Lower cardiac output is associated with neurodegeneration among older adults with normal cognition but not mild cognitive impairment

Abstract

Subclinical cardiac dysfunction is associated with smaller total brain volume on magnetic resonance imaging (MRI). To study whether cardiac output relates to regional measurements of grey and white matter structure, older adults (n=326) underwent echocardiogram to quantify cardiac output (L/min) and brain MRI. Linear regressions related cardiac output to grey matter volumes measured on T₁ and white matter hyperintensities assessed on T₂-FLAIR. Voxelwise analyses related cardiac output to diffusion tensor imaging adjusting for demographic, genetic, and vascular risk factors. Follow-up models assessed a cardiac output x diagnosis interaction with stratification (normal cognition, mild cognitive impairment). Cardiac output interacted with diagnosis, such that lower cardiac output related to smaller total grey matter (p=0.01), frontal lobe (p=0.01), and occipital lobe volumes (p=0.01) among participants with normal cognition. When excluding participants with cardiovascular disease and atrial fibrillation, associations emerged with smaller parietal lobe (p=0.005) and hippocampal volume (p=0.05). Subtle age-related cardiac changes may contribute to disrupt neuronal homeostasis and impact grey matter integrity prior to cognitive impairment.

1. Introduction

The brain receives a disproportionately large volume of cardiac output (Williams and Leggett, 1989), so subtly compromised cardiac function presumably could affect long-term brain health. Emerging animal (Bos et al., 2017;Schultz et al., 2015) and human evidence (Jefferson et al., 2017) suggests autoregulation, the process maintaining blood flow to cerebral tissue despite systemic perfusion variability, may be vulnerable to aging. Subclinical reductions in systemic perfusion are associated with reduced cerebral blood flow (CBF) on magnetic resonance imaging (MRI) (Jefferson et al., 2017). These associations localize to the temporal lobe, suggesting certain brain regions may be more vulnerable than others in the presence of subclinical cardiac dysfunction.

Prior work has shown both clinical and subclinical reductions in cardiac output are associated with smaller total brain volumes in middle aged (Jefferson et al., 2010) and older adults (Jefferson et al., 2010;Park et al., 2017;Sabayan et al., 2015). Given associations between reduced systemic perfusion and regional CBF (Jefferson et al., 2017;Loncar et al., 2011), neurodegeneration in certain regions may drive previously reported changes in total brain volume. Understanding if cardiac output has preferential effects on certain brain regions is necessary to better characterize cerebral vulnerability to subclinical cardiac dysfunction. Cerebral white matter may also be vulnerable to reduced systemic perfusion, as much of the white matter is located at watershed regions vulnerable to oligemia (Torvik, 1984). However, prior studies examining cardiac function and white matter changes focus on macrostructural damage (white matter hyperintensities (WMHs)) with mixed results (Jefferson et al., 2010;Jefferson et al., 2007;Sabayan et al., 2015). It is possible that a more sensitive measure of white matter microstructure, such as diffusion tensor imaging (DTI), may provide clarity on these associations.

Recent work suggests hemodynamic changes may precede abnormal protein accumulation, structural brain changes, and cognitive decline (Iturria-Medina et al., 2016). However, as mild cognitive impairment (MCI) develops, associations between subclinical cardiac dysfunction and brain outcomes may become obscured by growing pathological burden. Examining how associations between cardiac dysfunction and brain health differ across the cognitive aging spectrum will begin to clarify whether hemodynamic changes precipitate or exacerbate structural brain changes. However, it is currently unknown how associations between cardiac output and structural brain health may differ between individuals with normal cognition (NC) and MCI. Additionally, reductions in CBF contribute to AD pathology (Makinen et al., 2008;Wen et al., 2004), suggesting decreased cardiac output may lead to neurodegeneration in AD-specific regions. Thus, it is necessary to explore associations across the cognitive aging spectrum and specifically examine associations with brain volume in regions susceptible to AD pathology (Schwarz et al., 2016).

