Cardiovascular fitness and structural brain integrity: an update on current evidence

Abstract

An aging global population and accompanying increases in the prevalence of age-related disorders are leading to greater financial, social, and health burdens. Aging-related dementias are one such category of age-related disorders that are associated with progressive loss of physical and cognitive integrity. One proposed preventative measure against risk of aging-related dementia is improving cardiovascular fitness, which may help reverse or buffer age-related brain atrophy associated with worse aging-related outcomes and cognitive decline. However, research into the beneficial potential of cardiovascular fitness has suffered from extreme heterogeneity in study design methodology leading to a lack of cohesion in the field and undermining any potential causal evidence that may exist. In addition, cardiovascular fitness and exercise are often conflated, leading to a lack of clarity in results. Here, I review recent literature on cardiovascular fitness, brain structure, and aging with the following goals: (a) to disentangle and lay out recent findings specific to aging, cardiovascular fitness, and brain structure, and (b) to ascertain the extent to which causal evidence actually exists. I suggest that, while there is some preliminary evidence for a link between cardiovascular fitness and brain structure in older adults, more research is still needed before definitive causal conclusions can be drawn. I conclude with a discussion of existing gaps in the field and suggestions for how they may be addressed by future research.

Introduction

An aging global population is creating an unprecedented need to preserve and prolong both physical and mental health [1–6]. Increases in average population age, as well as in longevity, have put extreme stress on social welfare systems; the growing size of adult populations is one of the largest factors associated with the major increase in age-related social and financial burden between 1990 and 2017 [3]. As of 2017, over 92 diseases have been identified as age-related, accounting for 50% or more of global burden among adults [1, 3]. Age-related diseases are defined as those with incidence rates that increase quadratically with age among the adult population, leading to progressive loss of physical and cognitive integrity [3, 7, 8].

Aging-related dementias (ARDs) are one category of age-related disorders that have received a lot of attention in recent decades [4, 7, 9]. As the greatest risk factor for developing ARD is age, the most effective means to reduce ARD burden and to control costs is to extend years of life lived free of disease and associated disability [1, 7, 8]. However, no cure for ARD exists. Thus, there is a critical need for research into preventative measures or interventions to delay or prevent onset and progression of symptoms. Of particular importance are large-scale efforts to identify effective interventions against age-related cognitive decline. Such decline precedes clinical diagnosis of ARD and has a disproportionate negative impact on health [5, 10–12].

One potential intervention for cognitive decline associated with ARD is improving cardiovascular fitness [2, 5, 6, 13]. While the advantages of higher levels of cardiovascular fitness pertaining to physical health are well documented—increased bone mass, increased mobility, increased quality of life, and decreased risk of cardiovascular disease [1, 14–16]—emerging evidence also suggests beneficial effects on brain and cognitive function [1, 5, 7, 15, 17–20]. Better cardiovascular fitness has been associated with better executive function and decreased rates of cognitive decline [11, 17–22]. Higher cardiovascular fitness has also been linked with superior structural integrity of the brain including greater cortical thickness and surface area, as well as larger subcortical gray matter volume in areas like the hippocampal formation, which supports cognition and memory processes affected in ARD [6, 17, 19, 23, 24].

The links between better structural brain integrity, cognitive ability, and higher cardiovascular fitness are critical connections as deterioration in the brain is linked with declines in cognitive ability. Thus, individual differences in structural brain integrity may provide a unique window on age-related cognitive decline and risk of ARD. ARD often has a gradual onset, with structural atrophy occurring years in advance of observable clinical and cognitive symptoms [8]. For example, brains of individuals who later develop ARD exhibit accelerated age-related atrophy of gray matter volume in subcortical regions that often precedes the onset of cognitive impairments [25]. A similar acceleration of cortical thinning and reduction in surface area have also been observed in the brains of those who later develop ARD [25–27]. As such, changes in the structural integrity of the brain may represent part of the procession towards future age-related cognitive decline and risk of ARD. Consequently, improving cardiovascular fitness as a means of mitigating these negative age-related outcomes through slowing or reversing the rate at which brain structures decline offers a promising avenue of research.

However, when appraising a novel mechanism that may hold promise as a potential way to intervene or postpone symptoms of ARD, such as cardiovascular fitness, the mechanism must first be tested and evaluated at multiple levels of research. An ideal research program consists of a series of studies that build towards causal proof and justify the expense of implementing randomized clinical trials. The first stage is often animal research. Animal models are advantageous due to the ease in which test subjects can be randomly assigned to conditions, such as the heightened ability to control dose and any potential confounds, the short lifetime of the test subjects, and the relative affordability of the mode of research.

