Extreme Deep White Matter Hyperintensity Volumes Are Associated with African American Race

Abstract

Background

African Americans (AAs) have a higher prevalence of extreme ischemic white matter hyperintesities (WMH) on magnetic resonance imaging (MRI) than do European Americans based on the Cardiovascular Health Study (CHS) score. Ischemic white matter disease, limited to the deep white matter, may be biologically distinct from disease in other regions and may reflect a previously observed trend toward increased risk of subcortical lacunar infarcts in AA. We hypothesized that extreme deep WMH volume (DWMV) or periventricular volume (PV) may also have higher prevalence in AAs. Thus, we studied extreme CHS scores and extreme DWMV and PV in a healthy population enriched for cardiovascular disease (CVD) risk factors.

Methods

We imaged the brains of 593 subjects who were first degree relatives of probands with early onset coronary disease prior to 60 years of age. WMHs were manually delineated on 3T cranial MRI by a trained radiology reader the location and volume of lesions were characterized using automated software. DWMV and PV were measured directly with automated software and the CHS score was determined by a Neuro-radiologist. Volumes were characterized as being in the upper 25% versus lower 75% of total lesion volume. Volumes in the upper quartile vs. the remaining were examined for AA versus European American (EA) race using multiple logistic regression (GEE adjusted for family relatedness) and adjusted for major vascular disease risk factors including age ≥ 55 years vs. younger than 55, sex, current smoking, obesity, hypertension, diabetes, and LDL>160.

Results

Participants were 58% women and 37% AA, with a mean age of 51.5±11.0 years (range, 29-74 years). AAs had significantly higher odds of having extreme DWMV (OR, 1.8; 95% CI, 1.2-2.9; p=0.0076) independent of age, sex, hypertension, and all other risk factors. AAs also had significantly higher odds of having extreme CHS scores ≥3 (OR, 1.3; 95% CI, 1.1-3.6; p=0.025). Extreme PV was not significantly associated with AA race (OR, 1.3; 95% CI, 0.81-2.1; p=0.26).

Conclusions

AAs from families with early-onset CVD are more likely to have extreme DWMV (a subclinical form of cerebrovascular disease) and an extreme CHS score, but not extreme PV, independent of age and other CVD risk factors. These findings suggest that this AA population is at increased risk for DWMV and may be at increased risk for future subcortical stroke. Longitudinal studies are required to see if DWMV is predictive of symptomatic subcortical strokes in this population.

Introduction

Over 790,000 strokes occur annually in the United States, making stroke the fourth leading cause of death and the leading cause of disability in people over the age of 65 [1]. African Americans (AAs) have a higher risk of stroke and subcortical lacunes as well as attendant morbidity and mortality than do European Americans (EAs) [2,3]. Ischemic white matter hyperintensities (WMHs) are thought to represent ischemic small vessel disease of the brain and have been associated with stroke and dementia [4,5]. The Cardiovascular Health Study (CHS) score is an ordinal measure of WMH that ranges from 0 to 9 [6]. It is based on visual comparison of participant magnetic resonance images (MRIs) to standardized “scoring” MRIs and represents a qualitative assessment of WMH burden, including periventricular volume (PV) and deep WMH volume (DWMV), as well as ventricular size and atrophy. The Atherosclerosis Risk in Communities (ARIC) study reported that the prevalence of high CHS scores (≥3) is greater in AAs than in EAs [6,7]. In general, the difference between a CHS of 2 and a CHS of 3 involves higher burden of subcortical lesion, or DWMV [6]. PV represents the most critical portion of WMH volume measured by the CHS and is more predominant in elderly people than in younger individuals [8-10].

