Lower cardiac output is a risk factor for faster cerebral atrophy over a 11‐year follow‐up period in APOE‐ε4 carriers

Abstract

INTRODUCTION

Subclinical lower cardiac output (volume of blood pumped per minute) cross‐sectionally relates to smaller cerebral gray matter volumes in older adults. This study relates cardiac output to longitudinal gray matter volumes among middle‐aged and older adults over an 11‐year period.

METHODS

Linear regression (cross‐sectional) and mixed effects (longitudinal) models related baseline cardiac output to gray matter volume trajectory, adjusting for demographic factors. Secondary models tested cardiac output × apolipoprotein E (APOE) ‐ε4 status interactions.

RESULTS

Lower cardiac output related to smaller brain volumes (p‐values < 0.04) in APOE‐ε4 positive participants only. In longitudinal models, baseline cardiac output interacted with APOE‐ε4 status (p = 0.006). Lower baseline cardiac output related to greater increase in inferior lateral ventricle volume over time in APOE‐ε4 carriers (p = 0.01) only.

DISCUSSION

Results suggest among APOE‐ε4 carriers, subclinical cardiac dysfunction relates to greater neurodegeneration cross‐sectionally and longitudinally. However, results must be interpreted with caution and replicated.

Highlights

• This is the first longitudinal study to examine cardiac dysfunction and atrophy.

• Lower cardiac output is associated with smaller brain volumes in apolipoprotein E (APOE) ‐ε4 carriers.

• Results are most prominent in the medial temporal lobes.

• Subclinical cardiac changes may contribute to greater neurodegeneration over time.

1. BACKGROUND

While the effects of overt cardiac dysfunction and heart failure on brain health are well‐established, ¹ , ² a growing body of evidence now suggests subtle reductions in cardiac function can lead to adverse brain health outcomes even in the absence of cardiovascular disease (CVD). ³ , ⁴ Prior work has shown that lower cardiac output (representing the amount of blood pumped out the heart per minute), even in the normal range, is cross‐sectionally associated with decreased cerebral blood flow (CBF), ⁵ , ⁶ global structural brain changes, ² , ³ , ⁴ , ⁷ and worse cognitive performance, ⁸ , ⁹ , ¹⁰ among middle‐aged and older adults. Specifically, lower cardiac output is cross‐sectionally associated with smaller total, frontal lobe, and occipital lobe gray matter volumes, ¹¹ but not white matter changes. ³ , ⁴ Thus, it is likely that gray matter damage is a primary pathway through which subclinical cardiac dysfunction affects brain health over the lifespan. Given increasing evidence that age‐related changes in systemic vascular health contribute to cognitive decline and Alzheimer's disease (AD), ¹² , ¹³ , ¹⁴ longitudinal studies are needed to better understand how subtle reductions in cardiac output may precipitate or exacerbate abnormal structural brain changes.

The current study examines baseline cardiac output in relation to the cross‐sectional and longitudinal trajectory of gray matter volumes assessed on T₁‐weighted magnetic resonance imaging (MRI) among middle‐aged and older adults free of clinical stroke, dementia, and heart failure at study entry. Based on prior cross‐sectional associations, ³ , ⁴ we hypothesize that lower cardiac output at baseline is associated with faster atrophy in gray matter regions over the follow‐up period. Additionally, we hypothesize that associations are modified by the presence of the apolipoprotein E (APOE) ‐ε4 allele, a genetic risk factor for AD, ¹⁵ and a moderator of vascular damage ¹⁶ and cognitive status. ¹⁷ Past cross‐sectional studies have shown that associations between cardiac output and brain health are driven by APOE‐ε4 carriers, ¹⁰ likely due to the effect of APOE‐ε4 on cerebrovascular autoregulation. ¹⁸ Prior cross‐sectional work has shown that associations between cardiac output and gray matter volumes are driven by cognitively unimpaired (CU) participants, ¹¹ likely due to subtle changes in blood flow that occur with age, ¹⁹ prior to the onset of cognitive impairment. Based on these prior studies, we hypothesize that associations between baseline cardiac output and gray matter atrophy will be strongest among CU participants and APOE‐ε4 carriers.

