The Impact of Physical Activity on Brain Structure and Function in Youth: A Systematic Review

The conduct and reporting of this review adhere to the guidelines outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analysis statement.³⁰  The review protocol was registered with the International Prospective Register of Systematic Reviews (CRD42017081804).

1. Types of participants: participants were typically developing school-aged children (usually 5–18 years of age; however, children outside this age range were included if they were recruited within schools). Studies including populations with learning difficulties, cognitive deficits, and developmental disorders were excluded.

2. Types of studies: experimental studies were eligible if the authors reported statistical analyses of changes in brain structure or function before and after a physical activity intervention.

3. Measure of physical activity, cardiorespiratory fitness, or muscular fitness: studies with objective (eg, accelerometers and pedometers) or subjective measures of physical activity (eg, exercise session attendance and self-report questionnaires); cardiorespiratory fitness (eg, maximum oxygen consumption [VO_2max] test, Progressive Aerobic Cardiovascular Endurance Run, and predictive equations); and/or muscular fitness (eg, dynamometry, standing long jump, and push up test) were eligible.

4. Brain imaging techniques: studies that reported findings from MRI techniques (eg, functional MRI [fMRI], diffusion tensor imaging [DTI], and arterial spin labeling [ASL]) that have been used to identify structural and functional mechanisms that may explain the relationship between physical activity, cardiorespiratory fitness or muscular fitness, and cognition or academic achievement were eligible.

Six electronic databases (PubMed, PsychINFO, Scopus, Ovid Medline, SportDiscus, and Embase) were searched for studies published within the last 16 years (2002–March 1, 2019) (Supplemental Table 4). Additional searches of recently published systematic reviews in which the associations between physical activity, cardiorespiratory fitness or muscular fitness, and cognitive outcomes were examined were conducted, and the reference lists of all retrieved studies were reviewed. The search was restricted to articles published in the English language.

The study screening and selection process was performed on Covidence.³¹  One reviewer screened the titles and abstracts of records retrieved by the search strategy and classified these as possibly relevant or definitely irrelevant. The full-text articles of records classified as possibly relevant were retrieved and independently reviewed by 2 reviewers. Studies were classified as include or exclude. If there was disagreement between reviewers, consensus was sought through discussion. Reasons were provided for excluding any possibly relevant studies.

Both reviewers independently extracted data from included studies into a purpose-built data extraction template in excel. Data extraction included (1) sample data (including sample size, age, sex, and education); (2) study details (location, design, setting, duration, and assessment points); (3) assessment of physical activity, cardiorespiratory, and/or muscular fitness (objective, subjective, laboratory-based, or field-based); (4) neuroimaging data (MRI modality, analysis methods, regions of interest, etc); (5) data analysis (statistical methods used, confounders adjusted for, etc); and (6) study findings (quantitative and qualitative).

All studies were independently assessed by 2 reviewers and were scored as low, high, or unclear for 8 criteria according to the Cochrane collaboration risk of bias tool and scoring.³²  Any disagreement concerning risk of bias assessment between the 2 reviewers was resolved through discussion. Consensus was reached on all articles included in the review.

Figure 1 displays the flow of studies through the review process. After the exclusion of duplicates, the systematic search yielded 9508 potentially relevant citations, of which 153 were retained for full-text review. There was almost-perfect interrater agreement for the full-text review (κ = 0.97).³³  A total of 9 articles satisfied the inclusion criteria and were included in the review, reporting results for 5 distinct and 4 similar samples of participants ranging from 8.7 ± 0.6³⁴  to 10.2 ± 1.0³⁵  years of age and typically of relatively low socioeconomic status. The sample size ranged from 9³⁶  to 143.³⁴  The studies were conducted in North America (8 in the United States) and Asia (1 in China). Of the included studies, 7 were RCTs and 2 were acute before and after studies. Detailed information about each included study is presented in Table 1.

Detailed information about the risk of bias for the included studies is presented in Table 2. In summary, all 9 (100%) were deemed to be at unclear risk of selection bias, with unclear description of (1) sequence generation process, (2) concealed allocation processes, and [in 5 (56%) studies] (3) subgroup selection processes. Seven (78%) studies were deemed at unclear risk of reporting bias because of lack of availability of a protocol published by means of either an article or trial registration. Six (67%) studies were deemed at high risk of attrition bias because of significant dropout with inadequate analyses. Overall, only 2 (22%) studies scored as low risk of bias for ≥3 (of the 8 criteria.^34,35  There was substantial interrater agreement for the risk of bias assessment (κ = 0.61).³³ 

Four different MRI modalities were used across the 9 included studies. Four (44%) studies used task-based fMRI, 3 (33%) studies used DTI, 1 (11%) study used ASL, and 1 (11%) study used resting-state fMRI. Data for 12 regions were reported across the 9 included studies: anterior cingulate cortex, cerebellum, corpus callosum, frontal lobe, hippocampus, parietal lobe, superior longitudinal fasciculus, uncinate fasciculus, cognitive control network, default mode network, executive control network, and motor network.

