATT: arterial transit time; Tex: time of exchange; MoCA: Montreal cognitive assessment. Continuous variables were presented as mean (SD) and categorical as counts (percentage). Pearson's Chi-squared and Fisher's exact test were used for categorical test. Wilcoxon rank sum test (non-parametric) was used for continuous data. Difference between NEURO-BMC and SPACE were tested and p-value below 0.05 is considered to be significant. Sex was not assessed statistically between the two cohorts as SPACE did not include any females as participants. Values in bold indicate statistically significant p-values (p < 0.05).

Results

Characteristics of participants

Comparison between datasets performed for included and excluded participants showed that the excluded participants had more males (43 versus 68%, p = 0.006), higher BMI (23.7 versus 24.9, p = 0.033), hypertension, and higher SBP (123 versus 133 mmHg, p = 0.013), and DBP (77 mmHg versus 82 mmHg, p = 0.019) (Supplemental Table 1). Comparison between datasets performed for NEURO-BMC and SPACE participants showed that SPACE were older (57 versus 66, p ≤ 0.001) and had a higher proportion of cardiovascular disease (5.5 versus 18%, p = 0.035), higher BMI (23.1 versus 25.5, p = 0.003), SBP (120 versus 135 mmHg, p ≤ 0.001), DBP (76 mmHg versus 81 mmHg, p = 0.014), and ATT (1.15 versus 1.22, p = 0.023) (Table 1). SPACE participants were also observed to have higher proportions of hypertension and diabetes.

Supplemental Table 2 compares demographic characteristics and ME-ASL parameter values between sexes. In general, males were older (63 versus 56) and had higher BMI (24.5 versus 23.0 km/m²), SBP (133 versus 116 mmHg), and DBP (82 versus 74 mmHg) (p ≤ 0.001 for all characteristics mentioned). Only 9.7% of the participants had ever smoked, and a higher proportion were males. More males were found to be hypertensive (35% of males and 13% of females, p = 0.002). Differences between sexes were also observed in Figure 2 for ME-ASL parameters, with males having lower Tex (p ≤ 0.001), longer ATT (p ≤ 0.001), and higher CBF (p ≤ 0.001). When Tex (Male: β = 0.327, p = 0.470, Female: β = −0.183, p = 0.749) and CBF (Male: β = −0.108, p = 0.559, Female: β = 0.176, p = 0.479) were compared over age opposite trend was observed (Figure 3). ATT showed a similar trend between (Male: β = 0.002, p = 0.225, Female: β = 0.002, p = 0.399) over age. Males were also observed to have lower MoCA scores (p = 0.022). Diabetes, hyperlipidemia, and cardiovascular disease were not significantly different between the sexes.

Figure 2.

Differences in sex values of GM Tex (A), GM ATT (B) and GM CBF (C). Triangle: Male, Circle: Female. T-test was conducted to check for significance differences between sex for each ASL parameters (A-C).

Figure 3.

Age versus ASL by sex of GM Tex (A), GM ATT (B) and GM CBF (C). Triangle, Solid line: Male, Circle, Dash line: Female.

Relationship of potential risk factors with ASL parameters

Males had a higher BBB permeability, reduced CBF, and a prolonged ATT (Figure 2). Older age was significantly associated with longer ATT and a lower CBF (Figure 3). The univariate analysis showed that increased SBP, DBP, and BMI were associated with increased Tex, lower CBF, and longer ATT (Table 2). Smoking was not associated with any of the ASL parameters. After adjusting for age, sex, and hypertension, BMI was associated with CBF, and SBP and DBP were associated with ATT, and CBF (Table 3). For Tex models after adjusting for age, sex, hypertension, CBF, and ATT, only SBP was associated with DBP having a borderline association. The relationship between BMI and SBP with CBF was no longer observed when the analysis was repeated, including the excluded participants with poor-quality images (data not shown). VIF (Supplemental Table 7) did not have any significant multicollinearity present among all variables in the models.

Table 2.

Univariate linear regression, between potential risk factors and ASL parameters.

