Cerebrovascular reactivity in Alzheimer's disease signature regions is associated with mild cognitive impairment in adults with hypertension

Abstract INTRODUCTION Vascular risk factors contribute to cognitive decline suggesting that maintaining cerebrovascular health could reduce dementia risk. The objective of this study is to evaluate the association of cerebrovascular reactivity (CVR), a measure of brain blood vessel elasticity, with mild cognitive impairment (MCI) and dementia. METHODS Participants were enrolled in the Systolic Blood Pressure Intervention Trial Memory and Cognition in Decreased Hypertension (SPRINT‐MIND) magnetic resonance imaging substudy. Baseline CVR in Alzheimer's disease (AD) signature regions were primary variables of interest. The occipital pole and postcentral gyrus were included as control regions. RESULTS Higher AD composite CVR was associated with lower MCI risk. No significant associations between inferior temporal gyrus, occipital pole, or postcentral gyrus CVR and MCI risk, or any regional CVR–combined risk associations were observed. DISCUSSION CVR in AD signature regions is negatively associated with occurrence of MCI, implicating CVR in AD signature regions as a potential mechanism leading to cognitive impairment.


BACKGROUND
Dementia is a complex neurological condition often caused by neurodegenerative pathologies and exacerbated by cerebrovascular diseases.Dementia can be characterized by gradual progression from normal cognition to mild cognitive impairment (MCI) and then dementia.
Alzheimer's disease (AD) pathology and cerebrovascular risk factors, such as hypertension, high cholesterol, diabetes, and stroke, contribute to the risk of dementia through multiple possible pathways, [1][2][3] including impaired regional cerebrovascular reactivity (CVR), 4 a measure of cerebrovascular responsiveness.
Studies have shown that persons with AD, the most prevalent pathology in adults with dementia (65.3%), 5 and MCI have lower CVR compared to healthy older adults, [6][7][8] including a link between lower CVR and poorer cognitive performance. 9,10Similarly, transgenic mice overproducing amyloid and transforming growth factor (TGF)-β1 developed an early progressive decline of cerebral arterial function and displayed arterial dysfunction unrelated to oxidative stress.
While there is evidence that chronic exposure to increased amyloid beta (Aβ) causes microvascular dysfunction before development of cerebral amyloid angiopathy (CAA) in amyloid-overproducing mice, 11 independent of plaque accumulation, neurovascular changes occurred in parallel to age-modulated accumulation of amyloid plaques, indicating a correlation between decreased CVR and increased AD-like neuropathological changes. 12,13ndings from the Religious Orders Study and the Memory and Aging Project (ROSMAP) relate moderate and severe CAA (prevalence of 35.8%), atherosclerosis (33.2% prevalence), and arteriolosclerosis (31.3% prevalence) with cognitive decline. 5This supports the notion that neurodegenerative and cerebrovascular pathologies interact to drive cognitive impairment. 14Rodent models and human studies have shown that CAA may reduce CVR, 15,16 possibly through amyloid accumulation in brain vessels, resulting in loss of vessel integrity or loss of blood supply and ischemia. 14Older adults with higher carotid intimal medial thickness had a 2.5-fold increased risk of dementia, solidified by an estimated 2-fold increased risk among individuals with bilateral plaque. 17In two studies, CVR increased after carotid endarterectomy, indicating a close relationship between atherosclerosis and CVR. 18,19Similar to findings among participants with atherosclerosis, arteriosclerosis was associated with worse global cognition, 20 and 1.2-fold increase in risk of AD. 21CVR is a risk marker for incident symptomatic lacunar infarction, usually caused by arteriosclerosis, highlighting its role in cerebrovascular degeneration and potential implication in cognitive decline.While converging evidence implicates CVR as a correlate of cognitive decline, there is limited evidence elucidating the association of CVR and the development of MCI and dementia.
Imaging studies have shown that vascular contractility issues are spatially similar to amyloid imaging findings. 14,22Disrupted perivascular amyloid clearance seems to be a pathogenic mechanism in AD and CAA, 14 suggesting that deteriorating CVR might be an early indicator for progression to dementia.Among middle-aged adults, increased 3. Future directions: CVR could point to new biological pathways that may play an early role in dementia progression.Our findings need to be replicated in a larger cohort and for a longer follow-up period, allowing for a greater power to detect potential relationship with incident dementia.
cardiovascular risk profile was associated with lower CVR in regions underlying the default mode network, 23 which is implicated in risk of MCI and AD, with some specificity to the hippocampus, 4 a memory processing hub closely related to AD progression.These indicate a potential relationship between CVR of AD signature regions and subsequent cognitive decline.[26] The objective of this study is to investigate the relationship between CVR and MCI.We hypothesize that higher CVR in AD signature regions will be associated with lower rates of adjudication with MCI during the follow-up.To demonstrate that these associations are specific to AD signature regions, CVR-MCI risk relationships were assessed within the control regions of occipital pole (OP) and postcentral gyrus (PCG), regions typically not affected by neuropathological changes during early stages of AD. 23 To investigate the specificity of CVR as a predictor of MCI progression, exploratory analyses evaluated the relationships with dementia and a composite outcome of MCI and dementia.

