Abstract
Objective To test whether elevated blood pressure (BP) relates to grey matter volume (GMV) changes in young adults who had not previously been diagnosed as hypertensive (systolic BP (SBP)/diastolic BP (DBP)≥140/90 mmHg).
Methods We associated BP with GMV from structural 3 Tesla T1-weighted MRI of 423 healthy adults between 19-40 years (mean age=27.7±5.3 years, 177 women, SBP/DBP=123.2/73.4±12.2/8.5 mmHg). Data originated from four previously unpublished cross-sectional studies conducted in Leipzig, Germany. We performed voxel-based morphometry on each study separately and combined results in image-based meta-analyses (IBMA) to assess cumulative effects across studies. Resting BP was assigned to one of four categories: (1) SBP<120 and DBP<80 mmHg, (2) SBP 120-129 or DBP 80-84 mmHg, (3) SBP 130-139 or DBP 85-89 mmHg, (4) SBP≥140 or DBP≥90 mmHg.
Results IBMA yielded: (a) lower regional GMV was correlated with higher peripheral BP; (b) lower GMV with higher BP when comparing individuals in sub-hypertensive categories 3 and 2, respectively, to those in category 1; (c) lower BP-related GMV was found in regions including hippocampus, amygdala, thalamus, frontal and parietal structures (e.g. precuneus).
Conclusions BP≥120/80 mmHg was associated with lower GMV in regions that have previously been related to GM decline in older individuals with manifest hypertension. Our study shows that BP-associated GM alterations emerge continuously across the range of BP and earlier in adulthood than previously assumed. This suggests that treating hypertension or maintaining lower BP in early adulthood might be essential for preventing the pathophysiological cascade of asymptomatic cerebrovascular disease to symptomatic end-organ damage, such as stroke or dementia.
Author contributions
Study concept and design: Schaare, Villringer.
Statistical analysis: Schaare.
Acquisition or interpretation of data: All authors.
Drafting of the manuscript: Schaare, Villringer.
Critical revision of the manuscript: All authors.
Acknowledgements
We thank all volunteers for their participation in any of the studies. Furthermore, we thank all researchers, technicians and students who planned, collected, entered and curated data used in this manuscript.
Introduction
Hypertension (HTN) is highly prevalent and the leading single risk factor for global disease burden and overall health loss1,2. The risk for insidious brain damage and symptomatic cerebrovascular disease (CVD, e.g. stroke and vascular dementia) multiplies with manifestation of HTN3. Midlife HTN is a major risk factor for late-life cognitive decline and has been associated with risk for dementia, including late-onset Alzheimer’s disease (AD)3–5.
Importantly, HTN is also related to sub-clinical functional6,7 and structural5–14 brain changes, or asymptomatic CVD, including brain volume reductions in the medial temporal and frontal lobes5,6,9–11,15. Hippocampal volumes, in particular, have been consistently associated with HTN-related reductions5,9,10,15. Furthermore, computational anatomy has been employed to detect subtle cerebral changes, such as microstructural white matter (WM) alterations 13 or reductions in regional grey matter5,11, in middle-aged and older adults with elevated blood pressure (BP).
Recent statements suggest that symptomatic clinical disease, resulting from elevated BP, could be prevented by avoiding primary BP elevations and sub-clinical target organ damage (including brain damage) in early adulthood and middle-age3,16,17. However, effects of elevated BP on adult brains before the age of 40 are unclear. Preliminary evidence from 32 young, normotensive adults, showed that BP-reactivity correlated with lower amygdala volume18.
This study aimed to investigate if subtle structural brain changes occur in early adulthood (<40 years) at sub-hypertensive BP levels. We hypothesized that higher BP would relate to lower regional grey matter volume (GMV) and that this would predominantly affect frontal and medial temporal lobes, including amygdala and hippocampus.