The current study examines associations among older individuals between cardiac output and regional grey matter volumes assessed on T₁-weighted MRI, white matter macrostructure assessed on T₂-weighted fluid attenuated inversion recovery (FLAIR), and white matter microstructure assessed on DTI. Given prior work correlating cardiac output with total brain volume (Jefferson et al., 2010;Sabayan et al., 2015) and regionally-specific CBF reductions (Jefferson et al., 2017), we hypothesize lower cardiac output is associated with decreased grey matter volume, especially in the temporal lobe. We also hypothesize lower cardiac output may not correlate with WMHs but is associated with compromised white matter microstructure seen on DTI in watershed regions (Torvik, 1984). In light of prior work, (Iturria-Medina et al., 2016;Jefferson et al., 2017), we hypothesize associations between cardiac output and structural brain outcomes are modified by diagnosis.

2. Methods

2.1. Cohort

The Vanderbilt Memory & Aging Project (VMAP) is a longitudinal observational study investigating vascular health and brain aging (Jefferson et al., 2016). Inclusion required participants be ≥60 years, speak English, have adequate auditory and visual acuity, and have a reliable study partner. As part of a comprehensive screening, participants were excluded for a cognitive diagnosis other than NC, early mild cognitive impairment (eMCI) (Aisen et al., 2010), or MCI (Albert et al., 2011), MRI contraindication, history of neurological disease (e.g., multiple sclerosis, stroke), heart failure, major psychiatric illness, head injury with loss of consciousness>5 minutes, or a systemic or terminal illness affecting follow-up participation.

At enrollment, participants completed a comprehensive examination, including fasting blood draw, physical examination, clinical interview, medication review, echocardiogram, and brain MRI. A total of 326 participants were included in the present analysis. See Figure 1 for participant exclusion details. The protocol was approved by the Vanderbilt University Medical Center Institutional Review Board. Written informed consent was obtained from participants prior to data collection.

Figure 1. Participant Inclusion and Exclusion Details.

Missing data categories are mutually exclusive. An additional 5 participants were excluded from analyses examining the AD signature due to having non-usable imaging data, and an additional 1 participant was excluded from DTI analyses due to motion. AD=Alzheimer’s disease; CVD=cardiovascular disease; eMCI=early mild cognitive impairment; MCI=mild cognitive impairment; NC=normal cognition.

2.2. Echocardiogram

As previously reported (Jefferson et al., 2017), standard 2-dimensional, M-mode, and Doppler transthoracic echocardiography was performed by a single research sonographer on a Philips IE33 cardiac ultrasound machine (Philips Medical, Andover, MD). Digital images with measurements were confirmed by board-certified cardiologists (DKG, LAM) blinded to clinical information. Cardiac output was calculated as stroke volume times heart rate. Final measurements were from a single cardiac cycle for participants in normal sinus rhythm or the average of 3 cardiac cycles for participants in atrial fibrillation.

2.3. Brain MRI

Participants were scanned at the Vanderbilt University Institute of Imaging Science on a 3T Philips Achieva system (Best, the Netherlands) using an 8-channel SENSE reception coil array. T₁-weighted images (repetition time=8.9ms, echo time=4.6ms, spatial resolution=1×1×1mm³), T₂-weighted FLAIR images (repetition time=11000ms, echo time=121ms, spatial resolution=0.45×0.45×4mm³), and diffusion tensor images along 32 diffusion gradient vectors (TR/TE=10000/60ms, b-value=1000s/mm², spatial resolution=2×2×2mm³) were acquired as part of a larger multimodal protocol.