In population-based studies, AAs have been found to have a higher prevalence of subcortical small vessel disease in the form of silent lacunar strokes [8,11]. Deep white matter lesions are anatomically specific to the subcortical region of the brain and are thought to have different pathology and attendant risk factors than WMHs in other regions [4,12-14]. It is unknown whether the prevalence of extreme DWMV is greater in AAs than in EAs or if other unique epidemiologic characteristics are associated with DWMV. Because of the previously identified association between AA race and subcortical stroke, we hypothesized that subcortically located extreme DWMV, rather than extreme PV, may be increased in AAs.

To this end we studied the prevalence of severe WMH disease as represented by a CHS score of 3 or greater, as well as extreme DWMV and PV in the upper quartile of range, in an asymptomatic population of AAs and EAs enriched for vascular risk to determine if extreme WMH was associated with AA race and other risk factors. We applied updated 3.0 Tesla MRI volumetrics, which directly quantify and localize DWMV and PV, and compared the data to ordinal CHS scores. Our goal was to determine whether different regions, representing the components of the CHS score, would have different associations with previously identified risk factors, including race [15].

Methods

Sample and Recruitment

Participants were recruited from the ongoing prospective study called Genetic Study of Atherosclerosis Risk (GeneSTAR), which was designed to characterize genetic and biologic factors associated with incident cardiovascular and cerebrovascular disease in families of patients with early-onset coronary artery disease (CAD). This study was approved by the Johns Hopkins Medicine Institutional Review Board. All participants gave informed consent. Early-onset CAD was used as a marker for increased familial risk of vascular disease. Probands under the age of 60 (39.5% AA and 33.6% female) were identified at the time of hospitalization for an early-onset CAD event, including acute myocardial infarction or acute coronary syndromes with angiographic evidence of a flow-limiting stenosis of >50% diameter in at least one coronary artery. Apparently healthy, asymptomatic siblings and their offspring, and the offspring of the probands, were eligible for this study if they were 29 to 75 years of age and had no history of CAD, stroke, or transient ischemic attacks. Siblings and offspring were excluded if they had a history of chronic corticosteroid use; life-threatening diseases such as active AIDS, renal failure, or cancer; neurologic diseases that would preclude accurate MRI interpretation; or implanted metals that precluded MRI testing.

Participant Screening

Subjects underwent comprehensive screening for risk factors. Medical history, current medication use, and physical condition were assessed by physical examination and standard methods. Participants self-identified their racial group and were screened for traditional Framingham stroke risk factors, including hypertension, diabetes, smoking, and obesity [16]. Participants with atrial fibrillation or symptomatic heart disease were excluded. Anthropometric measures, including height in inches and weight in pounds, were determined with a fixed stadiometer and a balance scale while the participant was wearing light clothing and no shoes. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Obesity was defined as a BMI ≥ 30 kg/m², in accordance with the national obesity guidelines [17]. Current cigarette smoking was assessed by self-report of any smoking within the past month and/or two expired carbon monoxide levels of ≥8 ppm. Blood pressure was measured three times over the course of the day according to American Heart Association guidelines. The average was used to characterize resting blood pressure. Hypertension was defined as an average blood pressure ≥140 mmHg systolic, or ≥90 mmHg diastolic, and/or use of an antihypertensive drug. After participants fasted for 9-12 hours overnight, blood was taken for measurement of lipids and glucose. Type 2 diabetes was defined as a physician-diagnosed history, a fasting glucose ≥ 126 mg/dL, and/or use of hypoglycemic antidiabetic medications. Total cholesterol, high-density lipoprotein cholesterol, and triglyceride levels were measured according to United States Centers for Disease Control standardized methods [18], and low-density lipoprotein cholesterol (LDL) was estimated by using the Friedewald formula [19]. For persons with triglyceride levels > 400 mg/dL, ultracentrifugation methods were used. Hypercholesterolemia was defined as an LDL ≥ 160 mg/dL.