As humans are complex and exist in complicated environments, proof within animals cannot be directly translated as proof in humans. Thus, the second stage is to see if the same general paradigm exists in humans, often taking the form of cross-sectional observational studies of human populations. A third stage might be to investigate the temporal aspects of the mechanism through longitudinal observational studies. The fourth and most important stage is to actively manipulate the variable of interest during randomized clinical trials. These types of studies, while expensive and time intensive, offer researchers a way to isolate a single mechanism through randomly assigning subjects to either an experimental condition or a control condition and then measuring the mechanism’s effectiveness in changing some phenotype from pre to post intervention. Importantly, in order for change to be causally attributed to the mechanism of interest, any clinically relevant change in phenotype should be primarily within the experimental group.

Published research evaluating the effects of cardiovascular fitness on the brain does exist at all four stages of the proposed research program, and recent articles often state the relationship as robust and reliable—despite mixed findings [2, 17, 21, 34–37]. However, the extent to which each stage has actually been sufficiently addressed remains unclear—in terms of both physical activity (for a recent review, see [13]) and cardiovascular fitness. This is especially true within human research. As the four proposed stages build on each other, any holes within early stages of research can lead to faulty assumptions that may undermine both experimental design and conclusions in later stages. As such, it is important to critically evaluate research touted as reliable evidence.

An additional problem than has impeded the progress of neuroscientific research into cardiovascular fitness and the brain is the lack of cohesion in research methodology. High methodological variability across studies makes it difficult to assess both the utility of exercise as a way to increase cardiovascular fitness and the efficacy of cardiovascular fitness in changing brain structure. This, in turn, hinders the ability to isolate promising future avenues for further research as well as the ability to translate findings for clinical purposes.

Thus, the goals of this narrative review are (a) to disentangle recent findings relating to aging, cardiovascular fitness, exercise, and brain structure within each stage of the research program discussed above (i.e., within animal, observational cross-sectional or longitudinal, and intervention research); (b) to ascertain whether each stage of research has been sufficiently addressed; and finally, (c) to identify existing gaps in the field that could be addressed by future research.

Aging and the brain

Aging, decline, and risk of ARD

Structural decline in the brain, like associated decline in cognitive ability, is a normal part of the aging process. As the brain ages, it naturally deteriorates. Neural deterioration is due, in part, to cellular mechanisms similar to those that affect other organ systems throughout the body [38]. However, the brain is additionally subject to other aging processes, such as lower rates of cellular regeneration and accelerated breakdown of the blood–brain barrier, that make it particularly susceptible to age-related decline [31, 39, 40]. Because of this, individual differences in brain structure may be uniquely sensitive mechanisms for isolating predictive patterns of aging—normal or otherwise. For example, decreases in cortical thickness, surface area, and subcortical gray matter volume—three commonly used structural brain measures used in aging research (Fig. 1)—are associated with normal, age-related decreases in executive functioning, physical dexterity, and memory. Accelerated or exaggerated rates of decline over time in these facets of the brain are associated with clinically significant decreases in functional domains and greater overall risk of ARD [8, 41].

Fig. 1.

Anatomical representations of three main structural brain measures used in aging research. a Cortical thickness is a measure of the distance from the outer boundary of the white matter to the outer edge of the pial surface [28]. Thickness peaks during late childhood and declines as age increases, with rates sometimes accelerating after 50 years [29, 30]. Some research suggests that age-related decreases in cortical thickness are related to changes in the cellular integrity of the cortical mantle, although exact causal factors are still unclear. b Surface area refers to the outer gray matter surface of the brain and peaks during late adolescence before declining in a linear fashion. Surface area is highly correlated with total brain volume, and changes in surface area may be related to the size of intracortical elements or to volumetric changes in local subcortical regions, such as adjacent white matter volume [31]. c The hippocampal formation is a seahorse-shaped subcortical structure located in what is commonly referred to as the limbic region. Relative to other regions of the brain, hippocampal formation volume peaks late in life with some evidence showing increases as late as 40+ years of age [8, 32]. However, relative volumes of hippocampal subfields show independent heterogeneous trends in age-related atrophy and may reflect distinct neural processes [33]. Image created by Annchen Knodt

If rates of decline in the brain over time are represented as a continuum,1 normal brain aging refers to indications of age-related differences in the rate of decline in brain structure within the general population, concurrent with declining cognitive function but not necessarily clinically significant impairment [30]. Accelerated rates of structural deterioration—associated with less healthy aging trajectories, such as increased risk of ARD and higher incidence of clinically significant cognitive impairment—can be considered disordered neural aging trajectories [8, 30, 41] (Fig. 2). It should be noted that while the following review focuses on research concerning cardiovascular fitness and aging-related changes in gray matter, aging-related microstructural decline in white matter also contributes significantly to risk of cognitive decline and ARD and may also be associated with changes in cardiovascular fitness (e.g., [43]).