2. METHODS

2.1. Study cohort

The Vanderbilt Memory and Aging Project (VMAP), ²⁰ is a longitudinal study investigating vascular health and brain aging. Between 2012 and 2014, participants were recruited into the Legacy Cohort who were age 60 years or older and met cognitive status criteria for CU, early mild cognitive impairment (eMCI), ²¹ or MCI. ²² In 2021, the study enrolled additional CU participants into the Expansion Cohort who were age 50 years and older and met cognitive status criteria for CU. Inclusion criteria required participants to be proficient in English, have adequate auditory and visual acuity, and have a study partner. At eligibility, participants underwent a medical history and record review, clinical interview, and neuropsychological assessment for cognitive status by consensus. CU was defined as (a) Clinical Dementia Rating (CDR) = 0 (no dementia), (b) no deficits in activities of daily living directly attributable to cognitive impairment, and (c) no evidence of neuropsychological impairment defined as standard scores falling within 1.5 standard deviations of the age‐adjusted normative mean. ²¹ eMCI was defined as (a) CDR = 0.5 (reflecting mild severity of impairment), (b) no deficits in activities of daily living attributable to cognitive issues, and (c) no neuropsychological impairment defined as standard scores falling within 1.5 standard deviations of the age‐adjusted normative mean. ²¹ Finally, MCI was defined as (a) CDR = 0 or 0.5 (reflecting mild severity of impairment), (b) relatively spared activities of daily living, (c) neuropsychological impairment within at least one cognitive domain (i.e., performances falling greater than 1.5 standard deviations outside the age‐adjusted normative mean or pre‐morbid level of functioning), (d) concern of a cognitive change by the participant, informant, or clinician, and (e) absence of a dementing syndrome. ²²

RESEARCH IN CONTEXT

1. Systematic review: The authors used PubMed to review literature related to the impact of cardiac function on brain health over time. Prior work has identified cross‐sectional associations, but large, longitudinal studies are needed to further understand these complex associations.

2. Interpretation: Lower cardiac output at baseline is associated with small brain volumes over time in the medial temporal lobes in participants who are apolipoprotein E (APOE) ‐ε4 carriers. Results suggest that subclinical cardiac dysfunction may be a key risk factor for future neurodegeneration in individuals with a genetic predisposition to Alzheimer's disease (AD).

3. Future directions: Screening of cardiovascular function, even in middle age, may be a useful tool for risk quantification among individuals who are APOE‐ε4 carriers. Further research is needed to understand why APOE‐ε4 may lead to greater vulnerability in the setting of reduced cardiac function.

Participants were excluded from either cohort for MRI contraindication, history of neurological disease (e.g., stroke, dementia, head injury with loss of consciousness > 5 minutes), prevalent heart failure, major psychiatric illness, and systemic or terminal illness that could impact follow‐up participation. At enrollment, participants completed a comprehensive evaluation, including but not limited to fasting blood draw, physical examination, clinical interview with medication review, echocardiogram, and brain MRI. Assessments were repeated at 18‐month, 3‐year, 5‐year, 7‐year, and 9‐year follow‐up visits. Participants were excluded from the current study for missing cardiac output, brain MRI, or covariate data (Figure 1). The Vanderbilt University Medical Center Institutional Review Board approved the protocol. Written informed consent was obtained from participants prior to the commencement of data collection. Data, analytic methods, and study materials can be obtained by contacting the corresponding author.

FIGURE 1.

