Cardiorespiratory fitness is differentially associated with motor cortex laterality in middle-aged and older adults

Abstract

Tasks associated with unilateral patterns of functional magnetic resonance imaging (fMRI) activation often demonstrate bilateral activation with aging (hemispheric asymmetry reduction). We examined relationships between the modifiable lifestyle variable cardiorespiratory fitness (CRF), hemispheric asymmetry reduction, and visuomotor task performance in middle-aged and older adults. Sixty-four participants aged 35–86 years completed progressive, maximal cardiopulmonary exercise testing to assess VO₂peak and a standardized test of motor coordination, the Grooved Pegboard Test. fMRI was acquired during a visuomotor task requiring a right-hand motor response. The relationships between hemispheric asymmetry during the fMRI task, CRF, and performance on simple (fMRI task) and complex (Grooved Pegboard Test) motor tasks were examined. Age moderated the relationship between CRF (VO₂peak) and hemispheric asymmetry. Among middle-aged adults, greater VO₂peak was associated with more hemispheric asymmetry; no association was observed in older adults. Age marginally moderated the relationship between hemispheric asymmetry and Grooved Pegboard performance. Among middle-aged adults, greater hemispheric asymmetry was marginally associated with better performance; among older adults, reduced asymmetry showed a trending association with better performance. These findings highlight age-related differences in the relationship between CRF, behavioral performance, and fMRI activation and emphasize the importance of investigating brain function, cognition, and age across the adult lifespan.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-43466-z.

Introduction

Aging is associated with cognitive decline and concomitant brain changes^1,2. Activity during task-related functional magnetic resonance imaging (fMRI) often changes with age, such that activation patterns identified in younger adults may shift in midlife and older adulthood^3,4. There is evidence that tasks associated with unilateral patterns of fMRI activation in younger adults often shift to a more bilateral pattern in older adulthood³. For example, young adults pressing a button with the right hand is typically associated with fMRI activation in the contralateral (left) hemisphere motor cortex⁵ and fMRI deactivation in the ipsilateral (right) hemisphere^6,7. However, an older adult pressing a button with their right hand might show more bilateral (rather than unilateral) fMRI activation in the motor cortex³. This increase in bilateral activity in aging, or reduction in hemispheric asymmetry, has been observed in multiple task domains, including episodic memory encoding and retrieval^8–10, working memory^11–13, and inhibition¹⁴, and has been referred to as Hemispheric Asymmetry Reduction in Older Adulthood (HAROLD)³.

Although HAROLD has been identified among older adults, expression of this pattern may vary substantially between individuals due to several factors, including modifiable lifestyle variables. Cardiorespiratory fitness, for example, is a known source of variability in aging and positively associated with cognitive performance and brain structure in aging^15–17 and associated with reduced risk of Alzheimer’s disease^18,19. Moreover, there is evidence that older adults with higher cardiorespiratory fitness show patterns of activation more similar to young adults compared to lower-fit older adults, including greater activation in medial temporal lobe regions during associative memory tasks²⁰ and greater within-network connectivity during resting state^21,22. However, it remains unclear whether higher fitness levels are associated with broader age-related changes in neural organization, such as hemispheric asymmetry.

To our knowledge, one study has examined the relationship between cardiorespiratory fitness and patterns of hemispheric asymmetry²³(cf., McGregor et al. (2011)²⁴, who examined patterns of hemispheric asymmetry in active v. sedentary older adults compared to younger adults). McGregor et al. (2013)²³ asked younger (19–32 years) and middle-aged (40–60 years) adults to squeeze a button between their right-hand index finger and thumb in response to periodic visual prompt. Comparable to other unilateral motor tasks and in line with HAROLD, this task is known to elicit no activation in the right (ipsilateral) motor cortex in young adults and more positive activation with increasing age^25,26. In middle-aged adults, greater cardiorespiratory fitness was associated with less ipsilateral motor activity during the task, suggesting more hemispheric asymmetry. No relationship between cardiorespiratory fitness and ipsilateral motor activation in younger adults was observed²³. Further, in middle-aged adults, less ipsilateral motor activity was associated with faster performance; this relationship was again not observed in younger adults. Notably, older adults were not included in the McGregor et al. (2013) study. As both hemispheric asymmetry and motor performance continue to decrease beyond middle adulthood^3,27,28, exploring the relationship between cardiorespiratory fitness and hemispheric asymmetry in middle-aged and older adults may provide insight into the mechanisms underlying brain and cognitive change in aging.

The current study aimed to extend the literature by exploring the relationship between cardiorespiratory fitness and hemispheric asymmetry in middle-aged and older adults. fMRI was collected during completion of a visuomotor task that required a fractionated right-handed motor response. Outside of the scanner, participants completed a comparatively more complex standardized visuomotor task, the Grooved Pegboard Test²⁹, which requires fine distal motor dexterity and a higher level of visual-motor integration. Participants also completed a progressive maximal cardiopulmonary exercise test on a cycle ergometer to assess cardiorespiratory fitness (peak rate of oxygen consumption (VO₂peak)). Task performance was assessed during both the simple unilateral button-pressing task performed during functional MRI and the more complex Grooved Pegboard Task to elucidate the relationships between a modifiable lifestyle variable (cardiorespiratory fitness), fMRI activation, and behavioral performance. We hypothesized that cardiorespiratory fitness would be associated with greater hemispheric asymmetry (a more “youthful” pattern of brain activation). Furthermore, we hypothesized that greater hemispheric asymmetry would be associated with better task performance and that these relationships would vary with age.

