Vascular reactivity measured with BOLD fMRI upon visual stimulation: tested in elderly population

(1)

Vascular reactivity measured with BOLD fMRI upon visual stimulation: tested

in elderly population

F. van Baarzel BSc.

Leiden university medical center

Studentnumber: 10985611

Supervisor: Dr. S. van Rooden PhD & Dr. J. van der Grond PhD

Second examinator: Dr. H. Krugers PhD

Word count: 6689

01/09/2020

(2)

Objective: Sporadic cerebral amyloid angiopathy (sCAA) is a subtype of a small vessel disease (SVD) that is highly common in elderly. Due to a lack of early diagnostic tools for sCAA, early vascular damage caused by CAA is difficult to detect in the brain. Earlier findings showed that vascular reactivity measured with BOLD fMRI is a promising neuroimaging marker for detecting early vascular damage caused by CAA, even before the onset of symptoms. However, this neuroimaging marker has not been yet investigated regarding sensitivity towards other forms of SVD or ageing. In this study, we tested if vascular reactivity was associated with ageing, cardiovascular (CV) risk factors and general SVD-markers.

Method: 34 healthy elderly participants (51-84 years ) underwent 3T MRI scanning. A flickering checkboard was presented to all participants during fMRI. Vascular reactivity was computed from the BOLD signal change (%) in the occipital lobe upon visual stimulation, consisting of 3 parameters: time to peak, time to baseline and amplitude. Each participant was asked about the presence of CV risk factors, including hypertension, hyperlipidemia, smoking and diabetes. SVD-markers were scored for each participant on anatomical MRI scans according to preexisting criteria. These markers include: microbleeds, lacunar infarcts, enlarged perivascular spaces (EPVS), white matter hyperintensities (WMH), cortical superficial siderosis (cSS), brain atrophy, and intracerebral hemorrhage (ICH). Results: A decreased BOLD amplitude and time to peak was associated with increasing age. No age-effects were found for time to baseline. No association was found between CV risk factors and vascular reactivity. As well as no association between SVD-markers and vascular reactivity. Conclusion: This study suggests that vascular reactivity in the occipital region, measured with fMRI BOLD, is associated with vascular ageing but not with common SVD. This implies that vascular reactivity could be a possible marker for sCAA, as previously indicated, without contamination of other types of SVD.

(3)

Vascular reactivity measured with BOLD fMRI upon visual stimulation

2 Introduction

Vascular damage as a result of small vessel disease (SVD) is common in the elderly above the age of 551. This damage is mainly caused by: plasma proteins that leak into the vessel wall,

accumulation of lipid-containing macrophages in the vessel wall, or fibrosis of the vessel wall1. SVD

can present as a stroke and/or cognitive decline, or is unnoticed until revealed by neuroimaging2.

Often, SVD coexists with other neurodegenerative diseases, such as Alzheimer’s disease, and is regarded as main origin for vascular dementia2. In an initiative to describe the standards for reporting

vascular changes on neuroimaging, several SVD neuroimaging markers were documented: acute and subacute small subcortical infarcts, lacunar infarcts, white matter hyperintensities (WMH), enlarged perivascular spaces (EPVS), microbleeds (MB), and brain atrophy (STRIVE )2. Based on these

criteria, SVD in the brain can easily be detected. However, this is not the case for each subtype of SVD. Cerebral amyloid angiopathy (CAA) is a subtype of SVD, caused by the deposition of amyloid-Aβ (amyloid-Aβ) in small vessels3, 4, resulting in intra-cerebral hemorrhages (ICHs), cognitive decline and

transient focal neurologic symptoms5, 6.

It remains a challenge to investigate and detect CAA in a general population, although it has a high comorbidity with Alzheimer’s disease and a high prevalence of CAA in elderly. Post-mortem studies have shown that up to 98% of AD patients show CAA8, 9 and a little over 30% of elderly show

CAA without other clinical or pathological conditions10. Definite diagnosis of CAA can only be

performed post mortem. Until then, diagnosis can only be made based on the so-called Boston criteria, based on CT and/or MRI neuro imaging, resulting in a ‘possible-CAA’ or ‘probable-CAA’ diagnosis7. The actiual diagnosis of CAA up to this day can only be made in patients who already

express symptoms, based on the detection of specific SVD markers that are related to CAA and clinical data (Boston criteria)7. According to these criteria, a probable or possible CAA diagnosis can

be made upon presence of clinical data, lobar, cortical or cortical-subcortical hemorrhages (ICH, microbleeds; MB), and/or the presence of cortical superficial siderosis (cSS)7. Since CAA is

diagnosed after symptoms occur, it is hardly possible to detect the earliest (presymptomatic) signs of vascular damage caused by CAA. Presymtomatic CAA-patients are only found by co-incidence and referred to as sporadic CAA. Nevertheless, particularly presymptomatic CAA is interesting from pharmacological point of view regarding the development of CAA treatment in general, since

symptomatic CAA is already characterized by intracranial hemorrhage, and therefore therapeutically not well suited to develop treatment strategies. Still, because of the difficulty to include patients with sporadic CAA, also the detection of early CAA biomarkers is severely hampered.

In this respect, presymptomatic patients with a hereditary form of CAA (hereditary cerebral hemorrhages with amyloidosis-Dutch type (HCHWA-D)), offer a unique disease model to provide insight into the early manifestations of vascular damage caused by CAA. In an effort to find early markers of vascular damage in CAA, few years ago an fMRI study was done in presymtomatic and

(4)

symptomatic HCHWA-D patients3. In this study a visual stimulus paradigm was applied to measure

the BOLD response in the occipital lobe (figure 1).