Inclusion and exclusion criteria. Missing data categories are mutually exclusive

2.2. Echocardiogram

As previously reported, ⁶ standard two‐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) at baseline. Digital images with measurements were confirmed by a board‐certified cardiologist (D.K.G.) blinded to clinical information. Stroke volume was calculated from the left ventricular outflow tract velocity‐time integral and diameter. 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 three 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 eight‐channel SENSE reception coil array between 2012 and 2017. The system was upgraded to a 32‐channel coil in 2017. In the Legacy Cohort, T₁‐weighted images were acquired with the following parameters: repetition time (TR) = 8.9 ms, echo time (TE) = 4.6 ms, spatial resolution = 1 × 1 × 1mm³. In the Expansion Cohort and Legacy Cohort for 9‐year follow‐up, T₁‐weighted images were acquired with the following parameters: TR = 6.5 ms, TE = 2.9 ms, spatial resolution = 1 × 1 × 1mm³. Images were post‐processed with a Multi‐Atlas Segmentation pipeline. ²³ Images were registered to Montreal Neurological Institute space, appropriate atlases were registered to T₁‐weighted images, and labels were statistically fused. Gray matter volume regions of interest (ROIs) were calculated for total, frontal lobe, temporal lobe, parietal lobe, occipital lobe, hippocampal, and inferior lateral ventricle volumes. ²⁴

To evaluate gray matter changes associated with neurodegeneration due to AD pathology, ²⁵ an AD signature imaging variable was calculated. As previously described, ²⁶ T₁‐weighted images were also post‐processed using FreeSurfer (http://surfer.nmr.mgh.harvard.edu/). ²⁷ The AD signature imaging marker was calculated by summing bilateral cortical thickness measurements from regions shown to distinguish individuals with clinical AD from CU, including the entorhinal cortex, temporal lobe, parietal lobe, fusiform gyrus, and precuneus. ¹¹ , ²⁵ , ²⁶ , ²⁸

Because T₁‐weighted image sequence parameters differed between the VMAP Legacy and Expansion Cohorts, we used longitudinal ComBat, ²⁹ to harmonize T₁‐weighted image outputs across both cohorts. Longitudinal ComBat inputs included the features to be harmonized, the specification of batch variables, and the specification of a linear mixed‐effects model.

2.4. Analytical plan

As previously described, ⁶ covariates were collected at enrollment and selected a priori for their potential to confound analytical models. Covariates included age, sex, education, race/ethnicity, baseline body surface area, baseline Framingham Stroke Risk Profile (FSRP) score, ³⁰ (excluding points assigned for age), cognitive status, intracranial volume, and APOE‐ε4 status. Body surface area (m²) was calculated as weight^0.425 * height^0.725 * 0.007184. APOE‐ε4 status was defined as positive (ε2/ε4, ε3/ε4, ε4/ε4) or negative (ε2/ε2, ε2/ε3, ε3/ε3).

Linear regression models with ordinary least square estimates cross‐sectionally related baseline cardiac output and regional gray matter volumes, adjusting for age, sex, race/ethnicity, education, body surface area, cognitive status, FSRP (minus points assigned for age), APOE‐ε4 status, and intracranial volume (for example, frontal gray matter volume ∼ cardiac output + age + sex + race/ethnicity + education + body surface area + cognitive status + FSRP (minus age points + APOE‐ε4 status + intracranial volume). Models examining the AD signature did not include intracranial volume because the AD signature imaging marker is calculated from cortical thickness measurements, which are not correlated with intracranial volume. ³¹ Models also tested cardiac output × APOE‐ε4 status and cardiac output × cognitive status interactions on gray matter volumes, with follow‐up models stratified by APOE‐ε4 status (positive, negative) and cognitive status (CU, MCI), excluding the small subset of participants with eMCI.

Linear mixed effects regression with random intercepts and slopes and a follow‐up time interaction term related baseline cardiac output to longitudinal gray matter volumes (one ROI per model). Models adjusted for baseline age, sex, race/ethnicity, education, baseline body surface area, baseline FSRP (excluding points assigned for age as age was included as its own covariate), follow‐up time, baseline cognitive status, APOE‐ε4 status, and baseline intracranial volume (for example, frontal gray matter volume ∼ baseline cardiac output*follow‐up time + baseline age + sex + race/ethnicity + education + baseline body surface area + baseline cognitive status + baseline FSRP (minus age points + APOE‐ε4 status + baseline intracranial volume). Models examining the AD signature imaging marker omitted intracranial volume as a covariate. ³¹ Models also tested cardiac output × APOE‐ε4 status and cardiac output × cognitive status interactions with follow‐up time on longitudinal gray matter volumes, with follow‐up models stratified by APOE‐ε4 status (positive, negative) and cognitive status (CU, MCI). Cognitive status interaction models excluded the subset of participants with eMCI (n = 27) as the sample size is too low to be powered to detect a three‐way interaction with CU, eMCI, and MCI.