Methods
We applied voxel-based morphometry19,20 (VBM) to four previously unpublished independent datasets including young adults aged between 19-40 years without previous diagnosis of HTN or any other severe, chronic or acute disease. Results from each dataset were combined in image-based meta-analyses (IBMA) for well-powered, cumulative evaluation of findings across study differences (i.e. recruitment procedure, inclusion criteria and data acquisition, Figure 1, data available from bioRxiv (Supplementary Table 1) https://doi.org/10.1101/239160).
Participants
We included cross-sectional data of 423 young participants from four samples. The samples were drawn from larger studies that were conducted in Leipzig, Germany, between 2010-2015: 1. Leipzig Study for Mind-Body-Emotion Interactions (Babayan et al., under review), 2. Neural Consequences of Stress Study (Reinelt et al., in preparation), 3. Neuroanatomy and Connectivity Protocol21, 4. Leipzig Research Centre for Civilization Diseases (LIFE)22.
The objective of study 1 was to cross-sectionally investigate mind-brain-body-emotion interactions in a younger (20-35 years) and an older (59-77 years) group of 228 healthy volunteers. Study 2 aimed to investigate neural correlates of acute psychosocial stress in 79 young (18-35 years), healthy, non-smoking men. The study protocol for the baseline assessment of participants in study 2 was adapted from the protocol in study 1. In study 3, 194 healthy volunteers between 20-75 years of age participated in one session of MRI and completed an extensive assessment of cognitive and personality measures. This dataset aimed to relate intrinsic functional brain connectivity with cognitive faculties, self-generated mental experience, and personality features. Together, studies 1-3 constitute the MPI-Leipzig Mind-Brain-Body database. Study 4 (LIFE-Study) is a population-based dataset in the city of Leipzig, Germany, with the objective to investigate the development of major modern diseases. Overall, 10000 participants were randomly drawn from the local population of whom 2667 underwent MRI and detailed screening. With dementia being one of the key scientific topics in this study, most participants in the MRI-subcohort were adults above the age of 60 years. The exact inclusion procedure and numbers for the current investigation is depicted in Figure 1. Inclusion criteria for our study were age between 19-40 years, availability of high-resolution structural T1-weighted MRI and ≥1 BP measurements. Participants were excluded in case of previously diagnosed HTN, intake of antihypertensive drugs or severe diseases (data available from bioRxiv (Supplementary Table 1) https://doi.org/10.1101/239160).
Standard Protocol Approvals, Registrations, and Patient Consents
The studies were in agreement with the Declaration of Helsinki and approved by the ethics committee of the medical faculty at the University of Leipzig, Germany (ethics reference numbers study 1: 154/13-ff, study 2: 385/14-ff, study 3: 097/15-ff, study 4: 263-2009-14122009). Before entering the studies, participants gave written informed consent.
Blood pressure measurements
Systolic (SBP) and diastolic blood pressure (DBP) were measured at varying times of day using an automatic oscillometric blood pressure monitor (OMRON M500 (samples 1-3), 705IT (sample 4), OMRON Medizintechnik, Mannheim, Germany) after a seated resting period of 5 min. In sample 1, three measurements were taken from participants’ left arms on three separate occasions within two weeks. In sample 2, two measurements were taken from participants’ left arms on two separate occasions on the same day. In sample 3, blood pressure was measured once before participants underwent MRI. In sample 4, the procedure consisted of three consecutive blood pressure measurements, taken from the right arm in intervals of 3 minutes. In each sample, all available measurements per participant were averaged to one systolic and one diastolic blood pressure value. These averages were used for classification of BP.
