What Does Exercise Do for the Brain?

Blog post by Andrew Davies, PhD

The benefits of exercise on quality of life are well established, and include improved mental health and immune regulation, as well as reduced incidence of cardiovascular disease, cancer, and diabetes (for extended and current reviews, see 1, 2, and 3). The evidence in support of these claims is extensive, consistent, and widely accepted with little opposition. However, historically, and perhaps at least somewhat anecdotally, the benefits of exercise have also been extended to various measures of cognition, such as learning, memory, and problem solving. For example, the CDC (4) claim that “Physical activity can help you think, learn, problem-solve”, and the World Health Organization (WHO) state that “Physical activity enhances thinking, learning, and judgment skills” (5). The preponderance of evidence clearly demonstrates that exercise is both highly advisable and beneficial from a holistic perspective, but the question persists: what does exercise do for the brain? Two very recent publications have arrived at somewhat conflicting answers to this question.

Any assessment of human behavior and its effects is often complicated by confounding variables, and exercise studies are certainly no exception. When comparing and evaluating available data on the benefits of exercise, subject age, general health and cognitive status, the type, duration, frequency, and intensity of exercise, and the choice of outcome measures used to assess cognition may themselves be cause for debate. Randomized controlled trials (RCTs) are often considered the gold standard to assess the effects of interventions, and overcome some of these limitations, but blinding subjects to exercise treatment is extremely challenging (6). Data to be included are therefore inherently heterogeneous; fortunately, this is offset to some extent by the number of available published studies and participants.

In a recent umbrella review of meta-analyses, Ciria et al. sought to answer the question: does regular exercise conclusively improve cognition in healthy populations? Their answer was a somewhat surprising “no”. It is important to emphasize that their analysis is specific to healthy participants, and does not conclude that regular exercise does not improve cognition; rather, as the authors emphasize, it suggests that the currently available evidence is insufficient to establish a causal effect (7).

For the purposes of their umbrella review, Ciria et al. included 24 meta-analyses of randomized controlled trials (RCT), constituting 109 separate primary studies, and over 11,000 participants. As noted by the authors themselves, observational and epidemiological studies may also have merit, but were not included in this study. Their results suggest that variations in study sampling, influential variables, baseline measures, control groups (such as physical or non-physical activity), and publication bias, have collectively resulted in an overestimation of the magnitude and significance of positive effects in the existing meta-analyses. The authors note that “the particular conclusions from the different meta-analyses cannot be taken as the empirical evidence accumulated over years, but as selective slices of it”. Their ultimate conclusion is therefore not that exercise does not improve cognition, but rather that, when subjected to further scrutiny, the existing meta-analyses do not conclusively support the existence of such a causal effect (7).

An interesting additional observation from the authors is that, despite the numerous studies evaluating the potential effects of exercise on cognition, “the absence of a firm theoretical model of the mechanisms involved in exercise-induced cognitive improvements in humans is surprising” (6). As they note, foundational preclinical research on numerous possible mechanisms of action is available. More detailed human studies are precluded for innumerable reasons, but previously proposed theories fail to completely account for existing observations and published evidence. It remains to be seen if an adequate and sufficiently complex model can be developed to account for the effect of exercise on cognition in human populations given the inherent complexity of the behavior and associated interactions.

Umbrella reviews:
Overviews (or reviews) of reviews, that aggregate the results of systematic reviews and/or meta-analyses. As an ever-increasing number of systematic reviews and meta-analyses are published, umbrella reviews can provide a high level overview of existing evidence and help to resolve discrepancies amongst the conclusions of individual reviews.

Systematic reviews:
Summaries of existing evidence from studies that meet predetermined criteria for inclusion. By clearly stating the question and methods explicitly prior to selection and analysis of primary studies, the ultimate goal is to distill existing evidence in a thorough and objective manner to answer the research question of interest.

Meta-analysis:
The statistical analysis of data from multiple individual studies, with the goal of pooling results from primary research to arrive at a consensus estimate.

Randomized controlled trials:
Prospective studies in which subjects are randomly assigned to treatment/intervention or control groups, with the goal of determining the effect of the treatment or intervention.

Primary studies:
Involve direct collection of data from subjects, such as patients or populations.

In a separate study published only days after the Cilia et al. paper, Cheval et al. (8) arrived at a different conclusion; namely that moderate and vigorous physical activity improved cognitive function. The authors begin with an acknowledgement of the numerous putative mechanisms that may account for any exercise-induced changes in cognition, specifically emphasizing the need to evaluate causality and the bi-directional nature of this relationship. Cognition is inherently required for physical activity, leading to an important additional question: how does cognition influence physical activity?