Figure 1 Vascular reactivity in HCHWA-D patients. (A) The visual stimulus paradigm that was used in the

study of van Opstal et al (2017)3. During the fMRI scan, participants had to press a button every time the colored

dot in the center of the checkerboard changed color. After 20s, the checkerboard disappeared and only the colored dot remained for 28 s. The participants were instructed do keep pressing the button, also after the checkerboard disappeared. (B) The paradigm that was used was particularly designed in order for the occipital lobe to become active. The occipital lobe is a region of interest in CAA, because vascular damage caused by CAA is profoundly seen in the occipital area6. (C) The BOLD response after visual stimulation is different for

presymptomatic HCHWA-D patients versus control. This difference in BOLD response is used to calculate the parameters of vascular reactivity. The light grey area represents the ‘on’ stimuli, where the checkerboard is presented for 20s. The white area represents the ‘off’ stimuli, where the checkerboard disappears for 28s. (D) Depiction of a normal BOLD response. The x-axis represents the experimental time in seconds. The y-axis represents the change in BOLD amplitude in percentages. The red line represents the individual fitted trapezoid. From this trapezoid the individual BOLD parameters are derived: BOLD amplitude is calculated by subtracting

(5)

4 Note. (C) Reprinted from van Opstal, A.M., et al., Cerebrovascular function in presymptomatic and symptomatic individuals with hereditary cerebral amyloid angiopathy: a case-control study. Lancet Neurol, 2017. 16(2): p. 6

The results of this study indicated that that the BOLD response in HCHWA-D patients was significantly delayed and decreased compared to controls. More interestingly, this effect was already detected in presymptomatic HCHWA-D patients. All the parameters of vascular reactivity (time to peak, time to baseline and amplitude) were altered even before the onset of symptoms.

These findings ratified the results of an earlier study from Dumas et al. (2012)11 that demonstrated

that vascular reactivity in advanced sporadic

symptomatic CAA-patients was altered compared with controls (figure 2). The study from van Opstal et al. (2017)3 showed that vascular reactivity is a potential

good measure to detect early vascular damage caused by CAA even before onset of symptoms. The delayed and decreased BOLD response in presymptomatic HCHWA-D patients could indicate the earliest

functional loss of the vessel wall in reaction to a visual stimulus causing ischemic damage3, 11.

Figure 2 Vascular reactivity in advanced CAA-patients. (A) The figure shows the BOLD signal change

between CAA-patients and controls. The form of the BOLD response in CAA patients is significantly delayed and decreased compared to the controls. The x-axis represents the experimental time in seconds (light grey=’off’ stimulus, white=’on’ stimulus). The y-axis represents the BOLD amplitude change in percentages. (B) Trapezoid fits of controls and CAA-patients.

Note. Reprinted from Dumas, A., et al., Functional magnetic resonance imaging detection of vascular reactivity in cerebral amyloid angiopathy. Ann Neurol, 2012. 72(1): p. 76-81.

The presymtomatic HCHWA-D patients that were investigated by van Opstal et al. (2017)3

were relatively young (~34 years). As known, sporadic CAA (not hereditary) is most common in the elderly population. It is therefore important to investigate this potential early marker for CAA in the elderly population. However, using the same method to measure vascular reactivity before the onset of symptoms in elderly could result in some difficulties. First of all, as people age in general, their vasculature also undergoes ageing. Vascular ageing is associated with changing the mechanical and structural properties of vascular wall, resulting in a decreased arterial compliance (ability of the blood vessel to extend and contract with changes in pressure)12. Potentially this could mean that vascular

reactivity is sensitive for ageing, not particularly by underlying CAA in elderly.

The vascular wall is also susceptible to change in the presence of cardiovascular (CV) risk factors, which can be modifiable or non-modifiable. Modifiable CV risk factors include factors such as hypertension, obesity, smoking, and lifestyle. These modifiable risk factors in turn all contribute to arterial stiffness12. Early vascular ageing is therefore common in people with increased burden of CV

(6)

risk factors13. Since vascular ageing might interfere with the measurement of vascular reactivity, CV

risk factors can contribute to this effect.

Secondly, vascular damage in elderly is part of the normal ageing process of the brain. In a study to establish the prevalence of age related brain abnormalities (The

Rotterdam Scan study), they scanned over 1000 elderly participants. They found that 68% of participants between 60-70 years had WMH14, 17,8% of participants between 60-69 years

had microbleeds15, 20% of the participants had silent infarcts (of which 80% lacunar

infarcts)16 and they found an association between higher age and smaller brain volume17. The

prevalence of brain abnormalities also increased with age. This study shows that SVD is commonly found in the brains of elderly. The vascular damage caused by SVD could interfere with the measurement of vascular reactivity. This could potentially mean that vascular

reactivity is associated with vascular damage caused by SVD and not by CAA.

In summary, ageing, modifiable CV risk factors and SVD could have an effect on vascular reactivity measured in elderly. In this present study we investigate this by looking at the effect of (1) age, (2) CV risk factors (hypertension, hyperlipidemia, diabetes and smoking), and (3) SVD-markers on vascular reactivity in elderly. In line with previous studies3, 11, a visual paradigm was used to

activate the occipital lobe. After that, the parameters of vascular reactivity were derived from the BOLD response after visual stimulation.

As mentioned before, vascular ageing is associated with a decreased arterial12, therefore we

expect that age is associated with vascular reactivity. A higher age could result in a delayed time to peak, and time to baseline and a decreased amplitude of the BOLD response. The presence of CV risk factors might contribute to early vascular ageing, therefore we expect that CV risk factors are also associated with vascular reactivity. The presence of CV risk factors can result in a delayed time to peak, and time to baseline and a decreased amplitude of the BOLD response. As SVD-markers are indicators for vascular damage, a delayed time to peak, and time to baseline and a decreased amplitude of the BOLD response is expected to associate with SVD-mark.

(7)

Methods Participants and study design

This study is a observational study in 34 healthy adults from 51-84 years (table 1). The study took place at the Leiden University Medical Center (LUMC) and included a 3T MRI scan between January 16th and September 1st of 2020. Healthy participants were recruited from the data-center database at the Radiology department and various advertisements around Leiden.

