Cardiorespiratory fitness in relation to cerebral vascular and metabolic health in older adults with coronary artery disease

Abstract

This study investigated whether higher cardiorespiratory fitness (VO₂peak) is associated with better cerebral vascular and metabolic health—specifically cerebral blood flow (CBF), cerebrovascular reactivity (CVR), cerebral metabolic rate of oxygen (CMRO₂), and oxygen extraction fraction (OEF) across healthy older adults and patients with coronary artery disease (CAD). Thirty-seven healthy adults (65.3 ± 8.3 years old) and 35 CAD patients (66.4 ± 9.3 years old) underwent calibrated fMRI using hypercapnic and hyperoxic gas challenges (including carbogen) to quantify gray matter CBF, CVR, CMRO₂, and OEF. VO₂peak was obtained from a maximal cycle ergometer cardiopulmonary exercise test. Associations between VO₂peak and brain biomarkers were evaluated with group terms to test CAD-specific effects. Across all participants, VO₂peak was positively associated with CBF (β = 0.32, p = 0.02) and CVR (β = 0.002, p = 0.04) in gray matter, indicating an association between aerobic fitness and vascular health. Metabolic effects differed by group: in CAD patients, VO₂peak correlated positively with CMRO₂ (β = 0.08, p = 0.02), suggesting higher fitness may be associated with preserved oxidative metabolism, while in healthy controls, VO₂peak was negatively associated with OEF (β = −3.6, p = 0.02), consistent with aging-related adaptations driven by improved CBF without CMRO₂ changes. VO₂peak is positively associated with cerebral vascular function in older adults and shows group-specific metabolic benefits: in CAD, higher VO₂peak relates to preserved CMRO₂, whereas in healthy individuals it is linked primarily to enhanced perfusion and reduced extraction. These findings support aerobic exercise as a promising strategy to mitigate CAD-related brain alterations and highlight VO₂peak as a potentially modifiable target for prevention and rehabilitation.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s11357-026-02128-8.

Introduction

Coronary artery disease (CAD), caused by atherosclerotic narrowing of coronary arteries, is the most common form of cardiovascular disease and the first cause of death worldwide [1]. In addition to increased mortality risk, CAD also has extensive effects beyond cardiac health, significantly impacting the brain. CAD is associated with gray matter atrophy, reduced cerebral blood flow (CBF), and impaired cerebrovascular reactivity (CVR), which is the ability of brain vessels to dilate or constrict appropriately in response to metabolic demands [2, 3]. Collectively, these vascular and structural impairments contribute to an elevated risk of cognitive decline [4]. In our recent work [5], we demonstrated that individuals with CAD exhibit not only deficits in CBF and CVR, but also altered brain metabolism, with individuals with CAD showing reduced cerebral metabolic rate of oxygen consumption (CMRO₂), the rate at which brain tissue consumes oxygen to sustain neuronal activity, and elevated oxygen extraction fraction (OEF), a marker of the balance between oxygen delivery and consumption, reflecting the proportion of oxygen extracted from blood by brain tissue.

Structured cardiac rehabilitation programs aimed at improving cardiorespiratory fitness improve patient survival and quality of life in cardiovascular disease, with VO₂peak levels and their improvement being strong independent predictors of morbidity and mortality in CAD [6, 7]. Beyond cardiovascular improvements, higher cardiorespiratory fitness — typically measured by peak oxygen uptake (VO₂peak) — is associated with beneficial effects on brain structure and function. For instance, improved VO₂peak following rehabilitation correlates positively with increased brain volume [8]. Additionally, cardiac rehabilitation has been shown to directly enhance CBF, suggesting beneficial cerebrovascular adaptations in CAD [3, 8]. However, the relationship between VO₂peak and other affected brain physiological components in CAD, such as CVR and metabolic function, remains almost completely unexplored.

In normal aging, VO₂peak has been positively associated with better executive function and memory [9, 10], as well as with greater brain volume in the frontal and temporal lobes and the hippocampus [11, 12]. However, its relationship with physiological markers—such as CBF, CVR, CMRO₂, and OEF—is less well documented. The relationship between VO₂peak and vascular biomarkers such as CBF and CVR has been documented in a few populations. CBF was found to be positively associated with VO₂peak in children [13], young and older adults [14], and CAD patients [8], and in mild cognitive impairment (MCI) [15]. These studies typically show positive associations between VO₂peak and CBF, suggesting that greater aerobic fitness may enhance cerebral perfusion and support brain function [8, 13, 15]. However, a few studies have documented null or even negative associations (particularly in older populations), suggesting that age may moderate the relationship [14, 16].

Only a few studies have investigated the relationship between VO₂peak and CVR, showing a more complex relationship with fitness than CBF. DuBose et al. [17] have suggested that the relationship may follow an inverse U-shape, with a positive relationship between CVR and VO₂peak for individuals who are less physically active and a negative relationship for those who are more active than average. This is consistent with several studies in the aging literature [14, 18, 19], but the relationship in populations with heart disease remains to be established. Given that both CBF and CVR are affected in CAD [3] and show a relationship with cardiorespiratory fitness in normal aging, examining their relationship with VO₂peak in CAD may provide insight into mechanisms linking cardiorespiratory fitness to brain vascular health in this population.