To determine if outliers were driving the results, additional models were calculated excluding predictor or outcome values > 4 standard deviations from the group mean. 24 total outliers were identified across 1840 observations (1.3%) and they were excluded in a case wise fashion. Multiple comparison correction was performed across outcomes in each model using a false discovery rate (FDR) based on the Benjamini‐Hochberg procedure. ³² Significance was set a priori at p < 0.05, and analyses were conducted using R 4.2.2 (www.r‐project.org).

3. RESULTS

3.1. Participant characteristics

The sample included 756 adults aged 50‐92 years at baseline (67 ± 9 years, 47% males, 81% non‐Hispanic White individuals). Baseline cardiac output was 5.1 ± 1.4 L/min. The mean follow‐up time was 4.7 years, and 57% of the sample had more than one visit. See Table 1 for participant characteristics for the total sample and stratified by APOE‐ε4 status. 600 participants were CU, 129 were MCI, and 27 were eMCI.

TABLE 1.

Participant characteristics

Parameter	Total n = 756	APOE‐ε4 negative n = 488	APOE‐ε4 positive n = 268	p‐Value
Demographic and health characteristics
Age, years	67 ± 9	68 ± 9	67 ± 8	0.08
Sex, % male	47	48	46	0.68
Race, % White non‐Hispanic	81	81	80	0.72
Education, years	16 ± 2	16 ± 2	16 ± 2	0.79
Body surface area, m2	1.9 ± 0.2	1.9 ± 0.2	1.9 ± 0.2	0.98
Montreal Cognitive Assessment, total	26 ± 3	26 ± 3	26 ± 3	0.19
Framingham Stroke Risk Profile, total*	9.7 ± 4.9	9.7 ± 5.0	9.8 ± 4.8	0.88
Systolic blood pressure, mmHg	134 ± 17	134 ± 17	134 ± 17	0.62
Anti‐hypertensive medication usage, %	44	43	47	0.25
Diabetes, %	14	11	20	0.001
Current cigarette smoking, %	2	2	1	0.72
Prevalent CVD, %	5	4	6	0.20
Atrial fibrillation, %	3	3	3	0.71
Left ventricular hypertrophy, %	8	7	11	0.05
Cardiac output, L/min	5.1 ± 1.4	5.1 ± 1.3	5.1 ± 1.5	0.89
Follow‐up time, years	4.7 ± 3.0	4.7 ± 3.1	4.6 ± 3.0	0.52
Gray matter variables
Total brain volume annual change, cm3	−3343 ± 7856	−3369 ± 7551	−3296 ± 8423	0.57
Frontal lobe volume annual change, cm3	−1108 ± 3725	−1116 ± 3747	−1092 ± 3696	0.86
Temporal lobe volume annual change, cm3	−657 ± 1417	−621 ± 1362	−726 ± 1517	0.79
Parietal lobe volume annual change, cm3	−791 ± 2191	−821 ± 2005	−735 ± 2508	0.90
Occipital lobe volume annual change, cm3	−354 ± 1080	−365 ± 1016	−334 ± 1195	0.45
Hippocampal volume annual change, cm3	−59 ± 102	−51 ± 100	−75 ± 104	0.02
Inferior lateral ventricular volume annual change, cm3	108 ± 184	81 ± 145	158 ± 233	0.001
AD signature annual change, mm	−0.008 ± 0.029	−0.007 ± 0.028	−0.010 ± 0.030	0.16

Note: Values presented as mean ± standard deviation or frequency. Values in bold indicate p < 0.05. *A modified score was included in models excluding points for age (6 ± 3).

Abbreviations: AD, Alzheimer's disease; APOE, apolipoprotein E; CU, cognitively unimpaired; CVD, cardiovascular disease; MCI, mild cognitive impairment.