Methods

Participants

The present sample included 64 participants (35–86 years; mean age (SD) = 61.6 (11.6) years; n = 36 (56%) females; Table 1) from the Fitness, Aging, Stress, and TBI Exposure Repository (FASTER; S. Hayes & J. Hayes). FASTER is an ongoing data collection initiative that aims to deeply phenotype individuals across the adult human lifespan. Participant exclusion criteria include a history of seizures, stroke, or other major cerebrovascular accident; myocardial infarction; cancer treatment using radiation or chemotherapy; type I diabetes; HIV/AIDs; current DSM diagnosis of bipolar disorder or other psychotic disorders; formal diagnosis of a learning disability; and current use of illicit drugs or regular excessive alcohol consumption. Participation also requires native English speaking or formal instruction in an English classroom beginning before third grade; and having a primary care physician. All experimental procedures were approved by The Ohio State University Institutional Review Board and were performed in accordance with the Declaration of Helsinki and national and institutional ethical guidelines. Written informed consent was obtained from all participants, and all participants received financial compensation for participation.

Table 1.

Participant Demographics by Age Group.

	Middle-Aged Adults	Older Adults	Overall	p
	n = 37	n = 27	n = 64
Age (years)	54.1 (8.9)	71.9 (4.9)	61.6 (11.6)	< 0.001*
Sex
Male	n = 14 (37.8%)	n = 14 (51.9%)	n = 28 (43.8%)	0.389
Female	n = 23 (62.2%)	n = 13 (48.1%)	n = 36 (56.3%)
Mild TBI	n = 12 (32.4%)	n = 11 (40.7%)	n = 23 (35.5%)	0.674
Race (White)	n = 33 (89.2%)	n = 27 (100%)	n = 60 (93.8%)	0.214
Education (years)	17.2 (2.0)	17.1 (2.3)	17.1 (2.1)	0.87
VO2peak (ml/kg/min)	26.5 (7.6)	21.8 (7.6)	24.5 (7.9)	0.017*
ACSM VO2peak Percentile	73.5 (27.0)	73.3 (24.2)	73.4 (25.6)	0.978
fMRI Task Accuracy	94.6% (7%)	90.0% (10.5%)	92.7% (8.8%)	0.039*

Note: ACSM VO₂peak Percentile is derived from age-adjusted normative ranges provided by the American College of Sports Medicine^82,83. The middle-aged group includes participants aged 35–64 years and the older adult group includes participants aged 65 years and older.

At the time of analysis, 177 FASTER participants had available age, VO₂peak, and task fMRI data. The present study identified and removed participants with poor quality task fMRI or VO₂peak data (see Data Collection and Preprocessing; n = 23); a history of moderate or severe traumatic brain injury per the Boston Assessment of Traumatic Brain Injury, Revised (BATL-R; n = 9); persistent depressive disorder at time of testing per the computerized version of the Structured Clinical Interview for the DSM-5 (NetSCID-5; n = 6); or met criteria for mild cognitive impairment (n = 6). Mild cognitive impairment was defined as (a) scoring 1.5 standard deviations or more below the mean of same-age and education peers on two or more tests within the same cognitive domain (episodic memory, executive function, processing speed, or language) or (b) scoring 1.5 standard deviations or more below the mean on at least one test within each of the four domains³⁰.

Analyses used to identify age-associated patterns in task fMRI were performed with participants meeting the above criteria aged 18 years and older (n = 133). However, further analyses additionally excluded those without complete demographic data and task data (n = 24), those with invalid performance on the visuomotor fMRI task (< 50% compliance with button pressing, n = 7), and outliers (2 standard deviations above or below the mean, n = 7). In the present protocol, off-task motor movement such as finger, hand, or arm motion during rest periods was not directly assessed (e.g., via video or accelerometer). Given the brief task duration and the importance of high-quality motion-related BOLD contrasts, we instead accounted for potential off-task motor movement during the fMRI task by excluding participants if average beta coefficients within the left or right motor cortex ROIs (see Analysis – Visuomotor Task and Asymmetry Index) were less than zero (indicating greater motor BOLD response during rest than during task, n = 23). Finally, these analyses were also performed using only middle-aged and older adults (35 years and older). Young adults were not included in these analyses due to the small number of young adult participants within the sample at the time the analysis was completed (n = 8). Participants were included in the analyses regardless of self-reported handedness. However, Supplemental Table 2 reports results with left-handed participants (n = 4) excluded.

Table 2.

Hierarchical regression assessing relationships between VO₂peak, asymmetry index, and task performance.

Predictor variables	Step 1	Step 2	Step 3	Step 4
Model 1: Age X VO2peak on Asymmetry
Education (years)	−0.004	−0.082	−0.073	−0.048
Sex (male)	0.008	−0.095	−0.009	−0.089
Age (years)		−0.340*	−0.289+	−0.266+
VO2peak (ml/kg/min)			0.107	0.075
Age X VO2peak				−0.293*
Simple Slopes: Middle-Aged Adults				0.537*
Simple Slopes: Older Adults				−0.267
R2	0.000	0.107	0.115	0.192
ΔR2		0.107*	0.008	0.077*
Model 2: Age X Asymmetry on fMRI Task Performance
Education (years)	0.051	−0.024	−0.007	−0.005
Sex (male)	0.032	0.068	−0.048	−0.052
Age (years)		−0.328*	−0.258+	−0.287*
Asymmetry Index			0.207	0.205
Age X Asymmetry Index				0.102
R2	0.003	0.102	0.141	0.148
ΔR2		0.099*	0.039	0.007
Model 3: Age X Asymmetry on Grooved Pegboard Speed
Education (years)	−0.093	0.019	0.024	0.031
Sex (male)	−0.618*	−0.469*	−0.463*	−0.473*
Age (years)		0.489*	0.510*	0.442*
Asymmetry Index			0.064	0.062
Age X Asymmetry Index				0.238+
Simple Slopes: Middle-Aged Adults				−0.282+
Simple Slopes: Older Adults				0.278+
R2	0.103	0.324	0.328	0.369
ΔR2		0.221*	0.004	0.041+
Model 4: Age X VO2peak on Grooved Pegboard Speed
Education (years)	−0.093	0.019	−0.009	0.006
Sex (male)	−0.618*	−0.469*	−0.735*	−0.784*
Age (years)		0.489*	0.330*	0.335*
VO2peak (ml/kg/min)			−0.329*	−0.348*
Age X VO2peak				−0.181+
Simple Slopes: Middle-Aged Adults				−0.304*
Simple Slopes: Older Adults				−0.524*
R2	0.103	0.324	0.397	0.426
ΔR2		0.221*	0.073*	0.029+