While the relationship between cardiorespiratory fitness and CBF has often been reported, there is a lack of research on how cardiorespiratory fitness affects resting cerebral metabolic biomarkers (CMRO₂ and OEF). This gap persists despite evidence from functional near-infrared spectroscopy (fNIRS) studies suggesting that higher fitness levels may influence brain oxygenation dynamics during exercise and cognitive tasks, particularly in older adults [20, 21]. To date, only two recent studies have directly explored the relationship between VO₂peak and brain metabolism, both focusing exclusively on healthy older adults. Huppert and Qiao [22] reported a negative relationship between global CMRO₂ and VO₂peak, whereas a recent study from our group [23] found no significant association between VO₂peak and CMRO₂. Instead, this previous study identified a negative relationship between VO₂peak and oxygen extraction fraction (OEF), highlighting potential differences in how fitness impacts cerebral metabolism and oxygen availability. However, this relationship has not been explored in CAD, where metabolic impairments may be more pronounced [5]. This study aims to comprehensively evaluate — for the first time —the association between VO₂peak and cerebral metabolic markers (CMRO₂ and OEF) across the whole brain in both CAD patients and healthy controls. Documenting these relationships could help identify sensitive metabolic biomarkers linked to fitness, facilitating more targeted interventions to preserve brain health in clinical populations.

Therefore, the aim of this study is to investigate the relationships between VO₂peak and a comprehensive array of cerebral vascular and metabolic biomarkers in individuals with CAD as compared to healthy controls (matched by age and sex). We hypothesize that greater VO₂peak will be associated with better cerebral vascular and metabolic health, denoted by higher CBF, CVR, and CMRO₂ and lower OEF, especially in CAD patients given their larger deficits. This study will break new ground in our understanding of how cardiorespiratory fitness can improve brain physiological health in CAD patients.

Materials and methods

Participants

Ninety-nine participants (16 females) aged ≥50 were recruited; 87 completed the study (46 CAD, 41 controls). Twelve discontinued: five due to COVID-19 interruptions, two for claustrophobia, two for MRI discomfort, and three for personal reasons. Of the 87 completers, 12 provided only resting CBF data (5 minutes) due to hypercapnia discomfort, leaving 75 with full datasets (S1 Figure). Three with poor imaging quality were excluded, yielding a final sample of 35 CAD and 37 controls with gas challenge (Table 1). The study received approval from the Comité d’éthique de la recherche et du développement des nouvelles technologies (CÉRDNT) of the Montreal Heart Institute in accordance with the Declaration of Helsinki. Data collection was conducted at the Montreal Heart Institute. CAD inclusion criteria required a confirmed history of coronary artery disease, including prior acute coronary syndromes (unstable angina, NSTEMI, STEMI), coronary revascularization (PCI, CABG), or clinically documented angina; coronary angiography was usually performed as standard care (Table 2). Controls were free of cardiac disease, diabetes, hypertension, and cognitive impairment. All participants were fluent in English or French to ensure informed consent and accurate cognitive assessment. Written informed consent was obtained at the first visit.

Table 1.

Demographic data on participants with gas challenge data

	Controls (37)	CAD (35)	P-value
Age (years)	65.35 ± 8.31	66.42 ± 9.29	0.83
Sex	Female (10)	Female (6)	-
Education (years)	16.22 ± 2.45	16.03 ± 4.03	0.80
VO2peaka (ml/min/kg LBM)	27.95 ± 6.46	22.14 ± 5.11	< 0.01
MoCAb	26.73 ± 2.01	25.33 ± 3.04	0.02
MMSEc	28.25 ± 1.48	28.42 ± 1.17	0.97
Executive function (z-score)	0.314 ± 0.62	−0.32 ± 1.01	< 0.01
Working memory (z-score)	0.04 ± 0.83	0.033 ± 0.79	0.95
Processing speed (z-score)	0.26 ± 0.72	−0.34 ± 0.86	< 0.01
Verbal/episodic memory (z-score)	0.19 ± 0.96	0.041 ± 0.97	0.31
SBP (mmHg)d	127.12 ± 13.85	126.96 ± 14.54	0.96
BMI (kg/m2)e	26.11 ± 3.94	26.95 ± 3.57	0.36

Bold values indicate statistically significant differences

^aVO₂peak, peak oxygen uptake during maximal exercise; a measure of cardiovascular fitness

^bMoCA, Montreal Cognitive Assessment; a 30-point test screening for mild cognitive impairment

^cMMSE, Mini-Mental State Examination; a brief 30-point test to assess global cognitive function

^dSBP, systolic blood pressure

^eBMI, body mass index

Table 2.

Clinical information in CAD and HC participants

Clinical health characteristics in patients with CAD (n = 35), healthy controls (n = 37)
	CAD patients	Healthy controls
Cardiometabolic risk factors, n (%)
Diabetes	3 (8)	-
Hypertension	18 (51)	-
Dyslipidemia	28 (80)	3(8)
Obesity	5(14)	5(11)
Familial history of cardiovascular disease	21 (60)	6(16)
Cardiovascular history, n (%)
Previous MIa	14 (40)	-
STEMIb	8 (22)	-
NSTEMIc	6 (17)	-
Percutaneous coronary intervention (PCI)	22 (62)	-
Coronary artery bypass grafting (CABG)	10 (28)	-
Angina	10 (28)	-
Cardiovascular medication, n (%)
Beta-blocker	17 (48)	-
Aspirin	11 (31)	-
Antilipemic agents (e.g., statins)	26 (74)	1(2)
RAASd inhibitors	11 (31)	-
Calcium channel blockers	5 (14)	-

^aMI, myocardial infarction

^bSTEMI, ST-segment elevation myocardial infarction

^cNSTEMI, non-ST-segment elevation

^dRAAS, renin–angiotensin–aldosterone system

Exclusion criteria included stroke, neurological/psychiatric/respiratory or thyroid disease, Mini Mental State Examination (MMSE) < 25 (Folstein et al., 1975), tobacco use, high alcohol intake (> 2/day), MRI contraindications, recent surgery (< 6 months), recent acute coronary event (< 3 months), left ventricular ejection fraction < 40%, symptomatic aortic stenosis, severe CAD, pending bypass surgery, implanted defibrillator/pacemaker, malignant arrhythmias, arthritis/claudication, severe exercise intolerance, or hypercapnia discomfort > 5 on the Banzett scale.