Table 1 Demographics of participants M=man, F=female, SD=standard deviation

Image acquisition

Structural imaging was performed on a 3T Philips Achieva scanner for all participants and included three-dimensional T1-weighted MRI (echo time [TE] 3.5 ms, repetition time [TR] 7.9 ms, flip 8°, 155 slices, field of view [FOV] 250 x 196 x 171 mm, and scan duration ~4 minutes), T2-weighed (TE 80 ms, TR 4745 ms, flip 90°, 48 slices, FOV 220 x 175 x 144 mm, slice thickness 3 mm, and scan duration ~7.5 minutes), fluid attenuated inversion recovery (FLAIR) (TE 246 ms, TR 4800 ms,

inversion time [TI] 1650 ms, flip 90°, 326 slices, FOV 250 x 250 x 183 mm, matrix 224 x 224 voxels, and scan duration ~4.5 minutes ), and susceptibility weighted MRI (TE 7.2 ms, TR 31 ms, flip 17°, 130 slices, FOV 230 x 190 x 130, matrix 384 x 316 voxels, voxel size 0.6 x 0.6 x 2 mm, and scan duration ~3.5 minutes).

Functional scans were done for all the participants. The visually stimulated blood-oxygen-level-dependent (BOLD) fMRI scans were acquired with TE 30ms, TR 1500 ms, 25 slices, FOV 220 x 220 x 75 mm, matrix 80 x 77 voxels, 130 dynamics, and scan duration 201 s. BOLD-weighted echo-planar imaging (EPI) volumes positioned on the occipital lobes were acquired using the vendor-supplied 32-channel coil. During the functional MRI, a visual stimulus was chosen to elicit robust activation across the occipital lobe and demonstrated altered vascular reactivity. Participants were presented with 4 blocks of 20 second ‘’on’’ stimulus condition, which consisted of a flashing black-and-white checkerboard whose checks were scaled exponentially with eccentricity, subtending approximately 24° of visual angle (i.e., 12° of eccentricity). The checks were counterphase flickering

Characteristics Elderly participants (n=34) Sex (M/F) 9/25 Mean age ±SD, yr 64.2 ±7.6 Mean MMSE ±SD 28.7 ±1.1 Hypertension (%) 7(20.6) Diabetes (%) 1(2.9) Hyperlipidemia (%) 4(11.8)

(8)

at 8Hz to strongly drive neurons in primary visual cortex. Next, a 28 second ‘’off’’ condition consisting of a blank screen filled with gray or average luminance was presented. The participants were

instructed to press a button when the colored dot in the middle of the checkerboard changed color (from dark to light and vice versa) during ‘’on’’ and ‘’off’’ condition, this made sure that the participants remained focused during the task. Each scan was limited to 3 minutes 12 seconds, three scans (runs) were collected during each experimental session for a total of about 10 minutes of fMRI acquisition. Image processing

Vascular reactivity was established by measuring the percentage signal change in BOLD response during fMRI after discontinuation of visual stimulation on three aspects: to-peak, time-to-baseline and amplitude as previously described11.

Image preprocessing was done using the FMRIB Software Library 5.0.11 (FSL). Functional volumes from the 3 runs were concatenated, then skull-stripped, motion corrected, smoothed

(Gaussian kernel, FWHM = 8mm), and high-pass filtered with a cutoff of 100 seconds using the FEAT tool within the FSL analysis package (http://www.fmrib.ox.ac.uk/fsl). The pre-processed functional images were nonlinear registered to individual T1-weightes scans, which were registered to MNI-152 standard space. The registered scans were visually controlled for to ensure correct registration.

The preprocessed functional MRI scans were further processed to measure the cortical activity in response to the visual task, as described in Coppen et al. (2018)18. FEAT (FSL) with a

default (canonical) hemodynamic response function was used to yield volumetric statistical parametric maps of activation strength for each voxel, expressed in a Z-statistic map. The top 10% of the voxels with the most activation in the Z-statistic map was used to create a binary mask, in order to obtain the functional ROI for each participant. The average time course of the BOLD response to the visual stimulus for each study participant was calculated by averaging the time-courses of all voxels within this ROI. The resulting timeseries were cut up into block that each contained stimulus period (20s) and rest period (28s). The timeseries of each block were expressed as percentage BOLD change using the mean value of all blocks. Based on the method described in Dumas et al. (2012)11 a

trapezoidal function was fit to vascular reactivity response in the average BOLD timeseries to describe the time-to-peak, time-to-baseline and amplitude of the response. Time-to-peak was calculated from the time it took from the beginning of the block (t=0) to its trapezoid ceiling. Time-to-baseline was defined as the duration from the end of the stimulus (t=20s) to the Time-to-baseline. Response amplitude was defined as the distance from the baseline to the peak response.

SVD-markers

The SVD-markers were rated for all the participants by two trained independent raters. Consensus score between the raters was made by an experienced neuroimaging researcher with more than 15y experience. The visual markers were detected on 3T anatomical MRI scans. A clear description of all the visual markers can be found in figure 3-7. All the visual markers were scored according to preexisting criteria (table 2).

(9)

8 Figure 3 Example image of microbleeds and lacunar infarct Microbleeds are evaluated on SWI-scans

according to the current consensus criteria20. Microbleeds appear as round ‘black voids’ in either the cortex as

CMB (A) or in deeper brain structures as DMB (B). Lacunar infarcts are evaluated on FLAIR and T2w scans in the basal ganglia, thalamus and brain stem according to the STRIVE criteria2. A lacunar infarct appears as a

(10)

Figure 4 Example image of severity of EPVS in the brain EPVS are evaluated on T2-weighted scans. The top

row figures (B,D) show BG-EPVS in the brain. BG-EPVS was scored as previously described22. The slice and

side with the highest number of EPVS was used to score. (D) shows a participant who scored a 4 on BG-EPVS, which is considered severe. (B) shows a participant who scored a 1 on BG-BG-EPVS, which is considered not severe. The figures (A,C) show CSO-EPVS in the brain. CSO-EPVS was evaluated on a 4-point scale, described in Doubal et al.21. The number of CSO-EPVS refers to the anatomical left side of the brain. The slice

right above the ventricles was chosen to score CSO-PVS on. (D) shows a participant with severe CSO-EPVS (score 4), (B) shows a participant who had very little CSO-EPVS (score 2).