Note: Columns report beta coefficients for predictor variables at each step of the hierarchical model. In post-hoc simple slopes analyses, the middle-aged adult group includes individuals aged 35–64 years and the older adult group includes individuals aged 65 years and older. *significant at p < 0.05; ⁺trending at p < 0.1.

Data collection and preprocessing

Cardiopulmonary exercise testing

Progressive maximal cardiopulmonary exercise testing was performed using an Excalibur Sport cycle ergometer. Participants were asked to continue cycling to failure, with two-minute stages of increasing resistance and a pedal speed of 60 revolutions per minute. Peak rate of oxygen consumption (VO₂peak, ml/kg/min) and respiratory exchange ratio (RER) were measured across 30-second time windows. Heart rate, blood pressure, and ratings of perceived exertion (RPE) using the 20-item Borg scale³¹ were assessed every two minutes, at the time of each exercise intensity increase. VO₂peak was considered valid if at least two of the following criteria were met: (a) RER > 1.0; (b) RPE > = 17, which corresponds to an exertion level of ‘very hard’; (c) plateau in recorded VO₂, defined as no increase in VO₂ while workload continued to increase. This was operationalized as VO₂ increasing by less than 2.1 ml/kg/min across a minute span at peak workload³².

Magnetic resonance image acquisition and preprocessing

All images were collected using a Siemens Prisma 3 T scanner (Siemens Healthineers; Erlangen, Germany) with a 32-channel head coil and sequences adapted from the Human Connectome Project-Aging protocol³³. Multiecho T₁ MPRAGE (repetition time (TR) = 2500 ms; echo time (TE) = 1.8, 3.6, 5.4, 7.2 ms; inversion time (TI) = 1010 ms; spatial resolution = 0.8 × 0.8 × 0.8 mm), and T₂-weighted images (TR = 3200 ms; TE = 564 ms; spatial resolution = 0.8 × 0.8 × 0.8 mm) were collected and used for brain segmentation and spatial normalization. These images were preprocessed and segmented using FreeSurfer (version 7.2.0; http://surfer.nmr.mgh.harvard.edu/) recon-all.

fMRI was collected with an echo planar imaging sequence (TR = 800 ms; TE = 28 ms; spatial resolution = 3 mm x 3 mm x 3 mm; FOV = 216 mm; matrix = 72²; multiband = 4) while participants performed a visuomotor checkerboard task from the Human Connectome Project – Aging (VISMOTOR; for details, see Harms et al., 2018³³ and Bookheimer et al., 2019³⁴). Briefly, participants viewed blocks of a checkerboard flickering at 8 Hz interspersed with blocks of rest; each block lasted 30 s. There were four rest blocks and three checkerboard blocks. During checkerboard blocks, sections on the left or right side of the checkerboard turned red every 3 s, remaining red for 0.5 s. Participants indicated via right hand button press whether the red section appeared on the left (index finger) or right (middle finger) side (Fig. 1). EPI images were considered good quality if (a) median framewise displacement < 0.4 mm and (b) accuracy on the fMRI task > 50%.

Fig. 1.

fMRI was collected while participants performed a visuomotor task adapted from the Human Connectome Project – Aging³³. The task was designed with 30-second fixation and task blocks (total task time = 3.5 min). During task blocks, participants were asked to fixate on a center crosshair while presented with a checkerboard flickering at 8 Hz. Periodically, red boxes appeared on the left or right side of the checkerboard. Participants were asked to press a button with their right hand to indicate on which side of the checkerboard the boxes appeared.

MR images were stored, processed, and analyzed using the Owens and Cardinal High Performance Computing clusters from The Ohio Supercomputer Center, which provides High Performance Computing resources and expertise to academic researchers across the State of Ohio. Owens and Cardinal are Dell-built, Intel^® Xeon^® processor-based computing clusters, each with more than 20,000 cores and more than 350 nodes.

Task images were preprocessed using elements from the FMRIB Software Library (FSL; version 6.0.7; https://fsl.fmrib.ox.ac.uk/fsl/docs/#/)^35,36, ICA-AROMA³⁷, and Advanced Normalization Tools (ANTs; https://stnava.github.io/ANTs/). Images were motion and distortion-corrected using FSL MCFLIRT and topup. FSL feat was used to smooth images with a 5 mm FWHM kernel. ANTs antsRegistrationSyN.sh was used to train co-registration of unsmoothed functional images directly to EPI templates in MNI-152 standard space. These EPI templates were generated by averaging EPI images from the present study that were well-registered to MNI-152 space using FSL’s two-step registration process (EPI to high-resolution T1-MPRAGE, then T1-MPRAGE to MNI-152 standard space) into a single, average EPI image. Separate templates were created for each of three age groups – young adults (18–34 years old), middle-aged adults (35–64 years old), and older adults (65 years and older) – to optimize registration at older ages where anatomy may have changed³⁸. As ANTs registration files are formatted differently from those output by FSL, c3d_affine_tool (version 1.1.0; http://www.itksnap.org/pmwiki/pmwiki.php) and Connectome Workbench’s wb_command -convert-warp (version 1.3.2; https://www.humanconnectome.org/software/connectome-workbench)³⁹ were used to convert the registration files into a format usable by FSL and ICA-AROMA. Next, artifacts due to motion were removed from functional images using ICA-AROMA. White matter and CSF masks from FreeSurfer were converted into binary masks using FreeSurfer mri_binarize; these masks were eroded using a 2 mm x 2 mm x 2 mm kernel using FSL fslmaths; and average timeseries from each mask were extracted from the preprocessed functional images. These timeseries were used as regressors during first-level analysis to remove signal artifacts.