Data collection spanned four visits: eligibility screening (medical history, MMSE, 2-min hypercapnia tolerance test; two excluded), cognitive assessment, VO₂peak test, and MRI acquisition.

Maximal cardiopulmonary exercise test (CPET)

Incremental CPET to volitional exhaustion was performed on a cycle ergometer (E100, Cosmed, Italy) according to the latest recommendations and as previously published [24, 25]. A 3-min warm-up phase at an initial workload of 20 W was followed by an incremental exercise test, with increases of 10 to 20 W per minute depending on the participant’s physical capacity, performed at a pedaling cadence between 60 and 80 rpm. The recovery phase consisted of 2 min of active recovery at 20 W at pedalling speed between 50 and 60 rpm, followed by 3 min of passive recovery where participants remained seated quietly on the ergometer. Gas exchange parameters were continuously measured at rest, during exercise, and during recovery using a metabolic system (Cosmed Quark, Cosmed, Italy), capturing on a breath-by-breath basis and then averaged in 10-s increments as recently published [24]. There was continuous electrocardiogram (ECG) monitoring before and during the test and in the recovery (T12x, Cosmed, Italy). Diastolic (DBP) and systolic (SBP) blood pressures (Tango M2, Suntech, USA) and ratings of perceived exertion (RPE) were measured at rest and every 3 min throughout the test. Participants reached maximal effort when one of the following four criteria was met: (1) a plateau of the V̇O₂ (< 150 mL/min in the last 30 s of effort) despite increased power; (2) a respiratory exchange ratio > 1.10; (3) an inability to maintain 50 rpm; or (4) patient exhaustion due to fatigue or other clinical symptoms (ECG and/or blood pressure abnormalities requiring exercise cessation) [25].

The highest VO₂ value reached during the exercise phase, averaged over a 10-s interval, was considered VO₂peak and expressed in mL/min/kg of lean body mass (Krachler et al., 2015). Lean body mass was estimated using a body composition analyzer (BC 418, TANITA, Japan), which also provided measurements of fat mass.

MRI acquisition

Data were collected using a 3-T Skyra MRI system with a 32-channel head coil. The acquisition protocol included structural MRI, dual-echo pseudo-continuous arterial spin labeling (pCASL) for simultaneous acquisition of perfusion and bold oxygen level dependent (BOLD) signals, and a blood magnetization map (M0) for perfusion quantification. pCASL data were acquired with a voxel resolution of 3.4375 × 3.4375 × 7 mm, TR/TE1/TE2/alpha: 4000/10/30 ms/90°, and a labeling duration of 1517 ms with a post-labeling delay (PLD) of 1300 ms. The M0 acquisition had identical parameters, but with a TR of 10 s, to ensure fully recovered magnetization.

Because the MRI data was acquired within two sub-sessions to allow the participant to take off the mask for the rest of the data acquisition, two T1-weighted acquisitions were collected to ensure registration accuracy of the pCASL slab. One lower resolution acquisition was performed right before pCASL while the person had the mask on to facilitate pCASL slab registration, and a higher resolution acquisition was performed during the second sub-session, without the mask, to be used as the main source of anatomical information. The low-resolution T1-weighted acquisition was acquired using a magnetization prepared rapid gradient echo (MPRAGE) sequence with TR/TE/flip angle = 15 ms/3.81 ms/25° with a 1.5-mm isotropic resolution. The high-resolution T1-weighted structural images were acquired with a MPRAGE sequence, TR/TE/TI/Flip angle = 2300 ms/2.32 ms/900 ms/8° with a 0.9-mm isotropic resolution.

Respiratory manipulation

The RespirActTM system (RespirAct_TM, Thornhill Research, Toronto, Canada) was used during the breathing manipulation to target specific end-tidal partial pressures of CO₂ and O₂. Three conditions were sequentially targeted for 2 min, preceded and followed by 2 min of air inhalation: a hypercapnic condition (targeting 5 mmHg above baseline for end-tidal partial pressure CO₂ (PETCO₂)), an isocapnic hyperoxic condition (targeting 150 mmHg for end-tidal O₂), and a combined hyperoxia and hypercapnia condition (targeting 5 mmHg above baseline for PETCO₂ and 150 mmHg for PETO₂). A rebreathing face mask was used to deliver gases to the participants. The system delivered gas at a flow rate of 20 L/min, while the concentrations of exhaled CO₂ and O₂ were monitored. On their first study visit, a familiarization session was done with a 2-min hypercapnia manipulation to ensure participant comfort during the MRI. Breathing discomfort was assessed using the Banzett dyspnea scale, and those with a score > 5 were not invited to continue in the study (n = 2) [26].

Respiratory data analysis

CO₂ and O₂ were sampled continuously throughout the breathing manipulation by the RespirAct, which also generates a time series of end-tidal CO₂ and O2 partial pressures. The end-tidal partial pressure is used as a proxy for arterial gas partial pressure [27]. To combine these values with the MRI data, a MATLAB script was employed to smooth the data, remove outliers, and resize the data to match the durations of the BOLD and ASL signals.

