Blood pressure and heart rate measures associated with increased risk of covert brain infarction and worsening leukoaraiosis in older adults

Abstract

Objective

In people without prior stroke, covert findings on serial MRI of incident brain infarcts and worsening leukoaraiosis are associated with increased risk for ischemic stroke and dementia. We evaluated whether various measures of blood pressure and heart rate are associated with these MRI findings.

Approach and Results

In the Cardiovascular Health Study, a longitudinal cohort study of older adults, we used relative risk (RR) regression to assess the associations of mean, variability, and trend in systolic blood pressure (SBP), diastolic blood pressure (DBP), and heart rate (HR) measured at four annual clinic visits between two brain MRIs with incident covert brain infarction (CBI) and worsening white matter grade (WMG, using a 10 point scale to characterize leukoaraiosis). We included participants who had both brain MRIs, no stroke before the follow-up MRI, and no change in antihypertensive medication status during follow-up. Among 878 eligible participants, incident CBI occurred in 15% and worsening WMG in 27%. Mean SBP was associated with increased risk for incident CBI (RR per 10 mmHg 1.28; 95% confidence interval (CI) 1.12–1.47), and mean DBP was associated with increased risk for worsening WMG (RR per 10 mmHg 1.45; 95% CI 1.24–1.69). These findings persisted in secondary and sensitivity analyses.

Conclusions

Elevated mean SBP is associated with increased risk for CBI, and elevated mean DBP is associated with increased risk for worsening leukoaraiosis. These findings reinforce the importance of hypertension in the development of silent cerebrovascular diseases, but the pathophysiologic relationships to blood pressure for each may differ.

INTRODUCTION

In people without prior stroke, covert brain infarction (CBI) and leukoaraiosis are described as “silent” cerebrovascular diseases, but they have clinically important consequences. CBI is associated with a two-to-four fold increased risk of clinically defined ischemic stroke, independent of vascular risk factors, and a two-to-three fold increased risk of dementia.^1–4 They may also directly disrupt functional networks, leading to deficits affecting cognition, gait, and other functions.⁵ Similarly, leukoaraiosis is associated with an increased risk for ischemic stroke, worse outcomes after stroke, dementia, and mortality.^6–11 Leukoaraiosis prevalence in older adults exceeds 95%, and CBI is also common: the estimated prevalence of CBI in adults over age 50 is approximately 20% compared to 2–14% for overt ischemic stroke in the U.S.^2,11 While the American Heart Association/American Stroke Association recently published a scientific statement highlighting the importance of these conditions and the need for further studies to guide their management, optimal prevention strategies after detection of these conditions have not been established, in part due to uncertainties about their pathogenesis that may resemble or differ from ischemic stroke.¹² Furthermore, strategies for the prevention of incident CBI and worsening of leukoaraiosis have not been established. Given their high prevalence and clinical sequelae, improving understanding of the pathogenesis of CBI and leukoaraiosis may help guide strategies for prevention of cerebrovascular diseases.

Hypertension is associated with CBI and leukoaraiosis as well as clinically defined ischemic stroke.^2,13–15 Both primary and secondary stroke prevention guidelines focus on absolute reductions in systolic and diastolic blood pressure, but some studies suggest that very aggressive blood pressure reductions may lead to an increased risk of recurrent stroke.¹⁶ Accordingly, investigators have sought to improve current understanding of the role of cardiovascular physiology in brain infarction by studying other measures including variability in blood pressure and heart rate. Prior studies suggest associations of blood pressure and heart rate variability with ischemic stroke, but these associations are inconsistent and depend on the time interval of measurement (beat-to-beat, daily, weekly, or visit-to-visit).^17–19 Despite these potential associations with ischemic stroke, it is not known if visit-to-visit blood pressure and heart rate measures are associated with an increased risk of incident CBI or progression of leukoaraiosis.

In this study, we used data from two brain MRIs performed approximately five years apart in the Cardiovascular Health Study (CHS) to assess associations of visit-to-visit cardiovascular measurements, including mean, variability, and trend in systolic blood pressure (SBP), diastolic blood pressure (DBP), and heart rate (HR), with incident CBI and progression of leukoaraiosis, building upon findings from prior work based on these scans.^20–21

Materials and Methods are available in the online-only Data Supplement.

