Carotid Intima–Media Thickness and Markers of Brain Health in a Biracial Middle-Aged Cohort: CARDIA Brain MRI Sub-study

Abstract

Background

We investigated whether carotid intima–media thickness is associated with measures of cerebral blood flow (CBF), white matter hyperintensities, and brain volume in a biracial cohort of middle-aged individuals.

Methods

We performed a cross-sectional cohort study based on data from a multicenter, population-based study Coronary Artery Risk Development in Young Adults. Using linear and logistic regression, we estimated the association of the composite intima–media thickness measured in three segments of carotid arteries (common carotid artery, carotid artery bulb, and internal carotid artery) with volume (cm³) and CBF (mL/100 g/min) in the total brain and gray matter as well as volume of white matter hyperintensities (cm³).

Results

In the analysis, 461 participants (54% women, 34% African Americans) were included. Greater intima–media thickness was associated with lower CBF in gray matter (β=−1.36; p = .04) and total brain (β=−1.26; p = .04), adjusting for age, sex, race, education, and total brain volume. The associations became statistically nonsignificant after further controlling for cardiovascular risk factors. Intima–media thickness was not associated with volumes of total brain, gray matter, and white matter hyperintensities.

Conclusions

This study suggests that lower CBF in middle age is associated with markers of atherosclerosis in the carotid arteries. This association may reflect early long-term exposure to traditional cardiovascular risk factors. Early intervention on atherosclerotic risk factors may modulate the trajectory of CBF as people age and develop brain pathology.

The continuous increase in life expectancy is accompanied by rising prevalence of age-related conditions, such as dementia, which is a syndrome characterized by progressive decline in cognitive functions, resulting in patients’ dependency on caregivers (1). Early identification of individuals who are at risk for dementia is important to implement preventative strategies to decrease its burden (2). Dementia is on average diagnosed at the age of 80 years (3–5), but there is general agreement about its long preclinical period, when subtle pathological changes in the brain are present, but do not manifest with cognitive impairment (6). Several recent population-based studies indicated that individuals as young as 50 years old have already established functional or structural changes of the brain, such as white matter injury or decreased brain perfusion (7–9), which could indicate an incipient subtle vascular brain disease. This may be a consequence of previous long-term exposure to traditional cardiovascular risk factors (CVRFs), such as elevated levels of blood pressure, fasting plasma glucose, cholesterol, and body mass index (BMI) as well as sedentary behavior and smoking (7).

Markers of subtle injury of the white matter and lower cerebral blood flow (CBF) were found in middle-aged individuals from Coronary Artery Risk Development in Young Adults (CARDIA) study and the third generation of the Framingham Heart Study and were associated with hypertension, higher fasting blood glucose, obesity, smoking, and sedentary lifestyle (7–10). The pathophysiological foundation for the aforementioned observations could be atherosclerosis, which may reflect lifetime exposure to CVRFs (11). Atherosclerosis is associated with vascular brain disease, cognitive decline, and dementia in older adults (12,13). Through remodeling of vessels and narrowing of their diameters, increased oxidative stress and inflammation, atherosclerosis may contribute to reduced CBF, greater volume of white matter hyperintensities (WMH) and atrophy that eventually manifest with cognitive impairment (14).

Within the CARDIA study, we are well positioned to investigate whether these associations are already evident in midlife. Limited data from two well-described cohorts suggest that relatively poorer cognitive function is associated with markers of atherosclerosis, including calcified atherosclerotic plaques, which reflect advanced atherosclerotic processes (15), and carotid intima–media thickness (cIMT), which is a measure of early subclinical atherosclerosis (16). However, the mechanisms through which atherosclerosis influences brain health at this relatively young age have not been elucidated yet.

As carotid arteries form a major pathway for blood to flow into the brain, we investigated the association of cIMT to early markers of vascular brain disease, using state-of-the-art brain imaging acquired within the CARDIA Brain MRI Sub-study. Specifically, we aimed to study the relationship between cIMT and CBF, WMH and loss of brain tissue in a biracial cohort of middle-aged adults. Further, we sought to determine whether such associations are statistically explained by long-term exposure to traditional CVRFs, which may lie earlier in the causal pathway. As levels of atherosclerosis may vary by sex and race (17), we also assessed whether the associations of interest differ depending on these two factors.

