Association of Cardiovascular Risk Factors with Cerebral Perfusion in Whites and African Americans

Abstract

Background:

Midlife cardiovascular risk factors increase risk for Alzheimer’s disease (AD). Despite disproportionately high cardiovascular disease and dementia rates, African Americans are under-represented in studies of AD risk and research-based guidance on targeting vascular risk factors is lacking.

Objective:

This study investigated relationships between specific cardiovascular risk factors and cerebral perfusion in White and African American adults enriched for AD risk.

Methods:

Participants included 397 cognitively unimpaired White (n = 330) and African American (n = 67) adults enrolled in the Wisconsin Alzheimer’s Disease Research Center who underwent pseudo-continuous arterial spin labeling MRI. Multiple linear regression models examined independent relationships between cardiovascular risk factors and mean cerebral perfusion. Subsequent interaction and stratified models assessed the role for APOE genotype and race.

Results:

When risk factor p-values were FDR-adjusted, diastolic blood pressure was significantly associated with mean perfusion. Tobacco use, triglycerides, waist-to-hip ratio, and a composite risk score were not associated with perfusion. Without FDR adjustment, a relationship was also observed between perfusion and obesity, cholesterol, and fasting glucose. Neither APOE genotype nor race moderated relationships between risk factors and perfusion.

Conclusion:

Higher diastolic blood pressure predicted lower perfusion more strongly than other cardiovascular risk factors. This relationship did not vary by racial group or genetic risk for AD, although the African American sample had greater vascular risk burden and lower perfusion rates. Our findings highlight the need to prioritize inclusion of underrepresented groups in neuroimaging studies and to continue exploring the link between modifiable risk factors, cerebrovascular health, and AD risk in underrepresented populations.

Keywords

Aging Alzheimer’s disease blood pressure cerebrovascular circulation neuroimaging obesity

INTRODUCTION

As the population ages, a dramatic increase is expected in prevalence, cost, and morbidity due to Alzheimer’s disease (AD) [1]. Emerging evidence points toward the confluence of cardiovascular risk factors and cerebrovascular pathophysiology contributing to the clinical manifestation of AD [2]. Modifiable cardiovascular risk factors (e.g., hypertension, dyslipidemia) present at middle age significantly increase risk for clinical AD and have been proposed as potential targets for AD prevention and treatment [3 –6]; even a 5-year delay in AD onset is projected to reduce the prevalence of AD by nearly half [7]. This may be particularly relevant for African Americans who are disproportionately affected by AD [4, 8] and often have elevated vascular risk factors [9]. Moreover, African Americans may be disproportionately impacted by socioeconomic disadvantages that also have been implicated in increased risk for dementia and stroke [9 –11]. The mechanisms by which vascular risk factors increase risk for AD dementia remain unclear; however, one hypothesis proposes that vascular risk factors contribute to AD-related pathophysiology via blood-brain barrier dysfunction and hypoperfusion [12]. Specifically, vascular risk factors lead to blood-brain barrier dysfunction and a reduction in cerebral blood flow which can reduce amyloid-β (Aβ) clearance at the blood-brain barrier and increase production of Aβ from the amyloid-β protein precursor (AβPP) leading to Aβ accumulation. Capillary hypoperfusion may induce neuronal dysfunction and can induce hyperphosphorylation of tau leading to neurofibrillary tangle formation. Support for this hypothesis is demonstrated by evidence of perfusion dysfunction in adults at risk for AD via APOE ɛ4 genotype carriage [13 –15] or family history [16], and perfusion disruption prior to AD pathology development in animal models of AD [17].

Arterial spin labeling (ASL) magnetic resonance imaging (MRI) is a promising technique for detection and characterization of cerebrovascular health. The main physiological parameter measured by ASL is cerebral perfusion or the delivery rate of oxygen and nutrients to the capillary bed and is expressed as the volume of blood per volume of tissue per minute (mL/100 g/min) [18]. Previous studies have reported associations between reduced perfusion and diabetes mellitus [19], hypertension [20], and tobacco use [21]. Our prior studies demonstrated that increased age and components of metabolic syndrome or estimated 10-year cardiovascular disease risk score [22] were associated with lower perfusion in a sample of white, late middle-aged adults [23 –25].

However, it remains unclear if there are particular cardiovascular risk factors that have a stronger association with reduced cerebral perfusion in midlife and therefore may be more important targets for prevention efforts. Importantly, the unique contributions of specific cardiovascular risk factors to cerebral perfusion have not been explored in racially diverse populations, despite African Americans having a higher prevalence of cardiovascular risk factors. Therefore, the aims of this study were to: 1) investigate the unique relationships between specific cardiovascular risk factors and cerebral perfusion in a racially diverse cognitively unimpaired late middle-aged population, 2) examine if increased genetic risk for AD (APOE ɛ4 carrier status) moderates relationships between cardiovascular risk factors and cerebral perfusion, and 3) assess within-group relationships between vascular risk factors and cerebral perfusion among African American and White subsamples.

MATERIALS AND METHODS

Participants

This study was approved by the University of Wisconsin-Madison Health Sciences Institutional Review Board and the Wisconsin Alzheimer’s Disease Research Center (ADRC). The initial sample included data from 406 middle-aged and older adults enrolled in the ADRC Clinical Core. Three participants were excluded due to large metal artifact impacting signal and six participants were excluded due to poor signal-to-noise ratio on ASL, resulting in a final sample of n = 397. Participants included in this study were classified as cognitively unimpaired (e.g., no evidence of mild cognitive impairment or dementia) at their baseline study visit by consensus conference panel based on National Institute on Aging–Alzheimer’s Association (NIA-AA) criteria [26, 27]. Enrollment in the Core targeted individuals whose risk for AD was enriched due to a parental history of AD. Data from participants who self-identified as African American or non-Hispanic White as primary race and completed neuroimaging exams were included in analyses.