Fig. 2.

A stylized illustration of simulated data showing individual differences (thin grey lines) in aging-related rates of atrophy in the brain throughout adulthood. Normal aging trajectories are what, on average, the majority of the population experiences over time (thick red line). However, a subpopulation experiences comparatively accelerated rates of atrophy in the brain, leading to increased risk of ARD (dark purple line). According to this model, measuring change in brain structure over time may help identify those individuals with accelerated or pathological brain aging compared to the rest of the population; the relative acceleration of brain deterioration may be indicative of higher risk of future clinically significant impairment.

Cardiovascular fitness

What is cardiovascular fitness?

At the macro level, cardiovascular fitness is a measure of the physical work capacity of an individual, or the maximum rate at which the body can utilize oxygen [16, 44–46]. It reflects how efficiently the respiratory and circulatory systems are providing oxygenated blood to relevant musculature during active moments [44, 47]. In practice, cardiovascular fitness is most often estimated through calculating the volume of oxygen consumed during maximal aerobic exercise, a measurement referred to as VO₂Max [16, 44, 48]. VO₂Max is usually measured by monitoring respiratory and/or heart rates while participating in a standardized effort test on an exercise machine (e.g., a friction-brake standing bike or a treadmill) [1, 45, 49]. Measures of VO₂Max are considered a reflection of the combined working of cardiovascular, respiratory, and neuromuscular systems during maximal whole-body exercise and are often considered the gold standard for measuring cardiovascular fitness in health and fitness fields of research [16, 44].

VO₂Max declines as we age at a rate of approximately 1% per year: by the age of 65, VO₂Max capacity is 30–40% lower than it was as a young adult [48, 50]. Age-related declines in VO₂Max are, in large part, attributable to coinciding decreases in max heart rate [44, 50]. As max heart rate decreases, higher stroke volume (the volume of blood pumped by the ventricles with each heartbeat) is required to maintain cardiac output. Over time, stroke volume is less and less to compensate, leading to lower cardiovascular performance and lower fitness capacity [50]. Decreases in cardiovascular fitness can contribute to an increased risk of disability, loss of independence, and reduced quality of life [1, 16, 49]. Declines in general vascular health are associated with increased risk of ARD [7]. Research suggests that an effective way to improve vascular health as we age is through lifestyle modifications such as habitual aerobic exercise [7, 50]. In fact, some research suggests that of all the modifiable lifestyle risk factors for ARD (e.g., diabetes, hypertension, obesity, smoking, depression, physical and cognitive inactivity), increasing the proportion of the general population who are regularly physically active by 25% would have the highest statistical impact on rates of ARD [7].

However, VO₂Max is not an infallible estimate of cardiovascular fitness. For example, intraindividual day-to-day variation for VO₂Max measurement is between 4 and 6% ([51, 52], 2001); in people with various diagnoses, such as multiple sclerosis [51] and cardiomyopathy [53, 54], this variation can be up to 10% [51, 55, 56]. To account for this variation, some researchers suggest that clinical interventions need to see an increase in VO₂Max of at least 10% for it to be interpreted as real, meaningful change [51].

In addition, as standard VO₂Max measurements are normalized for body weight, even small shifts in weight can alter scores without any actual change in fitness [50, 55]. These measurement vulnerabilities indicate that cross-sectional studies utilizing VO₂Max may be especially vulnerable to measurement error leading to increased noise during analyses. In addition, the results of a VO₂Max test are only valid if an individual is able to attain their maximum effort without fatiguing first or being limited by other problems, such as musculoskeletal impairment which increases in prevalence as age increases [57]. Finally, the lack of general procedures and specific test selection can lead to understressing or overstressing participants, which can, in turn, result in invalid conclusions because of ceiling or floor effects [56, 57].