RESULTS

Study Population

As detailed in Figure 1, 878 participants were eligible for the primary analysis and 1844 participants for the sensitivity analysis. Baseline characteristics and mean (SD) for the exposure variables in the primary and sensitivity analysis samples are presented in Table 1 (additional descriptive statistics for each individual analysis and a comparison to all CHS participants undergoing the first MRI are provided in the Online Supplement Table I). Among 683 participants who completed both brain MRIs and demonstrated no infarcts on the initial MRI, 101 individuals (15%) had incident CBI. Among 793 participants with white matter grade measurements (WMG, a 10 point scale characterizing the extent of leukoaraiosis) on both MRIs and a grade of 8 or less on the initial MRI, 215 (27%) demonstrated worsening WMG. The distributions of number of CBI, WMG, and changes in WMG are shown in Figure 2. Evaluation of the exposure variables indicated minimal correlation for most pairs and at most modest correlations for a few pairs (Supplemental Table II).

Figure 1. Flow chart of participants.

CHS = Cardiovascular Health Study, CAD = coronary artery disease, CHF = congestive heart failure, AF = atrial fibrillation, MRI = magnetic resonance imaging, WMG = white matter grade, ARB = angiotensin receptor blocker.

Table 1.

Baseline characteristics (primary analyses and sensitivity analyses).

Characteristic	Primary analysis, mean SD or n (%)	Sensitivity analysis, mean SD or n (%)
Total number of participants	878	1844
Age (years, at follow-up MRI)	73.5 (4.2)	74.0 (4.4)
Sex (male)	315 (36.0%)	734 (39.8%)
Race (black)	111 (12.6%)	274 (14.9%)
Body mass index	26.1 (4.1)	26.7 (4.3)
Smoking status
Never smoker	429 (48.9%)	867 (47.0%)
Former smoker	365 (41.6%)	823 (44.6%)
Current smoker	84 (9.6%)	154 (8.4%)
Diabetes	69 (7.9%)	223 (12.1%)
CHS Clinic
North Carolina	197 (22.4%)	435 (23.6%)
California	281 (32.0%)	547 (29.7%)
Maryland	165 (18.8%)	362 (19.6%)
Pennsylvania	235 (26.8%)	500 (27.1%)
Antihypertensive medications	236 (26.9%)	711 (38.6%)
Interval between MRI scans (days)	1808.9 ± 215	1827.2 ± 217
MRI findings
Incident covert brain infarct	101/683 (14.8%)	N/A
Worsening white matter grade	215/793 (27.1%)	472 (28.1%)
Systolic blood pressure
Mean (mm Hg)	129.8 (15.0)	133.1 (16.2)
Variability	7.4 (4.5)	8.3 (5.4)
Trend	0.6 (3.4)	−0.1 (4.2)
Diastolic blood pressure
Mean (mm Hg)	68.8 (7.9)	68.9 (8.7)
Variability	4.5 (3.2)	4.9 (3.5)
Trend	−0.39 (2.2)	−0.8 (2.5)
Heart rate
Mean (beat per minute)	63.7 (8.2)	63.4 (8.7)
Variability	3.8 (2.6)	4.1 (3.2)
Trend	0.1 (2.5)	0.1 (2.6)

MRI = magnetic resonance imaging.

Figure 2. Incident covert brain infarcts (CBI) and white matter grade (WMG) on the follow-up MRI.

A. Incident CBI

B. WMG

C. Change in WMG

(A) Distribution of number of incident CBI on the follow-up MRI. (B) Distribution of WMG on the follow-up MRI. (C) Distribution of changes in WMG on the follow-up MRI.