(11)

10 Figure 5 Forms of white matter hyperintensities (WMH). (A) small caps, (B) large caps, (C) extending caps, (D) thin lining, (E) smooth halo, (F) irregular periventricular WMH, (G) punctuate deep WMH, (H) deep WMH with

beginning confluence, (I) confluent deep WMH, seen on fluid attenuated inversion recovery (FLAIR) scan23.

Note. Reprintend from Kim KW, MacFall JR, Payne ME. Classification of white matter lesions on magnetic

(12)

Figure 6 Example brain of cSS and ICH. The left figure (A) shows a brain with disseminated cSS (affects more

than one sulci). Superficial cortical hemosiderosis is defined as linear residues of blood in the superficial layers of the cerebral cortex25. Do not include cSS if it is contiguous with a ICH (often can be seen on other planes,

sagittal). The right figure (B) shows a brain with a large bleeding in the brain (ICH), seen often as a ‘black void’ with a structure inside5 Both cSS and ICH are detectible on susceptible weighting imaging scans (SWI).

(13)

Table 2 Overview visual SVD-markers The visual markers were scored accordingly to set criteria and

transformed to binary scores.

* WMH in the frontal lobe and occipital lobe were scored separately in three different areas: PVWM, DWM and juxtacortical WM, each on a scale from 0-2. The total score for frontal lobe (0-6) minus the score for occipital lobe (0-6) described the frontal-occipital (FO) gradient. A FO gradient of 0> indicates a frontal dominance and a FO gradient of <0 indicates an occipital dominance as previously described in Zhu et al. (2012)24

** Equal refers to a gradient of 0, meaning no dominance of a respective region. Frontal refers to a FO-gradient of >0, meaning a frontal dominance. Occipital refers to a FO-FO-gradient of <0, meaning an occipital dominance.

Note. MB=microbleed, LAC=lacunar infarcts, EPVS= enlarged perivascular spaces, WMH=white matter

hyperintensities. cSS=cortical superficial siderosis, ICH=intracerebral hemorrhage, DMB=deep microbleed, CMB=cerebral microbleed, CSO-EPVS=centrum semiovale enlarged perivascular space, BG-EPVS=basal ganglia enlarged perivascular space, DWM=deep white matter, PVWM=periventricular white matter,

JWM=juxtacortical white matter. SWI=susceptible weighting imaging, FLAIR=fluid attenuated inversion recovery.

Visual marke rs

Scan Score Value

MB SWI Number of microbleeds in subcortical regions (DMB) and

cortical regions (CMB)20 DMB 0=not present 1=>0 DMB CMB 0=not present 1=>0 CMB LAC FLAIR/T2w Number of lacunar infarcts in basal ganglia, thalamus

and brain stem2

0=no infarct 1=>0 infarcts EPVS T2w CSO-EPVS21 0=not present 1= 10 2=11-20 3=21-40 4=>40 BG-EPVS22 0=not present 1=<5 2=5-10 3=>10 4=innumberable CSO-EPVS 0= 2 (moderate) 1=≥ 3 (severe) BG-EPVS 0= 2 (moderate) 1=≥ 3 (severe)

WMH FLAIR Fazekas score23

DWM (0-3) PVWM: (0-3) Fazekas score=DWM+PVWM/2 FO-gradient24* FO-gradient=F-O Fazekas score 0=DWM <2 or PVWM <3 (moderate) 1=≥ DWM2 or PVWM ≥ 3 (severe) FO-gradient:** equal=0 frontal=>0 occipital=<0 Frontal DWM (0-2) PVWM (0-2) JWM (0-2) Occipital DWM (0-2) PVWM (0-2) JWM(0-2)

cSS SWI 0=not present25

1=focal (restricted to <3 sulci) 2=disseminated (4 sulci)

0=not present 1=present

ICH SWI Number of intracerebral hemorrhages5 0=not present

(14)

Additionally, with FMRIB Software library 5.0.11 (FSL) the total WMH volumes were estimated in mm3 with a segmentation threshold of 2,5 SD and the normalized grey matter (GM)

volume was estimated in mm3 by SIENAX19.

Statistical analysis

Statistical analysis was performed using IBM SPSS statistics 25. Any outliers in the data were removed from the analyses. All analyses were adjusted for age and gender. The time to peak, time to baseline and amplitude of the 3 runs were averaged for each participant, resulting in one measure of mean time to peak, time to baseline and amplitude each, for each participant. The WMH volume and GM volume were log transformed to generate a normal distribution.

To investigate the association between vascular reactivity and age, univariate analysis was performed with 3 separate models for each measurement of vascular reactivity. Vascular reactivity was included as dependent variable, gender as fixed factor and age as covariate. Univariate analysis were performed to investigate the association between vascular reactivity (dependent variable) and the visual SVD-markers (fixed factor), controlled for age and gender (covariates). All the SVD-markers were converted to binary variables, except for the FO-gradient (see table 2). The intraclass correlation (ICC) was calculated for each marker between the two individual raters with reliability analysis (two-way mixed model, absolute agreement), to determine the inter-rater reliability. The estimated WMH volumes and GM volumes were partial correlated with vascular reactivity, to investigate their association.

TTo investigate the association between vascular reactivity and CV risk factors, univariate analyses were performed. Vascular reactivity as dependent variable, CV risk factor as fixed factor, and age and gender as covariates. Only one participant had diabetes, therefore we excluded diabetes from the analysis. Lastly, the relationship between vascular reactivity and the relative and absolute head motion (mm) was investigated. This is to make sure that any found effects of vascular reactivity is not dependent of head motion. A Spearman correlation was performed to determine this relationship.

