Blood Pressure Variation and Subclinical Brain Disease

Abstract

BACKGROUND

Large blood pressure (BP) variability may contribute to stroke and dementia, but the mechanisms are largely unknown.

OBJECTIVES

This study investigated the association of BP variation, considering its magnitude and direction, with the presence and progression of subclinical brain disease in the general population.

METHODS

This study included 2,348 participants age ≥55 years from a prospective cohort study. BP was measured at each visit every 3 to 4 years from 1990 onward. Brain magnetic resonance imaging (MRI) was performed at all visits from 2005 onward. The authors primarily assessed variation as the absolute difference in BP divided by the mean over 2 sequential visits for both systolic BP (SBP) and diastolic BP (DBP), and further assessed the direction of the variation. The authors investigated the multivariate-adjusted associations of BP variation with subsequent measurements of MRI markers of cerebral small vessel disease, brain tissue volumes, and white matter microstructural integrity. Longitudinal changes in these markers also were assessed.

RESULTS

A large SBP variation (top vs. bottom tertiles), measured on average 7 years preceding brain MRI, was associated with higher odds of having severe white matter hyperintensities (WMH) (odds ratio [OR]: 1.32; 95% confidence interval [CI]: 1.21 to 1.43), lacunes (OR: 1.25; 95% CI: 1.04 to 1.48), and microbleeds (OR: 1.16; 95% CI: 1.03 to 1.31). Similarly, this variation was associated with smaller total brain volume and worse white matter microstructural integrity (all p < 0.001). A large SBP variation was also associated with the progression of WMH (rate ratio [RR]: 1.14; 95% CI: 1.02 to 1.27). Higher burdens of these brain imaging markers were observed with both large rises and falls in SBP. Similar findings were observed for DBP variation.

CONCLUSIONS

Elevated BP variation was associated with a wide range of subclinical brain structural changes, including MRI markers of cerebral small vessel disease, smaller brain tissue volumes, and worse white matter microstructural integrity. These subclinical brain changes could be the underlying mechanisms linking BP variation to dementia and stroke.

Subclinical brain vascular disease, such as cerebral small vessel disease, is prevalent in older people (1). In particular, cerebral small vessel disease, manifested in various forms such as white matter hyperintensities (WMH), lacunes, and microbleeds on magnetic resonance imaging (MRI), contributes significantly to stroke, cognitive impairment, and dementia (2). Reduced white matter microstructural integrity has also been suggested as a more sensitive imaging marker that may manifest before macrostructural changes such as WMH (3). The rapid development of neuroimaging techniques has enabled us to detect these early-stage cerebrovascular changes in vivo. Identifying modifiable risk factors of these subclinical changes will, therefore, facilitate early interventions to prevent stroke and dementia (4).

High blood pressure (BP) plays a pivotal role in the development of stroke, dementia, and subclinical brain disease (5). However, the positive associations of BP level with the risk of stroke, dementia, and subclinical brain disease somehow appear to attenuate or be reversed in late life (6–8). Emerging evidence suggests that BP variation over periods of hours, days, and years may contribute to the risk of stroke and dementia beyond BP level (9,10). The relationship between BP variation and subclinical cerebral vascular disease is less well-recognized, and previous studies are largely limited by small sample size and cross-sectional associations with inconsistent findings (11). The association of BP variation over a period of years with subsequent progression of subclinical brain disease remains unclear. We also do not know whether this putative association differs by the rise or fall in BP. Therefore, we investigated the association of BP variation with the presence and progression of subclinical brain disease measured by brain structural imaging markers. We assessed BP variation between sequential visits every 4 years on average, considering both its direction and magnitude.

METHODS

STUDY DESIGN AND PARTICIPANTS.

The study is embedded in the Rotterdam Study, a prospective cohort study initiated in 1990 in Rotterdam, the Netherlands. A detailed description has been published elsewhere (12). Briefly, 7,983 participants aged ≥55 years were recruited in 1990, and 3,011 participants aged ≥55 years were added in 2000. Follow-up visits are performed every 3 to 4 years and BP measurements are taken at each visit. Brain MRI has been implemented in the core protocol of the Rotterdam Study since 2005, and participants were invited to repeat MRI scans during follow-up visits every 3 to 4 years (13). As of September 2015, 2,680 participants (77.3% of 3,465 eligible invitees) completed ≥1 brain MRI examinations. The present study includes all participants who met all the following criteria: 1) completed ≥2 visits at which BP was measured; 2) free of dementia and stroke at the time of BP measurement; and 3) underwent ≥1 brain MRI scan following BP measurements repeated at 2 visits. Ultimately, 2,348 participants were eligible for the current study. A total of 4,127 MRI scans from all participants were assessed and the median scan interval was 3.4 years (interquartile range: 2.6 to 4.4 years). Changes of brain MRI markers over 2 sequential scans were assessed for 1,109 participants who underwent ≥2 MRI scans (Supplemental Figure 1).