Characteristic	GM Tex		GM ATT		GM CBF
	Beta (95% CI)	P	Beta (95% CI)	p	Beta (95% CI)	p
Sex
Male	−0.73 (−1.04, −0.42)	<0.001	0.67 (0.36, 0.98)	<0.001	−1.01 (−1.29, −0.72)	<0.001
Age	−0.02 (−0.04, 0.00)	0.092	0.03 (0.01, 0.05)	0.003	−0.03 (−0.05, −0.01)	0.01
Years of Education	0.01 (−0.03, 0.06)	0.6	−0.04 (−0.08, 0.01)	0.11	0.01 (−0.04, 0.05)	0.7
BMI (kg/m2)	−0.05 (−0.10, −0.01)	0.016	0.06 (0.01, 0.10)	0.011	−0.09 (−0.13, −0.05)	<0.001
Systolic Blood Pressure	−0.02 (−0.03, −0.02)	<0.001	0.02 (0.01, 0.03)	<0.001	−0.02 (−0.03, −0.01)	<0.001
Diastolic Blood Pressure	−0.04 (−0.06, −0.03)	<0.001	0.03 (0.02, 0.05)	<0.001	−0.04 (−0.06, −0.03)	<0.001
Ever Smoked
Yes	−0.3 (−0.85, 0.25)	0.3	0.4 (−0.15, 0.95)	0.2	−0.43 (−0.98, 0.12)	0.13
Hypertension
Yes	−0.63 (−1.01, −0.26)	0.001	0.58 (0.20, 0.96)	0.003	−0.54 (−0.92, −0.16)	0.006
Hyperlipidemia
Yes	−0.02 (−0.38, 0.33)	0.9	0.08 (−0.27, 0.44)	0.6	0.23 (−0.13, 0.58)	0.2
Cardiovascular Disease
Yes	−0.25 (−0.84, 0.34)	0.4	0 (−0.59, 0.59)	>0.9	−0.2 (−0.79, 0.40)	0.5
Diabetes
Yes	−0.35 (−0.90, 0.20)	0.2	0.48 (−0.06, 1.03)	0.086	−0.69 (−1.23, −0.15)	0.013
Anti-hypertensives
Yes	−0.47 (−1.02, 0.08)	0.093	0.48 (−0.07, 1.02)	0.091	−0.23 (−0.79, 0.32)	0.4

GM: grey matter; Tex: time of exchange; ATT: arterial transit time; CBF: cerebral blood flow; BMI: body mass index. Values in bold indicate statistically significant p-values (p < 0.05).

Table 3.

Multivariate linear regression adjusting for sex, age, and hypertension (and CBF and ATT only for TEX models), with ASL parameters.

Models		GM Tex			GM ATT			GM CBF
		Beta (95% CI)	p		Beta (95% CI)	p		Beta (95% CI)	p
	Age	0.24 (−0.46, 0.95)	0.5		0.01 (−0.01, 0.03)	0.282		0.00 (−0.02, 0.02)	0.722
Model 1	Sex	−21.56 (−32.47, −10.65)	<0.001		0.5 (0.15, 0.85)	0.005		−0.98 (−1.30, −0.65)	<0.001
	Hypertension	−14.57 (−26.50, −2.63)	0.018		0.37 (−0.02, 0.75)	0.063		−0.26 (−0.61, 0.10)	0.158
Model 2	BMI	−0.03 (−0.07, 0.02)	0.256	Model 5	0.04 (0.00, 0.08)	0.077	Model 8	−0.06 (−0.10, −0.02)	0.002
Model 3	SBP	−0.02 (−0.03, −0.01)	<0.001	Model 6	0.02 (0.01, 0.02)	0.002	Model 9	−0.01 (−0.02, 0.00)	0.039
Model 4	DBP	−0.03 (−0.05, −0.01)	<0.001	Model 7	0.03 (0.01, 0.04)	0.002	Model 10	−0.03 (−0.04, −0.01)	0.001

GM: grey matter; Tex: time of exchange; ATT: arterial transit time; CBF: cerebral blood flow; BMI: body mass index; SBP: systolic blood pressure; DBP: diastolic blood pressure. Model 1: Single ASL parameter ∼ age + sex + hypertension. Model 2: Model 1 + BMI, Model 3: Model 2 + SBP, Model 4: Model 3 + DBP. Values in bold indicate statistically significant p-values (p < 0.05).

Relationship of potential risk factors and ASL parameters with MoCA

Univariate logistic regression did not show any association of vascular risk factors with MoCA. (p > 0.05, data not shown). Univariate logistic regression showed borderline associations of Tex (OR: 1.51, p = 0.051) and significant associations of CBF (OR: 1.84, p = 0.006) with MoCA. However, there was no association observed between ATT and MoCA (OR: 0.75, p = 0.200). Higher Tex and higher CBF were associated with better cognition. However, multivariate analysis showed no significant associations after adjusting for confounders. VIF (Supplemental Table 7) indicated no significant multicollinearity among the variables in the models.