Methods

Source of Data

The analytical sample is based on the CARDIA study, as previously described in detail (18). Briefly, CARDIA is a prospective, longitudinal study that aims to investigate the development of cardiovascular disease beginning in young adulthood. It was established in 1985 and takes place in four CARDIA field centers in the United States (Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota; and Oakland, California). The original cohort consisted of 5,115 persons and was balanced with respect to race (52% of African Americans), gender (55% of women), age 18–30 years and educational level (40% had < 12 years of education at baseline). All participants provided a written informed consent at each exam and institutional review boards from each center annually approve this study. A separate written consent for participation in the CARDIA Brain MRI Sub-study was obtained.

Carotid Intima–Media Thickness

Ultrasound measurement of cIMT was performed at the follow-up exam in year 20 according to a standard protocol followed across all four CARDIA sites, as described in detail elsewhere (19). Briefly, high-resolution images were obtained by certified and centrally trained sonographers, using GE-Logiq-700. The thickness of the tunica media and intima was measured on the left and right side of the neck in three segments of the carotid arteries: the common carotid artery (CCA), the carotid artery bulb, and the internal carotid artery (ICA). Average values were calculated from measurements at the near and far walls of the arteries.

A composite measure of cIMT was obtained from the three segments of carotid arteries, as follows: First, a combined measurement of cIMT in the bulb and ICA was formed. Second, the z scores of cIMT in CCA and the bulb-ICA were averaged, giving the final composite score of cIMT, which was used in the present study. This method was previously used (20) and gives a normalized composite measure that captures overall subclinical atherosclerosis of carotid arteries, giving the specific segments of carotid arteries the same weight. The composite measure has been previously shown to be a more powerful predictor of cardiovascular events than the segment-specific cIMT (21,22).

The Ultrasound Reading Center was located at Tufts University and was responsible for training of technicians, reviewing ultrasounds tapes from the four field centers, providing quality rating for each scan and reporting results to the CARDIA coordinating center. Technician performance was evaluated for 3% of measurements in the field center and was regularly reported to the principal investigators of each field center.

Brain MRI Measures

At the follow-up exam in year 25, 719 individuals participated in the CARDIA Brain MRI Sub-study, as previously described (7). Briefly, participants underwent examination of the brain by magnetic resonance imaging (MRI) at three CARDIA field centers (n = 297 in Minneapolis; n = 252 in Oakland, and n = 170 in Birmingham). MRI was acquired on 3-T MR scanners (Siemens 3T Tim Trio/VB 15 platform was used in Minneapolis and Oakland; Philips 3T Achieva/2.6.3.6 platform was used in Birmingham).

As described previously (7), the MRI Reading Center at the University of Pennsylvania trained technologists to standardized protocols, and MRI data were transferred to a central archive and image processing. All three MRI field centers followed quality assurance protocols, which were previously developed for the Functional Bioinformatics Research Network (FBIRN), and the Alzheimer’s disease Neuroimaging Initiative (ADNI). The following cutoffs were used to ensure quality: FBIRN—Siemens scanners SFNR >220, RDC >3.1, Philips scanners SFNR >220, RDC >2.4; ADNI—SNR >300, Maximum Distortion >2.0. Performance across the scanners was acceptable for all sequences except the CBF measurements acquired in Birmingham, thus data from this field center were excluded from the analysis.

Images were checked for motion artifacts and other factors that may influence image processing; the successful scans were then processed with an automated pipeline. Visual quality check of parameter distributions was done to identify outliers. An index of technical errors was estimated from scans of three persons measured three times in the three centers. This was 1.2% for total brain volume (TBV), 27.8% for WMH, and 7.3% for gray matter CBF.

The parameters of interest for this study were CBF (of gray matter and total brain) and volumes (of total brain, gray matter, and WMH). CBF was measured using pseudo-Continuous Arterial Spin Labeling (pCASL) technique. The mean perfusion (mL/100 g/min) was quantified for the gray matter and the total brain. The measurement of white matter CBF is not used in this analysis because of its low accuracy (23). The volumes (cm³) of normal tissue (gray matter and total brain) were estimated from the sagittal 3D T1 sequence. Intracranial volume (ICV) as a measure of head size was estimated as the sum of gray matter, white matter and cerebral spinal fluid volumes from the sagittal 3D T1 sequence. The volume of WMH was estimated from the sagittal 3D FLAIR, T1, and T2 sequences and contains tissue damaged because of ischemia, demyelination or inflammation, and penumbra surrounding brain infarcts.