Authors of seven (78%) studies provided physical activity interventions and investigated effects on brain structure or function.^34,37–42  The duration of the interventions ranged from 3 to 9 months and generally consisted of moderate-to-vigorous physical activity (eg, 70%–80% maximum heart rate [HR_max]) either twice a week or each school day for 20 to 120 minutes. Of these, 4 studies measured cardiorespiratory fitness by means of oxygen uptake during a maximal graded treadmill test (modified Balke protocol).^34,37,39,42  Authors of 2 studies investigated changes in brain function in response to acute bouts of aerobic exercise at 60% to 70% HR_max.^35,36  Details of all interventions are outlined in Table 1.

Findings from each included study are presented by brain region below and Table 1, with effects further summarized in Table 3.

Authors of 3 RCTs with distinct but similarly aged samples reported results for changes in activation of the frontal lobe in response to physical activity interventions which ranged from 20 to 77 minutes each school day over 3 to 9 months. Authors of 2 of the RCTs assessed prefrontal activation during cognitive tasks (antisaccade [n = 2] and flanker [n = 2]) and found changes pre- and post- intervention but the effects were in opposite directions in both cases. Davis et al³⁸  reported that increased bilateral prefrontal (and decreased posterior parietal) cortex activity was observed during antisaccade performance in the physical activity group, whereas Krafft et al³⁹  reported decreased activation during antisaccade performance in several prefrontal (and parietal) regions including medial frontal gyrus, right inferior frontal gyrus, and bilateral precentral gyrus. Krafft et al³⁹  also observed increased activation of the superior frontal gyrus of the prefrontal cortex during incongruent trials of the flanker task in the physical activity group, whereas authors of the third RCT (Chaddock-Heyman et al³⁷ ) observed decreased activation in the right anterior prefrontal cortex during incongruent trials of the flanker task in the physical activity intervention group but no changes in the control group. Note that although both Chaddock-Heyman et al³⁷  and Davis et al³⁸  adjusted for baseline during their region of interest analyses, Krafft et al³⁹  did not report if/what covariates were adjusted for and employed a whole-brain analysis approach, which could contribute to the disparate results.

Authors of 3 studies reported results for the parietal lobe from task-based fMRI paradigms. Authors of 2 RCTs with similarly aged samples and relatively similar type, frequency, intensity, and duration of physical activity interventions found decreased parietal cortex activity during antisaccade performance after a physical activity intervention.^38,39  Both studies used comparable cluster size thresholds but it should be noted that while Davis et al³⁸  adjusted for baseline in their analyses, Krafft et al³⁹  did not report if and what covariates were adjusted for.

Chen et al³⁶  investigated the acute effects of a 30-minute bout of cycling (60%–69% HR_max) during task-based fMRI and reported improved n-back performance and increased activation of bilateral parietal cortices (as well as the left hippocampus and bilateral cerebellum).

Authors of 2 RCTs reported task-based fMRI results for the anterior cingulate cortex. Authors of 1 RCT found that participation in a physical activity intervention did not change activation of anterior cingulate cortex during neutral or incongruent conditions of a flanker task.³⁷  The other RCT found that although there were no significant correlations between changes in cardiorespiratory fitness and brain activation during task-based fMRI,³⁹  the physical activity intervention led to differential activation across 2 inhibition tasks, with decreased activation of the anterior cingulate cortex during an antisaccade task and increased activation of the cingulate gyrus during the incongruent condition of a flanker task.³⁹  Comparatively, the control group showed decreased activation during the flanker task.³⁹  Such differences across inhibition tasks highlights the complexity of brain activation during performance of tasks that tap different aspects of a similar cognitive construct.

Authors of 1 acute before and after study reported enhanced performance in an n-back task and increased brain activity (task-based fMRI) of the left hippocampus in response to an acute 30-minute bout of cycling (60%–69% HR_max).³⁶ 

Authors of 1 acute experimental study investigated the effects of a 30-minute bout of cycling (60%–69% HR_max) during task-based fMRI and reported improved n-back performance and increased activation of bilateral cerebellum.³⁶ 

Authors of 2 experimental studies reported results for specific functional brain networks. Authors of 1 RCT used an independent component analysis approach and reported that a physical activity intervention caused decreased synchrony between the default mode network and the cognitive control network with brain regions outside of those networks during resting-state fMRI.⁴⁰  There was no change in synchrony of the salience network, whereas the motor network had decreased synchrony with the left cuneus but increased synchrony with certain frontal regions.⁴⁰ 

Pontifex et al³⁵  investigated the acute effects of a 20-minute bout of fast walking and/or slow jogging (70% HR_max) on cerebral blood flow in 10.2 ± 1.0 year old (n = 41) and found no differences across any of the networks examined (frontoparietal, executive control, and motor networks).