The Rotterdam Study has been approved by the Medical Ethics Committee of the Erasmus MC and by the Dutch Ministry of Health, Welfare and Sport (Population Screening Act WBO). Written informed consent was obtained from all participants.

VARIATION IN BLOOD PRESSURE.

At each research center visit, after at least 5 min rest in a seated position, 2 BP measurements were taken on the right upper arm. The mean of these 2 measurements was used for that visit. Each participant contributed to a median of 2 visits (range: 2 to 5 visits) with valid BP measurements. Variation in BP over 2 sequential visits was assessed at the latter of the 2 visits using 2 measures: 1) BP variation (see later in this paper) as the primary measure; and 2) rise or fall in BP as the secondary measure to differentiate the direction of BP variation. The association of this variation with dementia risk has been reported elsewhere (10). Specifically, BP variation was calculated as the absolute difference in BP divided by the mean BP over 2 sequential visits (|difference|/mean). Rise or fall in BP was defined as the difference in BP between the 2 visits divided by the mean ([latter—earlier]/mean). To account for different visit intervals, both measurements were scaled to percentage per year, assuming a constant rate of change between the 2 visits. Both measures were assessed as repeated measurements, first assessed at the second visit using BP of the first 2 visits, and then updated at the third visit using BP of the second and third visits, and so on. A total of 4,410 repeated measurements on BP variation were derived for 2,348 participants. We assessed variation in both systolic blood pressure (SBP) and diastolic blood pressure (DBP) in the same way. Given that the variation in SBP and DBP were highly correlated (r = 0.60, p < 0.001) and the primary findings were similar, we mainly present the findings on SBP variation on both continuous and categorical scales. For categorical SBP variation, it was divided into tertiles to reflect small variation (<1.4%/year), moderate variation (1.4% to 3.3%/year), and large variation (>3.3%/year). For the direction of variation in BP, a large fall in SBP was defined as the bottom 20% (<−1.5% per year) and a large rise was defined as the top 20% (>3.5% per year) of all measurements.

BRAIN STRUCTURE AND MICROSTRUCTURE MEASURES.

Brain MRI was performed on a single 1.5-T MRI scanner (GE Healthcare) using a standardized protocol (13) in accordance with the widely accepted criteria (14). Four high-resolution axial sequences were obtained, including a T1-weighted sequence, a proton density–weighted sequence, a fluid-attenuated inversion recovery sequence, and a T2*-weighted gradient-recalled echo sequence. No contrast material was administered. Supratentorial brain tissue volumes, including gray matter, white matter, and WMH, were quantified using automated brain tissue segmentation and were inspected visually and corrected manually if needed (13,15). Total brain tissue volume was defined as the sum of the volume of gray matter, normal-appearing white matter, and WMH. As an important marker for Alzheimer’s disease (16), total hippocampal volume (defined as the sum of the left and right hippocampus) was also obtained by processing T1-weighted MRIs using FreeSurfer version 6.0 (Athinoula A. Martinos Center for Biomedical Imaging, Boston, Massachusetts) (17). For volumetric measures, 405 (10%) of 4,127 MRI scans with supratentorial cortical infarcts and suboptimal segmentation were excluded. Focal markers of cerebral small vessel disease were visually rated by trained research physicians (13). Specifically, lacunes were defined as subcortical lesions ≥3 mm and <15 mm with the same signal intensity as cerebrospinal fluid on all sequences and a hyperintense rim on fluid-attenuated inversion recovery sequence in the supratentorial region. Micro-bleeds were defined as focal round to ovoid areas <10 mm of low signal intensity on T2*-weighted imaging. From March 2006 onward, diffusion-tensor imaging (DTI) was included in the protocol. Data were pre-processed using a standardized pipeline (18). Global fractional anisotropy (FA) and mean diffusivity (MD) metrics in the normal-appearing white matter were assessed (19). FA measures the directionality of diffusion, whereas MD measures the overall magnitude of water diffusion. Lower FA and higher MD values indicate lower microstructural integrity of white matter.

We primarily assessed the presence of established markers of cerebral small vessel disease at a single MRI scan and its progression over 2 sequential scans an average of 3.4 years apart. The presence of cerebral small vessel disease markers includes: 1) high WMH burden, defined as white matter hyperintensity volume within the highest quartile; 2) the presence of (≥1) lacunes; and 3) the presence of (≥1) microbleeds. We assessed these markers both individually and as a composite outcome (20), the latter defined as the presence of any of the preceding 3 markers. Consistently, the progression of cerebral small vessel disease over 2 sequential scans was assessed among individuals who underwent 2 or more MRI scans. It includes: 1) rate of change in WMH volume (i.e., volumetric changes divided by scan intervals) within the highest quartile; 2) ≥1 new lacunes on the second scan; and 3) ≥1 new microbleeds on the second scan. Similarly, the progression of these markers was assessed both individually and as a composite outcome (defined as the progression of any of these 3 markers). We additionally examined the continuous measures of cerebral small vessel disease, including WMH volume, lacune count, and microbleed count.