Participants
Study participants were selected from the Systolic Blood Pressure Intervention Trial (SPRINT) Memory and Cognition in Decreased Hypertension (SPRINT-MIND) study magnetic resonance imaging (MRI) substudy.SPRINT-MIND was a multicenter randomized controlled trial testing the efficacy of intensive blood pressure control on cardiovascular, renal, and cognitive outcomes.8][29] Trial participants were at least 50 years old, had hypertension, had at least one additional risk factor for heart disease, but did not have diabetes or history of stroke.Additionally, persons residing in nursing homes, diagnosed with dementia, or treated with medication designed for dementia were excluded from the study.Participants were randomized to systolic blood pressure goal of < 120 mmHg (high intensity) or < 140 mmHg (low intensity).As a part of the study, eligible participants with access to the MRI sites were enrolled in the MRI study. 30Everyone with available MRI was included in the current study.The trial and the MRI substudy were approved by the institutional review board at participating centers and all participants provided written informed consent.

MRI measures and processing
Baseline MRI scans were used in the current study.MRI collection and processing techniques have been previously described. 28,30In summary, structural MRI of the brain included 1 mm isotropic T1 magnetization-prepared rapid acquisition gradient echo, T2, and A multi-atlas, multi-warp label-fusion method (MUSE) 31 was used to segment T1 scans into 145 anatomic regions of interest (ROI).CVR was derived from the breath-hold BOLD sequence. 30Breath holding elevated blood CO 2 and induced vasodilation to change the BOLD signal.The resulting images were motion corrected and smoothed, and a whole brain generalized linear model analysis was performed.The CVR brain map was then registered to T1 space and interpolated to 1 mm isotropic resolution. 28,30,32To reduce participant burden and discomfort, end tidal CO 2 was not measured.

Regions of interest
5][26] These regions were identified in the MUSE atlas 31 used for CVR quantification.
Hippocampus, parahippocampus, precuneus, FG, MTG, ITG, and AG were selected for the current study based on their regional association with AD pathologies and availability in the CVR atlas.An AD composite CVR was created as a volume-weighted sum of CVRs of AD signature regions.Additionally, the OP and the PCG were selected as reference regions because they demonstrate atrophy at late stages of AD progression. 33The PCG, a region in the sensorimotor cortex, is supplied by the anterior circulation like most components of the AD signature region, while OP is supplied by the posterior circulation, along with portions of the hippocampus and medial temporal lobe.