Cerebral perfusion acquisition and quantification

MRI scans were completed on a clinical 3T scanner (Discovery MR750, GE Healthcare, Waukesha, WI) using either a 32-channel (Nova Medical) or 8-channel head coil (Excite HD Brain Coil; GE Healthcare). All participants were instructed to abstain from food, tobacco, and caffeine at least four hours prior to the MRI scan. Cerebral perfusion was acquired using a background-suppressed pseudo-continuous arterial spin labeling (pcASL) sequence [28] utilizing a 3D fast spin-echo stack of spiral sequence. Scan parameters were as follows: echo spacing = 4.9 ms; TE = 10.5 ms with centric phase encoding; spiral arms = 8, spiral readout duration = 4 ms, FOV = 240×240×176 mm; 4 mm isotropic spatial resolution; reconstructed matrix size = 128×128×44; number of averages (NEX) = 3; and labeling RF amplitude = 0.24 mG. Immediately after each ASL scan, a proton density (PD) reference scan was performed with identical imaging acquisition parameters without ASL labeling but with a saturation pulse applied 2.0 seconds prior to imaging. This proton density (PD) image was used for ASL flow quantification as well as for imaging registration. Cerebral perfusion was quantified based on recommendations from the ASL consensus paper [18]. Specifically, the following equation was utilized: $C B F = \frac{6000 \cdot λ \cdot (S I_{c o n t r o l} - S I_{l a b e l}) \cdot e^{P L D / T 1_{b}}}{2 \cdot α \cdot T 1_{b l o o d} \cdot (1 - e^{- τ / T 1_{b l o o d}})} \frac{(1 - e^{- S T / T 1_{t i s s u e}})}{S I_{P D}} [m l / 100 g / m i n]$ where T1_blood is T1 of blood and is assumed to be 1.6 s at 3T. The partial saturation of the reference image (SI_PD) is corrected for by using a T1_tissue of 1.2 s (typical of gray matter). To avoid numerical errors and noise amplification in areas of low proton density, a maximum neighborhood filter was applied to the PD image prior to quantification (per [28]). ST is saturation time and is set to 2 s. The partial coefficientλ is set to the whole brain average, 0.9. The efficiency, α, is a combination of both inversion efficiency (0.8) and background suppression efficiency (0.75) resulting in an overall efficiency of 0.6. Post-labeling delay (PLD) was set to 2.025 s. Labeling duration (τ) was set to 1.5 s. Cerebral perfusion (or CBF) is reported in ml/100 gm/min units.

A T1-weighted structural scan (BRAVO) was acquired axially using the following imaging parameters: 3D fast spoiled gradient echo sequence; inversion time 450 ms; repetition time (TR) 8.1 ms; echo time (TE) 3.2 ms; flip angle 12 acquisition matrix 256 256; field of view (FOV) 256 mm; and slice thickness 1.0 mm. The T1-weighted volume was segmented into tissue classes using Statistical Parametric Mapping Version 12 (SPM12, www.fil.ion.ucl.ac.uk/spm) segmentation procedure to produce gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) probability maps. The GM tissue mask was smoothed and each participant’s PD image was coregistered to their smoothed T1-weighted GM image using an affine transformation procedure. The derived transformation matrix was applied to the ASL map. Each scan was visually inspected to confirm that the registration between the two scans was accurate; no subjects were excluded due to poor registration. The GM mask (thresholded at 0.3) was applied to the perfusion image to limit analyses to voxels with greater than 30% probability of containing gray matter and mean perfusion was computed across all voxels within the GM mask. Example perfusion images from this sample are depicted in Supplementary Figure 1. The primary outcome measure used in all analyses was mean GM cerebral perfusion.

Cardiovascular risk factors

Cardiovascular risk factor data were acquired during the physical examination at the WADRC clinical core study visit closest in time to the MRI scan date (mean = 3 months difference between study visit and MRI scan). Predictor variables included fasting total cholesterol, triglycerides, low-density lipoprotein cholesterol (LDL-c), HDL cholesterol (HDL-c), fasting glucose, waist circumference, body mass index (BMI), and systolic and diastolic blood pressures, tobacco use (binary: current or former smoker versus never smoker), tobacco exposure (total years smoked multiplied by the range of cigarettes/packs smoked daily [1 cigarette to <1/2 pack , or ≥2 packs]), and a composite vascular risk score (ACC/AHA atherosclerotic cardiovascular disease (ASCVD) 10-year risk score [22]). Blood pressure was measured up to three times within an examination visit to obtain a stable measure, with the participant seated using a random-zero sphygmomanometer. Waist circumference was measured with an anthropometric tape to the nearest centimeter with the participant standing. The waist circumference was taken at the level of the natural waist (narrowest part). Waist measurements were taken twice by trained staff, and the smallest measurement was used. All models included covariates of age at MRI scan, sex, years of education, genetic risk for late-onset AD (APOE ɛ4 carrier status), antihypertensive medication use, and self-reported presence/absence of diabetes. Additionally, post-hoc analyses substituted reading level (National Institutes of Health (NIH) Toolbox Oral Reading Recognition Test [29]; available for n = 348 participants) for years of education as prior studies suggest reading level may provide a more precise estimate of educational quality and thus better reflect potential socioeconomic status differences across racial groups [30].

Statistical analyses

All analyses were performed in R version 3.5.1 [31]. The relationships between vascular risk factors and mean perfusion (Aim 1) were examined using multiple linear regression models. Each model included one cardiovascular risk factor as a predictor (systolic or diastolic blood pressure, BMI, waist circumference, waist-hip ratio, total cholesterol, HDL-c, LDL-c, triglycerides, fasting glucose, tobacco use, tobacco exposure, or ASCVD 10-year score), along with covariates for age, sex, years of education (or alternately reading level), APOE genotype (ɛ4 carrier versus non-carrier), diabetes, and antihypertensive medication use, with mean cerebral perfusion as the outcome variable. Any model using tobacco exposure only included former or current smokers (never smokers were excluded), in order to focus in on differing exposure amounts without the large point mass of no smoking history, which is approximately two-thirds of the subjects. To help address the multiple comparisons, the Benjamini-Hochberg procedure was applied to risk factor results, at 5% false discovery rate control (FDR), to reduce the potential for Type 1 error.

A separate procedure compared the relative importance of 11 of the 13 vascular risk factors (fasting glucose and tobacco exposure were not included in models due to smaller sample sizes) through a Bayesian Information Criterion (BIC) backwards selection procedure. The starting model contained these 11 risk factors along with the other 6 covariates listed above (e.g., age, sex); covariate reduction was restricted to the set of 11 risk factors.

To test the hypothesis of APOE genotype moderating relationships between risk factors and cerebral perfusion (Aim 2), identical multiple linear regression analyses were conducted with the addition of an interaction term between APOE ɛ4 carrier status (ɛ4 carrier versus non-carrier) and the vascular risk factor of that model. To assess race-associated risk and potentially unique relationships between vascular risk factors and perfusion within African American and White samples (Aim 3), race was added as a covariate to Aim 1 models, and the new models were fit for the pooled sample. Then, the pooled sample was stratified by race and race-specific Aim 1 models were fit. To test for a difference in risk factor associations by racial group, the risk factor coefficients in race-specific models were compared and the difference quantified. Inference on this difference was conducted using non-parametric bootstrapping, sampling subjects with replacement within each stratum separately before refitting stratified models and comparing the risk factor coefficients. 2,000 iterations were performed for each difference analysis and the bootstrap percentile method was used to construct 95% confidence intervals around the difference estimates. Including a race-by-APOE interaction was also considered for inclusion when models included the race covariate. However, the interaction was non-significant in all but one of the risk factor models, and since the fundamental interpretation of this interaction is the same for all models, we chose not to include this interaction in any of the models presented here.