Primary, secondary, and exploratory analyses: association with incident CBI and worsening WMG

In primary analyses, mean SBP was associated with increased risk for incident CBI (RR 1.28, 1.12–1.47, p<0.001), and mean DBP was associated with increased risk for worsening WMG (RR 1.45, 1.24–1.69, p<0.001) (Table 2). An association between positive HR trend and increased risk for incident CBI was nominally significant (RR 2.03, 1.11–3.72, p=0.02, using a conservative Bonferroni correction of p=0.003 to account for multiple comparisons). In exploratory analyses, mean PP was nominally associated with increased risk for incident CBI (RR 1.22, 1.03–1.44, p=0.02), and mean MAP was nominally associated with increased risk for incident CBI (RR 1.42, 1.12–1.79, p=0.004) and associated with worsening WMG (RR 1.29, 1.12–1.48, p<0.001) (Table 3). In secondary analyses, we observed associations between mean SBP and increased number of incident CBIs (beta 0.06, 0.02–0.10, p=0.002) and between mean DBP and higher WMG (beta 0.13, 0.07–0.19, p<0.001) (Online Supplement, Table III). An association between positive HR trend and increased number of incident CBIs was again nominally significant (beta 0.18, 0.01–0.34, p=0.03). The findings of models incorporating both SBP and DBP and both PP and MAP demonstrated similar findings except that the nominal associations between mean PP and incident CBI and between mean MAP and incident CBI were no longer significant (Online Supplement, Tables IV–V).

Table 2.

Relationships of SBP, DBP, and HR to dichotomous outcomes: incident covert brain infarcts (CBI) and worsening white matter grade (WMG).

Incident CBI (n=683)	SBP (10 points mm Hg)		DBP (10 points mm Hg)		HR (10 bpm)
	RR (95% CI)	P value	RR (95% CI)	P value	RR (95% CI)	P value
Mean	1.28 (1.12, 1.47)	<0.001	1.30 (0.99, 1.71)	0.06	1.17 (0.93, 1.47)	0.19
Variability	0.89 (0.60, 1.31)	0.54	1.56 (0.95, 2.56)	0.08	0.81 (0.42, 1.56)	0.54
Trend	0.65 (0.39, 1.07)	0.09	0.94 (0.34, 2.62)	0.90	2.03 (1.11, 3.72)	0.02
Worsening WMG (n=793)	SBP (10 points mm Hg)		DBP (10 points mm Hg)		HR (10 bpm)
	RR (95% CI)	P value	RR (95% CI)	P value	RR (95% CI)	P value
Mean	1.09 (1.00, 1.20)	0.06	1.45 (1.24, 1.69)	<0.001	1.10 (0.95, 1.28)	0.19
Variability	0.97 (0.72, 1.31)	0.83	1.26 (0.85, 1.86)	0.25	1.19 (0.78, 1.82)	0.43
Trend	1.10 (0.77, 1.57)	0.62	0.98 (0.56, 1.72)	0.95	1.13 (0.73, 1.74)	0.58

Visit-to-visit mean was calculated from four annual measurements. Visit-to-visit trend was calculated using linear regression (slope). Visit-to-visit variability was calculated from the standard deviation of the residuals of the linear regression. Relative risks (RR) and 95% confidence intervals (CI) estimated from RR regression model adjusted for age, sex, race, clinic location, smoking status, BMI, diabetes, time between MRI scans, and antihypertensive medication status. RRs are presented for mean, variability, and trend per 10 points of mm Hg for systolic and diastolic blood pressure and 10 beats per minute of heart rate. The Bonferroni corrected p-value is 0.003.

SBP = systolic blood pressure, DBP = diastolic blood pressure, HR = heart rate, bpm = beats per minute.

Table 3.

Relationships of PP and MAP to dichotomous outcomes: incident covert brain infarcts (CBI) and worsening white matter grade (WMG).

Incident CBI (n=683)	PP (10 points mm Hg)		MAP (10 points mm Hg)
	RR (95% CI)	P value	RR (95% CI)	P value
Mean	1.22 (1.03, 1.44)	0.02	1.42 (1.12, 1.79)	0.004
Variability	1.13 (0.75, 1.71)	0.55	1.15 (0.67, 1.97)	0.61
Trend	0.63 (0.34, 1.14)	0.13	0.63 (0.26, 1.53)	0.31
Worsening WMG (n=793)	PP (10 points mm Hg)		MAP (10 points mm Hg)
	RR (95% CI)	P value	RR (95% CI)	P value
Mean	0.98 (0.88, 1.10)	0.74	1.29 (1.12, 1.48)	<0.001
Variability	1.04 (0.77, 1.41)	0.81	1.12 (0.74, 1.69)	0.60
Trend	1.15 (0.77, 1.71)	0.51	1.07 (0.62, 1.86)	0.80

Visit-to-visit mean was calculated from four annual measurements. Visit-to-visit trend was calculated using linear regression (slope). Visit-to-visit variability was calculated from the standard deviation of the residuals of the linear regression. Relative risks (RR) and 95% confidence intervals (CI) estimated from RR regression model adjusted for age, sex, race, clinic location, smoking status, BMI, diabetes, time between MRI scans, and antihypertensive medication status. RRs are presented for mean, variability, and trend per 10 points of mm Hg for pulse pressure and mean arterial pressure. The Bonferroni corrected p-value is 0.003.