Plasma biomarkers
Blood collection and plasma biomarker processing have been previously described. 34

Cognitive assessments
All participants underwent a cognitive test battery assessing memory, processing speed, language, and executive function, 35 including the Montreal Cognitive Assessment (MoCA).Testing occurred three to four times over the course of the study: at screening or randomization visit, at 2-year and 4-year visits, and at study close-out if it was > 1 year after the 4-year follow-up visit.An expert adjudication panel of neurologists, neuropsychologists, geriatricians, and geropsychologists reviewed participants' cognitive scores, functional status assessment, and demographic and medical history and classified the participants into three primary categories at each assessment visit: no cognitive impairment, MCI, or probable dementia. 3Procedures for adjudicating MCI and dementia are described in the study protocol. 3Only participants with two consecutive expert classifications of MCI were defined as having MCI. 28Time to the first of two consecutive occurrences of MCI was used as the time to outcome for MCI.Time to the first identification of adjudicated dementia was used as the time to outcome for probable dementia (PD). 3,28

Statistical analyses
All statistical analyses were conducted in Stata (v17).Distributions of all variables of interest and covariates were inspected visually.
Demographic and lifestyle differences between participants with no adjudication with MCI or PD and with adjudication with MCI were tested using a Student t test.Pearson's correlation matrix and the corresponding P values were used to quantify the relationships between the a priori selected ROI volumes and CVRs.Age at randomization (and MRI scan), sex, education, race, regional volumes, systolic and diastolic blood pressure measures, heart rate, and total brain white matter lesions were treated as potential confounders in our analyses.Confounding was established if adding the candidate variable to the model changed the coefficient of regional CVR by > 20%.All potential confounders met this criterion for confounding and were included in the analyses as covariates.
Cox proportional hazards models were used to assess the relationship between baseline regional CVR and MCI and dementia.The time since randomization to the adjudication of MCI or dementia was used in analyses.Visual inspection of Martingale residuals plotted against variables of interest was used to assess linearity.The correlation of follow-up time with Schoenfeld residuals was used to assess if proportional hazards assumption was violated.Any variables violating linearity or proportional hazards assumptions would be transformed to meet these assumptions.
Interactions with treatment assignments were used to test whether high intensity blood pressure control modified the association of baseline regional CVR on MCI, dementia, and combined MCI/dementia outcomes.CVR-race interactions were used to test whether race modifies the association of baseline regional CVR and MCI risk, dementia, and the composite MCI/dementia outcome.These analyses focused on differences between participants identifying as non-Hispanic White and non-Hispanic Black.Modification by Hispanic ethnicity was not evaluated due to underrepresentation of Hispanic participants in the sample.Associations were considered significant if the two-sided type 1 error rate was < 0.05.All association with regional CVR P values were adjusted using Benjamini-Hochberg procedure.
The analyses were repeated for participants with available plasma biomarkers.Aβ 42 /Aβ 40 , total tau, and NfL were treated as potential confounders.Additionally, E values 36 for the estimates and for the confidence limits are presented as measures of sensitivity to potential unmeasured confounding for all analyses.

Baseline participant characteristics
Five hundred eleven participants were selected for the current analysis of 673 participants with MRI (selected from the total sample of 9361 participants based on criteria described in the protocol [Figure 1]).
One hundred forty-six participants had missing occipital pole CVR or failed quality control.Sixteen more participants had missing CVR values in other ROIs.Participants selected for the MRI substudy had few demographic differences from those who were not selected.Significant differences were observed for race, such that the MRI substudy had a lower proportion of Hispanic-identifying participants (4.66% vs. 10.89%, P < 0.01).Additionally, participants in the MRI substudy had As shown in Table 1, compared to participants who remained cognitively normal, those adjudicated with MCI were significantly older (P = 0.01), had a higher proportion of people with high school degree or lower (P = 0.03), and had a significantly lower baseline MoCA score (P < 0.01).Significant differences in baseline CVR were observed for all AD signature regions, such that those who subsequently were adjudicated with MCI had a lower CVR in all regions.
Additionally, significant baseline differences were observed for hippocampal volume (P = 0.02) between participants adjudicated with MCI and those that were cognitively normal during follow-up.Each regional CVR measure was moderately to strongly correlated with other regional CVR measures; each regional volume was moderately to strongly correlated with other regional volumes (r > 0.42 for all regions, Figure 2A).Baseline CVR and volumes did not exhibit strong correlations (Figure 2B).Baseline systolic blood pressure was moderately correlated with regional CVRs (all rs ≤ 0.18).Baseline systolic and diastolic blood pressures, baseline heart rate, and total brain white matter lesions were not found to confound the relationship between regional CVR and MCI risk.F I G U R E 2 Relationships of regional CVR and volumes.A, Selected brain regions and control regions.B, Pearson correlations of regional volumes and CVRs with each other.C, Regional volume and CVR in the CVR-MCI hazard relationship (adjusted for age at randomization, treatment assignment, race, sex, and education).AG, angular gyrus; CVR, cerebrovascular reactivity; FG, fusiform gyrus; ITG, inferior temporal gyrus; MCI, mild cognitive impairment; MTG, middle temporal gyrus; OP, occipital pole; PCG, postcentral gyrus