The authors emphasize that exercise and cognition is a bidirectional relationship; therefore, in addition to accounting for confounding variables, a conclusive demonstration of the causal effects of exercise on cognition must account for the effects of cognition on exercise. Common to both the Cilia et al. and Cheval et al. studies is an acknowledgement of the limitations of currently available RCT’ on exercise and cognition, and recognition of bias in their estimates of effects. In particular, Cheval et al. note that existing RCT data on the effects of exercise and cognition are typically based on comparatively small sample sizes of 100 or less participants, and are thus more likely to yield biased outcomes. Furthermore, they do not account for the potential influence of cognition on physical activity (8).

To overcome these specific obstacles, the authors employed Latent Heritable Confounder Mendelian Randomization (LHC-MR) to assess the bidirectional relationship between physical activity and cognitive function. Included in their analysis were data from two large genome-wide association studies (GWAS) of activity measured by accelerometer in over 90,000 subjects, and cognitive function in over 250,000 subjects, respectively. As the authors state, one of the advantages of this approach is that the traits under consideration (activity and cognition) do not need to be assessed in the same overlapping sample (8).

Accelerometer measurements were used to derive three levels of physical activity: average, moderate, and vigorous. Adjusting for age, these measures were subsequently used in both standard Mendelian Randomization (MR) and LHC-MR analysis. In all three physical activity groups (average, moderate, and vigorous), standard MR methods revealed non-significant causal estimates of the effects of physical activity on cognition as well as the reverse effect of cognition on physical activity. Applying LHC-MR revealed a potential positive causal effect of both moderate and vigorous physical activity on improved cognitive function, with no apparent reverse causal effect, and no evidence of a heritable confounder. As the authors note, a somewhat surprising finding was that the coefficient (magnitude) of the causal effect of physical activity on cognitive function was greater in the moderate physical activity group than the vigorous activity group. Cheval et al. conclude that the difference in results obtained using a typical MR approach and the results of the LHC-MR analysis are due to differences in the statistical power of the methods. In addition, the authors emphasize that the results of the two methods are consistent in terms of their effect estimates, differing only in their statistical conclusions (8).

What is Mendelian Randomization?
Put simply, Mendelian randomization (MR) is a method of assessing the effects of a factor on an outcome measure, using data from non-trial designs and variation in genes with known functions. In some cases, randomized controlled trials (RCTs) are not practical or ethical, so observational studies must instead be used. Unfortunately, observational data may be skewed by confounding variables or reverse causation; MR can be used to overcome these limitations to derive meaningful and unbiased estimates of causal influence. In a typical RCT, subjects are randomly assigned to treatment or control groups; in MR, this assignment is done based on single nucleotide polymorphisms (SNPs). The advantage of using SNPs in this way is that they are determined randomly at birth, and are therefore not subject to subsequent environmental confounds.

What is Latent Heritable Confounder Mendelian Randomization (LHC-MR)?
Latent heritable confounder mendelian randomization (LHC-MR) is an extension of traditional mendelian randomization that accounts for the presence of unmeasured (latent) confounding factors that are both heritable and associated with the exposure and outcome of interest. LHC-MR attempts to provide a more robust causal estimation by effectively accounting for the unmeasured factors that may influence the relationship between the exposure and outcome.

As noted, this study was largely limited to white populations of European ancestry, and therefore extrapolation to other populations may not be valid. In addition, interpretation of the findings is limited by many of the complications inherent in studies evaluating the relationship between exercise and cognitive function, making comparison with previous studies difficult. For example, although this study used accelerometer data to objectively assess physical activity, many studies use self-reported activity levels. Similarly, subject age varied greatly across studies, and differed between the samples used to assess physical activity and cognitive function, respectively. Indeed, it is possible that age and other temporal factors may play an important role in any causal relationship (8).Accelerometer measurements were used to derive three levels of physical activity: average, moderate, and vigorous. Adjusting for age, these measures were subsequently used in both standard MR and LHC-MR analysis. In all three physical activity groups (average, moderate, and vigorous), standard MR methods revealed non-significant causal estimates of the effects of physical activity on cognition as well as the reverse effect of cognition on physical activity. Applying LHC-MR revealed a potential positive causal effect of both moderate and vigorous physical activity on improved cognitive function, with no apparent reverse causal effect, and no evidence of a heritable confounder. As the authors note, a somewhat surprising finding was that the coefficient (magnitude) of the causal effect of physical activity on cognitive function was greater in the moderate physical activity group than the vigorous activity group. Cheval et al. conclude that the difference in results obtained using a typical MR approach and the results of the LHC-MR analysis are due to differences in the statistical power of the methods. In addition, the authors emphasize that the results of the two methods are consistent in terms of their effect estimates, differing only in their statistical conclusions (8).