PP = pulse pressure, MAP = mean arterial pressure.

Sensitivity analysis

An inclusive sensitivity analysis corroborated results of the primary analyses (Table 4). Similar to the primary and secondary analyses, associations between mean SBP and increased number of CBI (beta 0.07, 0.04–0.09, p<0.001) and between mean DBP and worsening WMG (RR 1.32, 1.20–1.45, p<0.001) remained significant. Mean DBP was also associated with increased number of CBI (beta 0.13, 0.08–0.19, p<0.001). Several associations were nominally significant including mean HR and increasing number of CBI, mean SBP and worsening WMG, DBP variability and worsening WMG, and mean HR and worsening WMG.

Table 4.

Sensitivity analysis: Relationships of SBP, DBP, and HR to number of covert brain infarcts (CBI) and worsening white matter grade (WMG) without exclusions for prior CBI, changing antihypertensive status, or cardiovascular comorbidities.

Number of CBI (n=1842)	SBP (10 points mm Hg)		DBP (10 points mm Hg)		HR (10 bpm)
	Beta	P value	Beta	P value	Beta	P value
Mean	0.07 (0.04, 0.09)	<0.001	0.13 (0.08, 0.19)	<0.001	0.06 (0.009, 0.10)	0.02
Variability	0.05 (−0.04, 0.13)	0.28	0.12 (0.001, 0.24)	0.05	−0.04 (−0.16, 0.07)	0.47
Trend	−0.001 (−0.10, 0.10)	0.98	0.03 (−0.13, 0.20)	0.69	0.13 (−0.02, 0.27)	0.09
Worsening WMG (n=1679)	SBP (10 points mm Hg)		DBP (10 points mm Hg)		HR (10 bpm)
	RR (95% CI)	P value	RR (95% CI)	P value	RR (95% CI)	P value
Mean	1.06 (1.00, 1.12)	0.04	1.32 (1.20, 1.45)	<0.001	1.10 (1.00, 1.20)	0.04
Variability	1.03 (0.88, 1.20)	0.72	1.30 (1.05, 1.62)	0.02	1.09 (0.89, 1.32)	0.41
Trend	1.07 (0.88, 1.32)	0.49	1.06 (0.76, 1.47)	0.66	1.01 (0.76, 1.34)	0.96

Visit-to-visit mean was calculated from four annual measurements. Visit-to-visit trend was calculated using linear regression (slope). Visit-to-visit variability was calculated from the standard deviation of the residuals of the linear regression. CBI analysis: beta coefficients and 95% confidence intervals (CI) estimated from linear regression model adjusted for age, sex, race, clinic location, smoking status, BMI, diabetes, time between MRI scans, and antihypertensive medication status (always, never, variable). WMG analysis: relative risks (RR) and 95% confidence intervals (CI) estimated from RR regression model adjusted for age, sex, race, clinic location, smoking status, BMI, diabetes, time between MRI scans, and antihypertensive medication status (always, never, variable). RRs are presented for mean, variability, and trend per 10 points of mm Hg for systolic and diastolic blood pressure and 10 beats per minute of heart rate. The Bonferroni corrected p-value is 0.003.

SBP = systolic blood pressure, DBP = diastolic blood pressure, HR = heart rate, bpm = beats per minute.

Stratified analysis: antihypertensive medications

Stratification by antihypertensive medication use status corroborated the association between mean DBP and worsening WMG in non-users (RR 1.49, 1.23–1.79, p<0.001) (Supplemental Table VI). The association between mean SBP and increased risk for incident CBI in non-users was still present but slightly less robust (RR 1.24, 1.06–1.44, p=0.007). There were nominally significant associations between mean DBP and increased risk for incident CBI, DBP variability and CBI, HR trend and CBI, and mean SBP and worsening WMG. All associations were attenuated in participants taking antihypertensive medications. When testing for interactions with antihypertensive medication use, the interactions were not statistically significant.