Magnetic Resonance Imaging (MRI)

All participants underwent magnetic resonance scanning according to a standard protocol on a Philips 3.0 Tesla scanner. The series included the following imaging sequences: 1) Axial T1-weighted MPRAGE (magnetization prepared rapid gradient echo): TR (repetition time) 10 ms; TE (time to echo) 6 ms; TI (inversion time) voxel size 0.75 × 0.75 × 1.0 mm³; contiguous slices, with field of view imaging (FOV) 240 mm; matrix 256×256×160 mm. 2) Axial turbo spin echo FLAIR (fluid attenuation inversion recovery): TR 11000 ms; TI 2800 ms; TE 68 ms; voxel size 0.47 × 0.47 × 3.0 mm³; contiguous slices, FOV 240 mm; matrix 256 × 256 mm. All images were reviewed for clinical pathology, checked, stored first on the in-house reading system, and then transferred to an off-site permanent storage facility. Confirmatory clinical reading was completed by a trained neuroradiologist (DY) using the methods of the Cardiovascular Health Study to define CHS score on an ordinal scale ranging from 0 to 9 [6]. We considered a CHS score of >3 as extreme (provide REF or alternative explanation, e.g., top 15%). Image processing and volumetric analysis were completed by biomedical engineers and their technical staff.

Volumetric Assessment

MPRAGE images were skull-stripped and co-registered to FLAIR images. Spatial normalization of the co-registered MPRAGE and FLAIR images into MNI space was performed via affine transformation. A trained rater manually delineated WMHs on the normalized MPRAGE and FLAIR images using Medical Image Processing, Analysis, and Visualization (MIPAV) software [20]. We segmented the brain in native MPRAGE space using an automated probabilistic methodology that utilizes a topology-preserving algorithm; the resulting tissue mask was mapped to MNI space [21]. We measured total brain intracranial, cortical grey matter, and white matter volumes in native MPRAGE space, and WMH volumes in MNI space. Total brain volume, in cubic millimeters, was identified as the sum of white matter, WMH, and grey matter volume from the vertex of the brain to the foramen magnum. Intracranial volume was defined in cubic millimeters as the sum of all Dura mater, soft tissue, and sulcal and ventricular cerebrospinal fluid volumes, inferior to bone, from the vertex to foramen magnum [22].

Spatial characterization of WMHs was carried out with in-house software designed to determine their location in relation to the ventricles and the deep white matter region in three-dimensional space. We determined connected components of WMHs with digital 26 connectivity (by measuring all 26 adjacent voxels). We defined periventricular lesions as those that were contiguous with a lesion voxel that was within 4 mm of the ventricle and defined deep white matter lesions as those that were not contiguous (fig. 1).

Statistical Analysis

Extreme DWMV and extreme PV were defined as total DWMV or total PV greater than the 75^th percentile. Extreme CHS scores were defined as a CHS score of 3 or greater. Demographic and vascular risk factor distributions were tabulated by the presence or absence of extreme CHS score, DWMV, and/or PV. To test differences by group, we used t-tests for normally distributed variables, Wilcoxon rank sum tests for non-normally distributed continuous variables, and χ²statistics for categorical variables. The concordance between dichotomous variables CHS score, extreme DWMV, and extreme PV was estimated by using tetrachoric correlation. We used Generalized Estimating Equations (GEE) regression analyses to correct for intra-familial correlations and to model being in the highest quartile of DWMV, or PV, or having CHS score ≥3, after adjusting for traditional vascular risk factors, including age, race, sex, hypertension, diabetes, current smoking, and obesity.

Results

Study Sample

The study population consisted of 593 apparently healthy individuals identified from 324 families of probands with early-onset CAD (one proband per family). On average, the study population consisted of 1.8 ± 1.2 relatives per family (range, 1-8). Siblings of probands comprised 53.1% of the group and offspring of siblings and probands comprised 46.9%. Sample characteristics stratified by race are shown in table 1. Most participants had some white matter disease; 89.9% had deep white matter disease, 73.7% had periventricular disease, and 14.3% had a CHS score of 3 or greater.

Table 1. Demographic characteristics and risk factors of participants by race.