While a direct measurement of VO₂Max, using some sort of graded maximum effort test, is considered the ideal way to measure cardio fitness, this is not always feasible due to expense of equipment, time commitment required, or health concerns. For example, the maximal aerobic exercise test to measure VO₂Max often requires a physician on site due to the physical strain the test can put on participants. These considerations are especially necessary when studies involve populations who are more vulnerable to physical stress, such as older populations that are frequently limited by cardiopulmonary, musculoskeletal, and neuromuscular impairments. Maximal testing is often contraindicated or of limited value [57], adding a level of difficulty to data collection that researchers are not always able to overcome and additional complications for interventions targeting clinical populations. Thus, cardiovascular fitness is often estimated or derived using other methods.

Alternative methods are used to estimate cardiovascular fitness. Estimates of fitness can be generated from other physiological measures collected at rest, through measures of daily activity or through self-report questionnaires. These indirect methods used to estimate cardiovascular fitness can vary considerably in both the type of information and the amount of detail collected. In addition, the way the data are manipulated can differ, resulting in estimates of fitness that are based on very different metrics. These issues have broad repercussions when assessing results from fitness studies and introduce a considerable amount of noise into analyses, which makes cross-study comparison difficult.

Cardiovascular fitness and exercise are not the same

Cardiovascular fitness, daily activity, and exercise are routinely conceptualized as facets of a greater fitness construct. However, despite how often these terms are grouped together, cardiovascular fitness, activity levels, and exercise are not the same; in fact, sometimes they are unrelated entirely. Despite being most commonly associated with exercise and daily activity, aging-related decline in cardiovascular fitness can be due to a broad range of factors outside of decreases in physical activity level. Changes in maximum cardiac output [58], overall cardiac function [59], blood pressure and vascular aging [60], arterial thickening and stiffness (Liu [61, 62]; Scuteri [63]), and reduced muscle mass, all aspects of aging, are also relevant elements to consider and all contribute to declines in cardiovascular fitness and health over time and may also contribute to aging-related structural decline in the brain [16, 45, 48, 49].

In other words, while a number of studies have demonstrated that normal, age-related decline in VO₂Max can be at least partially, if not completely, attenuated by physical activity (for a recent review, see [64]), some physiological indexes of overall cardiovascular health are not influenced by activity but rather are under the control of different aging processes that may also be contributing to risk of ARD [16, 45, 48]. Therefore, when the goal is to support the efficacy of exercise on health improvements through targeting changes in cardiovascular fitness, associations between changes in VO₂Max and changes in phenotypes of interest (e.g., brain structure) are insufficient; in these cases, to best account for exercise effects, analyses must include measures of physical activity [16, 45]. Likewise, studies that only measure time spent exercising cannot be generalized to cardiovascular fitness.

Unfortunately, challenges remain even when studies incorporate both types of cardiovascular fitness and exercise measures to directly test the associations between cardiovascular fitness and physical activity as well as their corresponding effects on relevant phenotypes (e.g., changes in brain structure). One main issue is that exercise comes in many forms: it can include components of anaerobic or strength training, balance training, aerobic or endurance training, flexibility, etc. [16, 65]. These different types of physical activity are associated with different physiological changes, including cardiovascular fitness [7, 66].

Broadly speaking, aerobic training (as opposed to anaerobic or strength training) is traditionally considered to be the best form of physical activity through which to increase cardiovascular fitness, although its efficacy depends on both the intensity (e.g., difficulty level of exercise sessions) and length (both the number of sessions per week and the overall span of time spent consistently training) of the training regime [16, 17, 20, 45, 48, 65]. Regularly performed aerobic exercise is thought to increase cardiovascular fitness through two different mechanisms. First, aerobic exercise affects central pathways that control oxygen delivery to skeletal muscle by increasing both cardiac output and stroke volume [16, 48]. Second, aerobic exercise improves the capacity of skeletal muscle to extract and use delivered oxygen during physical movement [16, 48]. However, all aerobic exercises are not created equal. While some research suggests that it is the magnitude of change in cardiovascular fitness, rather than the exact exercise dose or type, that best predicts positive downstream changes, there is no gold standard for the exact quantity and quality of exercise necessary to improve cardiovascular fitness [17, 20, 49, 67].

The effects of heterogeneity within fitness intervention methodology and implementation become even more complicated when looking at secondary associations such as cognitive decline and brain structure in older populations. Differences in effect sizes or the lack of significant outcomes from anti-aging interventions targeting changes in brain structure could be due to sample characteristics; to insufficient intensity of, adherence to, or duration of the physical activity utilized (insufficient dose); or to other confounds, rather than to a lack of efficacy. Thus, the lack of exercise program standardization in research makes comparisons of associations between cardiovascular fitness, exercise, and age-related changes in brain structure more difficult to assess and more vulnerable to error. For these reasons, studies that took steps to address these issues in their methods, as well as those that compared the efficacy of different possible activities in an effort to ascertain a hierarchy of effective methods within cardiovascular fitness interventions, will be emphasized below.