3.2. Cardiac output and gray matter

In cross‐sectional models, cardiac output was not related to gray matter volumes (p‐values > 0.33, Table S1), consistent with prior published results from this cohort in a smaller sample size. ¹¹ In longitudinal models, baseline cardiac output was not related to a change in gray matter volume in any region (p‐values > 0.16). See Table 2 for details. When excluding outliers, associations remained unchanged (data not shown).

TABLE 2.

Cardiac output associations with longitudinal gray matter volumes

Gray matter region	β	95% Confidence intervals	p‐Value
Total brain volume	−244	−682, 194	0.28
Frontal lobe volume	−153	−364, 58	0.16
Temporal lobe volume	−32	−114, 50	0.44
Parietal lobe volume	−49	−161, 63	0.39
Occipital lobe volume	27	−24, 78	0.30
Hippocampal volume	1.23	−5.0, 7.5	0.70
Inferior lateral ventricular volume	−7.1	−18.6, 4.4	0.22
AD signature	0.001	−0.0007, 0.0018	0.36

Note: n = 756. Models adjusted for age, sex, race/ethnicity, education, Framingham Stroke Risk Profile (removing points assigned for age), APOE‐ε4 status, body surface area, cognitive status, follow‐up time, and intracranial volume. Models examining the AD signature outcome did not covary for intracranial volume and excluded 18 additional participants.

Abbreviations: AD, Alzheimer's disease; APOE‐ε4, apolipoprotein E‐ε4.

3.3. Cardiac output × APOE‐ε4 status and gray matter

In cross‐sectional models, cardiac output did not interact with APOE‐ε4 status on gray matter volumes. However, in stratified models, lower cardiac output was associated with smaller total brain (β = 5768, p = 0.02, p_FDR = 0.05), temporal ,   lobe (β = 1184, p = 0.02, p_FDR = 0.05), and occipital lobe (β = 884, p = 0.02, p_FDR = 0.05) in APOE‐ε4 positive participants only (Figure 2, Table 3). Associations with parietal lobe (β = 1105, p = 0.04, p_FDR = 0.06), and hippocampal volumes (β = 69, p = 0.04, p_FDR = 0.06) did not meet the threshold for significance after FDR correction. When excluding outliers, no additional associations met the threshold after FDR correction emerged. In post‐hoc analyses examining whether heart rate or stroke volume was driving these associations, lower heart rate was associated with smaller total brain (β = 876, p = 0.01), frontal lobe (β = 321, p = 0.03), temporal lobe (β = 143, p = 0.04), and parietal lobe volumes (β = 221, p = 0.004). Stroke volume was not associated with any gray matter volumes (p‐values > 0.07).

FIGURE 2.

Cardiac output total brain volume in APOE‐ε4 carriers. Shading reflects the 95% confidence interval. β = 5,768, p = 0.02. APOE‐ε4, apolipoprotein E‐ε4

TABLE 3.

Cardiac output × APOE‐ε4 status interactions on cross‐sectional gray matter volumes with stratification

Parameter	Cardiac output × APOE‐ε4 status interaction n = 756			APOE‐ε4 negative n = 488			APOE‐ε4 positive n = 268
	β	95% CI	p‐Value	β	95% CI	p‐Value	β	95% CI	p‐Value
Total brain volume	2,156	−3,447, 7,759	0.45	1,393	−2,669, 5,456	0.50	5,768	930, 10,610	0.02 *
Frontal lobe volume	−144	−2,432, 2,143	0.90	1,092	−555, 2,738	0.19	1,854	−148, 3,856	0.07
Temporal lobe volume	849	−279, 1,977	0.14	212	−607, 1,032	0.61	1,184	211, 2,157	0.02 *
Parietal lobe volume	334	−903, 1,571	0.60	185	−711, 1,082	0.69	1,105	36.6, 2,173	0.04
Occipital lobe volume	604	−317, 1,525	0.20	−85.5	−768, 598	0.81	884	127, 1,641	0.02 *
Hippocampal volume	70.3	−2.51, 143	0.06	−22.0	−73.7, 29.8	0.41	69.1	4.77, 133	0.04
Inferior lateral ventricular volume	−0.34	−88.9, 88.2	0.99	−2.01	−61.1, 57.1	0.95	27.6	−57.9, 113	0.53
AD signature	0.007	−0.004, 0.017	0.22	−0.003	−0.011, 0.004	0.38	0.0002	−0.009, 0.009	0.96

Note: Models adjusted for age, race/ethnicity, education, Framingham Stroke Risk Profile (removing points assigned for age), cognitive status, body surface area, and intracranial volume. * indicates findings that would survive correction for multiple comparisons. Models examining the AD signature outcome did not covary for intracranial volume and excluded 41 additional participants. Values in bold indicate p < 0.05.