Regression diagnostics were performed for all models. Moderate non-normal residuals were observed for all models (i.e., Q-Q plots of residuals had right skew). Cook’s distance yielded no concerns of influential points and residuals analysis did not indicate outliers. A moderate concern for increasing variance in models using fasting glucose were observed. Standard transformations of the response were examined, but these did not improve the above concerns; as such untransformed variables were used for all analyses. Raw relationships between perfusion and each risk factor were examined, and no major concerns for non-linearity were noted.

RESULTS

Sample characteristics

Table 1 summarizes cardiovascular risk factor prevalence across the whole sample and within racial groups. Participants were mean age 61.9 (8.7) years old, more likely to be female, completed post-secondary education, and about a third were APOE ɛ4 carriers. About 14% of the sample self-reported diabetes and about 30% were taking an antihypertensive medication. The sample included 17% (n = 67) African Americans and 83% (n = 330) Whites. No differences were observed between African Americans and Whites on age, sex, years of education, or APOE genotype.

Table 1

Sample characteristics

		Stratified by race
	Total sample n = 397	White n = 330	African American n = 67	p*
Age, m (SD)	61.87 (8.67)	61.59 (8.74)	63.23 (8.28)	0.12
Female Sex, n (%)	265 (66.8)	219 (66.4)	46 (68.7)	0.777
Education, y, m (SD)	16.12 (2.59)	16.17 (2.59)	15.88 (2.56)	0.367
NIH ToolBox Oral Reading, m (SD)	114.43 (14.57)	117.23 (12.06)	99.19 (17.50)	<0.001
Years between MRI &study visit, m (SD)	0.25 (0.17)	0.25 (0.17)	0.23 (0.16)	0.084
Perfusion, mL/100 g/min, m (SD)	50.91 (12.16)	52.01 (11.87)	45.52 (12.24)	<0.001
APOEɛ4 carrier, n (%)	136 (34.3)	118 (35.8)	18 (26.9)	0.204
Antihypertensive medication use, n (%)	121 (30.6)	81 (24.5)	40 (61.5)	<0.001
Presence of diabetes, n (%)	56 (14.1)	31 (9.4)	25 (37.3)	<0.001
ASCVD 10-year risk score, m (SD)	10.87 (12.94)	9.56 (12.80)	17.62 (11.61)	<0.001
Systolic blood pressure, m (SD)	128.22 (17.43)	125.83 (15.79)	139.96 (20.31)	<0.001
Diastolic blood pressure, m (SD)	76.61 (9.10)	75.19 (8.60)	83.61 (8.31)	<0.001
Waist-to-Hip Ratio, m (SD)	0.91 (0.09)	0.90 (0.09)	0.94 (0.08)	<0.001
Waist circumference, cm, m (SD)	96.80 (15.56)	94.90 (14.97)	106.20 (15.10)	<0.001
Body mass index, m (SD)	29.21 (6.00)	28.43 (5.65)	33.09 (6.24)	<0.001
LDL-c, m (SD)	113.42 (33.32)	114.59 (32.20)	107.45 (38.27)	0.226
HDL-c, m (SD)	62.24 (18.93)	63.17 (18.50)	57.51 (20.50)	0.024
Total cholesterol, m (SD)	196.53 (38.05)	198.75 (37.25)	185.29 (40.31)	0.018
Triglycerides, m (SD)	111.51 (88.21)	110.18 (89.82)	118.18 (79.86)	0.444
Glucose, m (SD)	100.39 (29.29)	97.30 (22.23)	113.65 (47.27)	0.026
Current or former smoker, n (%)	143 (36.0)	106 (32.1)	37 (55.2)	<0.001
Tobacco exposure, m (SD)^†	12.58 (26.03)	10.62 (24.51)	22.22 (30.94)	<0.001

m, mean; SD, standard deviation; NIH, National Institutes of Health; MRI, magnetic resonance imaging; APOE, apolipoprotein E; ASCVD, atherosclerotic cardiovascular disease; LDL-c, low-density lipoprotein cholesterol; HDL-c, high-density lipoprotein cholesterol. *p values resulted from independent t-tests (continuous) or chi-square (dichotomous) comparing racial groups, ^†Tobacco exposure = packs per day×number of years smoked for current or former smokers. Note: Missing data for the following variables: NIH Toolbox Reading (n = 50; 37 black, 13 white); Antihypertensive medication use (n = 2); ASCVD (n = 4); LDL-c (n = 6); HDL-c (n = 3); Total cholesterol (n = 3); Triglycerides (n = 4); Glucose (n = 111; 12 black, 99 white).

Relationships among vascular risk factors and mean perfusion

Regression results are displayed in Table 2. Age (all models), sex (92% of models), APOE genotype (46% of models), and diabetes (69% of models) were significantly associated with cerebral perfusion, whereas years of education and antihypertensive medication use were not significantly associated with perfusion.

Table 2

Model estimates: Relationships between risk factors and perfusion

Risk factor	N	Beta	SE	CI	p	Adjusted p*
ASCVD 10-year risk score	393	0.0647	0.0833	–0.0986 – 0.228	0.44	0.47
Systolic blood pressure	395	–0.0646	0.0354	–0.134 – 0.0047	0.068	0.13
Diastolic blood pressure	395	–0.215	0.0646	–0.341 – –0.0882	0.00096	0.013
Waist circumference	395	–0.0883	0.0412	–0.169 – –0.00757	0.033	0.071
Waist-to-Hip Ratio	395	–13.4	7.87	–28.9 – 1.99	0.089	0.14
Body Mass Index	395	–0.274	0.103	–0.476 – –0.0732	0.0078	0.051
LDL-c	390	–0.0403	0.0183	–0.0761 – –0.00453	0.028	0.071
HDL-c	393	0.0741	0.0346	0.00629 – 0.142	0.033	0.071
Total cholesterol	393	–0.0223	0.0165	–0.0547 – 0.0101	0.18	0.26
Triglycerides	392	–0.00554	0.00678	–0.0188 – 0.00776	0.41	0.47
Glucose	290	–0.0625	0.0276	–0.117 – –0.00844	0.024	0.071
Current or former smoker	395	0.753	1.23	–1.65 – 3.16	0.54	0.54
Tobacco exposure	142	0.0355	0.0309	–0.025 – 0.096	0.25	0.33

CI, confidence interval; ASCVD, atherosclerotic cardiovascular disease; LDL-c, low-density lipoprotein cholesterol; HDL-c, high-density lipoprotein cholesterol; FDR, false discovery rate. Note: Covariates of age, sex, years of education, APOE genotype, antihypertensive medication use, and diabetes were included in all models. Age (all models), sex (92% of models), APOE (46% of models), and diabetes (69% of models) were significantly associated with cerebral perfusion, whereas years of education or antihypertensive medication use were not significantly associated with perfusion. *Adjusted p, FDR-corrected p-value after applying Benjamini-Hochberg correction.