DISCUSSION

Mean BP is associated with incident CBI and worsening leukoaraiosis

In this prospective cohort study of older adults without prior stroke and cardiovascular disease, elevated mean SBP was associated with increased risk for incident CBI, and elevated mean DBP was associated with increased risk for worsening leukoaraiosis, extending prior findings from the CHS.^20–21 These associations remained robust across several secondary and sensitivity analyses incorporating adjustment for vascular risk factors and antihypertensive medication status, inclusion or exclusion of participants with prior imaging-defined infarcts (in the incident CBI analyses), and classification of the outcome as dichotomous or as counts. These findings are particularly important in the setting of potentially conflicting recommendations: the Joint National Committee advises less stringent control of hypertension in elderly adults, whereas a recent scientific statement from the American Heart Association/American Stroke Association highlights the clinical relevance of imaging-defined vascular brain injury and urges, at a minimum, initiation of stroke primary prevention measures in affected individuals.^12,22 In patients without a history of stroke, this study suggests that incident CBI and worsening leukoaraiosis could be treatment targets for control of systolic and diastolic hypertension.

Differential associations between measures of pulsatile and steady flow with CBI and leukoaraiosis

Across several analyses, measures of steady blood flow (DBP and MAP) were associated with worsening leukoaraiosis. There was also a suggestion that measures of pulsatile blood flow (SBP and PP) were associated with incident CBI, although MAP was also associated with incident CBI in contrast to DBP and to a greater degree than PP. While prior studies have reported variable associations of pulsatile and steady blood flow with different forms of imaging-defined vascular brain injury, this study provides evidence suggesting the roles of different types of hypertension in the pathogenesis of CBI and leukoaraiosis.^23–28 The more consistent finding is the relationship of measures of steady blood flow with leukoaraiosis: multiple pathophysiologic mechanisms for the development of leukoaraiosis have been postulated, but this study emphasizes that diastolic hypertension likely has a critical role. One theory is that increased peripheral arterial stiffness (represented by increased brachial DBP as a proxy for carotid DBP) in conjunction with increased aortic pulsatility may augment transmission of the effects of aortic pulsatility to the cerebral small vessels, resulting in increased endothelial shear stress and dysfunction.²⁴ Alternatively, diastolic hypertension may correspond with increased peripheral arterial stiffness and reduced carotid flow velocity, resulting in reduced blood flow to cerebral small vessels and local hypoperfusion, independent of atherosclerosis.²⁵ By contrast, regarding peripheral measures of pulsatile blood flow, it is hypothesized that the late-life development of atherosclerosis in intracranial arteries may arise in conjunction with increased peripheral pulse pressure.²⁹ This study may support that hypothesis and may implicate atherosclerotic pathologies in the development of CBI. In light of these differential associations and the clinical significance of these conditions, future studies in the treatment of hypertension in adults (such as those in the CHS) should include incident CBI and worsening leukoaraiosis as distinct outcomes from one another and from clinically defined ischemic stroke.

HR trend and increased risk for CBI

We did find an association between positive HR trend and incident CBI, though its risk estimate has a moderately wide confidence interval and its p-value did not reach significance with a conservative Bonferroni correction. This association persisted through almost all analyses, including the stratified analyses. To our knowledge, an association between visit-to-visit HR trend and cerebrovascular outcomes has not been previously described. While mechanisms of vascular injury due to hypertension have been rigorously studied, it is less certain how changes in HR would lead to vascular brain injury. It is possible that this association may reflect a downstream or compensatory response to other changes in vascular biology that more directly lead to increased risk for CBI.