Covariates

We selected seven traditional CVRFs based on literature search as factors that may reflect the atherosclerotic burden and are associated with markers of vascular brain diseases and dementia (9,24): systolic blood pressure, diastolic blood pressure, BMI, total cholesterol, fasting plasma glucose, smoking, and sedentary behavior. We do not adjust for cerebrovascular or cardiovascular events because there were only very few cases in this sample (9,16). Blood pressure was measured three times and the average of the second and third measurement of the systolic blood pressure and diastolic blood pressure was calculated at each visit. A Hawksley random-zero sphygmomanometer was used at years 0, 2, 5, 7, 10, and 15; a digital blood pressure monitor Omron HEM-907XL, calibrated to the sphygmomanometer, was used at years 20 and 25. BMI was calculated from the height and weight that were obtained at each exam. Fasting total cholesterol was measured at each exam, as described elsewhere (25). Fasting plasma glucose was measured at years 0, 7, 10, 15, 20, and 25, methods have been previously described (25).

Cumulative values of systolic blood pressure (mm Hg-years), diastolic blood pressure (mm Hg-years), BMI (kg/m²-years), total cholesterol (mg/dL-years) and fasting plasma glucose (mg/dL-years) were calculated as the area under the curve of the values by age in years until the follow-up year 20 (restricting to those who have participated in three or more exams beyond baseline), using the “trapezoidal rule.” The method has been previously described in detail (26); the values can be interpreted as the estimated mean daily value for the entire period. Sedentary time was assessed based on a questionnaire at year 25 and is expressed as hours per day. Smoking status is based on self-report at year 25 and is categorized as never, current, and former smoker. Other covariates include age, sex, race (Caucasian or African American), baseline CARDIA site and years of education reported at year 25.

Analytical Sample

From 719 participants in the CARDIA Brain MRI Sub-study, 170 participants from Birmingham were excluded because the CBF measurements could not be pooled with the data from the two other CARDIA field centers. From the 549 remaining individuals, 544 had successfully processed images. Further, we excluded 83 participants who did not have data on cIMT at year 20, giving a final analytical sample of 461 individuals. The final analytical sample consisted of fewer African Americans (34% vs 49%; p < .001) and had a lower burden CVRFs (systolic blood pressure p = .004; diastolic blood pressure p = .02; BMI p < .001; total cholesterol p = .05; fasting plasma glucose p = .001; Supplementary Table I), when compared to the rest of the cohort examined at year 25 (n = 3,038, including the fourth field center not included in the brain MRI Sub-study). Similar differences were found when the analytical sample was compared to other participants from the CARDIA Brain MRI Sub-study that were excluded from the analysis (n = 258; Supplementary Table II) and those with missing values on cIMT (n = 83; Supplementary Table III).

Statistical Analysis

Descriptive data of the analytical sample are presented as the mean ± standard deviation (SD), frequency (n; %), or median (25%; 75%), where appropriate. Linear regression models were used to examine the associations of the independent variable (cIMT) with the dependent variables (gray matter and total CBF as well as volumes of gray matter and total brain). As the volume of WMH was skewed (median value 0.3 cm³), the variable was categorized into three groups: no WMH (volume 0 cm³), little WMH volume (≤0.3 cm³), and high WMH volume (>0.3 cm³). Logistic regression was used to estimate the association of cIMT to high versus low WMH load.

Potential confounding factors were controlled for in several steps. The first set of models was adjusted for age, sex, race, years of education, and CARDIA site. Models with CBF as the dependent variable were controlled for the TBV, which reflects the total amount of perfused tissue that largely influences CBF measures (27). Controlling for TBV also controls for atrophy. Models with brain volume and WMH as dependent variables were adjusted for ICV (measure of head size). In the second set of models, the associations of interest were adjusted for seven traditional CVRFs as follows: first, each CVRF was entered into the model separately; second, all CVRFs were added into the final model. The fit of the models was assessed for homoscedasticity and normality of the error distribution. No multicollinearity was detected, using variance inflation factors less than 1 or more than 10 as indicators.

In addition, the interactions between cIMT and race or sex were tested by including cross-product terms of race or sex with cIMT in the first set of models. These models allowed the associations of cIMT with the brain outcome to vary depending on race and sex, as previous studies indicated that they could be effect modifiers (17). Sensitivity analyses examined the associations of cIMT assessed at three segments of carotid arteries (CCA, ICA, and bulb) with brain parameters of interest. Furthermore, the associations of cIMT with CBF were controlled for the ratio TBV to ICV. All tests were two-sided and p value less than .05 was taken to indicate statistical significance. We used Statistical Package for the Social Sciences software, version 22 (IBM Corporation, Armonk, NY).