For uncorrected results, significant relationships were observed among mean perfusion and diastolic blood pressure (p < 0.001), waist circumference (p = 0.03), BMI (p = 0.008), LDL-c (p = 0.028), HDL-c (p = 0.033), and fasting glucose (p = 0.024). Relationships between mean perfusion and systolic blood pressure, total cholesterol, triglycerides, waist-hip ratio, tobacco use or exposure, or the composite vascular risk score (ASCVD) were non-significant (all p > 0.068). Adjustments to control the FDR over these 13 comparisons resulted in only one model retaining a statistically significant relationship with perfusion: diastolic blood pressure (FDR-corrected p = 0.013). The models including BMI, LDL-c, HDL-c, and fasting glucose were no longer significant following FDR correction.

The BIC reduction procedure retained only diastolic blood pressure out of the 11 risk factors. Supplementary Table 1 displays the results for the order that risk factors were removed from the model, with systolic blood pressure being removed first, and HDL cholesterol being removed last.

Relationships among APOE and vascular risk factors on mean perfusion

The interaction of APOE genotype×vascular risk factors was non-significant across all models (all uncorrected p > 0.19), indicating that the relationships between vascular risk factors and perfusion did not vary based on APOE carrier status. F-tests comparing models without APOE main effect or interaction to models with the main effect and interaction terms were also non-significant (all p > 0.05).

Relationships among race and vascular risk factors on mean perfusion

African Americans exhibited higher vascular risk values and lower perfusion compared to Whites (see Table 1). For example, African Americans had higher self-reported diabetes (37% compared to 9%), antihypertensive medication use (62% compared to 25%), and tobacco use (55% compared to 32%), as well as elevated risk factors observed on physical exam (e.g., higher blood pressure, measures of obesity, cholesterol, fasting glucose) and overall 10-year risk for cardiovascular event (18% versus 10%).

When race was added to models, the main effect of race was significantly associated with perfusion in all models (all p < 0.04), indicating that African Americans exhibited lower perfusion than Whites after adjustment for covariates. Furthermore, with race included in the models, the cardiovascular risk factors were no longer significantly related to perfusion, with all FDR-corrected p values >0.1; diastolic blood pressure, BMI, LDL-c, HDL-c, and fasting glucose all had significant uncorrected p values (see Table 3). In separate models fitted for African American and White subgroups, there were no significant relationships between any vascular risk factor and perfusion after FDR-correction. In uncorrected results, diastolic blood pressure, LDL-c, tobacco use, and tobacco exposure were associated with perfusion in Whites; no risk factors were associated with perfusion in African Americans (see Table 4). The raw, unadjusted relationships between diastolic blood pressure and cerebral perfusion among Whites and African Americans is depicted in Fig. 1, with a Generalized Additive Model trend to account for any raw non-linearity. Despite between-group variation in statistical significance, bootstrapping procedures revealed no evidence for racial group differences in relationships between the vascular risk factors and perfusion. The confidence intervals for race-specific model estimate differences all contained zero.

Table 3

Model estimates: Relationships between risk factors and perfusion when race is included as a covariate

Risk factor	N	Beta	SE	CI	p	Adjusted p*
ASCVD 10-year risk score	393	0.0927	0.0829	–0.0698 – 0.255	0.26	0.31
Systolic blood pressure	395	–0.0406	0.0362	–0.112 – 0.0304	0.26	0.31
Diastolic blood pressure	395	–0.168	0.0681	–0.302 – –0.035	0.01	0.13
Waist circumference	395	–0.0696	0.0414	–0.151 – 0.0117	0.09	0.20
Waist-to-Hip Ratio	395	–10.4	7.88	–25.8 – 5.08	0.19	0.27
Body mass index	395	–0.225	0.104	–0.428 – –0.0221	0.03	0.13
LDL-c	390	–0.0389	0.0181	–0.0743 – –0.00341	0.03	0.13
HDL-c	393	0.0701	0.0343	0.00285 – 0.137	0.04	0.13
Total cholesterol	393	–0.0233	0.0164	–0.0554 – 0.0088	0.16	0.25
Triglycerides	392	–0.00588	0.00671	–0.019 – 0.00728	0.38	0.38
Glucose	290	–0.0539	0.0274	–0.108 – –0.000289	0.05	0.13
Current or former smoker	395	1.28	1.22	–1.12 – 3.68	0.30	0.32
Tobacco exposure	142	0.0432	0.0289	–0.0134 – 0.0997	0.14	0.25

CI, confidence interval; ASCVD, atherosclerotic cardiovascular disease; LDL-c, low-density lipoprotein cholesterol; HDL-c, high-density lipoprotein cholesterol; FDR, false discovery rate. Note: Covariates of race, age, sex, years of education, APOE genotype, antihypertensive medication use, and diabetes were included in all models. Age, race, sex, APOE, and diabetes were significantly associated with cerebral perfusion, whereas years of education or antihypertensive medication use were not significantly associated with perfusion. *Adjusted p, FDR-corrected p-value after applying Benjamini-Hochberg correction.