Neuroimaging
MRI was performed at the same 3 Tesla MAGNETOM Verio Scanner (Siemens, Erlangen, Germany) for all studies with a 32-channel head coil. Whole-brain 3-dimensional T1-weighted volumes with a resolution of 1 mm isotropic were acquired for the assessment of brain structure. T1-weighted images in sample 4 were acquired with a standard MPRAGE protocol (inversion time TI=900 ms, repetition time TR=2300 ms, echo time TE=2.98 ms, flip angle FA=9°, field of view FOV=256×240×176 mm3, voxel size=1×1×1 mm3), while T1-weighted images in samples 1-3 resulted from an MP2RAGE protocol (TI1=700 ms, TI2=2500 ms, TR=5000 ms, TE=2.92 ms, FA1=4°, FA2=5°, FOV=256×240×176 mm 3, voxel size=1×1×1 mm3). Grey and white matter contrast are comparable for the two sequence protocols1,2, but additional preprocessing steps were performed for MP2RAGE T1-weighted images (data available from bioRxiv (Additional Methods) https://doi.org/10.1101/239160). Fluid-attenuated inversion recovery (FLAIR) images were acquired in all samples for radiological examination for incidental findings and for Fazekas scale ratings for white matter hyperintensities (WMH, Tables 1 and 2).
Data Processing and Statistical Analysis
Details on all analysis methods can be found in Supplementary Methods (data available from bioRxiv (Additional Methods) https://doi.org/10.1101/239160).
Blood pressure classification
For statistical analyses, all available BP measurements per participant were averaged to one mean SBP and DBP, respectively. Based on these averages, we categorized BP according to the European guidelines for the management of arterial hypertension23: category 1 (SBP<120 mmHg and DBP<80 mmHg), category 2 (SBP 120-129 mmHg or DBP 80-84 mmHg), category 3 (SBP 130-139 mmHg or DBP 85-89 mmHg) and category 4 (SBP≥140 mmHg or DBP≥90 mmHg).
Voxel-based morphometry (VBM): association of regional GMV and BP within each sample
For each of the four samples, 3 Tesla high-resolution T1-weighted 3-D whole-brain images were processed using VBM and the diffeomorphic anatomical registration using exponentiated lie algebra (DARTEL) method19,20 within SPM12. Voxel-wise general linear models (GLM) were performed to relate BP and GMV within each sample: we tested for a continuous relationship between GMV and SBP or DBP, in separate models, with a multiple linear regression t-contrast. The overall effect of BP category on GMV was tested with an Analysis of Variance (ANOVA) F-contrast. To assess differences in GMV between BP categories, the following pairwise t-comparisons were tested: (a) category 4 vs. category 1, (b) category 3 vs. category 1, (c) category 2 vs. category 1. All analyses included total intracranial volume (TIV), sex and age as covariates. The influence of body mass index (BMI) did not significantly contribute to the models and was thus not included as covariate in the analyses. We considered a sample eligible for image-based meta-analysis if its F- contrast effects exceeded an uncorrected peak-level threshold of p<0.001. Effects within each sample were explored at cluster-level p<0.05 with family-wise error correction for multiple comparisons.
Image-based meta-analysis (IBMA): association of regional GMV and BP across samples
To evaluate cumulative results from all samples while considering their heterogeneities, we combined the VBM outcome of each sample in IBMA. Meta-analyses were performed on the unthresholded t-maps with SDM software using default parameters24. Statistical significance of mean effect size maps was evaluated according to validated thresholds of high meta-analytic sensitivity and specificity24: voxel threshold=p<0.005, peak height threshold=SDM-Z>1.0 and cluster extent threshold=k≥10 voxels.
Exploratory IBMA for positive associations were performed in analogy to negative associations as described above.
IBMA of regions of interest (ROI): association of regional GMV and BP across samples in hippocampus and amygdala
We performed IBMA within atlas-defined masks to test if regional bilateral hippocampal and amygdalar volumes related to SBP, DBP and BP categories, respectively. The statistical thresholds were defined as p<0.05, SDM-Z>1.0 and k≥1 voxels.
Volumetry: association of total brain volumes and BP within the pooled sample
In addition to VBM and IBMA, we explored if total brain volumes (average volume over all voxels within a region) differed between BP categories. Specifically, we tested if estimated total intracranial volume, total grey matter volume, total white matter volume (WMV), total amount of white matter hyperintensities (WMH), total cerebrospinal fluid volume (CSFV), total left and right hippocampal and amygdalar volume differed between BP categories. WMH was assessed by Fazekas scale ratings from Fluid-attenuated inversion recovery (FLAIR) images25. For these comparisons within the total sample, we defined correlation models (for SBP and DBP as independent variable, respectively) and ANOVA models for BP category as independent variable. The models included the respective volume as dependent variable, as well as TIV (where applicable), sex, age, and sample (where applicable) as covariates. We considered p-values<0.05 as significant. The analyses were performed with R (3.2.3, R Core Team, 2015, Vienna, Austria; https://www.R-project.org/).