Results

The analytical sample included 54% women and 34% African Americans (Table 1). They were on average 46 years old when cIMT was measured and 51 years old when the brain MRI was acquired. The mean cIMT was 0.71 mm (SD 0.15). Greater cIMT (per 1 SD) was significantly associated with lower gray matter CBF (β=−1.36; 95% CI −2.67 to −0.05) and total CBF (β=−1.26; 95% CI −2.44 to −0.08), when controlled for age, race, sex, years of education, CARDIA field center, and TBV (Table 2). Analyses stratified by race or sex did not suggest differences in direction or effect size between race or sex groups (p value for the interaction was .4/.7 for race and .5/.5 for sex with gray matter CBF/total CBF as the dependent variable; data not presented in Tables). There were no significant associations of cIMT with volumes of gray matter, total brain, or WMH.

Table 1.

Characteristics of the Study Sample: CARDIA Brain MRI Sub-study (n = 461)

Characteristic	Value
Age, y	50.6 ± 3.4
Women, n (%)	250 (54)
African Americans, n (%)	158 (34)
Education, y	15.2 ± 2.3
cIMT, mm	0.71 ± 0.15
cIMT, composite score	−0.06 ± 0.84
Brain characteristics
CBF, mL/100g/min
Gray matter	56.0 ± 11.2
Total brain	49.5 ± 10.3
Volume, cm3
Gray matter	519.3 ± 52.2
Total brain	983.7 ± 104.0
White matter hyperintensities, n (%)
None	96 (21)
Little volume (<0.3 cm3)	131 (29)
High volume (≥0.3 cm3)	230 (50)
Cardiovascular risk factors at year 25
Systolic blood pressure, mm Hg	116.9 ± 13.6
Diastolic blood pressure, mm Hg	72.8 ± 10.6
Body mass index	28.3 ± 5.4
Total cholesterol, mg/dL	195.3 ± 34.7
Fasting blood glucose, mg/dL	96.1 ± 21.9
Sedentary behavior, h/d	6.6 ± 3.8
Smoking, n (%)
Never	281 (61)
Former	113 (25)
Current	63 (14)

Notes: Values are mean ± standard deviation or frequency (%). CBF = cerebral blood flow; cIMT = carotid intima–media thickness.

Table 2.

Associations of cIMT With Brain Characteristics: CARDIA Brain MRI Sub-study

Brain Parameter	Effect Size
Cerebral blood flow†	β (95% CI)
Gray matter	−1.36 (−2.67 to −0.05)*
Total brain	−1.26 (−2.44 to −0.08)*
Brain volume‡
Gray matter	−0.92 (−3.35 to 1.50)
Total brain	−2.45 (−5.64 to 0.74)
White matter hyperintensities‡	OR (95% CI)
None	Reference
Little volume	1.18 (0.81 to 1.72)
High volume	1.40 (0.98 to 1.99)

Notes: β is a coefficient estimated from linear regression models for an association of cIMT (per SD) with measures of cerebral blood flow and brain volume. OR is estimated from multinomial logistic regression models for an association of cIMT (per SD) with a high and little volume of white matter hyperintensities, when compared to no volume. CI = confidence interval; cIMT = carotid intima–media thickness; OR = odds ratio; SD = standard deviation.

^†Adjusted for age, sex, race, CARDIA site, years of education and total brain volume.

^‡Adjusted for age, sex, race, CARDIA site, years of education and intracranial volume.

*p < .05.

The association of cIMT with the CBF measures became nonsignificant when we controlled for all CVRFs together (gray matter CBF: β=−0.77; 95% CI −2.21 to 0.67; total CBF: β=−0.83; 95% CI −2.13 to 0.47). When controlled separately, the variables that attenuated the associations the most were blood pressure, BMI and fasting blood glucose (Table 3). Specifically, systolic blood pressure and BMI diminished the associations of cIMT with gray matter CBF by 34%; diastolic blood pressure by 26% and fasting blood glucose by 21%. Similarly, systolic blood pressure attenuated the association of cIMT with total CBF by 31%, BMI by 29%, diastolic blood pressure by 24%, and fasting blood glucose by 18%. In sensitivity analyses, we did not find any significant associations to our MRI outcomes of any of the specific segments of carotid arteries (CCA, ICA, bulb; Supplementary Table IV). The association of cIMT with the CBF measures was similar in direction and effect size, when controlled for the ratio TBV to ICV (Supplementary Table V).