Imaging data processing

T1 processing

All structural images underwent preprocessing with the Brain Extraction Tool (BET) in FSL to remove the skull. Subsequently, FSL’s FAST was employed to segment the structural images into gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF).

Preprocessing of BOLD and pCASL

Both ASL and BOLD datasets were motion-corrected using FSL’s mcflirt. A brain mask was created from the motion-corrected BOLD images using BET and applied to both echoes to remove the scalp. The first echo of the dual-echo pCASL acquisition was used to compute the perfusion-weighted time series, from which CBF was estimated via surround subtraction. Volumes with voxel intensities exceeding 3 standard deviations from the mean were flagged, and those with > 50% outlier voxels were excluded to ensure sufficient temporal signal-to-noise ratio (tSNR) [28]. Participants were excluded if their pCASL data showed a tSNR below 0.5 or if they exhibited excessive motion during acquisition, defined as translational head movements exceeding 2 mm, to ensure overall data quality. The absence of macrovascular enhancement was ascertained by visual inspection and by quantifying the coefficient of variation as in [29], which was well below the level expected [30] for significant macrovascular contribution (see Supplementary Materials Section 2).

The M0 image was skull-stripped using BET and used for perfusion calibration. The second echo was used to extract the BOLD-weighted signal via surround addition, reflecting changes in blood oxygenation.

CBF and CVR maps

Perfusion was quantified from the preprocessed ASL time-series using FSL’s BASIL with kinetic modeling and M0 acquisition. Partial volume correction was applied via weighted averaging based on tissue contributions from GM, WM, and CSF segmentations of the T1-weighted data [31]. CVR was quantified from the BOLD time-series derived from the second echo of the pCASL data using surround addition. Due to dead space in the tubing, there is an individual-specific delay between gas sampling and the corresponding brain MRI signal, requiring adjustment of end-tidal traces. This lag was estimated via cross-correlation using a custom MATLAB script, selecting the delay where PETCO₂ and the mean GM BOLD signal were maximally correlated [32]. Absolute CVR was calculated as the percent BOLD signal change per mmHg CO₂ from a general linear model (GLM) using the ETCO₂ time course as the regressor.

OEF and CMRO₂ maps

OEF was estimated using the general calibration model (GCM), based on the deoxyhemoglobin dilution model of the BOLD signal [33]. Detailed equations and the full estimation process for OEF are provided in the supplementary material. Following OEF estimation, CMRO₂ was calculated using CBF maps and arterial oxygen content using Fick’s principle, as described in Eq. 3 of the supplementary material.

Registration to MNI space

The pCASL images were registered to MNI space in three separate steps. First, a rigid body registration with nearest neighbor interpolation was conducted using Advanced Normalization Tools (ANTs) to align the pCASL mean images with the low-resolution (1.5-mm) T1-weighted image. Next, the low-resolution T1-weighted images were registered to the high-resolution (0.9 mm) T1 space using a multistage approach that combined rigid, affine, and nonlinear (SyN) transformations with a multi-resolution strategy. In the final step, the high-resolution T1 images were registered to MNI space using the same transformation method applied for the registration from low to high resolution. Finally, the pCASL images were registered to MNI space by applying the three transformation matrices obtained from the previous registration steps, utilizing three warp images.

Statistical analysis

To examine whether the relationship between VO₂peak and imaging biomarkers varied by group, we conducted linear regression analyses in R. Each model included VO₂peak as the primary predictor, with age, sex, education, and group entered as covariates. A VO₂peak × group interaction term was added to assess whether the group moderated the association between VO₂peak and each biomarker. This allowed us to test whether the strength or direction of the VO₂peak–biomarker relationship differed between patients and controls. Because vasoactive medication use is common in CAD and is correlated with disease severity, we performed sensitivity analyses adjusting for beta-blockers, RAAS inhibitors, and calcium channel blockers as covariates. P-values were adjusted for multiple comparisons using a region-of-interest (ROI)–based false discovery rate (FDR) procedure.

For regions in which the interaction term was statistically significant after FDR correction, we conducted post hoc analyses by refitting the model separately within each group. This allowed us to estimate the VO₂peak–biomarker slope in patients and controls independently and identify which group was driving the observed interaction effect.

Results

Cerebrovascular health and VO₂peak

In the full sample, VO₂peak was positively associated with both CBF and CVR, with no significant moderation by group. Higher VO₂peak was linked to increased CBF across the whole GM, including extensive regions of the frontal, parietal, and temporal cortices, as well as subcortical regions (Table 3; Fig. 1(A, B)). Similarly, CVR demonstrated a significant positive association with VO₂peak in global GM and in a more localized set of regions, primarily within the frontal, temporal, and precentral cortices. Detailed regional results are presented in Table 3 and Fig. 1(C).

Table 3.