We also assessed brain tissue volumes, including total brain tissue, gray matter, normal white matter, and total hippocampal volume. To correct for head size, all these volumetric measures were expressed as the percentage of total intracranial volume (i.e., the sum of gray and white matter and cerebrospinal fluid volumes). We also assessed the rate of change in brain tissue volumes (calculated as the volumetric changes divided by scan intervals).

To further assess early-stage changes in cerebral white matter, we examined measures on white matter microstructural integrity (global FA and MD) and their rate of change over 2 sequential visits using methods consistent with those for brain tissue volumes as described previously.

OTHER MEASUREMENTS.

Information on demographic characteristics was collected at the first visit. During each visit, smoking habits, body mass index, total cholesterol, high-density lipoprotein cholesterol, and medication use were assessed with standardized protocols. Hypertension was defined as a resting BP exceeding 140/90 mm Hg or the use of BP-lowering medication. Diabetes mellitus was defined as a fasting glucose level of ≥7.0 mmol/l, or the use of antidiabetic medication. Cardiovascular disease, including stroke, coronary heart disease, heart failure, and atrial fibrillation, was assessed via interviews and verified by medical records (12,21). Dementia and stroke status were also screened at baseline and all follow-up visits according to standardized protocols. In addition, all participants were continuously monitored for dementia and stroke through electronic linkage with medical records from routine clinical care (22,23).

STATISTICAL ANALYSIS.

To examine the association of BP variation with MRI markers, we performed 2 sets of analyses referred to as “presence analyses” and “progression analyses” hereinafter (Supplemental Figure 2). Generalized estimating equations (GEEs) with a first-order autoregressive correlation matrix and robust (empirical) variance were used to account for the within-subject correlation of repeated observations (24). The use of empirical standard errors provides estimates that are robust to misspecification of within-subject variance (25). GEE models were implemented using the SAS GENMOD procedure. Specifically, presence analyses examined the relationship between BP variation and the presence of MRI markers at a later visit, whereas progression analyses examined the relationship between BP variation and the progression of (or change in) MRI markers between subsequent visits. For both analyses, we included only BP variation measured at visits before the assessment of MRI outcomes to clarify the temporal order of the associations. For the presence analysis, ultimately 2,348 participants contributed to 8,037 pairs of repeated measurements and the median time interval between the measurement of BP variation and MRI scan was 7.0 years (IQR: 5.2 to 12.5 years); for the progression analysis, 1,109 participants contributed to 5,585 pairs of repeated measurements and the median time interval between BP variation and MRI markers progression over 2 sequential visits (the earlier of the 2 visits) was 5.0 years (IQR: 1.2 to 10.9 years). Different link functions were used in GEE models as appropriate: logistic regression was used to estimate the odds ratios (OR) for the presence of MRI markers; Poisson regression was used to estimate incidence rate ratio (RR) of progression of MRI markers; linear regression was performed for brain volumetric measures as z-scores.

To control for potential confounding, we adjusted for age, sex, APOE genotype, smoking habits, weight status (assessed by body mass index), and history of diabetes in the final models. These covariates were updated simultaneously with the measurements of BP variation. For MRI measures on white matter microstructural integrity, we also performed an additional analysis with further adjustment for volumes of normal-appearing white matter and WMH to assess the association independent of brain macro-structural changes. All covariates, except age, were categorical; missing data (<10%) were handled by adding an additional category indicating missing values. For categorical measures of BP variation, the category with the smallest variation in BP was set as the reference group. Test for linear trends was performed by entering a single ordinal term into the models. In the presence of nonlinear trend, curvature was tested using restricted cubic splines (26).

In our secondary analyses, to examine the impact of antihypertensive medication on the observed associations, we stratified the analysis by antihypertensive medication use during the 2 sequential visits when BP variation was assessed (as continuous users, intermittent users, and nonusers). We also stratified the analyses by age, sex, and baseline hypertension to identify potential effect modifications. To further assess whether short-term associations differ from long-term associations, we stratified the analysis by the time windows of exposure (i.e., for BP variation assessed within ≤7 [median] years vs. >7 years of MRI scan). Interaction was formally tested on a multiplicative scale by adding a product term to the model.

The following analyses were performed to test the robustness of the main findings: 1) with additional adjustment for mean BP level. It was not adjusted in the primary model because BP variation is expressed as the percentage of mean BP level, which has considered BP level; and 2) further adjusting for history of cardiovascular disease and antihypertensive medication use.