Table 3.

Associations of Carotid Intima–Media Thickness With CBF Adjusted for Cardiovascular Risk Factors: CARDIA Brain MRI Sub-study

Model	Gray Matter CBF	Total CBF
Carotid intima–media thickness adjusted for	β (95% CI)	β (95% CI)
Systolic blood pressure†	−0.90 (−2.28 to 0.48)	−0.87 (−2.12 to 0.38)
Diastolic blood pressure†	−1.01 (−2.33 to 0.32)	−0.95 (−2.15 to 0.25)
Body mass index†	−0.90 (−2.24 to 0.44)	−0.89 (−2.10 to 0.32)
Total cholesterol†	−1.21 (−2.55 to 0.13)	−1.13 (−2.34 to 0.08)
Fasting blood glucose†	−1.07 (−2.40 to 0.26)	−1.03 (−2.24 to 0.17)
Smoking status	−1.39 (−2.70 to −0.08)*	−1.29 (−2.48 to −0.11)*
Sedentary behavior	−1.37 (−2.68 to −0.06)*	−1.30 (−2.49 to −0.12)*
All cardiovascular risk factors	−0.77 (−2.21 to 0.67)	−0.83 (−2.13 to 0.47)

Notes: β is a coefficient estimated from linear regression models for an association of cIMT (per SD) with measures of cerebral blood flow; Models 1–7 were adjusted for a single vascular risk factor, Model 8 is adjusted for all 7 vascular risk factors. All models were also controlled for age, sex, race, CARDIA site, years of education and total brain volume. CBF = cerebral blood flow; CI = confidence interval.

^†Cumulative exposure to systolic blood pressure, diastolic blood pressure, body mass index, total cholesterol and fasting blood glucose was calculated as area under the curve from baseline up to year 20.

*p < .05.

Discussion

In a biracial cohort of middle-aged men and women, subclinical atherosclerosis measured by cIMT was associated with lower CBF in gray matter and the total brain, but not with brain volume or WMH. The associations of cIMT with CBF did not differ by race or sex. They were attenuated with the addition of traditional CVRFs to the models. We detected these associations using a composite value of cIMT, in line with a view that it is a more powerful marker of atherosclerosis than cIMT measured at specific segments of carotid arteries (22). This finding is important in the context of finding early biomarkers of vascular brain disease. Together with previous studies based on CARDIA and the Third Generation Cohort of the Framingham Heart Study (7–10), this study supports the evidence that functional changes of the brain are already present in midlife.

Expanding on previous studies on the relation of cIMT to measures of CBF in older adults (28), we propose that lower brain perfusion is present in 50 years old individuals in the terrain of vessels changed by atherosclerosis. Several authors previously suggested that decrease in cerebral perfusion precedes the occurrence of white matter lesions, brain atrophy, and dementia (29–31). However, most studies have been based on older adults, which presents a challenge to establishing temporality. Reverse causality, or a situation where changes in the brain affect levels of CVRFs, is unlikely to play a major role at the age of this cohort. Even though our study was cross-sectional and no conclusions can be made about trajectories of brain pathologies, the lack of association of cIMT with brain volume or WMH is consistent with the hypothesis that these represent more advanced pathologies, which occur later during the life course, whereas lower brain perfusion is an early marker of abnormal brain aging (7).

Previous studies reported that during ageing, brain energy metabolism decreases, which is reflected by reduced gray matter perfusion (32). This was found particularly in brain regions most vulnerable to neurodegeneration (32). Here we show that lower brain energy metabolism may be detectable already in middle-aged adults and speculate that it could be a consequence of atherosclerotic processes. Atherosclerosis, characterized by thickening of the vessel walls and reduction of their diameters, may cause insufficient delivery of oxygen and thus energy depletion in the brain. This may initiate a cascade of pathological events: accelerate the emergence of age-related capillary aberrations, cause a damage to the blood–brain barrier, and precipitate the occurrence of ischemic lesions. Further, deposition of amyloid and tau proteins may represent terminal manifestation of the breakdown of energy metabolism (33). All these established pathologies may then lead to brain atrophy, evident vascular brain disease and eventually manifest as cognitive decline and dementia.