Characteristic	AA (n=220)	EA (n=373)	p
Age, mean (SD)	52 (11)	52 (10)	0.098
Total cholesterol (mg/dL), mean (SD)	192 (43)	195 (39)	0.38
HDL cholesterol (mg/dL), mean (SD)	58 (16)	57 (17)	0.36
LDL cholesterol (mg/dL), mean (SD)	115 (39)	114 (37)	0.69
Diabetes	21%	9%	<0.0001
Female sex	64%	55%	0.0398
Smoking currently	24%	13%	0.001
Hypertension	59%	35%	<0.0001
Obesity (BMI ≥ 30 kg/m2)	58%	39%	<0.0001
Hypercholesterolemia (LDL ≥ 160 mg/dL)	11%	10%	0.58
Periventricular lesion volume upper quartile	29%	23%	0.087
Total lesion volume upper quartile	31%	22%	0.013
Deep white matter lesion volume upper quartile	33%	21%	0.001

AA = African American; BMI = body mass index; EA = European American; HDL = high-density lipoprotein; LDL = low-density lipoprotein; SD = standard deviation.

Association of Extreme DWMV, PV, and CHS Score with Race, Controlling for Other Risk Factors

Results of multivariate regression analyses (GEE) to predict extreme CHS score, extreme DWMV, and extreme PV are shown in tables 2a, 2b, and 2c, respectively. Many of the variables were correlated with one another. DWMV upper quartile and CHS ≥ 3: tetrachoric correlation = -0.3338, p = 0.0868; PV upper quartile and CHS ≥ 3: tetrachoric correlation = 0.8564, p = 0.0325; DWMH upper quartile and PV upper quartile: tetrachoric correlation = -0.7485, p = 0.0592.

Table 2a. Fully adjusted logistic regression model predicting CHS ≥3 (n=593).

Characteristic	OR	95% CI	p
African American	1.9	1.08 to 3.5	0.025
Female sex	1.07	0.63 to 1.8	0.79
Diabetic	0.73	0.34 to 1.5	0.42
Smoking currently	2.1	0.1.0 to 4.2	0.036
Hypertension	1.7	0.99 to 2.9	0.055
Obesity	0.56	0.32 to 0.98	0.042
	Estimate	95% CI	p
Age	0.14	0.11 to 0.18	<0.0001

CHS = Cardiovascular Health Study Scale score; CI = confidence interval; OR = odds ratio.

Table 2b. Fully adjusted logistic regression model predicting extreme deep white matter lesion volume (n=593)^a.

Characteristic	OR	95% CI	p
African American	1.8	1.2 to 2.9	0.0076
Female sex	1.6	1.1 to 2.4	0.021
Diabetic	0.69	0.37 to 1.3	0.22
Smoking currently	1.3	0.77 to 2.3	0.29
Hypertension	1.23	0.77 to 2.0	0.39
Obesity	0.58	0.37 to 0.91	0.18
	Estimate	95% CI	p
Age	0.082	0.06 to 0.1	<0.0001

CI = confidence interval; OR = odds ratio.

^aLesion volume in the upper quartile, with age as a continuous variable.

Table 2c. Fully adjusted logistic regression model predicting periventricular white matter lesion volume (n=593)^a.

Characteristic	OR	95% CI	p
African American	1.3	0.82 to 2.1	0.26
Female sex	0.91	0.60 to 1.4	0.66
Diabetic	1.3	0.70 to 12.3	0.41
Smoking currently	1.73	0.95 to 3.1	0.068
Hypertension	0.98	0.62 to 1.6	0.94
Obesity	0.69	0.44 to 1.07	0.10
	Estimate	95% CI	p
Age	0.11	0.084 to 0.13	<0.0001

CI = confidence interval; OR = odds ratio.

^aLesion volume in the upper quartile, with age as a continuous variable.