All effect estimates are given with corresponding 95% confidence intervals. All p values presented are 2 sided, with a p value of 0.05 or less determined by false discovery rate considered statistically significant, after correcting for multiple testing (27). Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc.).

RESULTS

Of 2,348 participants, 1,312 (55.9%) were women and the mean (SD) age was 61.8 (5.0) years. Table 1 describes the participant characteristics. Table 2 summarizes characteristics of MRI markers and their changes over sequential visits on average 3.4 years apart.

TABLE 1.

Participant Characteristics at Baseline^*

	Presence Analysis† (n = 2,348)	Progression Analysis† (n = 1,109)
Age, yrs	61.8 ± 5.0	60.9 ± 4.4
Women	55.9	52.2
APOE genotype
ε3/ε3	58.0	58.4
ε2/ε2 or ε2/ε3	14.1	14.9
ε2/ε4 or ε3/ε4	23.3	22.3
ε4/ε4	1.7	1.9
Missing	2.9	2.5
Smoking status
Never	24.4	25.2
Past	47.9	48.8
Current	26.3	24.7
Missing	1.4	1.4
Weight status‡
Normal weight	36.5	37.8
Overweight	47.6	46.3
Obese	13.5	13.6
Missing	2.4	2.3
History of diabetes	8.3	6.5
History of cardiovascular disease	5.2	3.9
Hypertension at baseline	45.1	43.4
Antihypertensive treatment at baseline	19.0	17.8
SBP, mm Hg	134 ± 20	133 ± 20
DBP, mm Hg	76 ± 11	76 ± 11
Magnitude of SBP variation, %/yr	2.2 (1.0 to 4.0)	2.1 (1.0 to 3.8)
Rise or fall in SBP, %/yr	1.1 (−0.9 to 3.0)	1.1 (−0.8 to 2.9)

Values are mean ± SD, %, or median (interquartile range).

^*Characteristics at the first visit after cohort entry.

^†Presence analyses examine the relationship between blood pressure (BP) variation and the presence of cerebrovascular markers at a later magnetic resonance imaging (MRI) visit, while progression analyses examine the relationship between BP and the change in MRI markers between subsequent visits (see text and Supplemental Figure 2).

^‡Weight status was assessed by body mass index (BMI), with overweight defined as 25# BMI≤30 kg/m² and obesity defined as BMI ≥30 kg/m².

DBP = diastolic blood pressure; SBP = systolic blood pressure.

TABLE 2.

Description of MRI Scans and Imaging Marker Profiles

MRI Profiles	Overall
Participants	2,348
Total number of scans	4,127
Average scan intervals, yrs	3.4 (2.6–4.4)
Age at MRI scan, yrs	74 ± 7
Total ICV, ml	1,136 ± 112
Brain tissue volumes, % of ICV
Total brain tissue	79.9 ± 3.8
Gray matter	45.6 ± 2.9
White matter	34.3 ± 3.5
Hippocampus	0.57 ± 0.06
Markers of cerebral small vessel disease
White matter hyperintensities, % of ICV	0.5 (0.3–1.0)
Presence of lacunar infarcts	313 (11.8)
Presence of cerebral microbleeds	750 (28.4)
Measures of white matter microstructural integrity
Fractional anisotropy	0.34 ± 0.02
Mean diffusivity,10−3 mm2/s	0.77 ± 0.03
Change in brain tissue volumes, % of ICV per yr
Total brain tissue	−0.32 ± 0.5
Gray matter	−0.10 ± 0.9
White matter	−0.23 ± 0.8
Hippocampus	−0.003 ± 0.006
Progression of markers of cerebral small vessel disease
White matter hyperintensities, % of ICV per yr	0.02 (0.0030.06)
Lacunar infarcts	100 (5.1)
Cerebral microbleeds	189 (9.6)
Change in white matter microstructural integrity
Change in fractional anisotropy, 10−3 per yr	−1.3 ± 4.1
Change in mean diffusivity (10−6 mm2/s per yr)	3.8 ± 6.4

Values are n, median (IQR), mean ± SD, or n (%), unless otherwise indicated. ICV = intracranial volume; MRI = magnetic resonance imaging.

SBP VARIATION AND CEREBRAL SMALL VESSEL DISEASE.

Table 3 and Figure 1 show the multivariate-adjusted associations of SBP variation with the presence of cerebral small vessel disease markers. The odds of having any marker of cerebral small vessel disease at a subsequent scan examination was higher with large SBP variation (OR comparing the top vs. bottom tertile: 1.30; 95% confidence interval [CI]: 1.18 to 1.45). Similar association with large SBP variation was observed for a high burden of WMH (OR: 1.32; 95% CI: 1.21 to 1.43), the presence of microbleeds (OR: 1.16; 95% CI: 1.03 to 1.31), and the presence of lacunes (OR: 1.25; 95% CI: 1.04 to 1.48). An additional analysis using continuous measures of these MRI markers revealed consistent results, showing significant association of SBP variation with WMH volume and the count of lacunes (both p for trend <0.05) and borderline-significant association with the count of microbleeds (Supplemental Table 1).