Task performance

Performance during the fMRI visuomotor task was first quantified using the simple, unilateral button-pressing task performed in the scanner. To account for potential response time and accuracy trade-offs, motor performance was calculated as (inverse efficiency score^40,41).

Additionally, motor performance was evaluated with an independent test of complex motor coordination, the Grooved Pegboard Test²⁹. During this task, participants were asked to place pegs into the holes of a pegboard with a single hand as fast as possible. This task was administered at a separate appointment typically occurring 1–2 weeks prior to MRI scanning. To most closely align with the task performed in the fMRI scanner, only right-hand task performance was analyzed, even for left-handed participants (n = 4). Performance was assessed as total time to place all pegs.

Analysis

Visuomotor task and asymmetry index

Voxel-based analyses were performed using FSL’s feat. The task was modeled as a block design and convolved with the double gamma hemodynamic response function. BOLD signal was prewhitened to remove temporal autocorrelation. The model additionally covaried for white matter signal, CSF signal, 16 motion parameters (output by MCFLIRT during preprocessing; see Magnetic Image Acquisition and Processing), and accounted for temporal derivatives. Subject-level modeling was performed in native space. Prior to group-level analysis, subject-level outputs were co-registered to MNI-152 standard space using co-registration parameters trained by ANTS (see Magnetic Image Acquisition and Processing). To characterize data, we examined both Task > Rest (Fig. 2B, Supplemental Fig. 1 A) and Rest > Task contrasts (Supplemental Fig. 1B); however, only Task > Rest contrasts were used for group-level analysis. Group-level analysis, modeled using a mixed-effects model (FLAME1), included chronological age as a between-subjects regressor to identify voxels with BOLD activation that changed with age (Fig. 3A, Supplemental Fig. 2).

Fig. 2.

(a) Percent change in BOLD signal within the contralateral motor ROI during motor task blocks averaged across all participants, with shaded regions indicating one standard deviation around the average BOLD change. The horizontal green line indicates the task period. The small oscillations during the task period occur concurrently with button presses. (b) Mean activation for the Task > Rest contrast for all participants included in fMRI analyses (see Methods – Participants; threshold z > 7, p < 0.00001). See Supplemental Fig. 1B for negative BOLD activation.

Fig. 3.

(a) Group-level analyses were used to identify voxels showing increased ipsilateral (right) motor cortex fMRI activation with age during a visuomotor task that required a right-handed motor response (threshold z < 3.1, p > 0.001; only positive BOLD activation visualized; see Supplemental Fig. 2B for negative BOLD response). (b) This cluster was binarized and mirrored to the left hemisphere to create left and right hemisphere motor cortex masks, from which average beta coefficients were extracted from each mask for each participant. (c) The computed asymmetry index (calculated as Asymmetry Index = (R – L)/(R + L)) were significantly negatively associated with age (p = 0.011).

A cluster of age-related changes in fMRI activation in the motor cortex was identified using a cluster-forming threshold of p < 0.001 (z < 3.1). A region of interest was created from this cluster and mirrored to the contralateral hemisphere using fslmaths. Average beta values from each cluster were extracted from subject-level contrast of parameter estimate (COPE) maps using fslmeants. This approach was selected to target voxels in the ipsilateral motor cortex showing age-related change in fMRI activation and a contralateral ROI of equivalent extent. The majority of the contralateral ROI overlapped with voxels of peak activation in group average Task > Rest activation maps (z > 7, p < 0.00001; Supplemental Fig. 1A).

Degree of asymmetry was calculated using an asymmetry index: . The linear scaling factor was included in the calculation to assist with visualization and interpretation. With this asymmetry index, values closer to zero indicate less asymmetry, whereas greater values indicate greater asymmetry.

Hierarchical regression

Four hierarchical regression models were conducted in R (version 4.3.3). Model 1 assessed (a) the main effect of age on asymmetry index, (b) the main effect of VO₂peak on asymmetry index, and (c) the moderating effect of age on the relationship between VO₂peak and asymmetry index. Models 2 and 3 assessed (a) the main effect of age on task performance, (b) the main effect of task performance on asymmetry index, and (c) the moderating effect of age on the relationship between task performance and asymmetry index. Model 2 modeled task performance using the fMRI visuomotor task and Model 3 modeled task performance using the Grooved Pegboard Test. Finally, a post-hoc hierarchical regression was performed to assess (a) the main effect of age on the Grooved Pegboard Test, (b) the main effect of VO₂peak on the Grooved Pegboard Test, and (c) the moderating effect of age on the relationship between VO₂peak and Grooved Pegboard Test performance (Model 4).

Step 1 of all models included education and sex as covariates. Step 2 added age (years). Step 3 added VO₂peak in Models 1 and 4 and asymmetry index in Models 2 and 3; Step 4 added the interaction effect. Table 2 details each model and the variables entered in each step. Post-hoc simple slopes analyses on the moderating effect of categorical age groups were used to probe significant interaction effects. For analyses with categorical age groups, the middle-aged group included individuals aged 35–64 years (n = 37, mean (SD) age = 54.1 (8.9) years) and the older group included individuals 65 years and older (n = 27, mean (SD) age = 71.9 (4.9) years). Statistical significance was set at p < 0.05. In addition, p-values between 0.05 and 0.10 were described as trends or tendencies, following conventions in applied sports science where borderline findings may still provide relevant insights^42,43.