(15)

14 Results

Vascular reactivity & Age

Univariate analyses showed a main effect of age in time to peak (F(1)=8,61, p=0,006) and amplitude (F(1,30)=8,58, p=0,006). This indicates that age is a predictor for time to peak and amplitude. However, no main effect of age for time to baseline (F(1,30)=0,002, p=0,97) was found (figure 7). No main effects were found for gender in time to peak (F(1,30)=3,50, p=0,071), time to baseline (F(1,30)=0,23, p=0,63) and amplitude (F(1,30)=1,94, p=0,17). Also, no interaction effect between age and gender were found in the models for time to peak (F(1,30)=2,51, p=0,12), time to baseline (F(1,30)=0,080, p=0,78) and amplitude (F(1,30)=1,47, p=0,24). These results support the expectation that the amplitude is decreasing with age. However, time to peak seem to be shorter with age and no age-effect was found for time to baseline, which is against our hypotheses.

Figure 7 Relationship vascular reactivity and age These graphs show the relationship between age and (A)

time to peak, (B) time to baseline and (C) amplitude for male and female.

Vascular reactivity and CV risk factors

Univaiate analyses showed that there is no association between vascular reactivity and CV risk factors. No main effect was found of hypertension for time to peak (R2adjusted=0,223, F(1,28)=1,06,

p=0,31), time to baseline (R2adjusted=-0,042, F(1,28)=0,12, p=0,73) and amplitude (R2adjusted=0,16,

F(1,28)=0,70, p=0,41). As well as no main effect of hyperlipidemia for time to peak (R2adjusted=0,18,

F(1,28)=0,26, p=0,61), time to baseline (R2adjusted=-0,044, F(1,28)=0,054, p=0,82) and amplitude

(R2adjusted=0,13, F(1,28)=0,89, p=0,36). Lastly, no main effect was found of smoking for time to peak

(R2adjusted=0,24, F(1,28)=0,83, p=0,37), time to baseline (R2adjusted=-0,017, F(1,28)=1,18, p=0,29) and

(16)

CV risk factors for all the parameters of vascular reactivity (figure 8). These results indicate that CV risk factors are not associated with vascular reactivity, which is not in line with the expectations.

Figure 8 Mean difference of vascular reactivity between CV risk factors The left graph shows the mean

difference between absence and presence of hypertension, hyperlipidemia and smoking for time to peak (A), time to baseline (B) and amplitude (C) after adjusting for age and gender. CV=cardiovascular

Vascular reactivity & SVD-markers

The inter-rater reliability between rater 1 and rater 2 was calculated for each SVD-marker (table 3). Visual SVD-marker ICC (average measure) 95% CI F(df1, df2) Significance (p-value) Lower bound Upper bound DMB 1,00 1,00 1,00 No difference - CMB 0,96 0,92 0,98 F(33,33)=24,29 <.001 Lacunar infarcts 0,77 0,59 0,90 F(33,33)=4,90 <.001 BG-EPVS 0,80 0,59 0,90 F(33,33)=4,81 <.001 CSO-EPVS 0,83 0,66 0,91 F(33,33)=5,88 <.001 Fazekas score 0,86 0,71 0,93 F(32,32)=6,88 <.001 FO-gradient 0,48 -0,042 0,74 F(32,32)=1,93 .034

(17)

16 Note. DMB= deep microbleed, CMB= cerebral microbleed, BG-EPVS= basal ganglia enlarged perivascular

spaces, CSO-EPVS= centrum semiovale enlarged perivascular spaces, ICC=intra-class correlation coefficient, CI= confidence interval

Microbleeds

No significant results were found between time to peak, time to baseline and amplitude for microbleeds. No main effect for DMB was found for time to peak (R2adjusted=0,16, F(1,28)=0,40,

p=0,53), time to baseline (R2adjusted=0,14, F(1,28)=2,21, p=0,15) and amplitude (R2adjusted=0,11,

F(1,28)=0,19, p=0,67). Also, no main effect for CMB was found for time to peak (R2adjusted=0,17,

F(1,30)=0,20, p=0,66), time to baseline (R2adjusted=0,011, F(1,30)=0,28, p=0,60), and amplitude

(R2adjusted=0,19, F(1,30)=2,40, p=0,13). No significant mean differences were found between the

absence and presence of neither DMB nor CMB (figure 9). These results indicate that microbleeds are not a predictor for vascular reactivity, which is against our expectations.

Figure 9 Mean difference of vascular reactivity in presence of microbleeds

The graphs (ABC) show the mean difference for each measurement of vascular reactivity between absence and presence of CMB and DMB after adjusting for age and gender.

(18)

Lacunar infarcts

No significant results were found between vascular reactivity and lacunar infarcts. No main effect was found for lacunar infarcts for time to peak (R2adjusted=0,27, F(1,29)=2,45, p=0,13), time to

baseline (R2adjusted=0,008, F(1,29)=0,42, p=0,52) and amplitude (R2adjusted=0,12, F(1,29)=0,12, p=0,73.

No significant different was found between the absence and presence of lacunar infarcts for all the measurements of vascular reactivity (figure 10). These results indicate that lacunar infarct is not a predictor for vascular reactivity and do not support our hypothesis.

Figure 10 Mean difference of vascular reactivity in presence of lacunar infarcts The mean difference for

time to peak (A), time to baseline (B) and amplitude (C) between presence and absence of lacunar infarcts after adjustment of age and gender. These results are in contradiction with our expectations; vascular reactivity does not seem to be sensitive for lacunar infarcts.

Note. LAC=lacunar infarct

WMH

No significant results were found between the Fazekas score and vascular reactivity. No significant main effect of Fazekas score was found for time to peak (R2adjusted=0,21, F(1,28)=1,11,

p=0,30), time to baseline (R2adjusted=0,061, F(1,28)=0,88, p=0,36) and amplitude (R2adjusted=0,18,

F(1,28)=2,28, p=0,14). No significant mean difference was found between moderate and severe Fazekas score and vascular reactivity (figure 11). The WMH volume had one outlier, which was removed from the analysis. In this analysis, no significant correlation was found between volume

(19)

18 Figure 11 Association between WMH and vascular reactivity The top row graphs (ABC) show the mean

difference of Fazekas score (moderate and severe) for all the measurements of vascular reactivity after adjusting for age and gender. (DEF) The bottom row shows the correlation between volume WMH and the measurements of vascular reactivity.