Authors of 3 studies reported results of 2 RCTs that had examined the effects of physical activity on white matter tracts in similarly aged children using regions of interest analyses.^34,41,42  One large RCT (n = 143) revealed that 2 hours of physical activity each school day for 8 months improved white matter integrity (ie, increased fractional anisotropy, which indicates the orientation of diffusion and is higher along well-defined pathways) and decreased radial diffusivity (a marker of myelin disintegration) in the genu of the corpus callosum from pretest to post-test, with no changes in estimates of axonal fiber diameter (axial diffusivity).³⁴  There were no changes in the white matter integrity of the wait list control group, reflective of typical development. The other RCT (n = 18) also delivered an 8-month intervention consisting of a 40-minute session each school day. Authors of 1 study reported that the physical activity group showed greater increases in bilateral uncinate fasciculus fractional anisotropy and greater decreases in left uncinate fasciculus radial diffusivity compared with the control group.⁴²  In the second report from this RCT, the physical activity intervention did not significantly increase white matter integrity in the superior longitudinal fasciculus. However, higher attendance in the exercise intervention, higher intensity, and greater total dose of exercise were all associated with increased fractional anisotropy and decreased radial diffusivity of the superior longitudinal fasciculus in a dose-response manner.⁴¹ 

In this systematic review, we examined evidence of the impact of physical activity on brain structure and function in youth from MRI studies. Nine experimental studies were included in the review, of which 7 were RCTs and 2 were acute before and after studies, reporting data for 12 regions acquired with 4 MRI modalities. All 7 RCT studies (4 samples) reported significant changes in either brain structure or function after a physical activity intervention in young people.^37–42 

To date, the parietal cortex is the only specific region that has had >1 RCT report in which authors found an impact of physical activity on brain structure or function and for the effects to be in the same direction (ie, authors of both RCTs found decreased posterior parietal cortex activity during antisaccade performance after a physical activity intervention^38,39 ). Otherwise, RCT findings for the impact of physical activity on activation during task-based fMRI were inconsistent (ie, authors of 1 study found an association and another did not) for the anterior cingulate cortex,^37,39  or conflicting (ie, physical activity had an impact on activation, but authors of 1 study reported increased activation and authors of another study reported a decreased activation in the case of each task paradigm [antisaccade and incongruent condition of a flanker task]) for frontal regions.^37–39  It should be noted that although the sample ages and cluster size thresholds were similar in these studies, the interventions varied from 20 to 77 minutes per session over 3 to 9 months which presents considerable heterogeneity.

The desired direction of the effect of physical activity on activation will differ depending on the region and context (eg, task and rest) of interest. However, positive associations between physical activity and activation of task-positive regions during performance of task paradigms is interpreted as a greater ability to use resources in some studies,^36,43,44  whereas negative associations (ie, less activation) are considered to represent a more efficient use of resources in others.^37,45  There is evidence to support decreased activation of a task-positive region during task performance being reflective of a more mature and adult-like brain^46–48  but this should be interpreted with caution until the findings have been replicated by studies adequately powered to perform mediation analyses.⁴⁹ 

Authors of 2 RCTs found that physical activity caused decreased activation of the posterior parietal cortex during antisaccade task performance. Although this did not reflect a difference in antisaccade performance between the physical activity and control group in 1 study,³⁹  authors of the other study did not report data for antisaccade task performance.³⁸  The inferior parietal lobule, located within the posterior parietal cortex, forms part of the default mode (task-negative) network,^50–52  which is known to decouple from the cognitive control network during successful performance of a cognitive task.⁵³  Therefore, these results may indicate a more refined, adult-like pattern of activation in the exercise group while maintaining equivalent levels of task performance.^54–56  A recent meta-analysis revealed that deactivation of the default mode network is essential for processing information so that it can later be remembered.⁵⁷  This diversion of processing resources from the default mode network to brain regions involved in the task performance has previously been demonstrated in a cross-sectional pediatric physical activity study. Despite similar memory performance to their inactive peers, during encoding of later remembered versus forgotten word pairs, participants with high levels of physical activity displayed (1) robust deactivation of the default mode network, (2) strong negative coupling with the hippocampus, and (3) a more focal increase in activation of the left hippocampus only.⁴⁵ 