T₁-weighted images were post-processed with an established Multi-Atlas Segmentation pipeline (Asman and Landman, 2013). As previously published (Jefferson et al., 2016), T₁-weighted images were registered to Montreal Neurological Institute (MNI) space and corrected for inhomogeneity and motion artifact. Appropriate atlases were geodesically selected and atlas labels were statistically fused. The AdaBoost segmentation adaptor framework calculated total intracranial volume (Hecht et al., 2012) and grey matter volume in 7 regions of interests (ROIs), including total brain, frontal lobe, temporal lobe, parietal lobe, occipital lobe, hippocampal, and inferior lateral ventricle volumes.

To evaluate grey matter changes associated with AD pathology (Schwarz et al., 2016), an AD signature variable was calculated. T₁-weighted images were post-processed using FreeSurfer (http://surfer.nmr.mgh.harvard.edu/) (Fischl and Dale, 2000). T₁-weighted images were registered to MNI space, intensity corrected, and skull stripped. Subcortical structures, cortical structures, and white matter were segmented, and white and grey matter surfaces were constructed for each hemisphere. Surfaces were manually inspected and corrected for registration, topological, and segmentation defects. After manual correction, images were reprocessed to update the transformation template and segmentation information. The AD signature was calculated by summing bilateral cortical thickness measurements from the entorhinal cortex, temporal lobe, parietal lobe, fusiform gyrus, and precuneus (Schwarz et al., 2016).

FLAIR images were post-processed using the Lesion Segmentation Toolbox for Statistical Parametric Mapping (SPM8) (Schmidt et al., 2012). As previously published (Osborn et al., 2018), each T₁-weighted image voxel was defined as grey matter, white matter, or cerebrospinal fluid (CSF) using the SPM8 tissue probability map. Manual corrections were made by individuals blinded to participant information. FLAIR images were segmented into 5 regions of interest (ROIs), including total brain, frontal lobe, temporal lobe, parietal lobe, and occipital lobe WMH volume using an MNI305 lobe atlas (Toro et al., 2009).

DTI data were post-processed through an established tract-based spatial statistics (TBSS) pipeline using the FMRIB Software Library (FSL) version 4.1.4 (http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSL) (Smith, S.M. et al., 2006). As previously published (Moore et al., 2018), images were corrected for motion and eddy currents. All fractional anisotropy images were non-linearly registered, and a mean skeleton was built only including voxels overlapping among 80% or more of participants. Each participant’s fractional anisotropy, mean diffusivity, radial diffusivity, and axial diffusivity images were projected onto the original mean fractional anisotropy skeleton.

2.5. Covariates

Covariate definitions have been previously described (Jefferson et al., 2018;Jefferson et al., 2017). Briefly, systolic blood pressure was the mean of two measurements. Medication review determined anti-hypertensive medication use. Diabetes mellitus was defined as current fasting blood glucose ≥126 mg/dL, hemoglobin A1c ≥6.5%, or current oral hypoglycemic or insulin medication usage. Current cigarette smoking was ascertained by self-report. Left ventricular hypertrophy (LVH) was defined on echocardiogram. Self-report atrial fibrillation was corroborated by echocardiogram, cardiac magnetic resonance, or documentation of prior procedure/ablation or medication usage for atrial fibrillation. Self-report cardiovascular disease (CVD), including coronary heart disease, angina, or myocardial infarction, was confirmed utilizing available medical records (note, heart failure was a parent study exclusion). Framingham Stroke Risk Profile (FSRP) score applied points by sex for age, systolic blood pressure, anti-hypertensive medication usage, diabetes mellitus, current cigarette smoking, LVH, atrial fibrillation, and CVD (D’Agostino et al., 1994). Body surface area (m²) was calculated as weight^0.425 * height^0.725 * 0.007184. Apolipoprotein E (APOE) ε4 genotyping was performed on whole-blood samples and APOE-ε4 status was defined as carrier (ε2/ε4, ε3/ε4, ε4/ε4) or non-carrier (ε2/ε2, ε2/ε3, ε3/ε3).