EPVS

No significant results were found between BG-EPVS and vascular reactivity. Results show that there is no main effect of BG-EPVS for time to peak (R2adjusted=0,14, F(1,28)=0,60, p=0,45), time

to baseline (R2adjusted=-0,057, F(1,28)=0,046, p=0,83) and amplitude (R2adjusted=0,18, F(1,28)=0,61,

p=0,44). No main effect was found of CSO-EPVS for time to peak (R2adjusted=0,19, F(1,28)=2,32,

p=0,14) and time to baseline (R2adjusted=0,059, F(1,28)=0,96, p=0,34). Remarkably, a main effect was

found for CSO-EPVS and amplitude (R2adjusted=0,25, F(1,28)=5,24, p=0,030) and an interaction effect

was found between CSO-EPVS and age (F(1,28)=6,09, p=0,020). However, no significant results were found between moderate and severe EPVS (figure 12). These results indicate that BG-EPVS is not a predictor for vascular reactivity. CSO-EPVS is not a predictor for time to peak and time to baseline. However, CSO-EPVS is a predictor for amplitude, but there was no mean difference between moderate and severe CSO-EPVS for the measurement of time to peak.

(20)

Figure 12 Mean difference of vascular reactivity between moderate and severe EPVS These graphs show

the mean difference between moderate and severe EPVS for (A) time to peak, (B) time to baseline and (C) amplitude after adjusting for age and gender.

Note. BG=basal ganglia, CSO=centrum semiovale

GM volume

No significant correlations were found between GM volume and vascular reactivity after adjusting for age and gender (figure 13).

(21)

20 Figure 13 Association between GM volume and vascular reactivity Correlation between vascular reactivity

and volume WMH.

Note. GM=grey matter

FO-gradient

Univariate analysis showed no significant results between FO-gradient levels and vascular reactivity. No group main effect was found for FO-gradient for time to peak (R2adjusted= 0,14,

F(2,25)=0,789, p=0,465), time to baseline (R2adjusted=0,001, F(2,25)=2,10, p=0,14) and amplitude

(R2adjusted=0,12, F(2,25)=0,13, p=0,88). No significant mean difference was found between equal,

frontal and occipital gradient and vascular reactivity (figure 14). These results indicate that FO-gradient is not associated with vascular reactivity.

Figure 14 Mean difference of vascular reactivity between equal, frontal and occipital FO-gradient. These

figures shows the mean difference of the mean time to peak (A), time to baseline (B) and amplitude (C) after adjusting for age and gender between equal, frontal and occipital FO-gradient.

The results from the correlation analysis between motion and vascular reactivity showed no significant correlation between relative motion and vascular reactivity (figure 15). These findings support the case vascular reactivity found in elderly is not associated with head motion.

(22)

Figure 15 Relation head motion and vascular reactivity The top row graphs (A,B,C) show the relationship

between time to peak (A), time to baseline (B) and amplitude (C) and absolute motion (mm). No significant correlation was found between absolute motion and vascular reactivity. The bottom row graphs (D,E,F) show the relationship between time to peak (D), time to baseline (E) and amplitude (F) and relative motion (mm). Note. Rs=spearman’s rho

(23)

22 Discussion

Earlier findings showed that vascular reactivity might be an early marker for vascular damage in presymptomatic HCHWA-D patients3. Ideally, this marker can be applied as well in detecting early

signs of sporadic CAA. Sporadic CAA appears to be most common in elderly, therefore it is important to investigate how vascular reactivity is manifested in the elderly population. The aim of this present study was to determine the relationship between vascular reactivity, age (1), CV risk factors (2) and SVD-markers (3). Vascular reactivity was measured with BOLD fMRI upon visual stimulation as previously described3.

Firstly, our study provides evidence that ageing could interfere with the measurement of vascular reactivity in elderly. The results suggest that time to peak and amplitude are in particular sensitive to age. No age-effect was found for time to baseline. Surprisingly, both amplitude and time to peak seem to decrease with age, which is not entirely in line with the hypotheses. Secondly, this study shows that there is no association between SVD-markers and vascular reactivity. As well as no association between CV risk factors and vascular reactivity. It seems therefore that vascular reactivity is only sensitive to age and not vascular damage caused by SVD and CV-risk factors. This age-effect is not caused by head motion in the scanner.

Ageing

Multiple studies showed a reduced BOLD response (amplitude) in the occipital regions in older subjects compared to younger26-28. A reduced BOLD response amplitude presumably indicates

normal ageing processes. Our current findings on the amplitude of the BOLD response supports this theory.

No consisting evidence exists whether the time to peak is affected by ageing29, 30. An

increasing time to peak has been associated with the progression of AD disease31. The fact that this

current study did not find a delayed time to peak could be evidence that an increasing time to peak indicates a pathological process. The time to baseline also seems to be part of conflicting evidence and often not reported in BOLD fMRI studies in elderly. While Huettel et al. (2001)29 found a shorter

time to baseline, West et al. (2019)30 found a longer time to baseline in elderly compared to younger

subjects (figure 16). This is in conflict with the current results, which showed no association between age and time to baseline.

Even though a substantial amount of effort was put into investigating the effect of ageing on the hemodynamic response, it remains unclear what the effect of ageing is on the BOLD response. By far the best explanation for this resides from the observation that elderly have a lower signal:noise ratio (more noise per voxel) compared to younger persons. This theory is supported by other

studies29, 32, 33. Research from Aizenstein et al. (2004)33 indicates that group differences in the BOLD

response are sensitive to voxel selection, which refers to including so-called ‘negative voxels’ , i.e. voxels with negative amplitude, and/or low-significance voxels. A number of studies failed in finding a reduced BOLD response after correcting for negative voxels29, 33. In the study from West et al.

(2019)30 was described how elderly have more negative voxels than younger participants, which could

(24)

most activated voxels was used in the analysis, this was done to select voxels with the highest significance level in order to exclude nonresponding and negative voxels. Although this is a rather strict criteria, it is necessary when analyzing the BOLD response in elderly.