Data Sharing
Results (i.e. unthresholded whole-brain statistical maps) from VBM analyses of each sample and from all IBMAs can be found online in the public repository NeuroVault for detailed, interactive inspection (http://neurovault.org/collections/FDWHFSYZ/). Raw data of samples 1-3 available from https://www.openfmri.org/dataset/ds000221/.
Results
Sample characteristics
The characteristics of the total sample by BP category are reported in Table 1. The total sample included 423 participants between 19-40 years of whom 177 were women (42%). Mean (SD) age was 27.7 (5.3) years. SBP, DBP, and BMI differed between BP categories (all p<0.001). An effect of sex yielded that men were more frequent in higher BP categories (all p<0.001).
Table 2 shows differences in characteristics between the four included samples. The samples differed in almost all characteristic variables, specifically regarding sex, age, SBP, DBP, smoking status, and brain volumes (all p<0.001).
VBM: association of regional GMV and BP within each sample
Figure 2 shows differences in regional GMV between BP categories for each of the four samples tested with an ANOVA F-contrast. Results show significant clusters of various extents that were distributed heterogeneously between the samples. Exploration of sample-specific effects showed a cluster in the left posterior insula for the F-contrast (peak MNI coordinates [−38,−24,24], F=11.35, cluster size k=1239) as well as clusters in left inferior frontal gyrus ([−42,34,0], T=4.99, k=2039) and in right anterior cingulate cortex ([14,34,14], T=4.44, k=2086) for the contrast BP category 4<1 in sample 1. In sample 2, the contrast BP category 4<1 yielded a cluster in left planum polare ([−44,−22,−3], T=10.70, k=1151) and the contrast BP category 2<1 yielded a trend for a cluster in left middle temporal gyrus (pFWE=0.059, [−54,−28,−9], T=5.41, k=683). Furthermore, higher DBP was associated with a cluster of lower GMV in left middle temporal gyrus in sample 2 ([−57,−45,6], T=6.03, k=1180). All other comparisons yielded no suprathreshold voxels (all pFWE>0.05). The statistical maps for sample-specific effects can be inspected on NeuroVault.
IBMA: association of regional GMV and BP across samples
Meta-analytic parametric relations between lower GMV and higher BP
As expected, increases in systolic and diastolic BP were associated with lower GMV. Specifically, higher SBP related to lower GMV in right paracentral/cingulate areas ([8,−30,56], SDM-Z=−3.859, k=288), bilateral inferior frontal gyrus (IFG, left: [−40,30,0], SDM-Z=−3.590, k=49; right: [−40,30,0], SDM-Z=−3.394, k=16), bilateral sensorimotor cortex (left: [−58,−20,24], SDM-Z=−3.290, k=146; right: [48,0,48], SDM-Z=−3.196, k=127), bilateral superior temporal gyrus (left: [−52,−10,6], SDM-Z=−3.268, k=78; right: [64,−42,12], SDM-Z=−3.192, k=42), bilateral cuneus cortex (left: [−8,−76,18], SDM-Z=−3.019, k=27; right: [10,−68,26], SDM-Z=− 2.937, k=18), and right thalamus ([8,−28,2], SDM-Z=−2.977, k=45; Figure 3A, Table 3). Increases in diastolic BP were related to lower GMV in bilateral anterior insula (left: [− 36,26,6], SDM-Z=−3.876, k=266; right: [34,10,8], SDM-Z=−3.139, k=100), frontal regions ([−26,24,54], SDM-Z=−3.820, k=62), right midcingulate cortex ([4,−34,50], SDM-Z=−3.545, k=246), bilateral inferior parietal areas (left: [−46,−26,48], SDM-Z=−3.239, k=59; right: [44,- 44,50], SDM-Z=−3.188, k=18), and right superior temporal gyrus ([60,2,−12], SDM-Z=−2.991, k=35; Figure 3B, Table 3).