TABLE 3.

SBP Variation and Cerebral Small Vessel Disease

	SBP Variation (by Tertile)				SBP Variation (Continuous)
MRI Markers	Small (<1.4%/yr)	Moderate (1.4%–3.3%/yr)	Large (>3.3%/yr)	p Value for Trend	Per SD	p Value
Odds ratios for the presence of cerebral small vessel disease*
WMH	1.00 (ref)	1.09 (1.01–1.19)	1.32 (1.21–1.43)	<0.001	1.15 (1.11–1.20)	<0.001
Microbleeds	1.00 (ref)	1.09 (0.97–1.23)	1.16 (1.03–1.31)	0.017	1.07 (1.02–1.13)	0.012
Lacunes	1.00 (ref)	1.06 (0.89–1.27)	1.25 (1.04–1.48)	0.017	1.07 (1.00–1.15)	0.057
Composite outcome†	1.00 (ref)	1.13 (1.02–1.25)	1.30 (1.18–1.45)	<0.001	1.13 (1.08–1.19)	<0.001
Rate ratios for the progression of cerebral small vessel disease*
WMH	1.00 (ref)	1.03 (0.93–1.15)	1.14 (1.02–1.27)	0.023	1.04 (1.00–1.09)	0.081
Microbleeds	1.00 (ref)	1.11 (0.94–1.32)	1.09 (0.91–1.30)	0.339	1.05 (0.98–1.13)	0.176
Lacunes	1.00 (ref)	0.91 (0.66–1.26)	1.00 (0.73–1.37)	0.970	1.08 (0.92–1.27)	0.338
Composite outcome‡	1.00 (ref)	1.04 (0.95–1.14)	1.14 (1.03–1.25)	0.008	1.04 (1.00–1.08)	0.054

Values are odds ratios or rate ratios (95% confidence intervals).

^*Presence analyses examine the relationship between BP variation and the presence of cerebrovascular markers at a later MRI visit, while progression analyses examine the relationship between BP and the change in MRI markers between subsequent visits (see text and Supplemental Figure 2). Association estimates were adjusted for age, sex, smoking habits, body mass index, APOE genotype, and history of diabetes.

^†Defined as the presence of any of the following 3 markers: white matter hyperintensities (WMH), microbleeds, and lacunes.

^‡Defined as the progression of any of the following 3 markers: WMH, microbleeds, and lacunes.

Abbreviations as in Table 1.

FIGURE 1. The Association of SBP Variation With the Presence of Cerebral Small Vessel Disease.

*SBP variation was <1.4%/year (bottom tertile), 1.4% to 3.3% (middle tertile), and >3.3% (upper tertile). †Association estimates after adjusting for age, sex, smoking habits, body mass index, APOE genotype, and history of diabetes. CI = confidence interval; SBP = systolic blood pressure.

Consistently, as shown in Table 3, a higher risk of the progression of any of cerebral small vessel disease markers was observed with large SBP variation (RR: 1.14; 95% CI: 1.03 to 1.25). More specifically, large SBP variation was associated with a fast progression in WMH (RR comparing top vs. bottom tertile: 1.14; 95% CI: 1.02 to 1.27), while the associations of SBP variation with the risk of developing incident microbleeds and incident lacunes were not statistically significant (both p for trend >0.05).

Table 4 summarizes the association of the rise or fall in SBP with the presence and progression of established markers of cerebral small vessel diseases. The associations of SBP variation with the presence of cerebral small vessel disease markers were irrespective of the rise or fall in SBP. The directions of association estimates were consistent for the progression of cerebral small vessel disease as a composite outcome, as well as for the progression of WMH and microbleeds, although none of these associations was statistically significant.

TABLE 4.