Figure 16 Age-effect in BOLD response The top row figures (A,B) show two studies in which a visual paradigm

(checkerboard) was used to measure the BOLD response in the occipital lobe. (A) shows that the amplitude is significantly decreased and the time to peak is delayed in elderly (57-74) compared to younger participants (18-30). (B) shows that the amplitude is not significantly different between young (18-32) and old (57-74), but there is a shorter time to peak for elderly compared to younger participants. The down two figures (C,D) show that young participants have fewer negative responding voxels, compared to older participants.

Note. (A,C,D) Reprinted from West, K.L., et al., BOLD hemodynamic response function changes significantly with

healthy aging. Neuroimage, 2019. 188: p. 198-207.

(B) Reprinted from Huettel, S.A., J.D. Singerman, and G. McCarthy, The effects of aging upon the hemodynamic response measured by functional MRI. Neuroimage, 2001. 13(1): p. 161-75.

Nevertheless, interpreting differences in the BOLD response due to ageing is difficult, because of the complex nature of a BOLD response26. The change in magnitude of the BOLD

response could reflect many underlying changes, (e.g., changes in neural activity, vascular

responsiveness). Earlier studies indicated that age-differences in BOLD response in the occipital lobe might rather reflect changes in neural-vascular coupling than changes in neural activation30, 34.

Comparisons of the BOLD signal made between individuals relies on the assumption of comparable neurovascular coupling. However, alterations in the cerebrovascular dynamics (e.g., vascular reactivity) could affect neural coupling35. Our study provides evidence that an altered vascular

reactivity could indicate that neurovascular coupling in the occipital region is indeed altered in normal ageing persons. It is therefore recommended that when two groups of different ages are compared in BOLD response, the effect of vascular reactivity on the hemodynamic response is quantified35 and

(25)

24 CV risk factors

CV risk factors are a known contributor to early vascular ageing, induced by chronic

inflammation, oxidative stress, lipid deposition and the start of atherosclerotic process13. Our results

suggest that vascular ageing is associated with vascular reactivity. Therefore the presence of CV risk factors might also associate with vascular reactivity. In contradiction to this hypothesis, current evidence points that the occipital lobe is not affected by vascular damage caused by a greater CV risk factor burden. Earlier fMRI evidence showed that participants with a higher CV risk profile showed greater task-related activation in inferior parietal region, this increase in activation was associated with poorer executive functioning36. If this is the case, one would assume that CV risk factors are

contributing or causing vascular damage in parietal regions of the brain, hence sparing the occipital region. However, this study analyzed 33 participants, of which only a view had these CV risk factors. Future studies should include participants with a greater variety in CV risk burden and preferably of different ages, to determine the sole contribution of CV risk factors on vascular reactivity.

SVD

Our study showed that vascular damage caused by SVD did not interfere with vascular reactivity measured in the occipital lobe upon visual stimulation. This suggests that this particular method is only sensitive for CAA, not SVD in general, that often expresses in other cortical or subcortical regions. However, in this study we could not test this CAA-effect since no CAA-patients were included. CV risk factors and SVD are highly associated, as a higher burden of CV risk factors causes vascular damage37, 38. This would explain why no association between vascular reactivity was

found in the occipital lobe and CV risk factors, since vascular damage by SVD spares the occipital regions.

Not many studies have investigated the effect of SVD on vascular reactivity. A meta-analysis study only found 5 studies that investigated the effect of SVD on vascular reactivity. Each study used a different vasoactive stimulus and calculated vascular reactivity differently and in different regions39.

Therefore, the results from these studies are difficult to compare to each other and to the current study. From the 5 studies, two studies found a decreasing vascular reactivity with increasing WMH, the remaining studies did not found such an association. In three studies they found a decreasing vascular reactivity with increasing age. Collectively, these studies do not give clear evidence if vascular reactivity is impaired in SVD. Future studies into the relationship between vascular reactivity and SVD should include a larger SVD population, with more variety in SVD severity.

In conclusion, this study suggests that vascular reactivity is a sensitive measurement for ageing. Vascular reactivity measured in the occipital lobe could possibly be a marker for vascular ageing and therefore not specifically for sCAA. A decreasing time to peak and amplitude possibly indicate a normal ageing process. However, a large body of conflicting evidence about the ageing effect on the BOLD response suggests that we should take this conclusion with great caution. This study did not provide evidence for an association between vascular reactivity and SVD-markers or CV

(26)

risk factors, indicating that these pathological processes differ from sCAA. Using vascular reactivity as an early marker for CAA (as proposed by van Opstal et al. 20173) should take into account that this

measurement could be interfered by vascular ageing and therefore give a distorted view of the BOLD signal change.

(27)

References

1. Grinberg, L.T. and D.R. Thal, Vascular pathology in the aged human brain. Acta Neuropathol, 2010. 119(3): p. 277-90.

2. Wardlaw, J.M., et al., Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration. Lancet Neurol, 2013. 12(8): p. 822-38. 3. van Opstal, A.M., et al., Cerebrovascular function in presymptomatic and symptomatic

individuals with hereditary cerebral amyloid angiopathy: a case-control study. Lancet Neurol, 2017. 16(2): p. 115-122.

4. DeCarli, C., et al., Vascular Burden Score Impacts Cognition Independent of Amyloid PET and MRI Measures of Alzheimer's Disease and Vascular Brain Injury. J Alzheimers Dis, 2019. 68(1): p. 187-196.

5. Wermer, M.J.H. and S.M. Greenberg, The growing clinical spectrum of cerebral amyloid angiopathy. Curr Opin Neurol, 2018. 31(1): p. 28-35.

6. Charidimou, A., Q. Gang, and D.J. Werring, Sporadic cerebral amyloid angiopathy revisited: recent insights into pathophysiology and clinical spectrum. J Neurol Neurosurg Psychiatry, 2012. 83(2): p. 124-37.