Meta-analytic differences in regional GMV between BP categories
Meta-analytic results for category 4 (highest BP) compared to category 1 (lowest BP) yielded lower regional GMV in frontal, cerebellar, parietal, occipital, and cingulate regions (Figure 3C). Table 3 describes the specific regions with lower GMV, including bilateral IFG (left: [−52,- 28,12], SDM-Z=−3.473, k=107; right: [40,30,26], SDM-Z=−3.093, k=10), right midcingulate cortex ([12,−42,48], SDM-Z=−2.854, k=11), and right precuneus ([10,−52,18], SDM-Z=−2.836, k=21).
We also compared GMV of individuals at sub-hypertensive levels (category 3 and 2, respectively) to GMV of individuals in category 1. Figure 3D shows meta-analysis results for the comparison between category 3 and category 1. Compared to category 1, category 3 was associated with lower GMV in bilateral IFG (left: [−40,30,2], SDM-Z=−2.598, k=24; right: [36,6,34], SDM-Z=−3.474, k=179), sensorimotor cortices (left: [−60,−20,36], SDM-Z=−2.857, k=205; right: [6,−28,54], SDM-Z=−3.119, k=179), bilateral middle temporal gyrus (left: [−56,−64,16], SDM-Z=−2.222, k=28; right: [48,−50,20], SDM-Z=−3.119, k=179), right insula ([36,8,−18], SDM-Z=−2.523, k=123), right occipital regions ([42,−74,12], SDM-Z=−2.454, k=25), left parietal ([−60,−20,36], SDM-Z=−2.857, k=205), bilateral thalamus (left: [−12,−32,0], SDM-Z=− 2.264, k=133; right: [20,−32,6], SDM-Z=−2.384, k=133), left anterior cingulate cortex ([−10,36,- 6], SDM-Z=−2.384, k=102), and left precuneus ([−12,−54,14], SDM-Z=−2.187, k=20; Table 3).
Figure 3E illustrates brain regions that yielded lower meta-analytic GMV comparing category 2 to category 1. These include left frontal regions ([−54,−10,14], SDM-Z=−3.407, k=230), right inferior occipital gyrus ([30,−96,−8], SDM-Z=−3.290, k=102), bilateral temporal regions (left: [− 34,−16,−30], SDM-Z=−3.164, k=133; right: [46,−74,12], SDM-Z=−2.734, k=26), left precuneus ([−8,−54,22], SDM-Z=−3.084, k=433) and inferior parietal regions (supramarginal, [54,−24,32], SDM-Z=−2.968, k=31, and angular gyri, [−36,−64,42], SDM-Z=−2.827, k=30), as well as midcingulate cortex ([8,−18,46], SDM-Z=−2.647, k=32; Table 3).
Meta-analytic differences in regional hippocampal and amygdalar volumes between BP categories
In this IBMA ROI comparison, higher SBP was correlated with lower bilateral posterior medial hippocampal volume (Figure 4). Higher DBP was correlated with lower left hippocampal volume and lower right anterior hippocampal volume. Furthermore, all higher BP categories were associated with lower regional hippocampal volume when compared to the lowest BP category 1 (Figure 4). Compared to category 1, BP category 4 was predominantly associated with lower left medial posterior hippocampus volume and category 3 with lower bilateral posterior and left medial hippocampus volume across samples. Lower volume comparing categories 2 and 1 was predominantly located in left lateral anterior hippocampus. Category 4 vs. category 1 and the associations with higher SBP and DBP also yielded significantly lower regional volume in bilateral amygdala, respectively. Effect sizes highly varied across samples (Figure 4).