Results

Asymmetry index

Voxel-based analyses identified a cluster in the ipsilateral (right) motor cortex for which fMRI activation (Task > Rest) increased linearly with age (Fig. 3A). Beta values were extracted from this cluster and its mirrored cluster in the contralateral (left) hemisphere (Fig. 3B), and the asymmetry index was calculated (mean (SD) = 45.4 (24.1)). The asymmetry index was associated with age (p = 0.011). A post-hoc Davies’ test identified no significant change in slope at any value of age (p = 0.59), suggesting that the relationship between asymmetry index and age is consistently linear and negative across all examined values of age.

Associations between Age, VO₂peak, and asymmetry index

Results of all hierarchical regression models are reported in Table 2. Model 1 assessed the relationships between age, VO₂peak, and asymmetry index. In Step 1, no relationship was observed between asymmetry index and the covariates, sex and education. In Step 2, adding age captured significantly more variance in asymmetry index (ΔR² = 0.107, p = 0.007). This suggests a pattern consistent with hemispheric asymmetry reduction in aging, wherein asymmetry decreases with increasing age. In Step 3, adding VO₂peak did not capture more variance in asymmetry index (ΔR² = 0.008, p = 0.461). However, in Step 4, age moderated the relationship between VO₂peak and asymmetry such that as age increased, the relationship between VO₂peak and asymmetry became more negative (less hemispheric asymmetry; ΔR² = 0.077, p = 0.022) (Fig. 4A). Post-hoc simple slopes analyses using categorical age groups (Fig. 4B) showed that in middle-aged adults, greater VO₂peak was associated with greater asymmetry (simple slopes p < 0.001). In older adults, no association was found between VO₂peak and hemispheric asymmetry (simple slopes p = 0.196).

Fig. 4.

For all plots shown, greater values of asymmetry index indicate more hemispheric asymmetry, smaller values of asymmetry index indicate less hemispheric asymmetry/more hemispheric symmetry. (a) Continuous age moderated the relationship between cardiorespiratory fitness and asymmetry index such that as age increased, the relationship between cardiorespiratory fitness and asymmetry index became more negative. (b) Post-hoc simple slopes analyses using categorical age identified a similar interaction effect driven by a significant slope in middle-aged adults (35–64 years; p = 0.001) but no significant slope in older adults (65 years and older; p = 0.196). (c) A similar, marginally significant moderating effect of continuous age was found on the relationship between performance on the Grooved Pegboard Test and asymmetry index, such that as age increased, the relationship between Pegboard performance and asymmetry index became more positive (p = 0.056). (d) Post-hoc simple slopes analyses using categorical age identified a similar interaction effect, such that greater asymmetry in middle-aged adults was associated with better task performance with near significance (p = 0.053), and less asymmetry in older adults showed a trending but non-significant association with better task performance (p = 0.093).

Associations between VO₂peak, asymmetry index, and task performance

Model 2 assessed the relationship between age, asymmetry index, and performance on the visuomotor fMRI task. In Step 1, education and sex did not capture significant variance in visuomotor task performance. Adding age in Step 2 captured significantly more variance (ΔR² = 0.100, p = 0.012). Finally, neither adding asymmetry index in Step 3 nor adding the Age X Asymmetry Index interaction in Step 4 accounted for significantly more variance in visuomotor task performance (ps > 0.01).

In Model 3, the relationship between age, asymmetry, and performance on the Grooved Pegboard task was similarly examined. In Step 1, sex showed a significant negative association with performance on the Grooved Pegboard Task (p = 0.013), such that females completed the task more rapidly than males. In Step 2, adding age accounted for significantly greater variance (ΔR² = 0.221, p < 0.001); the partial effect of sex was maintained (p = 0.033). In Step 3, asymmetry index did not improve the model. However, the partial effect of sex was maintained (p = 0.036). In Step 4, the Age X Asymmetry Index accounted for additional variance that was marginally significant (ΔR² = 0.040, p = 0.057; Fig. 4C); the partial effect of sex was maintained (p = 0.029). Similarly, post-hoc simple slopes analysis showed that in middle-aged adults, greater asymmetry showed trending association with faster time to complete the Grooved Pegboard Task (simple slopes p = 0.053; Fig. 4D). In older adults, a trending association was observed between reduced asymmetry and faster time to complete the task (simple slopes p = 0.094).

Finally, Model 4 examined the relationship between age, cardiorespiratory fitness, and performance on the Grooved Pegboard Test. Steps 1 and 2 replicated those of Model 3, showing a significant effect of sex and age. In Step 3, adding VO₂peak to the model accounted for significant variance on the Grooved Pegboard Test (ΔR² = 0.073, p = 0.008); the partial effect of sex was maintained (p = 0.002). In Step 4, there was a trending moderation effect of Age X VO₂peak, such that as age increased, the relationship between VO₂peak and Grooved Pegboard Test performance became stronger and more negative (ΔR² = 0.029, p = 0.089); the partial effect of sex was maintained (p < 0.001). Post-hoc simple slopes analysis with categorical age demonstrated that this effect was significant in both middle-aged adults (simple slopes p = 0.039) and older adults (p = 0.004).