The Cilia study is an umbrella review of existing meta-analyses of published RCTs (7).  In contrast, the Cheval study is a Latent Heritable Confounder Mendelian Randomization (LHC-MR) study, and was precluded from inclusion in the Cilia study due to publication date, regardless of methodology (8). The Cilia study was based on pre-existing data, arriving at the conclusion that existing evidence did not conclusively support a causal relationship between exercise and cognition. Subsequent to its publication, the Cheval study adopted a different methodological approach, ultimately concluding that moderate and vigorous exercise do improve cognition. In effect, they are not directly comparable. Although superficially similar, and yielding seemingly incompatible results, the two conclusions are not necessarily mutually exclusive. It is possible that existing RCT evidence is insufficient to demonstrate that exercise improves cognition, and yet LHC-MR analysis supports the conclusion that moderate or vigorous exercise does in fact improve cognition in a specific population (7,8).

Ultimately both sets of authors identify the limitations of existing evidence, and advise caution in interpretation of these results, underscoring the need for further replication and study. When considering the implications of the findings, it is important to clarify how they are to be used. In regards to the question of whether existing evidence of the effects of exercise on cognition is sufficient to inform governmental policy, the answer is likely no; this is more apparent if the policies in question are to be applied to all populations. In the first study, no causal effect was reported, and in the second, although a causal effect was described, caution in the generalizability of the findings and the need for replication were emphasized. The answer is therefore still unclear, and is likely insufficient to cross the threshold of certainty required to inform policy at the governmental or societal level (7,8).

However, in regards to how this evidence may inform the decisions of the individual, the answer may be different. Exercise is good for you. The effects are well-documented, varied, and are demonstrably beneficial. Will it improve your cognitive function? Perhaps, but it will improve your health and well-being.  At the level of the individual, does the question of whether exercise conclusively improves cognition matter? To quote Cilia et al., “let us not forget the pleasure of doing something for its own sake. The value of exercising may lie simply in its enjoyable nature.”

1. Barbaresko J, Rienks J, Nöthlings U. Lifestyle indices and cardiovascular disease risk: a meta-analysis. Am J Prev Med. 2018;55(4):555-64. doi: 10.1016/j.amepre.2018.04.046
2. McTiernan A, Friedenreich CM, Katzmarzyk PT, et al. Physical activity in cancer prevention and survival: a systematic review. Med Sci Sports Exerc. 2019;51(6):1252-1261. doi: 10.1249/MSS.0000000000001937
3. Uusitupa M, Khan TA, Viguiliouk E, et al. Prevention of type 2 diabetes by lifestyle changes: a systematic review and meta-analysis. Nutrients. 2019;11(11):2611. doi: 10.3390/nu11112611
4. CDC (Centers for Disease Control and Prevention). Physical Activity Boosts Brain Health. Division of Nutrition, Physical Activity, and Obesity. Updated Februrary 24th, 2023. Accessed June 6th, 2023. https://www.cdc.gov/nccdphp/dnpao/features/physical-activity-brain-health/index.html#:~:text=Physical%20activity%20can%20help%20you,of%20cognitive%20decline%2C%20including%20dementia
5. WHO (World Health Organization). Physical Activity. World Health Organization. Updated October 5th, 2022. Accessed June 6th, 2023. https://www.who.int/news-room/fact-sheets/detail/physical-activity
6. Hecksteden A, Faude O, Meyer T, Donath L. How to construct, conduct, and analyze an exercise training study? Front Physiol. 2018;9:1007. doi: 10.3389/fphys.2018.01007. eCollection 2018.
7. Ciria LF, Román-Caballero R, Vadillo MA, et al. An umbrella review of randomized control trials on the effects of physical exercise on cognition. Nat Hum Behav. Published online March 27th, 2023. doi: 10.1038/s41562-023-01554-4
8. Cheval B, Darrous L, Choi KW, et al. Genetic insights into the causal relationship between physical activity and cognitive function. 2023;13:5310. Sci Rep. doi: 10.1038/s41598-023-32150-1