Regional CVR in AD signature regions is inversely associated with MCI risk
After adjusting for confounding by regional volumes, significant inverse regional CVR-MCI risk associations were observed for hippocampus (P = 0.02), parahippocampus (P = 0.03), MTG (P = 0.02), precuneus (P = 0.03), AG (P = 0.03), and FG (P = 0.03).No significant ITG CVR-MCI risk association was observed (P = 0.11; Table 2, Figure 1C).In the same models, only the hippocampal volume was significantly inversely associated with MCI risk (P = 0.04).There was no significant OP CVR-MCI risk association and a non-significant trend level PCG CVR-MCI risk association was observed (P = 0.72 and P = 0.10, respectively).All models were additionally adjusted for age at randomization, sex, education, treatment assignment, and race/ethnicity.Regional volume-weighted AD signature composite CVR was negatively associated with adjudication with MCI in the follow-up (hazard ratio [HR] = 0.61 per 1 unit change, 95% confi-dence interval [CI]: [0.43, 0.85], P = 0.02) after adjusting for these covariates.

TA B L E 2
Cox regression hazard rate ratio estimates and 95% confidence intervals (adjudication of MCI as the outcome).College educated non-Hispanic White females randomized into low intensity treatment arm were treated as the reference group.Regional volume for MTG, ITG, and precuneus were transformed (raised to the power of −2 and multiplied by −1) to meet the assumption of proportional hazards.As an estimate of unmeasured confounding that might explain these observed CVR-MCI risk associations, the minimum E value for the point estimate of these associations without adjustment for AD plasma biomarkers was 2.04 (the minimum E value for lower confidence interval value = 1.29) among statistically significant associations.After adjusting for plasma biomarkers, these numbers were 1.77 and 1.11, respectively (Table 4).Unmeasured confounding of these strengths would not suffice to explain away the effect estimates.

Regional CVR in AD signature regions is not associated with combined dementia and MCI risk
After controlling for false discovery rate using Benjamini-Hochberg procedure and age at randomization, sex, education, treatment assignment, race/ethnicity, and confounding by regional volumes, no significant association between regional CVR and the composite outcome of MCI and/or probable dementia was observed (P > 0.16 for all regions; Table 5).

Effect modification by treatment assignment and race
No significant randomization by regional CVR interactions was observed for any AD region of interest (all interaction P values > 0.45).
Black versus White race did not significantly modify the association of regional CVR on MCI risk (all interaction P values > 0.10).

DISCUSSION
In the current study we found that higher CVR in AD signature regions was associated with lower MCI risk.These findings implicate lower CVR as a biomarker of cognitive decline and conversion to MCI, TA B L E 3 Cox regression hazard rate ratio estimates and 95% confidence intervals (adjudication of MCI as the outcome, additionally adjusting for plasma biomarkers).College educated non-Hispanic White females randomized into low intensity treatment arm were treated as the reference group.while controlling for regional brain volume, an established indicator of dementia risk, [24][25][26] as a confounder.Additionally, CVR was a stronger predictor of MCI than regional brain volumes, indicating that CVR may be an earlier predictor of cognitive decline compared to brain volume/atrophy.These relationships were not observed for CVR in the OP and PCG, suggesting that the relationship is primarily apparent in AD signature regions.CVR associations were also not found for the composite of MCI or probable dementia as an outcome, possibly suggesting that CVR is more associated with earlier stages of decline.The associ-ations were robust.Any unmeasured confounder had to be associated with at least a 1.77-fold increase in MCI risk and standardized effect size change of 1.77 on CVR to explain away the observed relationships.