Discussion

The current study examined the relationship between cardiorespiratory fitness and hemispheric asymmetry across middle-aged and older adults. We identified three key findings. First, age moderated the relationship between cardiorespiratory fitness and hemispheric asymmetry. In middle-aged adults (35–64 years), cardiorespiratory fitness was associated with greater hemispheric asymmetry. The relationship became more negative (less asymmetry) with increasing age. In older adults (65 + years), cardiorespiratory fitness was not associated with hemispheric asymmetry. Second, hemispheric asymmetry was not associated with performance on the comparatively simple fMRI fractionated motor task, and this relationship was not modified by age. However, on the relatively more complex distal motor task (Grooved Pegboard), there was a trend indicating that age modified the relationship between hemispheric asymmetry and performance. In middle-aged adults, greater hemispheric asymmetry was associated with better Grooved Pegboard task performance (p = 0.053), whereas in older adults, reduced hemispheric asymmetry showed a trending but relatively weaker association with better task performance (p = 0.094). Finally, consistent with prior literature^44–46, greater cardiorespiratory fitness was associated with better performance on both the simple and the complex motor coordination task, regardless of age.

The observed effects align with several theories of functional brain aging. Most prominently, prior literature has established that cardiorespiratory fitness tends to be associated with patterns of functional activation consistent with better brain function. “Better” here typically refers to one of two phenomena. First, brain activation may remain unchanged despite external stressors or risk factors (e.g., aging), consistent with the notion of brain maintenance^47,48. Second, brain activation may change in ways associated with retained or improved cognitive performance, consistent with the notion of compensation^47,48. Prior research has shown that during an associative memory task, young adults and high-fit older adults demonstrated less brain activation compared to lower-fit older adults in some brain regions⁴⁹. This suggests that in some brain regions, high-fit older adults have patterns of brain activity similar to young adults, which may reflect brain maintenance. In other brain regions during the same task, however, high-fit older adults had increased activation compared to low-fit older adults and young adults⁴⁹. Greater activity in these regions was found to mediate the relationship between cardiorespiratory fitness and task performance, suggesting that cardiorespiratory fitness may support compensation via increased activity in these regions. As this example makes clear, brain maintenance and compensation are not mutually exclusive phenomena, and patterns of brain maintenance or compensation may be region-specific for a given task.

Notably, age-related increased activation in multiple brain regions across a variety of tasks has been consistently observed^4,50. However, increased brain activation has been interpreted as compensation, or alternatively, dedifferentiation. The theory of dedifferentiation hypothesizes that some increases in brain activation with age may be unrelated to cognitive performance and instead reflect a reduction in the brain’s ability to efficiently utilize resources^4,51–53. If brain activity were hypothesized to be compensatory, one would expect that increased activity would be associated with better cognitive performance. In contrast, from a dedifferentiation perspective, increased brain activation would be associated with poorer cognitive performance. Thus, in its most simple form, whether increased brain activation reflects compensation or dedifferentiation can be determined by whether activation is positively (compensation) or negatively (dedifferentiation) linked to cognitive performance or with known associates of cognitive health³. Prior findings regarding aging and hemispheric asymmetry, particularly in the motor cortex, have been limited but mixed, with some studies providing evidence of compensation^3,25 and others of dedifferentiation^26,54,55.

Within the present study, some findings were consistent with the notion of brain maintenance. For example, greater cardiorespiratory fitness in middle-aged adults was associated both with better simple and complex task performance and with greater hemispheric asymmetry, comparable to the findings from McGregor et al. (2013)²³. Better task performance and greater hemispheric asymmetry are more often seen in younger adults³. Though the present data are cross-sectional, the presence of these patterns in middle-aged adults may suggest retention of brain and cognitive health, or brain maintenance. On the other hand, in older adults, although cardiorespiratory fitness was associated with better complex task performance, cardiorespiratory fitness was not associated with hemispheric asymmetry, and hemispheric asymmetry showed only a trending association with better complex task performance. This suggests that cardiorespiratory fitness may impact cognition through a mechanism other than hemispheric asymmetry in older adults. Such mechanisms could include associations with other age-related patterns of fMRI activity that may support cognitive function, such as the posterior-to-anterior shift in aging^49,50 (PASA), or with structural brain health, such as white matter integrity⁵⁶.

In the current study, there was tentative support for activation representing dedifferentiation in the motor cortex in middle-aged adults and compensation in older adults, as evidenced by a trending association of increased hemispheric asymmetry with better performance on the Grooved Pegboard Task in middle-aged adults and poorer performance on the Grooved Pegboard Task in older adults. However, given the marginal significance, caution is warranted, and the results should not be overinterpreted. These trend-level results (p = 0.05–0.1) should be considered hypothesis-generating, rather than confirmatory. Our findings in older adults contrast with those reported by Knights et al. (2021)⁵⁴, who examined the relationship between hemispheric asymmetry and cognitive performance in a sample of young, middle-aged, and older adults. The authors used both behavioral and multivariate Bayesian inference approaches to examine whether ipsilateral motor activity during a unilateral button-pressing task was positively related to task response time inside or outside the scanner. No evidence of compensation was found, nor any moderating effects of age on the relationship between ipsilateral motor activity and task performance. As such, the authors concluded that hemispheric asymmetry in the motor cortex was likely evidence of dedifferentiation. Additional work may be necessary for more definitive conclusions, and longitudinal task-related fMRI studies will facilitate the distinction between compensatory versus dedifferentiation patterns of age-related changes, and the extent to which these patterns are task- or region-specific.

It is important to note that although cardiorespiratory fitness was not associated with hemispheric asymmetry in older adults in the present analyses, there was a significant interaction of age on the relationship between cardiorespiratory fitness and hemispheric asymmetry, such that as age increased, the relationship between cardiorespiratory fitness and hemispheric asymmetry became more negative (less asymmetry). This suggests that the association between cardiorespiratory fitness and hemispheric asymmetry may not be homogenous among all individuals categorized as older adults in the present study (aged 65 years and older) and that cardiorespiratory fitness may have a greater effect on hemispheric asymmetry with increasing age. To better understand this relationship, future research with larger samples of older adults of greater age is necessary.