Table 4

Model estimates of relationships between vascular risk factors and perfusion stratified by race

	African American						White
Risk factor	N	Beta	SE	CI	p	Adj p*	N	Beta	SE	CI	p	Adj p*
ASCVD	64	–0.205	0.261	–0.716 – 0.306	0.436	0.891	329	0.143	0.0874	–0.0282 – 0.314	0.103	0.185
Systolic BP	65	–0.0403	0.0865	–0.21 – 0.129	0.643	0.891	330	–0.0388	0.0411	–0.119 – 0.0417	0.345	0.374
Diastolic BP	65	–0.129	0.203	–0.528 – 0.269	0.527	0.891	330	–0.159	0.0725	–0.302 – –0.0174	0.0285	0.133
Waist cir	65	–0.0325	0.114	–0.255 – 0.19	0.776	0.917	330	–0.0584	0.0453	–0.147 – 0.0304	0.198	0.287
Waist-to-Hip Ratio	65	–16.5	27	–69.4 – 36.4	0.543	0.891	330	–8.13	8.21	–24.2 – 7.97	0.323	0.374
Body Mass Index	65	–0.11	0.269	–0.638 – 0.418	0.686	0.891	330	–0.213	0.114	–0.437 – 0.0105	0.0627	0.163
LDL-c	63	–0.0234	0.0465	–0.115 – 0.0678	0.617	0.891	327	–0.0407	0.0197	–0.0793 – –0.00206	0.0398	0.133
HDL-c	64	0.128	0.0858	–0.0404 – 0.296	0.142	0.891	329	0.0463	0.038	–0.0282 – 0.121	0.224	0.291
Total cholesterol	64	0.00579	0.0447	–0.0818 – 0.0934	0.897	0.972	329	–0.0294	0.0175	–0.0637 – 0.00491	0.094	0.185
Triglycerides	64	–0.0272	0.0226	–0.0714 – 0.017	0.233	0.891	328	–0.00322	0.00696	–0.0169 – 0.0104	0.644	0.644
Glucose	54	–0.0332	0.0477	–0.127 – 0.0604	0.49	0.891	236	–0.0593	0.0374	–0.133 – 0.014	0.114	0.185
Smoker	65	–3.11	3.31	–9.6 – 3.38	0.352	0.891	330	2.71	1.32	0.122 – 5.29	0.0409	0.133
Tobacco exposure	65	0.000409	0.0545	–0.106 — 0.107	0.994	0.994	330	0.0652	0.0255	0.0153 – 0.115	0.0109	0.133

CI, confidence interval; Adj P, Adjusted P value; ASCVD, atherosclerotic cardiovascular disease 10-year risk score; BP, blood pressure; Waist cir, Waist circumference; LDL-c, low-density lipoprotein cholesterol; HDL-c, high-density lipoprotein cholesterol; Smoker, current or former smoker; FDR, false discovery rate. *Adjusted p, FDR-corrected p-value after applying Benjamini-Hochberg correction. Note: Covariates of age, sex, years of education, APOE genotype, antihypertensive medication use, and diabetes were included in all models.

Fig.1

Relationships between blood pressure (x-axis) and cerebral perfusion (mL/100 g/min; y-axis) in African American (blue) and White (red) participants, with Generalized Additive Model trend, 95% confidence intervals, and raw data points.

A post-hoc analysis examined if results varied with substitution of reading level for years of education as a covariate in the models. Despite similar educational attainment in years, reading levels were lower in African American than in White participants. In these models, reading performance did not account for a significant amount of variance in perfusion in the majority of the original models (p values ranged from 0.22–0.78) nor in the models that included the main effect of race (p values ranged from 0.51–0.99), even though the race main effect was still consistently significant (12/13 models, uncorrected). These results indicate that substitution of a measure of reading level for years of education failed to account for racial differences in perfusion in this sample.

Sensitivity analyses

Due to change in MR hardware during the study, 68% of the sample (n = 268) had data acquired with an 8-channel MRI head coil and 32% of the sample (n = 129) with a newer 32-channel head coil. Sensitivity analyses were conducted to assess whether similar results were observed in the 8-channel and 32-channel samples. Results of these sensitivity analyses showed that diastolic blood pressure exhibited a statistically significant relationship with perfusion in the sample scanned with the 32-channel head coil (uncorrected p = 0.001, FDR-corrected p = 0.017); whereas, this relationship was non-significant in the 8-channel head coil sample (uncorrected p = 0.220, FDR-corrected p = 0.466). No other risk factors exhibited significant relationships with perfusion after FDR-correction in the 8- or 32-channel coil samples (See Supplementary Table 2).

Race group differed by head coil (32-channel coil was used on n = 86 [26.1%] of the White sample compared with n = 43 [64.2%] of the African American sample). When both head coil type and race were included in the models, the main effects of both race (p values <0.05) and coil type (p values <0.002) were significantly associated with perfusion in models for ASCVD score, total cholesterol, and tobacco use. The main effect of head coil (p values ≤0.003), but not race, was significantly associated with perfusion in models for blood pressure (systolic and diastolic), waist circumference, waist hip ratio, BMI, LDL-c, HDL-c, triglycerides, and glucose. Diastolic blood pressure and BMI were the only risk factors that remained significantly associated with perfusion (uncorrected) when both race and head coil were included in models (Diastolic blood pressure: β= –0.13, p = 0.050; BMI: β= –0.22, p = 0.034); no risk factors remained significant after FDR correction. These results indicate that head coil type significantly associates with the outcome measure of perfusion, but that diastolic blood pressure and BMI exhibited significant relationships with perfusion after accounting for the effect of coil type.

DISCUSSION

In a well-characterized sample of 397 middle-aged and older African Americans and Whites, we observed relationships between cerebral perfusion and diastolic blood pressure, obesity (waist circumference and BMI), cholesterol (LDL and HDL), and fasting glucose. After correction for multiple comparisons, only diastolic blood pressure was significantly associated with perfusion, and there was no evidence for relationship heterogeneity by APOE ɛ4 status. Although African Americans exhibited higher vascular risk factors and lower perfusion rates, relationships between risk factors and perfusion did not significantly differ between these subgroups.

The current findings on elevated diastolic blood pressure and perfusion are consistent with prior studies that report relationships between high blood pressure and lower perfusion in cognitively healthy elderly samples [20 , 32–34]. In midlife, higher blood pressure has been found to be associated with lower perfusion in an AD mouse model [35] and with longitudinal decline in perfusion in a middle-aged adult sample [36]. Typically, cerebral autoregulation helps to maintain a constant level of cerebral perfusion with fluctuations of blood pressure through adjustment of cerebral arteriole resistance [37]. However, with long-term hypertension or frequent fluctuations in blood pressure, arteriosclerosis may affect the cerebral autoregulation resulting in hypoperfusion. Maintaining adequate perfusion is essential for providing oxygen and nutrients to brain tissue and neurons; reduced perfusion may increase risk for dementia and dementia due to AD by inducing early neuronal dysfunction and increasing production of Aβ from the AβPP [12]. We did not observe relationships between genetic risk for AD and perfusion, but prior studies in a large racially diverse cohort (Atherosclerosis Risk in Communities; ARIC) found that elevated arterial stiffness was associated with increased cerebrovascular and AD pathology (e.g., Aβ) [5] suggesting possible connection between vascular disease and development of AD pathophysiology.