Brain regions showing significant correlations between VO₂peak and CBF, CVR with age, sex, and group as covariates

	CBF			CVR
	β	P value	Adjusted-R2	β	P value	Adjusted-R2
GMa	0.32	0.02	0.17	0.002	0.04	0.10
L Sup Frontal G	0.35	0.03	0.15	n. s	n. s	n. s
R Sup Frontal G	0.36	0.03	0.13	0.002	0.04	0.15
L Mid Frontal G	n. s	n. s	n. s	0.003	0.04	0.16
L Inf Frontal G	n. s	n. s	n. s	0.003	0.04	0.10
R Inf Frontal G	0.28	0.02	0.12	n. s	n. s	n. s
L Sup Parietal G	0.32	0.03	0.13	n. s	n. s	n. s
R Sup Parietal G	0.42	<0.01	0.19	n. s	n. s	n. s
L precuneus	0.28	0.03	0.12	n. s	n. s	n. s
R precuneus	0.27	0.02	0.14	n. s	n. s	n. s
R Sup Temporal G	0.32	0.01	0.15	n. s	n. s	n. s
L Sup Temporal G	n. s	n. s	n. s	n. s	n. s	n. s
R Mid Temporal G	0.29	0.02	0.12	0.003	0.04	0.09
L Mid Temporal G	0.32	0.02	0.13	n. s	n. s	n. s
R parahippocampus	0.27	0.01	0.13	n. s	n. s	n. s
R Insula	0.31	<0.01	0.19	n. s	n. s	n. s
L Insula	0.24	0.03	0.15	n. s	n. s	n. s
L Cingulate G	0.34	0.01	0.15	n. s	n. s	n. s
R cingulate G	0.32	0.01	0.14	n. s	n. s	n. s
L Caudate	0.38	0.03	0.16	n. s	n. s	n. s
R Caudate	0.44	<0.01	0.18	n. s	n. s	n. s
R Putamen	0.36	<0.01	0.19	n. s	n. s	n. s
R hippocampus	0.29	0.01	0.13	n. s	n. s	n. s
L precentral G	n. s	n. s	n. s	0.003	0.04	0.12
R precentral G	0.32	0.03	0.14	0.003	0.04	0.13
L postcentral G	0.33	0.03	0.13	0.002	0.04	0.10
R postcentral G	0.30	0.03	0.15	n. s	n. s	n. s
L Sup Marginal G	0.32	0.02	0.12	n. s	n. s	n. s
R Sup Marginal G	0.33	0.01	0.15	n. s	n. s	n. s

Bold values indicate statistically significant differences

^aGM, gray matter; L, left; R, right; Sup, superior; Mid, middle; G, gyrus

 Fig. 1.

Associations of VO₂peak with vascular and metabolic brain markers. (A–B) CBF (A) and CVR (B): Left—brain surface projections show β values from linear regressions adjusted for age, sex, and group; only significant regions are displayed. Right—linear regressions of VO₂peak with whole GM CBF/CVR for all participants (green) and by group (yellow, controls; red, CAD). (C) Metabolic markers in GM: Left—VO₂peak vs CMRO₂ showing a significant positive association in CAD (red) but not in controls (yellow) or all participants (green). Right—VO₂peak vs OEF showing a negative association in controls (yellow) and no association in CAD (red) or overall (green, dashed). Shaded areas denote 95% CIs. Together, these panels highlight group-dependent links between cardiorespiratory fitness and brain vascular/metabolic function

Cerebral metabolic health and VO₂peak

Although no significant association between VO₂peak and either OEF or CMRO₂ was observed when analyzing the overall sample, moderation analysis revealed that group significantly moderated both relationships. Specifically, the VO₂peak × group interaction was significant for multiple brain regions in both OEF and CMRO₂ models, indicating that the association between cardiorespiratory fitness and these brain biomarkers differed between patients and healthy controls.

To further explore the nature of these interactions, post hoc linear models were conducted separately within each group. For CMRO₂, a positive association with VO₂peak was observed only in the CAD group, with effects in whole GM and several gray matter regions including the left middle temporal gyrus, left superior temporal gyrus, left inferior temporal gyrus, and the left and right middle orbital gyri, as well as the middle frontal gyri. In contrast, for OEF, a negative association with VO₂peak was observed only in the healthy control group. Significant effects were found in the GM and left middle temporal gyrus, as well as in the superior and inferior temporal gyri, middle orbital gyrus, and middle frontal gyri. Full statistical results for moderation analysis are reported in Table 4 and can be visualized in Fig. 1(C). Post hoc analysis results are reported in Table 5.

Table 4.

Regions showing significant group moderation of VO₂peak–OEF and VO₂peak–CMRO₂ associations (CAD vs. controls)

CMRO2	VO2peak		VO2peak×group (CAD)
	β	P value	β	P value	Adjusted-R2
GMa	−0.05	>0.05	2.23	0.03	0.12
L Middle Temporal G	−0.03	>0.05	2.77	0.03	0.14
L Superior Temporal G	−0.03	>0.05	2.63	0.04	0.11
L Inferior Temporal G	−0.05	>0.05	2.25	0.04	0.18
L Middle Orbital G	−0.04	>0.05	2.75	0.04	0.16
R Middle Orbital G	−0.04	>0.05	1.99	0.04	0.16
R Middle Frontal G	−0.02	>0.05	2.33	0.04	0.10
L Middle Frontal G	−0.03	>0.05	1.98	0.04	0.12
OEF	VO2peak		VO2peak×group (CAD)
	β	P value	β	P value	Adjusted-R2
GM	−0.002	0.03	0.003	0.04	0.11
L Middle Temporal G	−0.003	0.01	0.004	0.02	0.20
L Superior Temporal G	−0.003	0.02	0.004	0.02	0.14
L Inferior Temporal G	−0.001	>0.05	0.003	0.02	0.10
L Middle Orbital G	−0.002	>0.05	0.005	0.02	0.10
R Middle Orbital G	−0.002	>0.05	0.004	0.04	0.10
R Middle Frontal G	−0.003	0.02	0.003	0.04	0.14
L Middle Frontal G	−0.003	0.02	0.003	0.04	0.10
R Superior Frontal G	−0.002	0.03	0.003	>0.05	0.10
L Superior Frontal G	−0.003	0.03	0.002	>0.05	0.09

Bold values indicate statistically significant differences

^aGM, gray matter; L, left; R, right; Sup, superior; Mid, middle; G, gyrus

Table 5.