Other BP and HR measures are not consistently associated with incident CBI or worsening leukoaraiosis

In this study, variability and trend of SBP and DBP and mean and variability of HR were not consistently associated with CBI or worsening WMG: some demonstrated nominal statistical significance in the most inclusive sensitivity analysis, but these findings were unstable across analyses with exclusions aimed at reducing confounding. While HR mean and visit-to-visit variability were associated with mortality in a prior CHS analysis, we did not find a consistent relationship between these measures and covert imaging-defined vascular brain injury.³⁰ The lack of association between SBP or DBP visit-to-visit variability and CBI is notable as it may suggest that CBI and clinically defined ischemic stroke have different relationships with blood pressure variability.^17–19

The role of antihypertensive medication use

We assessed the potential for effect modification by antihypertensive medication use by stratifying study participants into users and non-users of a stable medication regimen. While the interactions in the full primary analysis sample were not statistically significant, we found that the associations between our primary exposure variables and incident CBI and worsening leukoaraiosis appeared to persist in the stratum of non-users. Interpretation of the lack of association in the stratum of medication users is less certain due to a low sample size and an absence of medication dosage and administration timing data in this study, both of which would tend to increase the variability of exposure among medication users. Future clinical trials may be useful to determine whether there is a beneficial effect of antihypertensives on imaging-defined vascular brain injury.

Strengths and limitations

Our study has several strengths including standardized BP and HR measurements at structured annual study visits, central adjudication of clinical events, centralized analysis of MRI scans with characterization of infarcts and WMG, and high-quality data on vascular risk factors collected prospectively. Additionally, this study population of healthy older adults offers an opportunity to establish these associations with relatively minimal confounding.

Our study had several limitations. First, it was limited to participants who underwent two brain MRIs according to the CHS protocol, an exclusion that reduced our study sample size and likely introduced a selection bias. Having no MRI or only a single MRI was the most frequent reason for exclusion from this analysis, which resulted from loss to follow-up, death, and various other reasons. Prior studies of the CHS demonstrated that participants who underwent brain MRI scans were healthier than those who did not, and those who underwent both brain MRIs were healthier than those that only underwent a single scan.^20–21,31 Accordingly, our study may underestimate the frequency of incident CBI and worsening WMG.

Second, the sample size for the primary analyses was also limited due to the exclusions of participants with baseline CAD, CHF, and AF. We were concerned about the relationships between these cardiovascular conditions and our primary exposures of interest: all could be confounders or mediators of the association of these exposures on incident CBI and worsening WMG. We chose to exclude these individuals from the primary analyses and include them in a sensitivity analysis. The beta coefficients and risk estimates in the sensitivity analysis did not change substantially, suggesting that these conditions may be minor confounders or mediators. Nonetheless, the main findings of this study, the associations between mean SBP and incident CBI and between mean DBP and worsening leukoaraiosis, were not affected substantially by these conditions.

Third, our study focused on annual measurements and cannot detect potential associations between shorter intervals of BP and HR measurements and the outcomes.

Finally, our analysis involved multiple comparisons with potentially non-independent exposure variables and outcomes; if these were all independent, we would expect one significant finding by chance alone among 18 tests of interest. Since a Bonferroni correction may inadvertently obscure meaningful associations if the exposures or outcomes are not independent, we elected to present both uncorrected p-values and a formal Bonferroni threshold for comparison.

CONCLUSIONS

In summary, this study provides evidence that systolic hypertension is associated with increased risk for CBI, and diastolic hypertension is associated with increased risk for leukoaraiosis. Combining these findings with prior studies, these differential associations could suggest differences in pathophysiologic mechanisms underlying CBI, leukoaraiosis, and clinically defined ischemic stroke.^17–19 Given the clinical significance of silent cerebrovascular diseases, these findings suggest a need to reevaluate the balance of benefits and risks of controlling systolic and diastolic hypertension in elderly adults. These findings need to be confirmed in other populations, including younger adults, and clinical trials may better establish the potential effect of antihypertensive medications on the risk of incident CBI and worsening leukoaraiosis.

HIGHLIGHTS.

• Persistently elevated SBP is associated with incident CBI.

• Persistently elevated DBP is associated with worsening white matter disease.

• This suggests that hypertension is a major risk factor in the development of both forms of silent cerebrovascular diseases.

• The exact relationship of blood pressure levels and variability to cerebrovascular diseases may differ between CBI, white matter disease, and stroke, suggesting potential differences in pathophysiology.

Abbreviations

CBI

covert brain infarction

CHS

Cardiovascular Health Study

WMG

white matter grade