Rise or Fall in SBP and Cerebral Small Vessel Disease

	Rise or Fall in SBP (%/yr)
MRI Markers	Large Fall (<−1.5%)	Stable (−1.5% to 3.5%)	Large Rise (>3.5%)	p Value for Trend*
Odds ratio for the presence of cerebral small vessel disease markers (presence analysis)†
WMH	1.28 (1.18–1.40)	1.00 (ref)	1.20 (1.11–1.30)	<0.001
Microbleeds	1.16 (1.03–1.32)	1.00 (ref)	1.19 (1.05–1.34)	0.003
Lacunes	1.24 (1.05–1.48)	1.00 (ref)	1.15 (0.96–1.38)	0.056
Composite outcome‡	1.33 (1.19–1.47)	1.00 (ref)	1.22 (1.10–1.35)	<0.001
Rate ratio for the progression of cerebral small vessel disease markers (progression analysis)‡
WMH	1.12 (1.00–1.25)	1.00 (ref)	1.15 (1.03–1.28)	0.090
Microbleeds	1.17 (1.00–1.38)	1.00 (ref)	1.11 (0.92–1.33)	0.356
Lacunes	1.22 (0.90–1.65)	1.00 (ref)	0.85 (0.60–1.21)	0.662
Composite outcome§	1.14 (1.04–1.25)	1.00 (ref)	1.14 (1.04–1.25)	0.127

Values are odds ratios or rate ratios (95% confidence intervals).

^*p value for nonlinear trend.

^†Presence analyses examine the relationship between BP variation and the presence of cerebrovascular markers at a later MRI visit, whereas progression analyses examine the relationship between BP and the change in MRI markers between subsequent visits (see text and Supplemental Figure 2). Association estimates were adjusted for age, sex, smoking habits, body mass index, APOE genotype, and history of diabetes.

^‡Defined as the presence of any of the following 3 markers: WMH, microbleeds, and lacunes.

^§Defined as the progression of any of the following 3 markers: WMH, microbleeds, and lacunes.

Abbreviations as in Tables 1 and 3.

SBP VARIATION AND BRAIN TISSUE VOLUMES.

As shown in Table 5 and Figure 2, increased SBP variation was associated with a smaller total brain tissue volume, as well as volumes of white matter, gray matter, and hippocampus (all p for trend <0.05), with the magnitudes of association appearing larger for white matter and hippocampus than for gray matter. Increased SBP variation was also associated with a faster reduction in total brain tissue volumes over 2 sequential scans (p value for trend = 0.032), while the associations with changes in subregions (i.e., white matter, gray matter, or hippocampus) were less consistent and not statistically significant (Table 5). Smaller brain tissue volumes were observed with both large falls and rises in SBP (Supplemental Table 2).

TABLE 5.

Association of SBP Variation With Brain Tissue Volumes

	SBP Variation (by Tertile)				SBP Variation (Continuous)
MRI Markers	Small (<1.4%/yr)	Moderate (1.4%–3.3%/yr)	Large (>3.3%/yr)	p Value for Trend	Per SD	p Value
Brain tissue volumes (in z-score)*
Total brain tissue	1.00 (ref)	−0.02 (−0.05 to 0.01)	−0.20 (−0.24 to −0.17)	<0.001	−0.10 (−0.12 to −0.09)	<0.001
Gray matter	1.00 (ref)	−0.03 (−0.08 to 0.02)	−0.08 (−0.13 to −0.03)	0.004	−0.04 (−0.06 to −0.02)	<0.001
White matter	1.00 (ref)	−0.01 (−0.05 to 0.03)	−0.22 (−0.26 to −0.18)	<0.001	−0.10 (−0.12 to −0.08)	<0.001
Hippocampus	1.00 (ref)	0.00 (−0.04 to 0.03)	−0.15 (−0.18 to −0.11)	<0.001	−0.08 (−0.09 to −0.06)	<0.001
Change in brain tissue volumes between subsequent visits (in z-score)*
Total brain tissue	1.00 (ref)	−0.09 (−0.17 to −0.01)	−0.09 (−0.18 to 0.00)	0.032	−0.04 (−0.08 to −0.01)	0.024
Gray matter	1.00 (ref)	−0.07 (−0.15 to 0.01)	−0.01 (−0.10 to 0.08)	0.738	0.00 (−0.04 to 0.04)	0.960
White matter	1.00 (ref)	0.03 (−0.04 to 0.10)	−0.02 (−0.10 to 0.06)	0.659	−0.01 (−0.04 to 0.02)	0.413
Hippocampus	1.00 (ref)	0.02 (−0.06 to 0.10)	−0.07 (−0.16 to 0.03)	0.175	−0.07 (−0.10 to −0.03)	<0.001

Values are z-scores (95% CI).

^*Association estimates were adjusted for age, sex, smoking habits, body mass index, APOE genotype, and history of diabetes. Abbreviations as in Table 1.

FIGURE 2. The Association of SBP Variation With Brain Tissue Volumes.

*SBP variation was <1.4%/year (bottom tertile), 1.4% to 3.3% (middle tertile), and >3.3% (upper tertile). †Association estimates after adjusting for age, sex, smoking habits, body mass index, APOE genotype, and history of diabetes. Abbreviations as in Figure 1.

SBP VARIATION AND WHITE MATTER MICROSTRUCTURAL INTEGRITY.