Meta-analytic relations between higher GMV and higher BP
Exploratory analyses also revealed associations between higher BP and higher GMV (data available from bioRxiv (Supplementary Figure 1, Supplementary Table 2) https://doi.org/10.1101/239160, and NeuroVault maps). However, the cumulative positive effects are comparably weaker than the cumulative negative results (Table 3, negative: 17 out of 34 clusters from parametric analyses with SDM-Z>3.0; positive: 0 out of 28 clusters from parametric analyses with SDM-Z>3.0), they show greater heterogeneity across studies (negative: maximum I2=4.0; positive: maximum I2=56.3) and they seem to appear primarily in regions where standard preprocessing of brain tissue is suboptimal (e.g. in cerebellum/inferior occipital regions26). We therefore regard these findings as overall questionable. By also providing the results as statistical maps on NeuroVault, future investigations can use the data for reliability analyses of potential positive associations.
Volumetry in pooled sample: association of total brain volumes and BP
All associations of volumetric brain measures (TIV, total GMV, total, WMV, total CSFV, total hippocampal, total amygdalar volume and total WMH) with SBP or DBP in the correlation models, or with BP categories in the ANOVA models were below the statistical threshold (all p>0.05, Table 1).
Discussion
In this image-based meta-analysis of four previously unpublished independent samples, we found that elevated, sub-hypertensive BP was correlated with lower GMV in several brain regions, including parietal, frontal, and subcortical structures in young adults (<40 years). These regions are consistent with the lower regional GMV observed in middle-aged and older individuals with HTN5,6,9–11,15. Our results show that BP-associated GM alterations emerge earlier in adulthood than previously assumed and continuously across the range of BP.
Interestingly, we found that BP was associated with lower hippocampal volume. In older individuals, the hippocampal formation and surrounding structures are known to be affected by HTN5,8–10,15. In a meta-analytic evaluation of HTN-effects on total GMV and on hippocampal volume, lower volumes across studies were only consistently found for the hippocampus15. In analogy to those findings, our results showed that hippocampal volume was affected by higher BP in a considerably younger sample. It should be mentioned that the effects in hippocampus only exceeded statistical thresholds in ROI analyses, similar to previous reports of lower hippocampal volume in older samples with manifest HTN that were all ROI-based5,9,10,15. As potential pathophysiological explanations it has been proposed that medial temporal (and frontal regions) might be especially sensitive to effects of pulsation, hypoperfusion and ischemia, which often result from increasing pressure3,15.
We furthermore observed correlations between lower amygdalar and thalamic volumes and higher BP, notably already below levels which are currently regarded as hypertensive. Amygdalar and thalamic nuclei are substantially involved in BP regulation as they receive baroreceptor afferent signals via the brainstem and mesencephalic nuclei, relaying these signals to primary cortical regions of viscerosensory integration, such as anterior cingulate cortex and insula27. It has been shown that lower amygdalar volume correlates with increased BP-reactivity during cognitive demand among young normotensive adults18. Previous studies have reported lower thalamic volume in HTN5, heart failure28, asymptomatic carotid stenosis29, and aging30. Higher systolic BP has also been related to higher mean diffusivity of white matter thalamic radiations13. Our results are in line with accumulating evidence of amygdalar and thalamic involvement in cardiovascular (dys-) regulation but may also reflect early pathology in these regions. For example, occurrence of neurofibrillary tangles in thalamus has also been reported in the earliest stages of AD neuropathology34.
Beyond subcortical structures, we found lower volumes in cortical regions: cingulate volume and insular volume were markedly lower with higher DBP in the meta-analysis results and in the individual analyses of sample 1. As noted above, these regions constitute primary cortical sites of afferent viscerosensory integration and modulate homeostasis via efferents to brainstem nuclei27. Lesions in cingulate cortex and insula result in altered cardiovascular regulation, increased sympathetic tone31,32 and myocardial injury33. Both regions are also critical for the appraisal and regulation of emotion and stress27. Thus, structural alterations in these regions may contribute to insidious BP elevations via sympathetic pathways.