Cardiovascular fitness and brain structure

Stage 1: animal models

Animal models are organisms with induced pathological, physiological, or behavioral processes that closely parallel similar conditions in humans [66, 68]. In neurofitness research, the most commonly used organisms are rats or mice as their brain development bears some similarity with that of the human brain [68, 69]. Additionally, evidence suggests that their physiological and behavioral responses to exercise training mimic those of humans [66, 69, 70]. Rodents also have the benefit of short gestation periods and many offspring [69].

Rodent models offer a variety of advantages over human studies for initial investigations into associations between the brain and behavior. To start, the relative affordability (and proliferation) of the laboratory animals makes it more feasible for researchers to gather data from large sample sizes [69]. Additionally, as researchers have, to a certain extent, complete control over the environment and lifestyles of the laboratory rodents, animal models have more causal leverage. It is much easier to ensure random assignment between conditions, to control potential confounding factors, to increase measurement precision, and to standardize dosage in experimental conditions. In addition, the shorter lifespans of rodents allow researchers to directly measure changes in their brain within short stretches of time.

However, the very reason these advantages exist severely limits the translatability of any findings as associations between cardiovascular fitness, exercise, and animal brain structure are only relevant if relatable effects exist within humans [68, 71]. Human beings are far more complicated than animals; they do not exist in controlled environments, they do not adhere equally to or expend equivalent effort in exercise paradigms, and they interact with a wide variety of potential confounds on a daily basis. There are also aspects of neurological structure unique to the human brain that limit anything other than base comparisons. These and other differences make them starkly different subjects of research. While animal research and rodent models offer an important first step, burden of proof rests with human research. Thus, research focus should transition from animal models to human models as quickly as possible. With this in mind, the following summary of existing proof of associations between cardiovascular fitness, exercise, and brain structure in laboratory animals within the published literature will be brief.

General findings

The majority of findings within animal literature focus on changes in structural integrity within the hippocampal formation as a consequence of voluntary exercise [36, 66, 70–75]. Specifically, extensive research seems to support the utility of exercise in increasing rates of neurogenesis, cell proliferation, and dendritic spine density in the mouse brain leading to higher gray matter volume and density in the hippocampus [66, 70, 74–76]. Associations between exercise and other brain regions are less commonly published. This is due, in part, to the relative size and simplicity of the rodent brain. As researchers are limited in both the number and precision of postmortem slices, they often have to prioritize a priori a single region of interest [76].

The association between voluntary exercise and increased hippocampal volume has been found across the lifespan of mice, suggesting that benefits (to a mouse) from increasing exercise are not reliant on any particular stage of development [69–71, 73, 75, 76]. That these associations are found across the lifespan is important because the human brain’s ability for neurogenesis is not stable—neurogenesis rates peak early in life and then decline over time [40]. However, the human hippocampal region, in particular, maintains higher capacity for neuronal and dendritic generation even in the adult brain, suggesting that the hippocampus may be a region more receptive to targeted interventions even in older populations [40, 71].

Limitations

Outside of the general limitations that exist when considering evidence from animal models discussed above, there was also a surprising amount of methodological heterogeneity in experimental paradigms. This is most evident in exercise training paradigms induced in the mice—both mode and dose [66]. There are several established activity options in rodent literature, but not all outcomes are comparable (for a more comprehensive overview, see [69]). For example, treadmill running offers the most precise and reliable control for targeted exercise-induced changes in rodent cardiovascular fitness and is considered the preferred modality. However, motivating animals for prolonged sessions can be challenging and forced running can lead to physiological and psychological stress (and the accompanying neural stress response) that may confound results [69]. Still, despite some methodological variability, changes in the rodent brain due to exercise seem to be both a robust and replicable finding. The question then stands, can this effect be found in humans?

Stage 2: cross-sectional studies

Observational cross-sectional studies are almost always the first step when transitioning from investigating a phenotype in animals to investigating it in humans as they are easy to run. Participants only need to commit to one time point, making recruitment and subsequent data collection easier, cheaper, and more accessible to the general population. In addition, it is easier to attain larger sample sizes and to more inclusively recruit from diverse populations. However, cross-sectional study designs that rely on observational data alone are poor at leveraging causal information and establishing causal effect, in part due to a lack of randomization and variable control within study parameters and subsequent analyses. For example, observational cross-sectional studies have found brain differences between adults who have high cardiovascular fitness compared to those who have low cardiovascular fitness, suggesting that fitness may buffer against aging-related decline in the brain [23, 77–79]—a phenomenon often referred to as neuroprotection.