2.6. Analytical Plan

Data were visually checked for normality, and WMHs (cm³) were log-transformed due to skewed distribution. Ordinary least-square regressions related cardiac output to total and regional grey matter volumes, the AD signature, and total and regional log WMHs adjusting for age, sex, education, race/ethnicity, FSRP (excluding points assigned for age), body surface area, diagnosis, and APOE-ε4 status. Models examining grey matter volumes and WMHs covaried for intracranial volume. Multiple comparison correction was performed using Benjamini–Hochberg’s procedure (Benjamini and Hochberg, 1995) to control a false discovery rate of α=0.05. The total number of tests was based on each individual model. For example, the association between cardiac output and grey matter in the entire sample were corrected for 8 tests based on the total number of grey matter regions assessed. For the DTI data, voxelwise analyses using general linear models and FSL randomise with 5000 permutations related cardiac output to fractional anisotropy, mean diffusivity, radial diffusivity, and axial diffusivity, with similar covariates. Permutation testing is commonly used for TBSS analysis due to the distribution of DTI values within the skeleton (Winkler et al., 2014). Multiple comparison correction was performed using the established cluster enhancement permutation procedure in FSL (Smith and Nichols, 2009). In secondary analyses excluding the small eMCI sample (n=27), models were repeated evaluating a cardiac output x diagnosis interaction term on grey matter and white matter outcomes with identical covariates. Models were then repeated stratified by diagnosis (NC, MCI). Interaction and stratified models were repeated combining the eMCI and MCI samples, but results were unchanged (data not shown).

Statistical significance was set a priori at p<0.05. Sensitivity analyses excluded participants with prevalent CVD or atrial fibrillation (n=34) to assess whether these conditions accounted for results. Additional models excluded predictor or outcome values >4 standard deviations from the group mean, yielding consistent results (data not shown). Analyses were conducted using R version 3.4.2 (www.r-project.org).

3. Results

3.1. Participant Characteristics

Participants included 326 adults age 60 to 92 (73±7 years, 59% male, 87% non-Hispanic White). Cardiac output ranged 2.0 to 8.7 L/min. See Table 1 for more detailed participant characteristics.

Table 1.

Participant Characteristics

Demographic & Health Characteristics	Total n=326	NC n=169	eMCI n=27	MCI n=130	p-value
Age, years	73±7	72±7	73±6	73±8	0.64
Sex, % male	59	58	74	56	0.22
Race, % Non-Hispanic White	87	87	85	86	0.96
Education, years	16±3	16±2	16±3	15±3	<0.001 †‡
APOE-ε4, % positive	35	30	22	45	0.009 ‡
FSRP, total scorea	12±4	12±4	14±3	13±4	0.05
Systolic blood pressure, mmHg	143±18	140±17	150±18	145±19	0.004 * ‡
Antihypertensive medication usage, %	54	53	56	55	0.93
Diabetes, %	18	16	22	20	0.57
Current smoking, %	2	2	4	3	0.70
Atrial fibrillation, %	7	6	11	7	0.60
Prevalent CVD, %	5	6	4	3	0.50
Left Ventricular Hypertrophy, %	4	3	4	6	0.40
Body surface area, m2	1.9±0.2	1.9±0.2	2.0±0.2	1.9±0.2	0.09
Heart rate, beats/min	65±10	65±11	65±13	64±10	0.50
Stroke Volume, mL	77±18	77±18	80±21	77±17	0.83
Cardiac output, L/min	5.0±1.3	5.0±1.3	5.0±1.2	4.9±1.3	0.75
Grey Matter Variables
Total brain volume, mm3	682,085±78,681	690,235±80,386	686,008±79,931	670,675±75,314	0.15
Frontal lobe volume, mm3	224,284±31,354	227,221±31,863	226,848±36,865	219,933±29,136	0.19
Temporal lobe volume, mm3	134,582±15,713	136,626±15,779	135,682±13,608	131,695±15,695	0.04
Parietal lobe volume, mm3	128,274±16,831	130,045±17,791	126,658±15,374	126,307±15,661	0.27
Occipital lobe volume, mm3	90,699±11,035	91,249±11,171	90,831±10,633	89,955±10,979	0.59
Hippocampal volume, mm3	7188±883	7450±840	7111±623	6865±877	<0.001 ‡
Inferior lateral ventricular volume, mm3	2004±1238	1683±882	1983±1242	2425±1491	<0.001 ‡
AD signature, mm	2.30±0.14	2.34±0.12	2.29±0.10	2.26±0.15	<0.001 * ‡
Intracranial volume, mm3	1,505,682±150,437	1,502,222±151,568	1,540,779±123,174	1,502,890±154,141	0.37
White Matter Variables
Total raw WMHs, cm3	13±20	11±21	9±9	17±20	<0.001 ‡
Total WMHs, log-transformed cm3	2.10±1.03	1.88±1.00	2.0±0.82	2.40±1.02	<0.001 ‡
Frontal lobe WMHs, log-transformed cm3	1.56±0.95	1.35±0.89	1.53±0.76	1.83±0.99	<0.001 ‡
Temporal lobe WMHs, log-transformed cm3	0.31±0.48	0.25±0.45	0.17±0.26	0.41±0.53	0.002 ‡
Parietal lobe WMHs, log-transformed cm3	0.87±0.93	0.75±0.90	0.71±0.74	1.07±0.98	0.005 ‡
Occipital lobe WMHs, log-transformed cm3	1.05±0.67	0.92±0.65	1.01±0.51	1.22±0.69	<0.001 ‡
Intracranial volume, cm3	1381±142	1378±142	1406±106	1379±150	0.48