Abbreviations: AD, Alzheimer's disease; APOE‐ε4, apolipoprotein E‐ε4; CI, confidence interval.

In longitudinal models, baseline cardiac output interacted with APOE‐ε4 status on inferior lateral ventricular volume (β = ‐30.5, p = 0.006, p_FDR = 0.04). Among APOE‐ε4 positive participants only, stratified models did not meet the threshold for significance after FDR correction (p_fdr > 0.07). Among APOE‐ε4 negative participants, baseline cardiac output was not related to gray matter volume trajectory (p‐values > 0.16). See Table 4 and Figure 3 for details. When excluding outliers, associations persisted (data not shown). In post‐hoc analyses examining whether heart rate or stroke volume were driving associations, both heart rate and stroke volume interacted with APOE‐ε4 status on inferior lateral ventricle volume (p‐values = 0.05). Lower heart rate was associated with larger inferior lateral ventricle volume over time in APOE‐ε4 participants (β = ‐3.5, p = 0.04), but stroke volume was not (β = ‐1.3, p = 0.11).

TABLE 4.

Cardiac output × APOE‐ε4 status interactions on longitudinal gray matter volumes with stratification

	Cardiac output × APOE‐ε4 status interaction n = 756			APOE‐ε4 negative n = 488			APOE‐ε4 positive n = 268
Parameter	β	95% CI	p‐Value	b	95% CI	p‐Value	β	95% CI	p‐Value
Total brain volume	466	−434, 1,366	0.31	−307	−810, 196	0.23	15	−813, 842	0.97
Frontal lobe volume	58	−367, 484	0.79	−176	−419, 67	0.16	−101	−506, 303	0.62
Temporal lobe volume	97	−80, 275	0.28	−50	−144, 45	0.30	18	−141, 177	0.82
Parietal lobe volume	100	−128, 329	0.39	−62	−189, 64	0.33	16	−204, 236	0.89
Occipital lobe volume	91	−18, 199	0.10	6.5	−52, 65	0.83	82	−21, 184	0.12
Hippocampal volume	5.20	−8.1, 18.5	0.44	−0.58	−7.9, 6.8	0.88	5.0	−6.4, 16.5	0.39
Inferior lateral ventricular volume	−30.5	−52.3, ‐8.7	0.006 *	3.4	−8.9 15.6	0.59	−30	−53, ‐7.4	0.01
AD signature	0.001	−0.001, 0.004	0.29	0.0002	−0.001, 0.002	0.71	0.001	−0.001, 0.004	0.26

Note: Models adjusted for age, race/ethnicity, education, Framingham Stroke Risk Profile (removing points assigned for age), cognitive status, body surface area, follow‐up time, and intracranial volume. *indicates findings would survive correction for multiple comparisons. Models examining the AD signature outcome did not covary for intracranial volume and excluded 18 additional participants. Values in bold indicate p < 0.05.

Abbreviations: AD, Alzheimer's disease; APOE‐ε4, apolipoprotein E‐ε4; CI, confidence interval.

FIGURE 3.

Cardiac output and APOE‐ε4 status interactions on longitudinal inferior lateral ventricle volume. Shading reflects the 95% confidence interval. Interaction p = 0.006, APOE‐ε4 positive p = 0.01, APOE‐ε4 negative p = 0.59. APOE‐ε4, apolipoprotein E‐ε4

In post‐hoc analyses to determine if age was driving these results (though age is similar between APOE‐ε4 positive and negative participants), cardiac output did not interact with age on cross‐sectional or longitudinal gray matter volumes (p‐values > 0.10). Stratified results by median age split were also null (p‐values > 0.15), suggesting older age is not driving these results.