Decreased synchrony between a given network and regions outside of that network is usually an indication of a more focal, coherent, and specialized pattern of activation.^58,59  Authors of 1 RCT in this review examined deactivation and activation of functional networks during resting-state fMRI and found that physical activity may be conducive of a more mature efficient brain by causing decreased synchrony of the default mode network and cognitive control network with brain regions outside of those networks during resting-state fMRI.⁴⁰ 

In terms of structural changes, 1 large RCT (FITKids2; n = 143) revealed that participation in physical activity can improve white matter integrity of the corpus callosum; a region important for cognitive processing.³⁴  A second RCT investigated effects of physical activity on white matter integrity and detected significant improvements in the bilateral uncinate fasciculus (which usually matures later than many other tracts⁶⁰ ).⁴²  This was particularly evident in the left uncinate fasciculus, which is linked with auditory-verbal memory proficiency, verbal IQ, and full-scale IQ.^42,61,62 

In a second study from the same RCT,⁴¹  changes in white matter integrity of the superior longitudinal fasciculus were not significantly different between the groups. However, higher attendance at exercise sessions, higher intensity, and greater total dose of exercise were positively associated with changes white matter integrity.⁴¹  Similarly, white matter integrity did not change among adults participating in a 1-year exercise intervention, but changes in fitness were positively associated with white matter integrity of prefrontal and temporal regions (which are linked by the uncinate fasciculus).⁶³  Improvements in fitness were also associated with changes in short-term memory, but increases in white matter integrity were not associated with short‐term memory improvement. In another larger-scale study involving adults, white matter integrity in multiple tracts (including those that connect medial temporal and prefrontal cortices) mediated the relationship between fitness and spatial working memory.⁶⁴  Additional support for the importance of fitness in terms of white matter integrity also exists in pediatric cross-sectional studies, which have found positive associations between fitness and fractional anisotropy in several of the same white matter tracts in children.⁶⁵ 

To date, no RCT has examined the impact of a physical activity intervention on volumes of brain regions in children or adolescents. This is surprising given that a recent meta-analysis on the effect of aerobic exercise on hippocampal volume in adults included 14 studies.⁶⁶  This review revealed a significant effect of aerobic exercise on both left and right hippocampal volume in comparison with control conditions in healthy older adults. The effects were driven by exercise attenuating normal age-related neurodegeneration, which has been shown to precede and lead to cognitive decline and Alzheimer disease.^67,68  Whether exercise can increase the volumetric growth of the hippocampus and whether these increases in volume subsequently confer benefits to cognition, memory, and/or academic performance during childhood and adolescence has not been established.

More studies in adolescents are needed because all experimental studies included in this review were conducted with children. Future researchers should also measure cardiorespiratory and muscular fitness so that (1) baseline fitness can be adjusted for in analyses and (2) changes in fitness due to physical activity interventions can be analyzed for correlations with changes in brain structure or function. There is considerable scope for different intensities, frequencies, and types of physical activity such as high-intensity interval training, resistance exercise, exergaming, and cognitively demanding physical activity to be explored.⁶⁹ 

Although this is the first systematic review of MRI studies in the area of pediatric physical activity, there are some limitations that should be noted. Most notably, because of the small number of RCTs and considerable heterogeneity of included studies, we were unable to conduct meta-analyses. In addition, we did not check for a file drawer effect so the risk of publication bias cannot be ruled out.

There are a number of common study limitations that should be noted. The majority of the included studies included small samples and/or relied on statistical significance analyses. The P values do not provide an indication of the size of an effect nor the importance of a result and by themselves are not a good measure of evidence regarding a model or hypothesis.⁷⁰  As such, the field needs to progress to promote the reporting of effect estimates in addition to the corresponding statistics.⁷¹  Risk of bias was largely unclear across all domains and studies. Researchers are encouraged to adhere to the Consolidated Standards of Reporting Trials guidelines⁷²  to reduce the risk of bias, particularly in terms of selection bias and reporting bias.⁷³  Findings need to be interpreted with caution until additional RCTs can (1) replicate findings and (2) establish whether exercise-induced changes in brain structure or function mediate the cognitive and/or academic benefits of physical activity.

There is some evidence from RCTs that participation in physical activity may enhance brain structure and function in terms of white matter integrity and activation of regions key to cognitive processes, respectively. No RCT researchers have reported on the impact of physical activity on volumes of brain regions in children or adolescents.

ASL

arterial spin labeling

BDNF

brain-derived neurotrophic factor

CI

confidence interval

DTI

diffusion tensor imaging

fMRI

functional MRI

HR_max

maximum heart rate

RCT

randomized controlled trial

VO_2max

maximum oxygen consumption