Discussion

This study unveils associations between potential risk factors and ME-ASL parameters. Our findings indicated that older age is linked to decreased CBF and prolonged ATT across both sexes. Notably, females have higher CBF and Tex compared to males. Elevated BMI was also associated with reduced CBF, but no associations with Tex and ATT were observed. Only diastolic blood pressure was associated with Tex, ATT, and CBF. Interestingly, ASL parameters were not associated with cognition as measured by the MoCA. These findings agree with previous studies where correlations of several vascular risk factors were associated with BBB degradations.^3,34

This study only observed significant associations between age and CBF and between age and ATT when both sexes were assessed as a whole cohort (Supplemental Figure 2). These results were consistent with previous findings from the Human Connectome Project-Ageing study which showed that with increased age CBF decreases and ATT increase. ²² We were only able to observe a trend in the association between Tex and age, but it was not significant. While Tex was not statistically significant, the trend for Tex aligns with the previous findings from studies and review conducted in animal models or humans where BBB permeability was observed to increase with age.^34–37 Given that Tex is the proxy marker for BBB permeability, it was expected to be lower as age increases, illustrating an increase in BBB permeability. Several explanations can be suggested for the lack of associations with Tex: First, the age distribution between both sexes was quite different, with a median age difference of eight years. This difference may have attenuated the associations observed in this study when the participants were stratified by sex. Ideally, if there were more older female participants, it may help strengthen the associations observed in this study. Second, we used mean total GM Tex in this study, which do not account for the variability across different brain regions. When considered together, differences between regions can perhaps nullify the effects present in the result, as described in a review, that the variability within individual regions is large. ²⁰

In our study, females were observed to have elevated blood perfusion rates, prolonged ATT, and reduced BBB permeability compared to males (Figure 2). Sex emerged as the most robust determinant, consistent with trends in other studies.^38,39 As reported in previous studies, females were often observed to have higher CBF and shorter ATT in specific brain regions. ¹¹ Adding to this finding, females typically exhibit lower hematocrit levels, resulting in higher CBF to compensate for this reduced haematocrit. ⁴⁰ Additionally, from a measurement perspective, lower hematocrit increases the T1-time of the labelled spins and thus leads to seemingly higher measured CBF values, which can hinder direct comparisons between sexes. However, such differences tend to diminish among those with mild cognitive impairments. ¹¹ The outcome observed in this study is consistent with evidence from previous in vitro and animal model studies that delved into the mechanistic behavior of BBB between sexes. ⁴⁰ Here, previous results have shown that sex hormones, especially estrogen, can influence or protect BBB integrity via the mechanisms involving transport proteins and brain pharmacokinetics, which could protect BBB integrity. ⁴¹ However, it is essential to note that these observations have yet to be thoroughly studied in vivo. In a separate investigation using DCE-MRI and diffusion-prepared ASL (DP-ASL), females also had lower BBB permeability, indicated by lower K^trans.^11,17,42

Higher diastolic blood pressure was associated with decreased Tex values, which indicates increased BBB permeability. This result aligns with previous findings showing that hypertension can increase BBB permeability by triggering a cascade of mechanisms.^3,34 In addition, both SBP and DBP were associated with longer ATT. Given that higher blood pressure can cause arterial stiffening, this may have also indirectly influenced the blood velocity, leading to longer ATT. In the past, this parameter has been challenging to measure and requires invasive methods that were only used in animal models. To our knowledge, few studies assessed the relationship between BP and ATT and most studies depend on proxy markers to measure ATT.^42,43 One such study used pulse transit time (PTT) as a proxy marker of ATT and showed that PTT increases along with BP. ⁴⁴ Similarly, in another study, ATT was also observed to increase as mean arterial pressure increases, which aligns with what we observed. ⁴⁵

Interestingly, the association of BMI was only observed with CBF after adjusting for confounding factors. As suggested by other studies, such confounding variables may contribute differently to the deterioration of BBB integrity, and this might explain why the effect observed from BMI might have been nullified after adjustment.^46–48 In particular, an elevated BMI above a defined cut-off point serves as an indicator for obesity. This is associated with increased vascular risk and compromised BBB integrity. Existing studies revealed that obesity contributes to cerebral vasodilation/hyperemia during reperfusion, exacerbating early post-ischemic BBB disruption. In addition, metabolic syndrome, including obesity, induces systemic inflammation that can cause BBB breakdown and increased immune cell infiltration. ⁴⁷ Given that BMI is highly variable, it may not be a sensitive measure to detect changes in Tex and ATT given that BMI is highly variable. Some studies found that obesity after 65 years may protect against cognitive impairments. ⁴⁸ Another explanation could be described by MacIntosh's work, where the association of ATT among obese individuals was only observed in specific brain regions. ⁴⁹ This study used only the mean ATT of the whole brain region, limiting sensitivity to detect any associations. Although our Tex and ATT did not significantly affect BMI, CBF-based cerebral hemodynamics align with those commonly observed in obese participants.