Vascular risk factors were more prevalent among the African Americans in this sample compared to their White counterparts, as seen in previous reports [9]. However, a composite vascular risk score (ASCVD 10-year risk estimate) was not associated with perfusion, which differs from our prior findings in a smaller, predominately White sample that was otherwise similar to the current participant group [24]. These results echo prior findings that Framingham stroke risk composite score significantly predicted perfusion in a primary sample but not within a more diverse validation sample [38] and the current findings are consistent with another study that did not find a relationship between Framingham scores and perfusion [39]. Composite scores may be useful in gauging overall vascular risk factor load, but may obscure the salience of individual cardiovascular markers for cerebrovascular health, particularly in racially diverse populations.

We observed that African American race predicted lower rates of cerebral perfusion independent of vascular risk load. A small body of work suggests that African Americans may exhibit lower cerebral perfusion prior to onset of other vascular symptomatology, even during young adulthood [40]. Given the strong negative relationship between age and perfusion reported here and elsewhere, our finding that African American race is an independent predictor of lower perfusion may also be consistent with studies suggesting that African Americans experience the onset of age-related morbidity earlier than their White counterparts. The “weathering hypothesis” developed by Geronimus and colleagues [41] is strongly implicated in cardiovascular health disparities [42] and suggests that premature deterioration is a consequence of physiological dysregulation provoked by accumulated exposure to social disadvantage over the life course. In support of this hypothesis, Hackman and colleagues recently demonstrated that life course socioeconomic position was positively associated with global and regional cerebral perfusion [43]. Relatedly, within African American cohorts, measures of adult literacy have been utilized as a proxy for quality of education and other influential early life exposures, and are a powerful explanatory factor for racial disparities in later-life cognitive function [30]. In this sample, we did not find that observed racial differences in educational attainment or quality (e.g., reading level) associated with perfusion or accounted for the racial differences in perfusion. These results suggest the possibility that increased risk for AD in African Americans may be related to earlier disease secondary to end-organ damage (e.g., lower cerebral perfusion) associated with chronic, long-term vascular risk factors. However, there are important limitations that can impact the race-specific results from this study. First, hardware change during data collection for this study (e.g., 8-channel to 32-channel head coil) was unfortunate as differences in signal-to-noise ratio between coils increased variability in our perfusion estimates. The relatively high collinearity between coil type and race (greater proportion of data from African Americans were collected using 32-channel coil) introduces a degree of uncertainty to conclusions about racial differences in perfusion. Secondly, there is recent evidence that hematocrit differences by sex and race/ethnicity can influence the quantification of perfusion estimates from ASL [44]. Unfortunately, we did not measure individual hematocrit or relaxation time of blood (T1_blood) in this sample and therefore used a fixed T1_blood estimate [18] in our perfusion quantification. Third, while our African American sample is growing, relatively small numbers still limit our power to detect significant relationships within this group and examine within-race social and vascular health gradients. It will be important to replicate these findings in larger samples and explore mechanisms underlying perfusion differences.

Finally, the current study’s sample characteristics and design constrain generalizability and causal inference. The sample is enriched for AD risk, and sampling on a family history of AD may impact generalizability despite our findings that the most well-established heritable risk factor for AD, APOE ɛ4, is not a key moderator of relationships between risk factors and perfusion. We also were not able to assess the role for timing as a modifier of relationships between vascular risk and perfusion. There is evidence that midlife may be a sensitive period for risk [45], while some markers such as hypertension and elevated diastolic pressure may actually be protective in late life, and future studies should continue integrating a life course approach to these questions [46]. Finally, there are constraints on our statistical conclusions. Our necessary adjustment for multiple comparisons increased the probability that associations between risk factors and perfusion may not have been detected. Specific processing procedures and model selection choices (e.g., examination of mean instead of regional perfusion, no partial volume correction due to lack of significant atrophy expected) are important to consider when comparing these results to other studies. Finally, this study was cross-sectional and included vascular risk factor data from just one point in time (e.g., visit closest to MRI scan); therefore, these results do not confirm hypotheses about associations between longitudinal trajectories of vascular risk factors and perfusion which will need to be investigated in future studies. Future work must also explore relationships between vascular risk factors and additional neuroimaging markers of cerebrovascular health (e.g., white matter hyperintensities, ASL spatial coefficient of variation), as well as cognitive decline and impairment.

Despite limitations, this study contributes to a growing body of literature exploring systemic vascular risk factors as markers of cerebrovascular health, and suggests elevated diastolic blood pressure may be an important prevention target for maintaining cerebrovascular health in middle-aged and older adults. Inconsistent findings between cohorts and within-cohort samples point to the importance of considering data collection and processing strategies as well as sample demographic distributions, particularly age and race, in interpretation of results. Our findings highlight the need to prioritize inclusion of African Americans and other underrepresented populations within cognitive aging study cohorts, and to continue exploring the link between potentially population-specific modifiable vascular risk factors, cerebrovascular health, and AD and other dementia risk.

Footnotes

ACKNOWLEDGMENTS

The authors gratefully acknowledge the assistance of researchers and staff at the Wisconsin Alzheimer’s Disease Research Center for assistance in recruitment and data collection. We would like to acknowledge Fabu Carter, MS and Brieanna Harris, BA for their assistance in recruitment of under-represented populations into the study. We would like to thank Kevin Johnson, PhD, Heather Schoul, BS and Jennifer Oh, BS for assistance in data acquisition, processing, and database organization. We also thank Jamie LaMantia, BS for manuscript formatting. Most importantly, we thank the dedicated Wisconsin ADRC participants for their continued support and participation in research. Research reported in this manuscript was supported by the National Institutes of Health, NIH-NIA grant number P50-AG033514, a Wisconsin Alzheimer’s Disease Research Center pilot award (HMJ, LRC), NIH-NIA R01 AG054059 (CEG), Alzheimer’s Association Research Fellowship award AARF-18-562958 (MZ), and the Veterans Affairs Advanced Fellowship in Women’s Health (MFW). The authors would like to acknowledge the clinical and neuropathology diagnostic support provided by the Wisconsin ADRC’s Clinical, Neuropathology and Biomarkers Cores, and biostatistical support provided by the Data Management and Biostatistics Core. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. A portion of the findings described were presented by Lindsay Clark at the Alzheimer’s Association International Conference in Chicago on July 22, 2018.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Kochanek

, Murphy

, Xu

, Arias

(2017) Mortality in the United States, 2016. NCHS Data Brief, 1–8.