Frontal and parietal volumes were affected in all our statistical comparisons. The precuneus cortex, especially, was associated with lower GMV in BP categories 4, 3, and 2 compared to category 1. Our results of lower BP-related GMV in regions such as hippocampal, frontal and parietal areas highlight specific brain regions which are known to be vulnerable to putative vascular or neurodegenerative damage mechanisms5,6,8–11,15,35. Raised midlife BP is not only known to be a major risk factor for vascular dementia, but some reports suggest a link between HTN and AD-type pathophysiology3,4. For example in neuropathological studies, raised midlife BP has been associated with lower post-mortem brain weight, increased numbers of hippocampal neurofibrillary tangles, and higher numbers of hippocampal and cortical neuritic plaques8. Similarly, a potential pathophysiological link between HTN and AD has been supported by noninvasive MRI studies: regions referred to as AD-signature regions (including inferior parietal, precuneus cortices, and medial temporal structures) have been associated with cortical thinning years before clinical AD-symptoms arise35 and with brain volume reductions predicted by increasing BP from middle to older age5. In light of these previous results, our findings of lower BP-related GMV in AD-signature regions may be indicative of a link to AD-pathology at an even earlier age; however, this cannot be causally inferred from our cross-sectional data. In the study by Power et al.5, BP also predicted volume loss in non-AD-typical brain regions, such as frontal lobe and subcortical gray matter, which may relate to other (than AD-related) pathophysiological mechanisms. A similar pattern seems to be reflected in our findings of lower GMV related to higher BP in non-AD-typical regions.
Some previous studies did not find relations between HTN and lower brain volumes, but associated HTN with other forms of structural or functional brain alterations, such as white matter injury36 or reduced cerebral perfusion37. A key aspect of diverging results is the heterogeneity of methods used to assess brain volumes. Earlier investigations of BP effects on brain tissues have applied manual or automated volumetric methods to quantify total brain volumes in pre-selected ROIs9,10,12. The focus of this study was to employ computational anatomy methods to assess regional GM differences across the whole brain. We found significant differences between BP groups using VBM but not in the analysis of total brain volumes. This supports the view that VBM is a sensitive measure to quantify regional morphological differences38 which might be undetected from the analysis of total brain volumes alone. In addition, we employed random-effects IBMA which results in effects that are consistent across studies and that may otherwise be neglected at sub-threshold. Investigating effects of BP on regional vs. total brain volumes at all stages of health and disease thus warrants further research with standardized methods to identify neuropathological mechanisms.
Our data, however, do not allow inference on causality between lower brain volumes and HTN, which likely involves complex interactions of different pathophysiological mechanisms that still need to be fully elucidated. It is assumed that vascular stiffness, endothelial failure and a dysfunctional blood-brain barrier are precursors of cerebral small and large vessel disease that reduce cerebral blood flow, disturb autoregulatory adjustment and decrease vasomotor reactivity, which may impair perivascular central nervous waste clearance systems3. These mechanisms have also been suggested to potentially underlie the epidemiological connection between vascular risk factors, such as HTN, and AD3. The similarities between our findings and AD-signature regions (see above) would also be consistent with this putative link. Consequently, demyelination, apoptosis and intoxication of neurons and glial cells, as well as grey and white matter necrosis accumulate and may be reflected in neuroimaging on a macroscopic scale. Lower GMV assessed by VBM, as reported in our study, can thus arise from neuronal loss, but also from alterations of glial cells or composition of microstructural or metabolic tissue properties 39. Our findings point to an early effect of such mechanisms on GM integrity which is present in the absence of overt disease, such as HTN, and in young age. Indicators of early atherosclerosis in major peripheral arteries can already be detected in youth40. Recently, arterial stiffness has also been associated with WM and GM alterations among adults between 24 and 76 years of age41. Thus, already early and subtle vascular changes, deficient cerebral perfusion and impaired perivascular clearance systems may initiate and sustain neuropathology from early to late adulthood.