However, it is possible that a number of unstudied factors may also be influencing or may even attenuate these findings, such as variables relevant since childhood. In other words, while better adult cardiovascular fitness may be correlated with increased structural brain integrity in adulthood, this correlation cannot be assumed to indicate causation. Instead, early life factors like childhood brain integrity, childhood IQ, or childhood SES—often referred to as neuroselective influences—may underlie the associations observed later in life [1, 5, 35]. According to the neuroselection model, children who begin life with better cognitive function or higher structural brain integrity may have made certain choices (e.g., healthy diet and nutrition, regular exercise, further educational attainment) or have had certain advantages (e.g., better school systems, access to healthier foods, better healthcare) that have carried forward to adulthood. As adults, they may obtain higher-salary occupations, have more leisure time for recreational physical activity, and may be more health-minded, resulting in higher levels of cardiovascular fitness. This scenario would result in a cross-sectional correlation between brain structure and cardiovascular fitness in adulthood; however, the association between cardiovascular fitness and the brain started with higher brain structure in childhood and that influence has merely carried forward.

Cross-sectional studies are snapshots of the brain and body at a single time point. As such, it is inherently difficult to discern temporal order—i.e., if differences in brain structure led to differences in cardiovascular fitness, or vice versa. Additionally, cross-sectional measures of cardiovascular fitness ignore the possibility of temporal trends within relationships, such as changes in cardiovascular fitness across time, which may have on associations with age-related changes in brain structure. Given that the most coherent way to assess individual differences in neurological aging could be through measuring change over time, the constrained temporal resolution further inhibits the relevancy of cross-sectional findings [42]. Thus, results from observational cross-sectional studies need to be interpreted with caution.

It should also be noted that inferences drawn from observational cross-sectional data about changes in some phenotype of interest are based on the assumption that the sample lacks any cohort effects and that any sample bias that may exist is not correlated with age or other factors [29]. For example, older age groups recruited by convenience may be biased due to inherent differences in the type of people who are motivated to volunteer to participate in studies. Superiorly functioning older adults may also be more likely to volunteer or to be accepted into studies with strict exclusion criteria which can lead to a sample skewed by recruitment bias [8]. There is also the issue of selective survival—those with the poorest cardiovascular fitness may have already died, and thus, the population no longer reflects this variance. Issues like these can make it hard to determine whether the associations observed in observational studies regarding cardiovascular fitness and brain reflect real ongoing change within individuals or if they are actually driven by methodological artifacts [8, 80].

General findings

There is a large amount of published literature reporting evidence of a positive cross-sectional association between cardiovascular fitness and brain structure. The majority of positive results are localized within the frontal and/or medial temporal lobes [47, 77–79, 81–88]; however, other regions, such as parts of the parietal lobe, the precuneus, and the striatum, have also shown positive associations with fitness [77, 87, 88].

Research has found associations between higher cardiovascular fitness in older individuals and better brain integrity in areas other than the prefrontal cortex and the hippocampal region, although these findings are less common. These areas include the caudate nucleus [89], regions associated with the default mode network [90], and global measures of brain volume and cortical thickness [2, 91, 92]. Interestingly, only one observational cross-sectional study [34] published null results, a surprising result considering the modesty of effect sizes within the literature. The lack of null findings could be indicative of preferential publication for cross-sectional fitness and brain structure studies that find significant results—known as the file-drawer problem.

The majority of studies utilize samples of healthy older adults free of neurological diseases, head injuries, depression, or cognitive impairment. However, given that interest in associations between cardiovascular fitness and the brain are driven, in part, by the hope for clinical translatability, the question of whether these associations exist in populations with more disordered aging trajectories remains pertinent. There does seem to be at least preliminary evidence that the relationship between increased cardiovascular fitness and higher structural integrity in the brain is even more pronounced in individuals with certain age-related phenotypes, such as current symptoms of ARD or relevant risk biomarkers [2, 78, 81, 87].

For example, associations between greater levels of cardiovascular fitness and higher distributed gray matter volume have been found in older adults diagnosed with early-onset Alzheimer’s disease (AD), but not in comparable healthy controls [2, 78]. Similar results have been found in populations with mild cognitive impairment and physiopathological biomarkers indicative of future risk of ARD (e.g., low levels of Aβ42,2 BDNF, and high tau protein counts within the cerebral spinal fluid) [81, 87]. While this evidence remains provisional as published studies are limited and sample sizes are quite small, it is of note that dementia symptom severity was not associated with VO₂Max, suggesting that these ARD-specific findings were not merely due to individuals with more severe dementia also having lower levels of cardiovascular fitness [78].