Post hoc linear regression results of the association between VO₂peak and brain metabolic biomarkers (CMRO₂ and OEF) within each group separately

CMRO2in CAD	VO2peak
	β	P value	Adjusted-R2
GMa	2.37	0.03	0.22
L Middle Temporal G	2.47	0.03	0.21
L Superior Temporal G	2.40	0.04	0.16
L Inferior Temporal G	2.16	0.03	0.23
L Middle Orbital G	2.35	0.04	0.14
R Middle Orbital G	2.14	0.04	0.13
R Middle Frontal G	1.87	0.03	0.19
L Middle Frontal G	1.75	0.04	0.12
OEF in HC	VO2peak
	β	P value	Adjusted-R2
GM	−0.003	<0.01	0.29
L Middle Temporal G	−0.003	<0.01	0.28
L Superior Temporal G	−0.003	<0.01	0.25
L Inferior Temporal G	−0.001	0.04	0.09
L Middle Orbital G	−0.003	0.03	0.15
R Middle Orbital G	−0.003	<0.01	0.16
R Middle Frontal G	−0.004	0.03	0.27
L Middle Frontal G	−0.003	0.03	0.15

^aGM, gray matter; L, left; R, right; Sup, superior; Mid, middle; G, gyrus

Sensitivity analysis

After additional adjustment for vasoactive medication use (beta-blockers, RAAS inhibitors, and calcium channel blockers), VO₂peak remained positively associated with CMRO₂, but the effects were attenuated (temporal GM: puncorr < 0.05, pFDR = 0.06; whole GM: puncorr < 0.05, pFDR = 0.08) and did not survive FDR correction. For CBF, inclusion of medication covariates did not materially alter the results; the VO₂peak–CBF associations remained unchanged, likely reflecting the larger sample size available for resting ASL analyses. For CVR, the associations with VO₂peak remained statistically significant after FDR correction in gray matter and frontal regions, including whole gray matter (estimate = 0.0088, pFDR = 0.044), left superior frontal gyrus (estimate = 0.006, pFDR = 0.043), right superior frontal gyrus (estimate = 0.005, pFDR = 0.043), and left middle frontal gyrus (estimate = 0.007, pFDR = 0.025). For OEF, the VO₂peak associations were attenuated and did not survive FDR correction after medication adjustment.

Discussion

In this study, we examined the associations between VO₂peak, a measure of cardiorespiratory fitness, and critical cerebral vascular and metabolic biomarkers in healthy controls and individuals with CAD. Our analysis revealed significant positive relationships between VO₂peak and both CBF and CVR, across the entire study cohort. These findings emphasize the beneficial effects of higher cardiorespiratory fitness on cerebrovascular health across the cardiovascular health spectrum. In contrast, no significant associations were observed between VO₂peak and either CMRO₂ or OEF in analyses combining both groups. However, moderation analyses revealed distinct associations in CAD patients compared to controls. We found VO₂peak to be associated with greater CMRO₂ in CAD participants within gray matter regions of the temporal lobes and orbital frontal cortex. This suggests that enhanced fitness may be beneficial for metabolic oxygen utilization in these brain regions, especially in individuals with CAD. Conversely, fitness was found to be associated with lower OEF exclusively in healthy controls within temporal and frontal gray matter regions. This result implies that in healthy individuals, the effects of cardiorespiratory fitness are predominantly reflected in higher CBF, with an unchanged CMRO₂, leading to lower OEF. This results in greater oxygen availability at rest. In contrast, in patients, concurrent improvements in CMRO₂ and CBF cancel out to result in unchanged OEF.

We observed widespread positive associations between VO₂peak and CBF in frontal, parietal, temporal, and key subcortical regions (insula, hippocampus, caudate, putamen). Also, medication-adjusted sensitivity analyses yielded virtually unchanged VO₂peak–CBF associations, suggesting that these results are relatively robust to vasoactive medication use despite the modest sample size. This pattern echoes previous findings in CAD cohorts [8] and in several studies in healthy participants showing higher CBF in gray matter and the hippocampus with greater fitness [13, 34, 35]. It is interesting to note, however, that these findings may reflect the physiological processes at the normal or lower range of fitness, since some studies in very highly trained adults report null or negative relationships [14, 16, 36, 37]. These discrepancies in highly fit adults are potentially due to a ceiling effect or greater oxygen extraction efficiency, which may reduce the need for high resting perfusion [16, 36, 38]. In our sample, the positive association may reflect fitness-related vascular adaptations, including increased capillary density shown in animal studies [39, 40] and enhanced perfusion in metabolically active regions.

Our CVR findings add important nuance to the existing literature by demonstrating a positive relationship between VO₂peak normalized by lean body mass (LBM) and CVR across both healthy individuals and patients with CAD. Specifically, significant associations were identified in key cortical regions such as the superior and middle frontal gyri, parietal lobes, middle temporal gyrus, precentral gyrus, and supramarginal gyrus—areas particularly susceptible to aging-related changes and known targets for exercise-induced neuroplasticity [41, 42]. The VO₂peak associations in gray matter and frontal regions remained significant after medication adjustment, indicating that the main CVR findings are robust even when accounting for vasoactive drugs. This positive relationship aligns with prior findings showing aerobic exercise’s beneficial impacts on cerebrovascular health, likely reflecting vascular adaptations such as increased arterial compliance and enhanced endothelial function [43–45]. Regular aerobic activity is known to induce structural remodeling and biochemical improvements, including upregulated endothelial nitric oxide production and reduced oxidative stress [46, 47], likely collectively enhancing vasodilatory capacity and CVR.