3.4. Cardiac output × cognitive status interactions and gray matter

In cross‐sectional models, cardiac output did not interact with cognitive status on gray matter volumes (p‐values > 0.08, Table S2). In stratified models, lower cardiac output was not cross‐sectionally related to gray matter volumes after FDR correction. In CU participants only, associations with lower total brain (β = 3167, p = 0.05, p_FDR = 0.17) and frontal lobe volume (β = 1438, p = 0.03, p_FDR = 0.17) approached significance, consistent with prior published work. ²⁶ In post‐hoc analyses examining whether heart rate or stroke volume were driving associations, neither heart rate nor stroke volume was associated with brain volumes (p‐values > 0.06).

In longitudinal models, baseline cardiac output did not interact with cognitive status on gray matter volumes (p‐values > 0.07). In stratified models, baseline cardiac output was not associated with a change in gray matter volume in any region in CU or MCI participants after FDR correction (p_FDR > 0.07). See Table S3 for details. Results were similar when excluding outliers (data not shown). In post‐hoc analyses examining whether heart rate or stroke volume were driving associations, heart rate was associated with smaller occipital lobe volume over time (β = 17, p = 0.04) over time, but stroke volume was not (β = 6.5, p = 0.17).

4. DISCUSSION

This study is among the first to examine the impact of subclinical cardiac dysfunction on neurodegeneration over time, building on previous cross‐sectional work, ¹¹ and among the first to comprehensively examine the impact of APOE‐ε4 on these associations. Among community‐dwelling older adults, this study found for the first time that subclinical cardiac dysfunction at baseline was associated with greater cross‐sectional and longitudinal rates of atrophy in individuals who are APOE‐ε4 carriers. In longitudinal models, associations are strongest in the medial temporal lobe. Importantly, this study includes the largest sample size to date (n = 756), examining how baseline cardiac output relates to neurodegeneration over time. Further, this study has a robust mean follow‐up time of 4.7 years, ranging up to 11 years, allowing for novel observations of how subtle changes in cardiac health in middle and older age impact brain health over time. Collectively, findings suggest that subclinical cardiac dysfunction, even prior to overt cardiac disease, is a key risk factor for exacerbating age and AD‐related neurodegeneration among individuals with a genetic predisposition for AD.

First, we found that among APOE‐ε4 carriers, lower cardiac output was cross‐sectionally associated with smaller volumes in the whole brain, temporal lobe, and occipital lobe. Associations were not present among APOE‐ε4 non‐carriers, suggesting that APOE‐ε4 carriers are more vulnerable to small, sustained reductions in cardiac function. Cardiac output likely impacts neurodegeneration through reduced CBF, as previously reported. ⁶ Subtle reductions in CBF lead to impaired neuronal regulation, ³³ and hyperexcitability, ³⁴ resulting in neurodegeneration. APOE‐ε4 is expressed in astrocytes, ³⁵ and causes greater blood–brain barrier (BBB) permeability through initiating an inflammatory cascade in pericytes. ³⁶ Therefore, in APOE‐ε4 carriers, the BBB response to subtle changes in CBF is likely less effective, making them more vulnerable to subtle changes in cardiac output and leading to greater atrophy. Cross‐sectional results largely implicate the entire brain, suggesting there is no regional specificity. However, it should be noted that there was not a significant APOE‐ε4 interaction in cross‐sectional models, so these findings must be replicated in other cohorts.