Limitations

One of the most obvious limitations within the cross-sectional literature is the extensive variability in study parameters. While many studies collected VO₂Max—the gold standard—as an estimate of cardiovascular fitness (e.g., [2, 47, 67, 84, 86–89]), an equivalent number of studies utilized alternative measurement criteria that assessed different facets of physical activity (e.g., [34, 77, 79, 81–83, 85, 90, 91]). As discussed in the section “Cardiovascular fitness and exercise are not the same,” physical activity and cardiovascular fitness, though correlated, are not the same thing.

For example, utilizing a step monitor to calculate daily activity [79] or a questionnaire to assess engagement in a running, walking, or jogging exercise program over the previous 10 years [82] can both be considered measuring some aspect of general fitness. However, average daily steps of an individual are not necessarily comparable to estimates of an individual’s average engagement in recreational exercise. Research suggests that people with active jobs (e.g., nurse, waiter, construction worker) do not necessarily have higher levels of cardiovascular fitness and, on average, may actually engage in lower levels of leisure-time exercise [93–95]. Further, both measures of daily steps and average recreational exercise per week may be entirely unassociated with measures of cardiovascular fitness. For example, individuals can increase their average weekly physical activity without seeing any corresponding changes in their cardiovascular fitness and vice versa [96–98]. Thus, studies that measure VO₂Max against average steps taken or self-reported physical activity questionnaires may be measuring different facets of fitness, limiting any possible cohesive summary.

Stage 3: longitudinal studies

Longitudinal study designs (or prospective/retrospective studies) do address some of the issues seen in observational cross-sectional research. For example, longitudinal observational data offer the benefit of measuring a single variable across multiple time points. Access to multiple time points of data reduces the effects of measurement error and noise in analyses, allowing for potentially more accurate effect sizes. Discrete time points rarely capture the nature of human change and growth. Therefore, repeat measurements over time are often a more ecologically valid way of constructing a coherent picture of a phenotype. Longitudinal data also enable researchers to investigate individual differences within subjects over time and establish estimates of temporal order. Access to measurements of change adds additional levels of detail in relationships (e.g., is current cardiovascular fitness associated with brain structure or is it actually more important to have higher sustained levels of cardiovascular fitness for a certain period of the lifespan?).

It is important to note, however, that having multiple time points does not necessarily offer additional leverage towards causal interpretations especially when using observational data. Longitudinal designs can, to some degree, manipulate observational data to approximate a test of cause and effect. For example, researchers can match characteristics in two groups at one time point and then test differences in a specific phenotype at a later time point with the goal to isolate change in a single variable while holding others constant. However, this is, at best, an approximation as the lack of controlled randomization means there is no way to ensure unmeasured variables do not exert influence on or potentially change concurrently with the phenotype of interest, confounding the relationship.

Another relevant issue with longitudinal designs is, as discussed above with cross-sectional studies, their vulnerability to selective recruitment, which can increase bias and limit the generalizability of findings. As participants must return for multiple waves of testing, there are also the possibilities of selective attrition and practice effects [99]. Such effects could have undue influence on the targeted phenotype, resulting in false estimates of true effect sizes. Finally, few longitudinal studies sample the entire adult agespan, which impedes estimations of rates of change as a function of age [8].

General findings

Positive findings from published longitudinal studies of cardiovascular fitness and brain structure follow the same general trends seen cross-sectionally: the majority of associations reported seem to be localized within frontal and medial temporal regions [67, 100–102] with less common associations being found between cardiovascular fitness and changes in global measures of brain volume [101, 103]. In addition, similar to the cross-sectional literature, published null findings are few and far between. Again, considering the modest effect sizes seen in longitudinal studies of fitness and brain structure, the lack of null results is surprising.

The most novel information offered by observational longitudinal studies is the ability to analyze changes in cardiovascular fitness and brain structure across time. In other words, how measures of fitness at one time point can be associated with brain structure years later. There is some evidence that better cardiovascular fitness in midlife is associated with higher structural integrity over time (e.g., [100, 101, 103]). However, the lack of baseline MRI measures in many of these studies makes it impossible to discern whether those with better brain structure measures in later life may have started with better structural measures at baseline regardless of cardiovascular fitness status. Thus, findings could be explained by neuroselection and reverse causation.