Kapasi

, DeCarli

, Schneider

(2017) Impact of multiple pathologies on the threshold for clinically overt dementia. Acta Neuropathol 134, 171–186.

Kivipelto

, Ngandu

, Laatikainen

, Winblad

, Soininen

, Tuomilehto

(2006) Risk score for the prediction of dementia risk in 20 years among middle aged people: a longitudinal, population-based study. Lancet Neurol 5, 735–741.

Gottesman

, Albert

, Alonso

, Coker

, Coresh

, Davis

, Deal

, McKhann

, Mosley

, Sharrett

, Schneider

ALC

, Windham

, Wruck

, Knopman

(2017) Associations between midlife vascular risk factors and 25-year incident dementia in the Atherosclerosis Risk in Communities (ARIC) Cohort. JAMA Neurol 74, 1246–1254.

Hughes

, Wagenknecht

, Craft

, Mintz

, Heiss

, Palta

, Wong

, Zhou

, Knopman

, Mosley

, Gottesman

(2018) Arterial stiffness and dementia pathology: Atherosclerosis Risk in Communities (ARIC)-PET Study. Neurology 90, e1248–e1256.

Gorelick

, Scuteri

, Black

, Decarli

, Greenberg

, Iadecola

, Launer

, Laurent

, Lopez

, Nyenhuis

, Petersen

, Schneider

, Tzourio

, Arnett

, Bennett

, Chui

, Higashida

, Lindquist

, Nilsson

, Roman

, Sellke

, Seshadri

; American Heart Association Stroke Council, Council on Epidemiology and Prevention, Council on Cardiovascular Nursing, Council on Cardiovascular Radiology and Intervention, and Council on Cardiovascular Surgery and Anesthesia (2011) Vascular contributions to cognitive impairment and dementia: a statement for healthcare professionals from the American Heart Association/American Stroke Association. Stroke 42, 2672–2713.

Zissimopoulos

, Crimmins

, St Clair

(2014) The value of delaying Alzheimer’s disease onset. Forum Health Econ Policy 18, 25–39.

Husaini

, Sherkat

, Moonis

, Levine

, Holzer

, Cain

(2003) Racial differences in the diagnosis of dementia and in its effects on the use and costs of health care services. Psychiatr Serv 54, 92–96.

Bravata

, Wells

, Gulanski

, Kernan

, Brass

, Long

, Concato

(2005) Racial disparities in stroke risk factors: the impact of socioeconomic status. Stroke 36, 1507–1511.

10.

Marden

, Tchetgen Tchetgen

, Kawachi

, Glymour

(2017) Contribution of socioeconomic status at 3 life-course periods to late-life memory function and decline: early and late predictors of dementia risk. Am J Epidemiol 186, 805–814.

11.

McHutchison

, Backhouse

, Cvoro

, Shenkin

, Wardlaw

(2017) Education, socioeconomic status, and intelligence in childhood and stroke risk in later life: a meta-analysis. Epidemiology 28, 608–618.

12.

Zlokovic

(2011) Neurovascular pathways to neurodegeneration in Alzheimer’s disease and other disorders. Nat Rev Neurosci 12, 723–738.

13.

Wierenga

, Clark

, Dev

, Shin

, Jurick

, Rissman

, Liu

, Bondi

(2013) Interaction of age and APOE genotype on cerebral blood flow at rest. J Alzheimers Dis 34, 921–935.

14.

Fleisher

, Podraza

, Bangen

, Taylor

, Sherzai

, Sidhar

, Liu

, Dale

, Buxton

(2009) Cerebral perfusion and oxygenation differences in Alzheimer’s disease risk. Neurobiol Aging 30, 1737–1748.

15.

Thambisetty

, Beason-Held

, An

, Kraut

, Resnick

(2010) APOE epsilon4 genotype and longitudinal changes in cerebral blood flow in normal aging. Arch Neurol 67, 93–98.

16.

Okonkwo

, Xu

, Oh

, Dowling

, Carlsson

, Gallagher

, Birdsill

, Palotti

, Wharton

, Hermann

, LaRue

, Bendlin

, Rowley

, Asthana

, Sager

, Johnson

(2014) Cerebral blood flow is diminished in asymptomatic middle-aged adults with maternal history of Alzheimer’s disease. Cereb Cortex 24, 978–988.

17.

ElAli

, Theriault

, Prefontaine

, Rivest

(2013) Mild chronic cerebral hypoperfusion induces neurovascular dysfunction, triggering peripheral beta-amyloid brain entry and aggregation. Acta Neuropathol Commun 1, 75.

18.

Alsop

, Detre

, Golay

, Gunther

, Hendrikse

, Hernandez-Garcia

, Lu

, MacIntosh

, Parkes

, Smits

, van Osch

, Wang

, Wong

, Zaharchuk

(2015) Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: A consensus of the ISMRM perfusion study group and the European consortium for ASL in dementia. Magn Reson Med 73, 102–116.

19.

Rusinek

, Ha

, Yau

, Storey

, Tirsi

, Tsui

, Frosch

, Azova

, Convit

(2015) Cerebral perfusion in insulin resistance and type 2 diabetes. J Cerebr Blood Flow Metab 35, 95–102.

20.

Beason-Held

, Moghekar

, Zonderman

, Kraut

, Resnick

(2007) Longitudinal changes in cerebral blood flow in the older hypertensive brain. Stroke 38, 1766–1773.

21.

Launer

, Lewis

, Schreiner

, Sidney

, Battapady

, Jacobs

, Lim

, D’Esposito

, Zhang

, Reis

, Davatzikos

, Bryan

(2015) Vascular factors and multiple measures of early brain health: CARDIA brain MRI study. PloS One 10, e0122138.

22.

Goff

Jr , Lloyd-Jones

, Bennett

, Coady

, D’Agostino

, Gibbons

, Greenland

, Lackland

, Levy

, O’Donnell

, Robinson

, Schwartz

, Shero

, Smith

Jr , Sorlie

, Stone

, Wilson

, Jordan

, Nevo

, Wnek

, Anderson

, Halperin

, Albert

, Bozkurt

, Brindis

, Curtis

, DeMets

, Hochman

, Kovacs

, Ohman

, Pressler

, Sellke

, Shen

, Smith

Jr , Tomaselli

(2014) 2013 ACC/AHA guideline on the assessment of cardiovascular risk: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. Circulation 129, S49–73.

23.