Note. Values denoted as mean±standard deviation or frequency.

^aa modified FSRP score was included in statistical models excluding points assigned to age (7±3)

^*NC different from eMCI

^†eMCI different from MCI

^‡NC different from MCI

AD=Alzheimer’s Disease; APOE=apolipoprotein E; CVD=cardiovascular disease; eMCI=early mild cognitive impairment; FSRP=Framingham Stroke Risk Profile; MCI=mild cognitive impairment; NC=normal cognition; WMHs=white matter hyperintensities.

3.2. Cardiac Output & Grey Matter

In main effect models, cardiac output was unrelated to total or regional grey matter volumes (p-values>0.10, Table 2). Cardiac output interacted with diagnosis on total grey matter (β=−8881, p=0.02), frontal lobe (β=−4041, p=0.04), occipital lobe (β=−2089, p<0.001), and hippocampal volumes (β=−150, p=0.02). See Figure 2 and Supplemental Table I for details. Results persisted when excluding participants with prevalent CVD and atrial fibrillation, and the interaction on occipital lobe volume would survive correction for multiple comparisons. In stratified models among NC participants, lower cardiac output related to smaller total grey matter (β=7311, p=0.01), frontal lobe (β=4064, p=0.009), and occipital lobe volumes (β=1112, p=0.007), reflecting increased neurodegeneration. Frontal lobe volume was lower, on average by 4064 mm³ per one unit decrease in cardiac output (defined as 1.3 L/min), a magnitude similar to the mean decrease in frontal lobe volume with an increase in age of more than 13 years. These associations would all survive correction for multiple comparisons. When excluding participants with prevalent CVD and atrial fibrillation, additional associations emerged. Lower cardiac output related to smaller total grey matter (β=11,345, p<0.001), frontal lobe (β=5829, p<0.001), and occipital lobe volumes (β=1594, p<0.001) and also to smaller parietal lobe (β=2568, p=0.005) and hippocampal volumes (β=94, p=0.047). In stratified models among MCI participants, lower cardiac output counter-intuitively related to higher occipital lobe (β=−1117, p=0.02) and hippocampal volumes (β=−137, p=0.02), though the latter observation was attenuated (β=−109, p=0.08) when excluding participants with prevalent CVD and atrial fibrillation. See Table 2 for details.