Variables independently associated with higher odds of extreme CHS score (table 2a) included AA race, older age, current smoking, and non-obesity. Variables independently associated with higher odds of extreme DWMV included AA race, older age, female sex, and non-obesity (table 2b). AA race was significantly associated with higher odds of extreme DWMV (OR, 1.8; 95% CI, 1.2-2.9; p = 0.0076), independent of all other risk factors, including hypertension (table 2b). When race-specific upper quartiles were used for race-stratified analysis, older age and female sex were associated with higher odds of extreme DWMV in AAs (p < 0.0001 and p = 0.0021, respectively; table 3a), whereas older age and thinner body habitus were associated with higher odds of extreme DWMV in EAs (p < 0.0001 and p = 0.026, respectively; table 3b).

Table 3a. Fully adjusted logistic regression model predicting extreme deep white matter lesion volume in African Americans (n=220)^a.

Characteristic	OR	95% CI	p
Female sex	2.7	1.4 to 5.3	0.0021
Diabetic	0.54	0.26 to 1.14	0.11
Smoking currently	1.7	0.76 to 3.8	0.20
Hypertension	1.3	0.59 to 2.7	0.56
Obesity	0.64	0.33 to 1.23	0.18
	Estimate	95% CI	p
Age	0.084	0.048 to 0.12	<0.0001

CI = confidence interval; OR = odds ratio.

^aLesion volume in the upper quartile, with age as a continuous variable.

Table 3b. Fully adjusted logistic regression model predicting extreme deep white matter lesion volume in European Americans (n=373)^a.

Characteristic	OR	95% CI	p
Female sex	1.1	0.67 to 1.9	0.68
Diabetic	1.1	0.43 to 2.7	0.87
Smoking currently	1.1	0.48 to 2.5	0.83
Hypertension	1.2	0.65 to 2.2	0.55
Obesity	0.47	0.24 to 0.91	0.026
	Estimate	95% CI	p
Age	0.082	0.054 to 0.11	<0.0001

CI = confidence interval; OR = odds ratio.

^aLesion volume in the upper quartile, with age as a continuous variable.

Age was the only variable independently associated with higher odds of extreme PV. AA race was not associated with higher odds of extreme PV (OR, 1.3; 95% CI, 0.81-2.1; p = 0.26), independent of all other risk factors, including hypertension (table 2c).

Discussion

Our results show that the prevalence of extreme DWMV is higher in AA than in EA family members of individuals with premature CAD. This association of AA race with extreme DWMV was independent of age, hypertension, and other known cardiovascular disease risk factors. This study confirms the observations of the previous ARIC study, which found an association between AA race and extreme CHS score [7]. Additionally, it builds on the results of the ARIC study by using modern, validated, direct measurements of DWMV with automated white matter segmentation. These methods allow for the attribution of lesions to brain region and analysis of associations between different risk factors and regions.

African American race has been associated with increased prevalence of many vascular disease phenotypes, including symptomatic and asymptomatic subcortical lacunar stroke and vascular disease in other organ systems, such as the coronary arteries and peripheral vasculature [23-29]. Furthermore, racial differences in white matter disease burden have been reported to be related to smoking and increased rates and severity of hypertension [7,30]. Although other studies have emphasized smoking as a risk factor, in our study, smoking was not associated with extreme DWMV in the stratified or combined analysis. Likewise, hypertension did not appear to drive the association between DWMV and AA race. Our analysis of DWMV predictors controlled for the diagnosis of hypertension as well as age. However, we cannot exclude the possibility that genetic differences that affect hypertension, such as angiotensin converting enzyme polymorphisms, might be more prevalent in this AA population [9]. Additionally, it is possible that an unidentified inheritable trait may be associated with the AA population in this study of related individuals.