Variable
Cerebrovascular dysfunction is among the key pathologic factors of AD and dementia. 37Studies have established that dysfunction characterized by blood-brain barrier (BBB) breakdown and cerebral blood flow (CBF) reduction contribute to AD. 38 CBF dysregulation and reduction is known to be associated with increased Aβ production. 39wever, BBB imaging studies have shown that older adults with cognitive dysfunction have BBB breakdown regardless of amyloid/tau biomarker changes, [40][41][42] suggesting that CVR reduction may be a contributor to cognitive decline due to cerebrovascular disease that  23 among a cohort of younger subjects, indicating that CVR-related changes in cognition may be independent from the progression of AD-related pathologies.The present study identifies an association between regional CVR and MCI, suggesting a possible protective effect of CVR in key AD brain regions.This finding opens the possibility that maintaining higher CVR may lower the risk of conversion to MCI, and subsequently to dementia although the present study was not powered to examine the association with incident dementia.
Participants for the current study were selected from the SPRINT-MIND trial, which evaluated the effect of intensive blood pressure control on incident dementia.It was important to assess whether the treatment modified the findings of this study.The relationship between CVR and MCI risk was not modified by treatment group (intensive vs. standard therapy).This could be due to differing underlying mechanisms through which CVR and blood pressure influence cognition or MCI, due to insufficient power to detect such an effect, or because the treatment may not have been of sufficient duration to observe a modifying effect.
Testing if race modifies the effect of CVR on adjudication with MCI is important.While cardiovascular diseases are more prevalent among non-Hispanic Black individuals, 45 research is needed to understand racial differences in cerebrovascular health.Although we may be underpowered to detect such differences, we did not find any significant racial/ethnic differences in the relationship between CVR and MCI risk.

LIMITATIONS
Although the data used in this study come from a clinical trial, the MRI substudy participants were a non-random subgroup of the overall trial population.The data analyzed were observational, such that no directed modification to CVR was conducted in the context of a clinical trial or intervention.This approach limits conclusions of causality of CVR effects on MCI.Additionally, this observational study used CVR at the baseline only and thus the combined effect of treatment and CVR could not be evaluated.The small number of cases of probable dementia (n = 9) in the sample limits our ability to study the relationship between CVR and dementia risk.Additionally, the etiology of MCI and dementia was not determined, and there is a possibility that the CVR-MCI relationship is driven by other causes of dementia.
To minimize participant burden, the CVR measurement protocol did not involve measuring end tidal CO 2 and the study did not have extrinsic measures of compliance.These decisions may result in underestimated values for vascular reactivity.Some populations with respiratory illnesses may have more biased results from this technique; however, we do not expect our study population to have disproportionally affected results due to the protocol.Additionally, CVR using the breath-hold technique was found to be correlated with controlled deep breathing CVR, with CO 2 inhalation CVR, and with resting state fluctuation amplitude; comparing BOLD versus arterial spin labeling (ASL), studies have found that ASL had one third of the activation, and we conclude that our results should be reproducible with another CVR measuring technique. 46 Apolipoprotein E (APOE) genotyping was performed on the study sample; however, given the E values calculated for the associations, APOE4 carrier status had to increase the risk of MCI by at least 1.77 in the presence of the other AD biomarkers to explain away the observed associations.In the ROSMAP study, APOE4 was associated with a 1.25fold increase in risk of incident MCI (compared to APOE ε3/ε3, 95% CI: 1.05, 1.49). 47In a population-based study set in rural China, APOE4 had an odds ratio of 1.14 (95% CI: 0.95, 1.36). 48APOE4 was also associated with 1.4-fold increase in incident MCI risk among Black participants enrolled in National Alzheimer's Coordinating Center (95% CI: 1.10, 1.78). 49While there is no current meta-analysis or population-wide study restricting the estimate of the effect of APOE4 on incident MCI, we believe it is unlikely that the estimate will be greater than 1.77 after adjusting for age, sex, education, race, and AD biomarkers and explain away our effect estimates.