Table 2.

Cardiac Output Associations with Grey Matter Volumes and Stratification by Diagnosis

	Main Effects (n=326)*			NC only (n=169)†			MCI only (n=130)†
Grey Matter	β	95% CI	p-value	β	95% CI	p-value	β	95% CI	p-value
Total brain volume	2984	−1185, 7154	0.16‡	7311	1540, 13,082	0.01	−650	−7075, 5776	0.84
Frontal lobe volume	1843	−327, 4013	0.10‡	4064	1045, 7083	0.009	363	−2887, 3614	0.83
Temporal lobe volume	170	−645, 985	0.68	340	−660, 1340	0.50	−108	−1602, 1386	0.89
Parietal lobe volume	874	−243, 1991	0.12	1430	−266, 3125	0.10	726	−794, 2247	0.35
Occipital lobe volume	130	−467, 727	0.67‡	1112	305, 1918	0.007	−1117	−2056, −177	0.02
Hippocampal volume	−18	−84, 48	0.59‡	53	−32, 137	0.22	−137	−250, −24	0.02
Inferior lateral ventricular volume	−15	−115, 85	0.77	−94	−194, 7	0.07	126	−74, 326	0.22
Schwarz AD signature	0.002	−0.010, 0.014	0.76	0.006	−0.009, 0.021	0.44	−0.008	−0.030, 0.014	0.47
White Matter
Total WMHs	−0.027	−0.113, 0.059	0.54	−0.010	−0.131, 0.111	0.87	−0.037	−0.176, 0.102	0.60
Frontal lobe WMHs	−0.046	−0.127, 0.034	0.26	−0.045	−0.154, 0.065	0.42	−0.037	−0.175, 0.101	0.60
Temporal lobe WMHs	−0.014	−0.059, 0.030	0.52	0.015	−0.043, 0.072	0.61	−0.041	−0.123, 0.042	0.33
Parietal lobe WMHs	−0.020	−0.103, 0.062	0.63	−0.007	−0.118, 0.104	0.90	−0.002	−0.143, 0.139	0.98
Occipital lobe WMHs	0.009	−0.048, 0.066	0.77	0.044	−0.031, 0.119	0.25	−0.038	−0.135, 0.058	0.43

Note:

^*Models were adjusted for age, sex, education, race/ethnicity, Framingham Stroke Risk Profile (excluding points assigned for age), body surface area, intracranial volume, diagnosis, and APOE-ε4 status.

^†Models were adjusted for age, sex, education, race/ethnicity, Framingham Stroke Risk Profile (excluding points assigned for age), body surface area, intracranial volume, and APOE-ε4 status.

^‡Indicates models where the cardiac output × diagnosis interaction term was significant. An additional 5 participants were excluded from the AD signature analyses.

β indicates the degree of change in volume per 1 L/min increase in cardiac output. AD=Alzheimer’s disease; APOE=apolipoprotein E; CI=confidence interval; MCI=mild cognitive impairment; NC=normal cognition; WMHs=white matter hyperintensities.

Figure 2. Scatterplots of Cardiac Output x Diagnosis on Grey Matter Volumes.

Lines reflect raw values of total grey matter (A), frontal lobe (B) occipital lobe (C), and hippocampal volume (D) corresponding to cardiac output. Shading reflects the 95% confidence interval. A: Interaction p value = 0.02, β = −8881; B: Interaction p value = 0.04, β = −4041; C: Interaction p value<0.001, β = −2089; D: Interaction p value = 0.02, β = −150. NC = normal cognition; MCI = mild cognitive impairment.