Table 6 and Figure 3show the association of SBP variation with white matter microstructural integrity. A larger SBP variation was associated with both lower global FA and higher MD in the primary multivariable adjusted models (beta [95% CI] comparing extreme tertiles of SBP variation: −0.08 [−0.13 to −0.04] for FA and 0.16 [0.12 to 0.20] for MD). After additional adjustment for normal-appearing white matter volume and WMH volume, the association remained statistically significant for both FA and MD measures (beta = −0.05 [−0.08 to −0.01] for FA and beta = 0.11 [0.07 to 0.16] for MD). When the changes in FA and MD over 2 sequential scans were assessed, their associations with SBP variation were attenuated substantially (p > 0.05; Table 6). The associations with lower FA and higher MD were observed for both large falls and rises in SBP (Supplemental Table 2).

TABLE 6.

Association of SBP Variation With White Matter Microstructural Integrity

	SBP Variation (by Tertile)				SBP Variation (Continuous)
MRI Markers	Small (<1.4%/yr)	Moderate (1.4%–3.3%/yr)	Large (>3.3%/yr)	p Value for Trend	Per SD	p Value
White matter microstructural integrity (in z-score)
Fractional anisotropy*	1.00 (ref)	−0.04 (−0.08 to 0.01)	−0.08 (−0.13 to −0.04)	<0.001	−0.04 (−0.06 to −0.02)	<0.001
Mean diffusivity*	1.00 (ref)	0.02 (−0.02 to 0.06)	0.16 (0.12 to 0.20)	<0.001	0.08 (0.07 to 0.10)	<0.001
Fractional anisotropy†	1.00 (ref)	−0.02 (−0.05 to 0.01)	−0.05 (−0.08 to −0.01)	0.006	−0.03 (−0.04 to −0.01)	<0.001
Mean diffusivity†	1.00 (ref)	0.01 (−0.04 to 0.05)	− (0.07 to 0.16)	<0.001	0.06 (0.04 to 0.08)	<0.001
Change in white matter microstructural integrity between subsequent visits (in z-score)*
Fractional anisotropy	1.00 (ref)	−0.03 (−0.22 to 0.17)	0.00 (−0.25 to 0.24)	0.952	0.02 (−0.12 to 0.16)	0.807
Mean diffusivity	1.00 (ref)	0.03 (−0.15 to 0.21)	0.07 (−0.17 to 0.31)	0.566	−0.02 (−0.24 to 0.20)	0.856

^*Association estimates were adjusted for age, sex, smoking habits, body mass index, APOE genotype, and history of diabetes.

^†With further adjustment for normal-appearing white matter volume and WMH volume in addition to the covariates adjusted in primary models.

Abbreviations as in Tables 1 and 3.

FIGURE 3. The Association of SBP Variation With White Matter Microstructural Integrity.

*SBP variation was <1.4% per year (bottom tertile), 1.4% to 3.3% (middle tertile), and >3.3% (upper tertile). †Association estimates after adjusting for age, sex, smoking habits, body mass index, APOE genotype, and history of diabetes. Abbreviations as in Figure 1.

SECONDARY AND SENSITIVITY ANALYSIS.

The primary results did not differ significantly by antihypertensive medication, age, sex, or baseline hypertension or by lengths of follow-up (Supplemental Figure 3). The associations of DBP variation with the presence and progression of cerebral small vessel disease were similar to and also appeared to be stronger than those for SBP variation (Supplemental Table 3). Additional adjustment for mean BP level revealed essentially the same results (Supplemental Table 4). Further adjustment for history of cardiovascular disease and antihypertensive medication showed similar findings (Supplemental Table 5).

DISCUSSION

In this population-based cohort study of older adults, we found that a large BP variation between sequential measurements on average 4 years apart was associated with the presence of established markers of cerebral small vessel disease, smaller brain tissue volumes, and worse white matter microstructural integrity (Central Illustration). Increased BP variation was also associated with a faster progression of WMH. These associations were similar irrespective of the direction of BP variation, and were also observed after accounting for mean BP level over the same period, with consistent associations observed for both SBP and DBP variation.

CENTRAL ILLUSTRATION. The Association of Systolic Blood Pressure Variability With the Subsequent Magnetic Resonance Imaging Measurements of Subclinical Brain Disease.

*Represents the absolute difference in SBP divided by mean SBP over 2 sequential visits on average 4 years apart scaled to percentage per year. SBP = systolic blood pressure; WMH = white matter hyperintensities.

COMPARISON WITH OTHER STUDIES.