Nevertheless, our results should be interpreted within the context of varied literature findings. Many of the previously documented positive relationships between CVR and VO₂peak were obtained by measuring CVR with trans-cranial Doppler imaging [48–51]. MRI-based studies investigating the relationship between cardiorespiratory fitness and CVR have produced inconsistent findings, and our study is among the first to demonstrate a positive association. Previous studies using MRI have reported negative [14, 18, 52], quadratic [17], and null relationships [53]. Most notably, Intzandt and colleagues (2020), whose methodology closely resembles ours (using hypercapnic BOLD-CVR and VO₂peak), reported negative associations between fitness and both CVR and CBF in healthy older adults. The critical difference between their study and ours is the study population; whereas Intzandt included only healthy older individuals, our study comprised a mixed group of CAD patients and controls. The positive association we observed may therefore reflect a particularly beneficial impact of cardiorespiratory fitness on cerebrovascular function within CAD patients—a notion consistent with previous evidence showing improved cerebral perfusion following cardiac rehabilitation in similar populations [3]. Additionally, differences in participant demographics, particularly sex composition, could contribute to this discrepancy. Our sample included a smaller proportion of females (22%), while Intzandt et al. included 66% females, which is important given prior evidence of sex-specific cerebrovascular responses to physical activity [18, 54]. Other studies such as DuBose et al. [17] further emphasize the complexity of the CVR–fitness relationship. DuBose et al. proposed an inverted-U (quadratic) relationship, suggesting optimal CVR at moderate fitness levels. Collectively, these varied outcomes highlight the nuanced, population-dependent nature of the fitness–CVR association, emphasizing the need to consider population characteristics and methodological factors when interpreting findings across studies.

Notably, although no significant relationship between VO₂peak and CMRO₂ was found in the overall study population, we observed a group interaction, showing that group moderates the relationship between fitness and oxygen usage. More specifically, VO₂peak was positively related to CMRO₂ within the CAD group across gray matter, particularly in frontal and temporal regions—areas known to be susceptible to the effects of aging and vascular disease [41, 55]. This association represents a novel finding exclusive to patients with CAD. Although medication-adjusted sensitivity analyses did not survive FDR correction, the association remained directionally consistent and near-significant in temporal regions (pFDR≈0.06), likely due to limited power for interaction analyses in our small sample. Our results contrast with a recent study [22] which reported a negative relationship between global CMRO₂ and VO₂peak in healthy older adults. This discrepancy could be due to differences in participant demographics, especially given that the same group of individuals with CAD were previously shown to have CMRO₂ deficits [5]. Alternatively, some of these discrepancies could be due to different methodological approaches. Gay et al. measured global CMRO₂ using TRUST, a T2-based method. Studies using TRUST have typically shown increased CMRO₂ with aging [56, 57], in contrast to the decrease documented with the T2*-based approaches such as the one used here [58]. Because all other parameters tested are improved with fitness in our healthy sample, our findings suggest that CMRO₂ may be at a ceiling in healthy older adults and only when CMRO₂ is affected by disease can it be improved through higher cardiorespiratory fitness. In CAD, the positive effect of VO₂peak on CMRO₂ might indicate improved mitochondrial function, which plays a central role in oxygen metabolism and is known to be compromised at least outside the brain in CAD [59]. However, given the cross-sectional design, reverse causation and residual confounding by disease severity are also possible as individuals with better preserved cerebral metabolism (higher CMRO₂) may be less affected by CAD and therefore exhibit higher VO₂peak.

To our knowledge, this is the first study to investigate the effect of VO₂peak on OEF and CMRO₂, in a population that includes both healthy individuals and patients with CAD. The negative association between OEF and VO₂peak, observed in healthy participants across multiple brain regions including the whole gray matter, suggests a compensatory mechanism in response to lower cardiorespiratory fitness. In healthy individuals with lower VO₂peak, the brain compensates for reduced oxygen delivery by increasing OEF, extracting more oxygen from the blood to meet its metabolic demands and maintain CMRO₂. This indicates that cerebral oxygen metabolism remains stable despite reduced fitness levels due to the brain’s ability to adjust oxygen extraction. In contrast, the absence of a significant relationship between OEF and VO₂peak in patients implies that this compensatory mechanism may be exhausted, having already led to metabolic impairment either through cell loss or mitochondrial dysfunction. This is consistent with our previous results [5] showing lower CMRO₂ in individuals with CAD.

The negative relationship between OEF and VO₂peak in healthy individuals suggests enhanced oxygen extraction to compensate for lower cardiorespiratory fitness. Medication-adjusted sensitivity analyses attenuated the VO₂peak associations such that they no longer survived FDR correction, and these effects should therefore be interpreted cautiously in light of limited power. The absence of this relationship in CAD patients indicates compromised compensatory capacity, highlighting the importance of fitness interventions to maintain cerebral metabolic integrity. Importantly, CAD likely reflects a clinical manifestation of systemic atherosclerosis and vascular aging rather than an isolated cardiac condition. Thus, the group differences and VO₂peak–brain relationships observed here likely reflect a spectrum of systemic vascular health. Because CAD typically co-occurs with cardiometabolic risk factors, our CAD group had a substantial comorbidity burden (e.g., hypertension, dyslipidemia, and obesity), while controls were stringently screened to be low risk. Accordingly, some CAD group-specific effects may reflect cumulative lifetime vascular risk burden rather than coronary pathology alone. These findings suggest a continuum of adaptive responses, where metabolic flexibility diminishes with disease progression, but can be rescued by higher cardiorespiratory fitness.