3.3. Cardiac Output & White Matter

Cardiac output was unrelated to total or regional WMHs (p-values>0.26, Table 2 but was related to DTI metrics. Lower cardiac output was counter-intuitively related to lower mean (corrected p-values<0.05) and radial diffusivity (corrected p-value=0.03), indicating healthier white matter microstructure. See Supplemental Table II. Results persisted when excluding participants with prevalent CVD or atrial fibrillation. Cardiac output marginally interacted with diagnosis on regional WMHs (p-values>0.04, Supplemental Table I), but stratified results were null (p-values>0.25, Table 2). Cardiac output did not interact with diagnosis on DTI metrics (corrected p-values>0.26). However, stratified analyses revealed counter-intuitive associations between lower cardiac output and lower mean, radial, and axial diffusivity among MCI participants only (corrected p-values<0.05). Results persisted when excluding participants with prevalent CVD or atrial fibrillation. See Figure 3 and Supplemental Table III for details. In stratified models among NC participants, cardiac output did not relate to DTI metrics (corrected p-values>0.10.)

Figure 3. Cardiac Output and DTI in MCI Participants.

Panel A: Mean skeleton shows regions where lower cardiac output is associated with lower mean diffusivity. Scatterplot shows least squares regression relating mean diffusivity values for every participant at one specific cluster and cardiac output. Panel B: Mean skeletons show regions where lower cardiac output is associated with lower radial and axial diffusivity. Parametric p-values, β, and region listed only represent the cluster with the lowest p-value for each DTI metric. All images were taken at z=65. DTI=diffusion tensor imaging; MCI=mild cognitive impairment.

4. Discussion

Among cognitively normal community-dwelling older adults, lower cardiac output was associated with neuroimaging markers of neurodegeneration, including smaller frontal lobe, occipital lobe, parietal lobe, and hippocampal volumes. By contrast, among MCI participants, associations were counter-intuitive, such that lower cardiac output related to larger occipital lobe and hippocampal volumes. While cardiac output was unrelated to measures of white matter macrostructure, unexpectedly, lower cardiac output was associated with greater white matter microstructure measured on DTI. Again, associations were driven by MCI participants.

This study is among the first to show region-specific associations between subclinical cardiac dysfunction and MRI evidence of neurodegeneration among cognitively normal aging adults. Furthermore, results extend prior work reporting associations between lower cardiac output and smaller total brain volume in middle age and older adults (Jefferson et al., 2010;Sabayan et al., 2015). In the setting of chronically reduced cardiac function, even at subclinical levels, an aging cerebrovasculature may be unable to adequately respond, resulting in modest CBF decreases as reported previously (Jefferson et al., 2017). Subtle reductions in CBF (oligemia) can lead to disruptions in cellular homeostasis (Hall et al., 2014), synaptic dysfunction, and subsequent neurodegeneration (Charriaut-Marlangue et al., 1996;Saura et al., 2004). Therefore, age-related changes in cardiac output may lead to neurodegeneration, visualized as grey matter atrophy on neuroimaging.

However, there are alternative explanations for the findings beyond oligemia in the setting of reduced cardiac output. The reported associations may be due to a third variable, such as intrinsic properties of aerobic metabolism (Smith, C.S. et al., 2006;Sullivan et al., 2003), affecting both cardiac output and brain volume. Alternatively, the cross-sectional association with smaller brain volumes may not reflect neurodegeneration. Rather, results could reflect normal variation in brain volumes and subsequent decreased metabolic tissue demand. Longitudinal studies are needed to understand the temporal nature of observations reported here.