Birdsill

, Carlsson

, Willette

, Okonkwo

, Johnson

, Xu

, Oh

, Gallagher

, Koscik

, Jonaitis

, Hermann

, LaRue

, Rowley

, Asthana

, Sager

, Bendlin

(2013) Low cerebral blood flow is associated with lower memory function in metabolic syndrome. Obesity (Silver Spring) 21, 1313–1320.

24.

Clark

, Berman

, Rivera-Rivera

, Hoscheidt

, Darst

, Engelman

, Rowley

, Carlsson

, Asthana

, Turski

, Wieben

, Johnson

(2017) Macrovascular and microvascular cerebral blood flow in adults at risk for Alzheimer’s disease. Alzheimers Dement (Amst) 7, 48–55.

25.

Hoscheidt

, Kellawan

, Berman

, Rivera-Rivera

, Krause

, Oh

, Beeri

, Rowley

, Wieben

, Carlsson

, Asthana

, Johnson

, Schrage

, Bendlin

(2017) Insulin resistance is associated with lower arterial blood flow and reduced cortical perfusion in cognitively asymptomatic middle-aged adults. J Cereb Blood Flow Metab 37, 2249–2261.

26.

Albert

, DeKosky

, Dickson

, Dubois

, Feldman

, Fox

, Gamst

, Holtzman

, Jagust

, Petersen

, Snyder

, Carrillo

, Thies

, Phelps

(2011) The diagnosis of mild cognitive impairment due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 270–279.

27.

McKhann

, Knopman

, Chertkow

, Hyman

, Jack

, Kawas

, Klunk

, Koroshetz

, Manly

, Mayeux

, Mohs

, Morris

, Rossor

, Scheltens

, Carrillo

, Thies

, Weintraub

, Phelps

(2011) The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 263–269.

28.

, Rowley

, Wu

, Alsop

, Shankaranarayanan

, Dowling

, Christain

, Oakes

, Johnson

(2010) Reliability and precision of pseudo-continuous arterial spin labeling perfusion MRI on 3.0 T and comparison with 150-water PET in elderly subjects at risk for Alzheimer’s disease. NMR Biomed 23, 286–293.

29.

Gershon

, Cook

, Mungas

, Manly

, Slotkin

, Beaumont

, Weintraub

(2014) Language measures of the NIH Toolbox Cognition Battery. J Int Neuropsychol Soc 20, 642–651.

30.

Sisco

, Gross

, Shih

, Sachs

, Glymour

, Bangen

, Benitez

, Skinner

, Schneider

, Manly

(2015) The role of early-life educational quality and literacy in explaining racial disparities in cognition in late life. J Gerontol B Psychol Sci Soc Sci 70, 557–567.

31.

Team RC (2018) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

32.

Claus

, Breteler

, Hasan

, Krenning

, Bots

, Grobbee

, van Swieten

, van Harskamp

, Hofman

(1996) Vascular risk factors, atherosclerosis, cerebral white matter lesions and cerebral perfusion in a population-based study. Eur J Nucl Med 23, 675–682.

33.

Waldstein

, Lefkowitz

, Siegel

, Rosenberger

, Spencer

, Tankard

, Manukyan

, Gerber

, Katzel

(2010) Reduced cerebral blood flow in older men with higher levels of blood pressure. J Hypertens 28, 993–998.

34.

Beason-Held

, Thambisetty

, Deib

, Sojkova

, Landman

, Zonderman

, Ferrucci

, Kraut

, Resnick

(2012) Baseline cardiovascular risk predicts subsequent changes in resting brain function. Stroke 43, 1542–1547.

35.

Wiesmann

, Capone

, Zerbi

, Mellendijk

, Heerschap

, Claassen

, Kiliaan

(2015) Hypertension impairs cerebral blood flow in a mouse model for Alzheimer’s disease. Curr Alzheimer Res 12, 914–922.

36.

Muller

, van der Graaf

, Visseren

, Mali

, Geerlings

(2012) Hypertension and longitudinal changes in cerebral blood flow: the SMART-MR study. Ann Neurol 71, 825–833.

37.

Van Beek

, Claassen

, Rikkert

, Jansen

(2008) Cerebral autoregulation: an overview of current concepts and methodology with special focus on the elderly. J Cereb Blood Flow Metab 28, 1071–1085.

38.

Jennings

, Heim

, Kuan

DC-H

, Gianaros

, Muldoon

, Manuck

(2013) Utility of total cerebral blood flow as an imaging biomarker of known cardiovascular risks. Stroke 44, 2480–2485.

39.

Glodzik

, Rusinek

, Brys

, Tsui

, Switalski

, Mosconi

, Mistur

, Pirraglia

, de Santi

, Li

, Goldowsky

, de Leon

(2011) Framingham cardiovascular risk profile correlates with impaired hippocampal and cortical vasoreactivity to hypercapnia. J Cereb Blood Flow Metab 31, 671–679.

40.

Hurr

, Kim

, Harrison

, Brothers

(2015) Attenuated cerebral vasodilatory capacity in response to hypercapnia in college-aged African Americans. Exp Physiol 100, 35–43.

41.

Geronimus

(1996) Black/white differences in the relationship of maternal age to birthweight: a population-based test of the weathering hypothesis. Soc Sci Med 42, 589–597.

42.

Thorpe

Jr , Fesahazion

, Parker

, Wilder

, Rooks

, Bowie

, Bell

, Szanton

, LaVeist

(2016) Accelerated health declines among African Americans in the USA. J Urban Health 93, 808–819.

43.

Hackman

, Kuan

, Manuck

, Gianaros

(2018) Socioeconomic position and age-related disparities in regional cerebral blood flow within the prefrontal cortex. Psychosom Med 80, 336–344.

44.

Smith

, Melbourne

, Owen

, Cardoso

, Sudre

, Tillin

, Sokolska

, Atkinson

, Chaturvedi

, Ourselin

, Hughes

, Barkhof

, Jäger

(2019) Cortical cerebral blood flow in ageing: effects of haematocrit, sex, ethnicity and diabetes. Eur Radiol 29, 5549–5558.

45.

Gottesman

, Schneider

, Albert

, Alonso

, Bandeen-Roche

, Coker

, Coresh

, Knopman

, Power

, Rawlings

, Sharrett

, Wruck

, Mosley

(2014) Midlife hypertension and 20-year cognitive change: the atherosclerosis risk in communities neurocognitive study. JAMA Neurol 71, 1218–1227.

46.

Yano

, Griswold

, Wang

, Greenland

, Lloyd-Jones

, Heiss

, Gottesman

, Mosley

(2018) Long-term blood pressure level and variability from midlife to later life and subsequent cognitive change: The ARIC Neurocognitive Study. J Am Heart Assoc 7, e009578.