We show that associations of cIMT with CBF were statistically explained by the addition of CVRFs to the models. In two previous studies of this middle-aged cohort from CARDIA, the associations of atherosclerosis measured by cIMT and calcified plaques to cognition were independent of CVRFs (15,16), indicating that atherosclerosis provides additional information in predicting brain outcomes. However, in both studies, authors adjusted for CVRFs that were assessed only at one time point, whereas we used a more robust measure of cumulative exposure to CVRFs over young adulthood. As the long-term exposure to CVRFs was independently associated with CBF, even when cIMT was included in the model, we propose that CVRFs, in particular hypertension, obesity and hyperglycemia, may contribute to a lower CBF also through other mechanisms independent of atherosclerosis. Hypertension induces structural alterations of cerebral blood vessels and impairment in cerebrovascular function that may, in concert with loss of microvessels, result in CBF reduction (34). Obesity and hyperglycemia may induce endothelial dysfunction and failure in cerebral vessel dilation, leading to impaired regulation of the blood flow (35). Particularly detrimental is the simultaneous presence of CVRFs that may have additive effects on the state of brain perfusion (36).

The association between cIMT and lower CBF is modest in this relatively young population, but several previous studies support its clinical relevance. Zeki Al Hazzouri and colleagues (16) found that cIMT is related to lower cognitive functioning in this middle-aged cohort and the association is consistent with the link between cIMT in older adults and the future risk of dementia (37). In addition, lower CBF in brain regions related to cognitive processes was found associated with smoking in this middle-aged population (8) and another study suggested that decreased CBF in midlife predicts the occurrence of dementia (38). Thus, it may be speculated that these subtle vascular pathologies will proceed to accumulate into old age and possibly form the structural basis for dementia, if there would not be a reduction in CVRFs.

The strengths of this study include its uniquely large well-characterized cohort of participants who have been followed-up for 25 years, have state-of-the-art brain imaging and high quality measurements of cIMT obtained in midlife. Another strength is participation of African Americans that have not been well represented in previous studies on the brain. Some limits in this study should be taken into account when interpreting the results. The lack of association of cIMT with brain matter volumes may reflect insufficient variability in these measures and low statistical power. Our analytical sample included generally healthier participants, so more severe levels of atherosclerosis may be missing (9). Compared to participants that did not receive MRI or had missing cIMT data, the final analytical sample consisted of fewer African Americans and had a lower burden of CVRFs; this may result in selection bias if the relationships between these factors and CBF differed between those included and not included from this analysis.

Further, the measures of CBF were made 5 years after the cIMT assessment. Owing to the cross-sectional design of the study, we cannot establish the temporal directions of the associations among cIMT, CVRFs, and CBF. Reverse causality is often an issue in studies of old people. As our sample was relatively young, we do not expect that higher levels of cIMT reflect the influence of low CBF. There are also some methodological issues relating to the interpretation of the pCASL measure. Thickening of the arteries due to atherosclerosis could lead to reduced blood velocity and a longer arterial transit time in the supplying small vessels, which could artificially lower the overall CBF. This possible confound should be taken into account when interpreting the data as it could bias our results, leading to overestimation of the association of higher cIMT with lower CBF that we found.

To conclude, this study suggests that lower CBF is associated with subclinical atherosclerosis in midlife, which may reflect previous exposure to traditional CVRFs. Given our relatively young cohort, this study emphasizes that early identification and intervention to delay CVRFs, in particular hypertension, obesity, and hyperglycemia, may modulate the trajectory of CBF as people age and develop brain pathology that may manifest with dementia. Future follow-up of this cohort will provide insights into how the association of CVRFs, atherosclerosis, and CBF evolve in relation to cognitive decline and dementia.

Funding

This study was supported by the Swedish Research Council (523-2012-2291 to P.C. and D.R.), the Alzheimer Foundation–Czech Republic (P.C.), project “Sustainability for the National Institute of Mental Health” (grant LO1611), with a financial support from the Ministry of Education, Youth and Sports of the Czech Republic (P.C.) and PRIMUS (247066) conducted at Charles University (P.C.). CARDIA is conducted and supported by the National Heart, Lung, and Blood Institute (NHLBI) in collaboration with the University of Alabama at Birmingham (HHSN268201300025C and HHSN268201300026C), Northwestern University (HHSN268201300027C), University of Minnesota (HHSN268201300028C), Kaiser Foundation Research Institute (HHSN268201300029C), and Johns Hopkins University School of Medicine (HHSN268200900041C). CARDIA is also partially supported by the Intramural Research Program of the National Institute on Aging (NIA) and an intra-agency agreement between NIA and NHLBI (AG0005). This manuscript has been reviewed by CARDIA for scientific content.

Conflict of interest statement

None declared.