7. Charidimou, A., et al., Advancing diagnostic criteria for sporadic cerebral amyloid angiopathy: Study protocol for a multicenter MRI-pathology validation of Boston criteria v2.0. Int J Stroke, 2019. 14(9): p. 956-971.

8. Attems, J., F. Lauda, and K.A. Jellinger, Unexpectedly low prevalence of intracerebral hemorrhages in sporadic cerebral amyloid angiopathy: an autopsy study. J Neurol, 2008. 255(1): p. 70-6.

9. Jellinger, K.A., Alzheimer disease and cerebrovascular pathology: an update. J Neural Transm (Vienna), 2002. 109(5-6): p. 813-36.

10. Esiri, M.M. and G.K. Wilcock, Cerebral amyloid angiopathy in dementia and old age. J Neurol Neurosurg Psychiatry, 1986. 49(11): p. 1221-6.

11. Dumas, A., et al., Functional magnetic resonance imaging detection of vascular reactivity in cerebral amyloid angiopathy. Ann Neurol, 2012. 72(1): p. 76-81.

12. Jani, B. and C. Rajkumar, Ageing and vascular ageing. Postgrad Med J, 2006. 82(968): p. 357-62.

13. Nilsson, P.M., E. Lurbe, and S. Laurent, The early life origins of vascular ageing and cardiovascular risk: the EVA syndrome. J Hypertens, 2008. 26(6): p. 1049-57.

14. de Leeuw, F.E., et al., Prevalence of cerebral white matter lesions in elderly people: a population based magnetic resonance imaging study. The Rotterdam Scan Study. J Neurol Neurosurg Psychiatry, 2001. 70(1): p. 9-14.

15. Vernooij, M.W., et al., Prevalence and risk factors of cerebral microbleeds: the Rotterdam Scan Study. Neurology, 2008. 70(14): p. 1208-14.

16. Vermeer, S.E., et al., Prevalence and risk factors of silent brain infarcts in the population-based Rotterdam Scan Study. Stroke, 2002. 33(1): p. 21-5.

(28)

17. Ikram, M.A., et al., Brain tissue volumes in the general elderly population. The Rotterdam Scan Study. Neurobiol Aging, 2008. 29(6): p. 882-90.

18. Coppen, E.M., et al., Structural and functional changes of the visual cortex in early Huntington's disease. Hum Brain Mapp, 2018. 39(12): p. 4776-4786.

19. Smith, S.M., et al., Accurate, robust, and automated longitudinal and cross-sectional brain change analysis. Neuroimage, 2002. 17(1): p. 479-89.

20. Greenberg, S.M., et al., Cerebral microbleeds: a guide to detection and interpretation. Lancet Neurol, 2009. 8(2): p. 165-74.

21. Doubal, F.N., et al., Enlarged perivascular spaces on MRI are a feature of cerebral small vessel disease. Stroke, 2010. 41(3): p. 450-4.

22. Zhu, Y.C., et al., Severity of dilated Virchow-Robin spaces is associated with age, blood pressure, and MRI markers of small vessel disease: a population-based study. Stroke, 2010. 41(11): p. 2483-90.

23. Kim, K.W., J.R. MacFall, and M.E. Payne, Classification of white matter lesions on magnetic resonance imaging in elderly persons. Biol Psychiatry, 2008. 64(4): p. 273-80.

24. Zhu, Y.C., et al., Distribution of white matter hyperintensity in cerebral hemorrhage and healthy aging. J Neurol, 2012. 259(3): p. 530-6.

25. Linn, J., et al., Subarachnoid hemosiderosis and superficial cortical hemosiderosis in cerebral amyloid angiopathy. AJNR Am J Neuroradiol, 2008. 29(1): p. 184-6.

26. Ances, B.M., et al., Effects of aging on cerebral blood flow, oxygen metabolism, and blood oxygenation level dependent responses to visual stimulation. Hum Brain Mapp, 2009. 30(4): p. 1120-32.

27. Buckner, R.L., et al., Functional brain imaging of young, nondemented, and demented older adults. J Cogn Neurosci, 2000. 12 Suppl 2: p. 24-34.

28. Raemaekers, M., et al., Effects of aging on BOLD fMRI during prosaccades and antisaccades. J Cogn Neurosci, 2006. 18(4): p. 594-603.

29. Huettel, S.A., J.D. Singerman, and G. McCarthy, The effects of aging upon the hemodynamic response measured by functional MRI. Neuroimage, 2001. 13(1): p. 161-75.

30. West, K.L., et al., BOLD hemodynamic response function changes significantly with healthy aging. Neuroimage, 2019. 188: p. 198-207.

31. Rombouts, S.A., et al., Delayed rather than decreased BOLD response as a marker for early Alzheimer's disease. Neuroimage, 2005. 26(4): p. 1078-85.

32. D'Esposito, M., et al., The effect of normal aging on the coupling of neural activity to the bold hemodynamic response. Neuroimage, 1999. 10(1): p. 6-14.

33. Aizenstein, H.J., et al., The BOLD hemodynamic response in healthy aging. J Cogn Neurosci, 2004. 16(5): p. 786-93.

34. Attwell, D., et al., Glial and neuronal control of brain blood flow. Nature, 2010. 468(7321): p. 232-43.

(29)

28 36. Chuang, Y.F., et al., Cardiovascular risks and brain function: a functional magnetic resonance

imaging study of executive function in older adults. Neurobiol Aging, 2014. 35(6): p. 1396-403.

37. Khan, U., et al., Risk factor profile of cerebral small vessel disease and its subtypes. J Neurol Neurosurg Psychiatry, 2007. 78(7): p. 702-6.

38. Palacio, S., et al., Lacunar strokes in patients with diabetes mellitus: risk factors, infarct location, and prognosis: the secondary prevention of small subcortical strokes study. Stroke, 2014. 45(9): p. 2689-94.

39. Blair, G.W., et al., Magnetic resonance imaging for assessment of cerebrovascular reactivity in cerebral small vessel disease: A systematic review. J Cereb Blood Flow Metab, 2016. 36(5): p. 833-41.