Many of our observed associations and correlations support the idea that extreme DWMV is an independent lesion type that conveys risks that differ from those of PV and CHS score. Extreme PV and extreme CHS shared similar associations, including age, smoking, and decreased obesity. It is important to note that HTN in AAs may represent an under treated disease with an earlier age of onset as compared to EA's which could have contributed to the increased predominance of extreme DWMV in this group. We were unable to designate the age of onset and control for this factor in our analysis. The correlation between CHS and PV was stronger than the correlation between CHS and DWMV. Extreme DWMV was associated with AA race, as was extreme CHS. However, in our stratified analysis, DWMV was independently associated with female sex but lacked associations with smoking, hypertension, and obesity, which were associated with extreme CHS, despite the fact that the CHS scale emphasized periventricular confluence. In the higher grades of CHS score (>6), involvement of the centrum semiovale is emphasized. These observations would support those of Fazeka's et al. [31], who reported different pathologic substrates in the periventricular and DWMV regions. In particular we did include the PV caps into the PV calculation. Lesions in this region are reported to be non-ischemic in nature. Thus, the lower PV relative to DWMV may reflect a proclivity for ischemic lesions and higher extreme DWMV as compared to PV.

Small vessel vascular disease in the subcortical region has been reported to be more prevalent in the AA population than in either the EA or Hispanic population [23-29]. The DWMV and periventricular lesions lie in the subcortical and periventricular regions, respectively. These regions have very different small vessels, with long cortically based small vessels serving the periventricular region and shorter small vessel perforators serving the subcortical DWMV. In past studies, different locations of WMH have been associated with different risk factors and clinical outcomes [32,33]. Investigators who have analyzed the pathologic nature of deep white matter lesions have reported that risk factors such as endothelial activation and inflammation are more prominent in the short subcortical vessels of the deep white matter than in the long periventricular vessels [4,12,32,34]. Our observations may result from enrichment of these risk factors in our study population [35].

Our findings have similarities to those of the ARIC study of severe CHS score [7]. In both studies, extreme CHS was associated with AA race, as well as age, hypertension, and smoking. The association of extreme CHS with smoking and hypertension appeared to be stronger in AAs than in EAs in both studies. However our population differed from that in the ARIC study. Our population was by definition asymptomatic, whereas the ARIC group included both symptomatic and asymptomatic individuals. All of our participants were relatives of a proband, whereas for the most part, ARIC participants were unrelated. The average age in our population was lower— 52, as compared to 62 in the ARIC cohort—and unlike the ARIC population, our population was notably enriched with vascular risk factors [6,35]. Interestingly, however, the percentage of participants with extreme CHS score was very similar in the two studies: 14.7% in our population and 12.5% in ARIC [36]. Our study is unusual in that obesity appeared to be associated with reduced risk of WMH of all types by all measures. Although this finding has been reported in other studies, such as The Woman's Health Initiative MRI Study [37], in the preponderance of literature, increased BMI is strongly associated with increased risk of WMH and WMH progression [5,16,18,19,24,38,39]. As shown in other studies, age was the predominant characteristic associated with WMHs of all types [40].

Other unique aspects of our study include the use of an MRI scanner with a 3.0 Tesla field strength. This MRI scanner has greater sensitivity for the detection of WMHs than does the 1.5 Tesla scanner and provides better assessment of lesion volume through improved signal-to-noise ratios [13,15]. In general, prevalence and volume measured for deep white matter lesions and periventricular lesions were significantly higher than those obtained with similar 1.5 Tesla methodologies [10]. We also used advanced programming that has been extensively validated to separate, localize, and measure WMH volumes. This program also has been extensively validated in other white matter diseases such as multiple sclerosis [20].

Conclusion

African American race is an independent risk factor for extreme DWMV and extreme CHS score in a population enriched for vascular risk factors. In contrast, PV is not associated with AA race. Extreme DWMV had unique epidemiological associations, suggesting that it represents a unique lesion type that differs from PV lesions. The association between extreme DWMV, which is a subclinical form of ischemic stroke, and AA race suggests that this population may have an increased risk of future subcortical stroke. This risk must be verified in future longitudinal studies.