Previous prospective cohort studies on long-term BP variability and brain imaging markers observed mixed findings, with small sample size and limited evidence on subsequent progression of MRI markers (11). Our study examined the association of BP variation with both the presence and the progression of a wider range of brain structural and microstructural measures in a larger sample, allowing for a better understanding of the integrated relationship between BP variation and brain health, especially of the early-stage vascular etiology of dementia and stroke. Our study also provides novel evidence on the association of long-term BP variation with white matter microstructural integrity. These findings are in line with a previous report linking increased SBP variation, measured by the same approach, to elevated risk of dementia and mortality (10). Our results are also in concert with the body of evidence relating long-term BP variability to stroke, coronary heart disease, and renal disease (28,29).

Our findings suggest a potentially detrimental effect of BP variation on cerebral white matter. First, the association of BP variation with both the presence and progression of WMH, in line with previous reports (11), seemed to be more pronounced compared with lacunes and microbleeds. Second, the association with BP variation also appeared stronger for white matter volume than for gray matter volume, consistent with the evidence showing that white matter could be more susceptible to vascular impairment (30). Third, increased BP variation was associated with worse white matter microstructural integrity, even after accounting for total white matter volume and for burden of WMH, suggesting that BP variation may contribute to white matter injury before brain macrostructural abnormality is present. The associations of BP variation with the changes in these brain imaging markers were attenuated substantially, possibly due to limited statistical power in progression analysis or short scan intervals. In addition, our study also observed worse subclinical brain disease profiles for both large rises and falls in BP, suggesting that it is the magnitude of BP variation rather than the direction (i.e., rises or falls) that plays a critical role, consistent with our previous report on BP variation and dementia risk (10).

POTENTIAL MECHANISMS.

The underlying mechanisms whereby BP variability is linked to subclinical brain vascular disease are largely unknown (31). First, large BP variability could increase pulsation of blood flow and dampen the smoothing of blood flow as it progresses to small arteries, particularly in highly perfused organs such as the brain, causing damage to brain microvasculature (32). Second, endothelial dysfunction may also play a part (33). As also suggested by animal studies, large BP variability could inhibit nitric oxide production and impair endothelial function, contributing to “neurovascular unit” injuries and blood-brain barrier abnormality, thereby contributing to cerebral small vessel disease (34). Third, extreme low and high BP levels resulting from excessive BP variability may escape the manageable limits of cerebral autoregulation and contributes to vasculature impairment (35). For example, BP below the lower limit of autoregulation may lead to hypoperfusion and related brain ischemic injury (36), whereas BP beyond the upper limit may lead to hypertensive encephalopathy (37). An alternative noncausal explanation is also possible. If subclinical brain impairment causes central autonomic dysfunction, it could also result in excessive BP variability (38). Notably, chronic BP variation over periods of years may have underlying mechanisms that differ from those of hour-to-hour and day-to-day BP variability (9), which warrant future investigation.

STUDY LIMITATIONS.

First, we assessed BP variation over 2 sequential visits using a practical and simple approach without statistical modeling. This approach minimized the issue of potential model misspecification, but substantial measurement error is possible, which may lead to underestimated associations (39). Second, the progression analyses were restricted to those who have at least 2 sequential visits with MRI scans; survival bias is thus possible. Third, to assess the progression of microbleeds and lacunes, side-by-side rating of repeated MRI scans of each participant would be ideal, but the process is labor-intensive and time-consuming. Our study, assessing progression by comparing measurements from different time points by independent readers, is prone to nondifferential misclassification, which may have also attenuated the true relationships. Similarly, measurement error in the changes in brain volumetric measures is also possible, although standardized protocol has been used across study visits. Finally, the findings on BP variation over periods of years may not be generalizable to variability over periods of days and other shorter periods.

CONCLUSIONS

Increased BP variation was associated with the presence of established markers of cerebral small vessel disease, smaller brain tissue volumes, worse white matter microstructural integrity, and faster progression of WMH. These findings suggest that BP variation may play an important role in etiology of subclinical brain changes, thereby contributing to stroke and dementia. If the observed association is causal, it further suggests the promising opportunity to prevent stroke and dementia at an early stage through targeting BP variation.

PERSPECTIVES.

COMPETENCY IN MEDICAL KNOWLEDGE:

Increased BP variation over a period of years may contribute to cerebral small vessel disease, smaller brain tissue volumes, and worse white matter microstructural integrity, independent of BP level.

TRANSLATIONAL OUTLOOK:

BP variation may play an important role in the early-stage etiology of stroke and dementia. Future studies are needed to investigate whether targeting BP variation, in addition to the conventional control of BP level, provides further protection against subclinical brain vascular disease and thereby stroke and dementia.

ABBREVIATIONS AND ACRONYMS

BP

blood pressure

CI

confidence interval

CSVD

cerebral small vessel disease

DBP

diastolic blood pressure

DTI

diffusion-tensor imaging

FA

fractional anisotropy

MD

mean diffusivity

MRI

magnetic resonance imaging

OR

odds ratio

RR

rate ratio

SBP

systolic blood pressure

WMH

white matter hyperintensities