Second, in longitudinal models, we found that cardiac output interacted APOE‐ε4 carrier status on inferior lateral ventricle volume. Though stratified results were not significant, they suggest that lower cardiac output is associated with expansion of the inferior lateral ventricle in APOE‐ε4 carriers but not among APOE‐ε4 non‐carriers. Expansion of the inferior lateral ventricle is attributable to atrophy in adjacent brain structures, namely the medial temporal lobe. The medial temporal lobe is where AD pathology first develops. ³⁷ It is possible that, as AD pathology builds up over time, this region becomes more susceptible to damage from subclinical cardiovascular changes, leading to greater atrophy. Another possibility is that this region is uniquely susceptible to vascular dysfunction due to a lower density of capillaries compared to other parts of the brain, ³⁸ which is consistent with recent work showing that the medial temporal lobe specifically is susceptible to small fluctuations in blood pressure, ³⁹ and hypoperfusion. ⁴⁰ More work is needed to understand the selective vulnerability of this region over other portions of the temporal lobe. Taken together, our results are among the first to suggest that the association between subclinical cardiac dysfunction and inferior lateral ventricular volume is modified by APOE‐ε4 status. Results highlight the need to take genetic information into account for precise prevention and treatment approaches for AD. Importantly stratified results did not meet the significance threshold after FDR correction, so findings must be interpreted with caution, highlighting that future studies are needed with larger APOE‐ε4 sample sizes.

Results also show that diagnostic interactions and stratification by cognitive status were largely null after correction for multiple comparisons. This is contrary to prior work, as we previously published cross‐sectional associations between cardiac output and lower gray matter volumes in CU participants. ¹¹ This difference may be due to changes in the sample size. Longitudinal results were also null, suggesting that the association between cardiac output and gray matter volume trajectory overtime is not modified by cognitive status. There was a mild association between lower baseline cardiac output and faster rates of atrophy in the occipital lobe in MCI participants only, but this did not survive correction for multiple comparison. More studies are needed to explore this trend, but this region may be more susceptible to subclinical cardiac dysfunction in the context of cognitive impairment. The occipital lobe is thought to be preserved until the later stages of AD, ⁴¹ so it is possible that the occipital lobe is most susceptible to changes to cardiac output compared to other regions of the brain with evolving AD pathology. This finding also aligns with other evidence suggesting the occipital lobe is more vulnerable to vascular risk in aging, ⁴² , ⁴³ but more work is needed to understand the susceptibility of the occipital lobe in normal aging and AD.

Our study has several strengths, including a clinically well‐characterized cohort, comprehensive ascertainment of potential confounders, and the use of core laboratories to analyze all echocardiography and brain MRI measurements in batches with technicians blinded to clinical information. Additionally, this study is among the largest to examine associations between cardiac output and brain health with a very robust follow‐up time of up to 11 years, greatly expanding our knowledge on how subtle alterations in cardiovascular health impact brain aging. Limitations include the predominantly White, well‐educated, and relatively healthy cohort, limiting generalizability to other races, ethnicities, and medical conditions. However, we might expect associations to be stronger among individuals with a greater burden of vascular risk or CVD. A single cohort was also used, so replication is needed. Additionally, cardiac output was measured during a single assessment, which may be influenced by acute fluctuations in metabolism not fully reflecting resting cardiac function. Finally, it should be noted that many findings did not survive correction for multiple comparisons, significant interaction results did not necessarily yield significant stratified results, and significant stratified results were not couple with a significant interaction, so results must be interpreted with caution.

Among older adults who are APOE‐ε4 carriers, lower baseline cardiac output is associated with smaller brain volumes cross‐sectionally and larger inferior lateral ventricle volumes over time. In individuals with a genetic predisposition to AD, subclinical cardiac dysfunction may be one pathway through which individuals have worse brain health outcomes. Reduced cardiac output may lead to oligemia, placing vulnerable neurons at risk for dysfunction and eventual neurodegeneration. Results suggest that cardiac screening may be an important component of future screening and prevention efforts in this population.

CONFLICT OF INTEREST STATEMENT

T.J.H. serves on the Scientific Advisory Board for Circular Genomics and serves as Deputy Editor for Alzheimer's and Dementia: TRCI and Senior Associate Editor for Alzheimer's and Dementia. A.L.J. has served as an advisor for Lantheus – Diagnostic and Therapeutic Innovations. D.L. serves as a member of the Independent Data Monitoring Committee for an investigational drug developed by Regeneron Inc. This role was independent, and the company had no influence on the content or conclusions of this manuscript. The remaining authors have nothing to disclose. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

Written informed consent was obtained from participants prior to the commencement of data collection.

Supporting information

Supporting information: alz71598‐sup‐0001‐Table S1‐S3.docx

Supporting information: alz71598‐sup‐0002‐ICMJE.pdf