There are several potential and likely co-occurring mechanisms underpinning the relationship between cardiorespiratory fitness and fMRI activation. Cardiorespiratory fitness has been associated with improved neurovascular function^57,58, which supports cellular health and the integrity of brain structure^59,60. Cardiorespiratory fitness has also been linked to higher cortical thickness and enhanced white matter microstructure in older adults^{15,49,56,61,62}. For instance, greater white matter integrity of the corpus callosum, a region important for hemispheric transfer of motor information, was observed in high-fit compared to low-fit older adults⁵⁶, which may be particularly relevant to the observed functional asymmetries in the motor cortex observed in the current study^63–65. In sum, these positive associations with cerebral perfusion, neurovascular function, and structural integrity may be contributing to the relationship between CRF and fMRI activation. Further, cardiorespiratory fitness and aerobic exercise have been associated with several growth factors including brain-derived neurotrophic factor (BDNF), insulin-like growth factor (IGF), and vascular endothelial growth factor (VEG-F), which have been linked to cardiovascular function and functional and structural markers of brain health^66–69. Acute exercise has also been associated with the activation of noradrenergic pathways, including the locus coeruleus and associated regions^70,71. Promotion of noradrenergic activity has been proposed to support patterns associated with cognitive and brain reserve^72–74. Finally, acute exercise has also been proposed as a neurovascular mimetic for the aging process; that is, acute exercise may stress the vasculature of the brain and induce neural inflammation in ways similar to the traditional aging processes⁷⁵. For instance, aging has been associated with reduced cerebral perfusion, and acute exercise has been associated with reduced cerebral perfusion^76,77. Habitual exposure to acute bouts of exercise may therefore prepare the brain for detrimental age effects and build resistance to age-associated neurovascular stress, limiting the negative impact of the aging process in later life. Further research is required, however, to fully understand the multitude of mechanisms that may underpin the relationship between cardiorespiratory fitness and patterns of functional brain activity, particularly as these mechanisms may vary across the lifespan.

Although not the focus of the present analyses, it is interesting to note that an effect of sex was found on Grooved Pegboard performance in both Model 3, which included the Age X Asymmetry Index interaction, and Model 4, which included the Age X VO₂peak interaction. In both models, female participants showed better task performance (faster time to complete the Grooved Pegboard task) than male participants. Sex differences in Grooved Pegboard are well-established in the literature. It has been hypothesized that these differences can be attributed to the smaller average hand size of women⁷⁸. However, further exploration of sex effects on of these relationships should be explored in future work, for example using triple interactions in studies with larger sample sizes.

Our current analysis focused on age-related differences in asymmetry in the motor cortex and their association with cardiorespiratory fitness. There were age-related differences in BOLD activation that occurred outside of the motor cortex, including BOLD signal decreases in the hippocampus and occipital lobe. These differences may also provide insight into the relationship between cardiorespiratory fitness and patterns of functional brain activity and may therefore be relevant to explore in future work.

The decision to focus the current analysis on age-related differences in asymmetry in the motor cortex also informed our ROI selection procedure. Prior literature examining HAROLD and other functional asymmetries utilized varying methods for ROI selection. For the present study, we chose to utilize a mirrored ROI approach. This involved deriving an ROI based on age-related increases in BOLD activation limited to the ipsilateral motor cortex. We then mirrored this ROI to the contralateral motor cortex. The contralateral ROI therefore captures a more specific extent of the contralateral motor cortex than might be identified using group mean peak BOLD activation assessed independently. The mirrored ROI overlaps almost entirely with group mean peak BOLD activation (Supplemental Fig. 1A). As such, the selected ROI procedure not only targets regions of strongest asymmetry across the sample but also creates study-specific ROIs that are relevant to the subjects of the analysis and are matched on spatial extent and size between hemispheres. Alternative approaches may also be appropriate and could provide differing results. For example, while our ROIs resulted in a measure of asymmetry that was driven by increased positive BOLD response in the ipsilateral motor cortex and relatively constant activation in the contralateral motor cortex (Supplemental Fig. 4), other ROIs might identify reduced magnitude negative BOLD response in the ipsilateral motor cortex. One approach that may allow for the greatest optimization of asymmetry across participants in future research might include using a separate task as a functional localizer to derive subject-specific motor ROIs in each hemisphere⁷⁹.

Due to the brief duration of the visual-motor fMRI task and the intermittent button-pressing during active blocks of the task, some methodological decisions were made to prioritize data quality. Prior research has shown that unilateral motor tasks, similar to the task performed in the present study, have been associated with increased BOLD signal during task relative to rest in the contralateral motor cortex (motor activation) and decreased BOLD signal during task relative to rest in the ipsilateral motor cortex (motor deactivation)^6,7. This pattern is more pronounced in younger adults and in tasks with greater motor demand⁶, but has been observed with varying spatial extent and strength of association across the lifespan⁷. These motor deactivations have been attributed to suppression of the ipsilateral motor cortex^6,7. However, a similar pattern of deactivation could be observed for other reasons – namely, off-task hand or arm movement during rest periods of the task. For this reason, in the current analyses we chose to exclude participants who showed an average beta coefficient less than zero in either the ipsilateral or contralateral ROI (n = 23). Our analysis sample included only participants aged 35 years and older, and suppression-related BOLD signal is less likely in this age range, particularly within the selected ROI. Post-hoc comparisons (Supplemental Table 1) of excluded participants indicated they do not significantly differ on key demographic variables from participants retained in analyses, though the average age of those excluded is trending younger than those included. Nevertheless, it is possible that deactivations among excluded participants were related to suppression, which may have influenced the present findings. Future research investigating motor cortex asymmetry would benefit from more direct measures of quality control – for example, hand-worn accelerometers or video of hand motion – to definitively separate suppression-related BOLD activity from BOLD activity related to off-task hand motion.