While the cardiac function and neurodegeneration association reported here among NC participants appears to be a global process affecting frontal, parietal, and occipital lobes, there may be some regional specificity within the frontal lobe where effects corresponded to over 13 years of brain aging. These largely global findings provide novel insight that the pathways through which cardiac output affects brain structure are not regionally dependent. While an association emerged with the hippocampus, we did not detect any association with temporal lobe volume, despite previous work showing the effects of reduced cardiac output on CBF are most robust in the temporal region (Jefferson et al., 2017). While blood flow changes may occur across the entire lobe, it is plausible that subsequent structural damage first occurs in deep temporal lobe structures, such as the hippocampus. Additionally, null white matter results in NC participants suggest subclinical cardiac dysfunction preferentially affects grey matter. Though white matter is traditionally considered especially vulnerable to ischemia (Pantoni et al., 1996), grey matter is more metabolically active (Pan et al., 2000) and less resistant to ischemia (Falcao et al., 2004;Fünfschilling et al., 2012), perhaps making it more susceptible to very subtle cerebral hemodynamic changes. Since the cohort studied here consists of minimal CVD, CBF reductions due to subclinical cardiac dysfunction may not be robust enough to yield white matter damage as previously reported in a clinical sample of aging CVD patients (Jefferson et al., 2007).

Contrary to expectation, lower cardiac output related to larger grey matter volumes among MCI participants though results did not survive multiple comparison correction. These findings are among the first to suggest that associations between cardiac function and brain health are modified by cognitive diagnosis. Similarly, though cardiac output did not interact with diagnosis on white matter microstructure, lower cardiac output counter-intuitively related to greater white matter microstructural integrity, particularly among MCI participants. Thus, these findings should be interpreted with caution. It is plausible an enrollment bias explains the unexpected results, especially given heart failure was a participant exclusion, and MCI participants likely have more neurodegeneration driving clinical impairment than their NC counterparts. If MCI participants have extensive neurodegeneration and are also susceptible to the effects of subclinical cardiac dysfunction, one might expect a more severe clinical phenotype, such as frank dementia, rather than MCI. An alternate hypothesis is among MCI participants, cardiac output increases in a compensatory manner to maintain adequate oxygen delivery in response to neural injury, aligning with regionally specific CBF fluctuations seen in dementia (Ding et al., 2014). This hypothesis, though speculative, suggests that systemic vascular health has a dynamic association with brain health and may play an important role in the response to brain pathology and injury. Additional work is needed to understand whether these counterintuitive associations persist or vary by clinical severity.

Study strengths include a clinically well characterized cohort, comprehensive ascertainment of potential confounders, the use of core laboratories to analyze all echocardiography and brain MRI measurements in batch with technicians blinded to clinical information, and the application of a cluster enhancement permutation procedure for multiple comparison correction in voxelwise analyses. Limitations include the cross-sectional methods, limited generalizability of findings given the older and predominantly non-Hispanic White cohort, and the single assessment of cardiac output, which may be influenced by acute fluctuations in metabolism not fully reflecting resting cardiac function.

5. Conclusions

The current study provides novel insight into how the effect of systemic cardiac health on brain heath may differ by cognitive diagnosis. Cardiac output is associated with regional neurodegeneration, including lower frontal, occipital, parietal, and hippocampal volumes, among older adults with normal cognition. Oligemia in the setting of subclinical cardiac dysfunction may alter neuronal homeostasis and contribute to neurodegeneration prior to cognitive decline. However, once cognitive decline begins, associations between cardiac output and brain structure yield a different pattern, suggesting cardiac output may change in response to brain health. Future research is needed to understand the mechanism behind these effects and how the presence of underlying brain pathology affects the association between cardiac output and brain structure.

Glossary

AD

Alzheimer’s disease

APOE

Apolipoprotein E

CBF

Cerebral blood flow

CSF

Cerebrospinal fluid

CVD

Cardiovascular disease

DTI

Diffusion tensor imaging

eMCI

Early mild cognitive impairment

FLAIR

Fluid attenuated inversion recovery

FSL

FMRIB Software Library

FSRP

Framingham Stroke Risk Profile

LVH

Left ventricular hypertrophy

MCI

Mild cognitive impairment

MNI

Montreal Neurological Institute

MRI

Magnetic resonance imaging

NC

Normal cognition

ROI

Region of interest

VMAP

Vanderbilt Memory & Aging Project

WMH

White matter hyperintensities