Future directions
The present study focused on AD signature regions due to AD being one possible etiology for dementia.Future studies should investigate

RESEARCH IN CONTEXT 1 . 2 .
Systematic review: The authors reviewed the literature using traditional sources.Relevant publications describing the rationale and methodology of the current study as well as studies interpreting and contextualizing findings of the current study are appropriately cited.Interpretation: The current study establishes a relationship between cerebrovascular reactivity in Alzheimer's disease (AD) signature regions and cognitive decline, characterized by adjudicated mild cognitive impairment, reaffirming the role of CVR and vascular dysfunction in dementia etiology.Identification and treatment of lower CVR might prevent or delay cognitive decline.
T2 fluid-attenuated inversion recovery images.Participants also underwent a breath-hold blood oxygen-level-dependent (BOLD) functional MRI (fMRI) sequence, in which imaging occurred during a block of four 16 second breath holds interspersed with free breathing (repetition time = 2000 ms, echo time = 25 ms, field of view = 224 mm, thickness = 3.5 mm, slices = 35, native resolution = 3.5 mm isotropic).
higher baseline MoCA scores (23.5 [standard deviation (SD) = 4.16] vs. 22.75 [SD = 4.35], P < 0.01).Thirty-four participants from the selected MRI cohort were adjudicated with MCI during the follow-up period (median follow-up = 5.36 years, interquartile range [IQR] = 2.12 years) and nine were adju-F I G U R E 1 Flow diagram of study sample selection.MRI, magnetic resonance imaging; SPRINT-MIND, Systolic Blood Pressure Intervention Trial Memory and Cognition in Decreased Hypertension dicated with possible dementia (median follow-up = 5.5 years, IQR = 2.05).
Participant demographic, cognitive, and anatomical characteristics.Participants were compared using the Student t test for continuous variables and the chi-square test for equal proportions for categorical variables.Bilateral regional measures are presented.Cohen D was used to assess the effect size of continuous variable comparisons and Cramer V was used for the categorical variables.
TA B L E 1 44 B L E 4 E values for point estimates and the lower bound of confidence interval needed to explain the observed associations.Cox regression hazard rate ratio estimates and 95% confidence intervals (adjudication of MCI or dementia as the outcome).College educated non-Hispanic White females randomized into low intensity treatment arm were treated as the reference group.Regional volume for MTG, ITG, and precuneus were transformed (raised to the power of −2 and multiplied by −1) to meet the proportional hazards assumption.or may not interact with AD pathologies.Additionally, in a large population study, cerebral hypoperfusion (measured through CBF) preceded the onset of clinical dementia.43ImpairedCVRindicatesloss of cerebrovascular integrity, accompanied with CBF reduction and BBB function disruptions.39Givenrecentmethodologicaladvancements in CVR mapping using MRI,44wider adoption of CVR mapping might help elucidate underlying vascular mechanisms of dementia pathology.
10breviations: AD, Alzheimer's disease; AG, angular gyrus; CVR, cerebrovascular reactivity; FG, fusiform gyrus; ITG, inferior temporal gyrus; MCI, mild cognitive impairment; MTG, middle temporal gyrus; OP, occipital pole; PCG, postcentral gyrus.may6Anotherstudyfoundthatcognitivedecline in amnestic MCI and AD, assessed by the Mini-Mental State Examination (MMSE) score for global cognition, was correlated with reduced CVR velocity, and that the change in velocity was not related to general or focal gray matter atrophy, microangiopathy, or motion.8Inanotherstudy,CVR in both gray matter and white matter of the whole brain was positively associated with cognitive performance, assessed with MMSE, among cognitively normal participants and participants with MCI.10These findings are aligned with other imaging efforts to relate CVR to the default mode network, a network implicated in AD progression,