On the other hand, in the present study, we chose not to exclude left-handed participants from analyses. Neural motor responses and task performance may differ between those who report left-handedness v. right-handedness⁸⁰; however, only 4 participants in the present analyses self-reported left-hand dominance. To assess the extent to which results were potentially biased by these participants, analyses were repeated after excluding participants self-reporting left-handedness (Supplemental Table 2). The directionality of results was maintained, though some model outcomes were slightly weakened (for example, the continuous Age X VO₂peak interaction on asymmetry was significant with all participants included, but trending with left-handed participants excluded, likely due to a weaker effect in older adults; the categorical interaction in this model remained unchanged). This suggests that findings were not substantially changed due to the inclusion of left-handed participants, but future work should repeat similar analyses with larger sample sizes to confirm the strength of observed effects.

Finally, cohort effects related to functional movement dynamics should also be considered when interpreting the results of motor tasks of varying complexity. For example, older adults may require distal unimotor dexterity more often for medication management, whereas middle-aged adults may be more likely to use similar fine motor skills during work-related activities⁸¹. Therefore, the fMRI and Grooved Pegboard tasks may differ both in the level of relative complexity and cohort-specific behavioral functions. However, the goal of the current study was limited to better understanding mechanisms of cognitive aging by studying the relationship between hemispheric asymmetry and cardiorespiratory fitness. Thus, future studies should examine how these mechanisms relate to functional movements.

The current study had several limitations. Limited data were available in young adults, precluding a comprehensive adult lifespan assessment of age-related differences in fMRI activation and associations with cardiorespiratory fitness. The study was cross-sectional, limiting evidence for causal relationships between aging, brain activation, and cardiorespiratory fitness. FASTER is, however, a longitudinal study, and we plan to examine longitudinal relationships when the 3-year follow-up data collection is complete. Our study sample was predominantly White (93.8%), well-educated (mean years of education = 17.1 years), and healthier than the general population given safety requirements for MRI and maximal cardiopulmonary exercise testing; for instance, participants with pacemakers or other contraindications for MRI were excluded. Although participants with moderate or severe TBI were not included in the present analyses, participants with mild TBI may be overrepresented compared to similar datasets. When mild TBI was included as a covariate in the main analyses, it was not significantly associated with the outcome variable (all p’s > 0.70). Finally, several findings were at the statistical trend level, p ≤ 0.10, and therefore warrant caution in their interpretation. Longitudinal research in more diverse samples is needed to clarify the extent to which patterns of hemispheric asymmetry in midlife are maintained in older adulthood and provide additional insight to the contributions of brain maintenance, compensation, and dedifferentiation.

Despite these limitations, the current study had multiple strengths. The gold standard for assessment of cardiorespiratory fitness, progressive maximal exercise testing, was implemented, rather than estimating VO₂ based on a submaximal test or a calculation. The study sample included middle-aged adults, who are often difficult to recruit for research studies, given that many are in the prime of their careers and may have caregiving obligations. Finally, the study implemented a simple design to provide a straightforward assay of hemispheric asymmetry and linked hemispheric asymmetry and cardiorespiratory fitness to two tasks, administered on different days, providing a more robust assessment of motor function than many extant fMRI studies.

In summary, the present study identified differential relationships between hemispheric asymmetry, cardiorespiratory fitness, and task performance in middle-aged and older adults. In middle-aged adults, greater cardiorespiratory fitness was associated with better task performance and with greater hemispheric asymmetry, which aligns with theories of brain maintenance. In older adults, greater cardiorespiratory fitness was associated with better task performance but not with hemispheric asymmetry. Finally, there were trending associations found between hemispheric asymmetry and task performance. Greater hemispheric asymmetry was marginally associated with better task performance in middle-aged adults and greater asymmetry was marginally associated with poorer task performance in older adults. This provides tentative evidence for differential mechanisms underlying hemispheric asymmetry reduction in middle-aged and older adults, wherein hemispheric asymmetry reduction may indicate dedifferentiation in middle adulthood but support compensation in older adulthood. These findings highlight the importance of examining modifiable lifestyle variables and their association with cognitive and brain measures prior to older adulthood. Furthermore, the observed relationship between cardiorespiratory fitness and hemispheric asymmetry in middle-aged adults, but not in older adults, may suggest a critical period during which cardiorespiratory fitness impacts specific mechanisms associated with cognitive and brain function.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

J.A.C: Conceptualization, Methodology, Formal Analysis, Writing – Original Draft, Writing – Review and Editing, Visualization; Inola Howe: Writing – Reviewing and Editing; W.J.K: Writing – Reviewing and Editing, Funding Acquisition; J.S.V: Writing – Reviewing and Editing; J.P.H: Writing – Reviewing and Editing, Project Administration; S.M.H.: Conceptualization, Methodology, Resources, Writing – Reviewing and Editing, Supervision, Project Administration, Funding Acquisition.

Funding

This work was supported by the National Institute on Aging (NIA) from the National Institutes of Health (NIH) under grant R01 AG068882 (awarded to S.M.H.) and The Ohio State University Center for Brain Injury Recovery and Discovery  (S.M.H.).

Data availability

Code used for reported processing and analyses are available via GitHub at https://github.com/osubbal/fMRI-HAROLD-CRF_JCloud. Behavioral datasets generated and analyzed during the current study are available via Zenodo at https://zenodo.org/records/16498040. Imaging datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.