Reduced CBF was observed as BMI increased in this study. Several explanations may account for this observation. First, obese individuals often have other existing metabolic syndromes (e.g., diabetes and hypertension) that cause a reduction in CBF. ⁴⁹ Second, obese individuals will have a higher level of adipose tissues in their body that compete for blood supply. Given that an individual's cardiac output is limited and with competing needs, it is possible that blood flow towards the brain was reduced and supplied to the visceral fats instead.^50,51

Unfortunately, in this study, the majority of the participants scored above 27 (76%) leading to an imbalance distribution which may not well represent those that score below 27. As hypothesized, we found that CBF and MoCA scores were positively associated, and this aligns with what was observed in another study higher CBF being associated with higher scores for MoCA (Table 4). ¹⁶ Unexpectedly, higher Tex was associated with scores above 27, suggesting reduced BBB permeability to have lower MoCA scores. This did not align with observations in other studies, which showed that lower MoCA scores are associated with increased BBB permeability. ⁵² While adjusting for confounders did not change the trends observed among all ASL-derived parameters, we no longer observed the associations of CBF with MoCA scores (Table 4).

Table 4.

Multivariate logistic regressions for ASL parameters and moCA scores adjusted for age, sex, hypertension.

	Multivariate
	Models	OR (95% CI)	p
Age		0.96 (0.90, 1.01)	0.148
Sex	Model 1	0.46 (0.18, 1.16)	0.098
Hypertension		0.65 (0.26, 1.65)	0.349
GM Tex	Model 2	1.24 (0.80, 1.98)	0.347
GM ATT	Model 3	0.96 (0.61, 1.51)	0.864
GM CBF	Model 4	1.48 (0.91, 2.45)	0.120

GM: grey matter; Tex: time of exchange; ATT: arterial transit time; CBF: cerebral blood flow; MoCA: Montreal Cognitive Assessment; Model 1: MoCA ∼ Age + Sex + Hypertension; Model 2: MoCA ∼ + GM Tex + adjusted covariates in Model 1; Model 6: MoCA ∼ GM ATT + adjusted covariates in Model 1; Model 7: MoCA ∼ GM CBF + adjusted covariates in Model 1. ASL parameters in these models were fitted as main independent variable predicting for MoCA.

This study has a series of strengths which included having participants with a broader age range, which allows us to observe changes that may take place from middle-aged participants to older adults. Using identical ASL protocols and scanning sites for two studies also allowed us to combine the cohort to enhance the analysis with a greater sample size. However, this study also had some limitations. First, we did not have a similar distribution of age range within each sex. Given that sex can strongly influence ME-ASL parameters, both sexes should contain similar age distributions in order to observe a stronger relationship. Second, we did not have hematocrit measures to correct the quantification of CBF, although the effects of hematocrit on Tex calculation are not known yet. Third, both studies were conducted cross-sectionally, and the temporal effect on BBB integrity could not be observed. Fourth, both cohorts were scanned with the identical scanner, ASL protocol, same group of radiographers, and within the same time period, we do not expect any MRI methods or instrumental bias between the cohorts. However, NEURO-BMC had younger participants and SPACE included only men. Although no significant differences were observed between the cohorts when comparing the male sub cohorts only (results not shown), these cohort inclusion differences should be acknowledged when interpreting our results. Fifth, currently, we do not know to what extent anti-hypertensives can affect Tex or ATT. Sixth, the SPACE study did not have an extensive cognition test that was the same as that of NEURO-BMC. Thus, we were only limited to MoCA for cognition assessment which we had observed a ceiling effect with most participants scoring well in MoCA leading to a small variability in the cohort. Seventh, it is yet validated if the relatively high WM ATT and relatively low WM CBF are sufficient for reliable ME-ASL WM Tex measurements. WM-CBF remains challenging to measure with ASL due to issues with higher ATT and low signal-to-noise ratio, especially in older adults with poorer cerebrovascular health. Hence, WM Tex, CBF and ATT were not included in our main analysis as the results may be spurious and make the WM ASL analysis misleading. Lastly, the quality checks conducted visually may have introduced selection bias. Future research can further investigate if there can be a quantifiable way to minimize selection issues such that it may improve the exclusion from quality checks. In addition, as observed in Supplemental Table 1, the total number of participants excluded amounted to 41 and had significant differences.

Furthermore, it is essential to acknowledge that the signal-to-noise ratio of ASL remains relatively low. This inherent limitation in ASL imaging, particularly when investigating microscopic components like the BBB, may reduce the quality and precision of the acquired data.

Our study highlights the potential of integrating ME-ASL to assess BBB permeability through a non-invasive method. Tex being a potential marker to assess BBB integrity, provides an opportunity to examine BBB changes in future studies. The sample size of this study is small, and further investigation with a larger cohort could enhance the observed effects. Considering that blood pressure has a significant relationship observed in our studies, it will also be worthwhile to explore the relationship between blood pressure variability and its impact on BBB integrity.