Limitations

The imaging technique used, arterial spin labeling (ASL), is limited by limited temporal signal-to-noise ratio (tSNR), which can be further reduced in individuals with vascular risk due to prolonged arterial transit times. The use of a fixed post-labeling delay (PLD) may therefore have underestimated CBF in participants with delayed blood transit. In this context, the PLD used here is shorter than the one recommended by Alsop et al. [60], which could bias the results toward greater macrovascular contribution, though our data quality check using visual inspection and CoV indicates that this is unlikely to be a main driver of our results. Future studies should employ multiple PLDs to improve accuracy. Signal quality could also be enhanced with background suppression techniques, though these were not applied here to preserve the BOLD signal for CVR, OEF, and CMRO₂ estimation.

Another limitation is the limited characterization of habitual physical activity and exercise behavior. Without detailed and objective measures of daily activity patterns and training history, we cannot determine whether specific exercise habits independently relate to cerebrovascular or metabolic function, or whether they modify the observed associations between VO₂peak and brain measures. Future studies should incorporate objective activity monitoring (e.g., accelerometry) and detailed exercise phenotyping (frequency, intensity, modality, and long-term training exposure) to disentangle the respective contributions of cardiorespiratory fitness and habitual physical activity to cerebral vascular and metabolic health.

Further limitations include the homogeneity of our single-site sample, which was predominantly male with insufficient female representation to examine sex-specific effects, excluded severe CAD, and lacked diversity in socioeconomic status, education, and ethnicity—factors that restrict generalizability. Although cognitive outcomes were not assessed in the present study, this limits our ability to directly link fitness-related vascular and metabolic differences to functional brain consequences. Future longitudinal and interventional studies should incorporate detailed cognitive testing to determine whether improvements in cerebrovascular and metabolic function translate into preserved or enhanced cognition in older adults with CAD. The absence of direct measures of cardiac function also limits interpretation of heart–brain mechanisms; incorporating such assessments would provide clearer insight. Finally, the cross-sectional design prevents causal inference. Intervention studies, particularly exercise-based trials in CAD patients, are needed to test whether fitness can improve cerebrovascular and metabolic markers.

Conclusion

This study provides novel evidence linking VO₂peak with vascular and metabolic markers of brain health in older adults with and without CAD. VO₂peak showed widespread positive associations with CBF and CVR, indicating aerobic fitness supports vascular function regardless of cardiovascular pathology. Metabolic responses, however, differed: CAD patients exhibited a positive association between VO₂peak and CMRO₂ in regions vulnerable to aging and disease, suggesting fitness helps preserve oxidative metabolism through improved mitochondrial function, neurogenesis, or neurotrophic mechanisms. In contrast, healthy participants showed elevated OEF with lower fitness, likely compensating to maintain CMRO₂ despite reduced CBF. These findings highlight VO₂peak as a modifiable factor for vascular and metabolic resilience in the brain and support aerobic interventions to enhance both cardiovascular and brain health, particularly in CAD patients at risk of cognitive decline and stroke.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

Study design and conceptualization: CJG, LB. Data acquisition: SS, SAT, ZPJ, AR, DS, CG, BI, AMB, LW. Imaging data processing and analysis: SS. Statistical analysis: SS. Manuscript preparation: All authors.

Funding

This work was supported by funding awarded to Claudine J. Gauthier from the Natural Sciences and Engineering Research Council of Canada (NSERC Discovery Grant: RGPIN-2015-04665; 2024-06455), Fonds de recherche du Québec (FRQ 5232), the Heart and Stroke Foundation of Canada (G-17-0018336), the Heart and Stroke Foundation New Investigator Award, the Henry J.M. Barnett Scholarship, the Michal and Renata Hornstein Chair in Cardiovascular Imaging and the Mirella and Lino Saputo research chair in cardiovascular health and the prevention of cognitive decline. Additional support was provided by the Canadian Institutes of Health Research (FRN: 175862, to Stefanie A. Tremblay), the Vascular Training Platform (VAST) (to Ali Rezaei), the Heart and Stroke Foundation of Canada and Brain Canada (to Zacharie Potvin-Jutras), and the Alzheimer Society Research Award (to Brittany Intzandt.).

Data Availability

Anonymized and defaced data that support the findings of this study will be made openly available on OpenNeuro (https://openneuro.org) within 6 months of publication. A persistent DOI will be provided upon release.

Declarations

Ethics approval and consent to participate

This study was approved by the Comité d’éthique de la recherche et du développement des nouvelles technologies (CÉRDNT) of the Montreal Heart Institute and was conducted in accordance with the Declaration of Helsinki. All participants were screened for eligibility and provided written informed consent prior to participation. Human data collection and MRI procedures were performed at the Montreal Heart Institute. No animal procedures were involved in this research.

Consent for publication

This manuscript has been reviewed and approved by all authors prior to submission. All participants provided written informed consent, which included consent for publication of anonymized group-level results.

Competing interests

The authors report no competing interests.