The cross-sectional design of our four study samples limits the interpretation frame for the results presented. Causality between BP and potential brain damage cannot be assessed with these data but is crucial for implications of early signs of cerebrovascular disease. Furthermore, the study samples differed regarding recruitment, sex distribution, sample size, prevalence of high BP, and data acquisition methods (BP and MRI) which might not represent the general population or standard acquisition protocols: similar to German prevalence42, men had higher BP in our study. We thus included sex as covariate in all our analyses to adjust for sex effects. We did not perform separate analyses for men and women given that one of the four samples included only men. However, the topic of sex differences in brain structure related to BP is a very interesting open question for future investigations. In sample 3, only one BP measurement was recorded which could be biased due to white coat hypertension or BP variability. Practice guidelines recommend an average of ≥2 seated readings obtained on ≥2 occasions to provide a more accurate estimate of an individual’s BP level23,43,44. By combining the samples in random-effects IBMA, we considered the limitations of each sample and accounted for within- and between-sample heterogeneity and evaluated effects cumulatively. Moreover, this approach enabled us to investigate the expected small effects of BP-related GM alterations in a well-powered total sample of over 400 young adults. To further ensure that the results are not substantially influenced by the heterogeneity of BP measurements across studies, we recalculated the parametric SBP analysis (Figure 3A) with only the first SBP reading in each study. The results of this additional analysis are strikingly similar to the results reported here (data available from bioRxiv (Supplementary Table 3, Supplementary Figure 2) https://doi.org/10.1101/239160). HTN is also the most important risk factor for WM damage3,12 and sub-clinical WM injury in relation to elevated BP levels has recently been reported in 19- to 63-year-old adults13. As our study included only GM measures, we cannot assess mediating effects of WM injury on GMV differences. We did not observe any significant differences in Fazekas scores for WMH between BP categories, likely due to their lower sensitivity and poorer specificity as a proxy for vascular disease in a sub-clinical sample of young adults with (mostly) normal BP.
Our study shows that BP-related brain alterations may occur in early adulthood and at BP levels below current thresholds for manifest HTN. Contrary to assumptions that BP-related brain damage arises over years of manifest disease our data suggest that subtle pressure-related GM alterations can be observed in young adults without previously diagnosed HTN. Considering our results, large-scale cohort studies should investigate whether sub-hypertensive BP and related brain changes in early adulthood increase the risk for subsequent development of CVD later in life. Gaining insights whether and how the brain is globally affected by vascular changes or if these are specific to susceptible regions could help identifying neuroimaging biomarkers for the earliest stages of CVD. Such data would provide evidence for future guidelines to formulate informed recommendations for BP-management in young adults, which are critical for the prevention of CVD. Lifestyle interventions and neurobehavioral therapy have recently been suggested to benefit CVD prevention17. Our results highlight the importance of taking BP levels as a continuous measure into consideration which could help initiate such early preventive measures.
Footnotes
Statistical analysis
Statistical analysis conducted by H. Lina Schaare, Max Planck Institute for Human Cognitive and Brain Sciences
Study funding
Max Planck Institute for Human Cognitive and Brain Sciences
Disclosures
Ms. Schaare reports no disclosures.
Ms. Kharabian Masouleh reports no disclosures.
Ms. Beyer reports no disclosures.
Ms. Kumral reports no disclosures.
Ms. Uhlig reports no disclosures.
Mr. Reinelt reports no disclosures.
Dr. Reiter reports no disclosures.
Dr. Lampe reports no disclosures.
Dr. Babayan reports no disclosures.
Ms. Erbey reports no disclosures.
Ms. Roebbig reports no disclosures.
Dr. Schroeter reports no disclosures.
Dr. Okon-Singer reports no disclosures.
Dr. Müller reports no disclosures.
Dr. Mendes reports no disclosures.
Dr. Margulies reports no disclosures.
Dr. Witte reports no disclosures.
Dr. Gaebler reports no disclosures.
Dr. Villringer reports no disclosures.
Author contributions
Study concept and design: Schaare, Villringer.
Statistical analysis: Schaare.
Acquisition or interpretation of data: All authors.
Drafting of the manuscript: Schaare, Villringer.
Critical revision of the manuscript: All authors.