Studies that do have multiple waves of MRI data can leverage both changes in cardiovascular fitness and change in brain structure over time in analyses, helping researchers to answer questions about temporal aspects of associations, such as whether cardiovascular fitness (and/or change in cardiovascular fitness) is associated with rates of atrophy in brain structure as well as with cross-sectional measures of brain integrity. Such studies—although few in number—have found evidence to suggest that both higher average weekly physical activity and upward change in cardiovascular fitness are independently associated with lower or reversed rates of atrophy in medial and temporal regions of the brain over time, suggesting that lower fitness may be a risk factor for faster age-related structural decline [67, 102].

It is important to note that longitudinal findings do diverge somewhat from cross-sectional findings. Whereas published cross-sectional studies report associations between overall brain integrity and cardiovascular fitness, most published longitudinal studies report associations only between rates of atrophy in the brain and cardiovascular fitness over time—with no significant associations between cardiovascular fitness and overall brain integrity. The discrepancy between cross-sectional and longitudinal findings can be interpreted in several ways. It could be that both aspects of cardiovascular fitness (both contemporary and change over time) are important to consider when investigating effects on the brain [67, 102]. However, it is also quite possible that some of the findings from cross-sectional designs were an artifact of reverse causation as they did not appear longitudinally in more rigorous change analyses, which were able to utilize baseline measures as a control.

Limitations

As seen in the cross-sectional literature, there is extensive variability in study parameters between observational longitudinal studies extending across all aspects of methodology. Temporal characteristics, such as the time between waves of data collection, differ drastically from study to study. For example, one study collected data at baseline and 2 years later [67] while another study looked at data collected at baseline and at 21 years after [101]. In addition, the characteristics of the sample population as well as each study’s method for measuring cardiovascular fitness showed considerable disparity.

Outside of measurement variability, as seen in the cross-sectional section, no studies have addressed the possibility that potential neuroselective confounds could be the underlying mechanism through which both brain structure and cardiovascular fitness change over the lifespan. For example, even though some longitudinal cohorts did have multiple MRI scans, no study accounted for childhood measures of brain integrity [23, 101, 103]. This is important because the end goal of prospective aging studies is to identify factors during early life to midlife that may increase some facet of brain reserve, potentially moderating the expression of age-related brain atrophy. Without controlling for confounds like childhood brain integrity, there stands the possibility that any atrophy or structural decline seen later in life may be due to structural anomalies present from the start of life.

This is nicely illustrated in a recent study by Belsky et al. [1]. Motivated by recent findings linking higher cardiovascular fitness with better cognitive ability, Belsky et al. [1]B6 looked to replicate the association in a robust, population-representative sample of midlife adults. As expected, they were able to find an association between adult IQ and cardiovascular fitness at age 38. However, Belsky et al. [1]B6 also had access to longitudinal data on the cohort, allowing them to test whether the relationship found between higher adult IQ and higher cardiovascular fitness may just be a downstream effect of having higher IQ during childhood. The addition of childhood IQ to their analyses eliminated the significant association between adult IQ and cardiovascular fitness, suggesting that rather than adult cardiovascular fitness having a neuroprotective effect on adult IQ, both higher adult cardiovascular fitness and higher adult IQ may be attributable to neuroselection. These findings are relevant for associations between cardiovascular fitness and brain structure, as both are independently associated with neuroselective variables such as childhood IQ.

A recent study by our research group of the same population at age 45 found more promising results when looking at brain structure rather than cognition. They tested associations between cardiovascular fitness, the rate of change in cardiovascular fitness throughout adulthood, and various measures of brain structure at age 45. We found that better cardiovascular fitness at age 45 was associated with thicker cortex in frontal and temporal lobes and smaller hippocampal fissure volumes at age 45. These associations were unaffected by the inclusion of childhood IQ in the analyses. However, as they did not have access to childhood brain imaging, which is a more relevant neuroselective variable when assessing adult brain structure,3 any causal associations drawn by them (and observational longitudinal studies in general) about the neuroprotective effects of cardiovascular fitness on brain structure over time (particularly studies lacking multiple imaging time points) must be approached with caution.

Stage 4: intervention studies

Intervention study designs, or randomized control studies, are often considered the best option for establishing causal relationships. Ideally, these allow for isolating and testing specific mechanisms that may lead to change in a phenotype of interest (e.g., testing whether manipulating cardiovascular fitness in older adults through exercise leads to changes in structural integrity in the brain). It is a more statistically rigorous way to isolate the true effect of the phenotype of interest—fitness—from other possible confounding mechanisms.

However, while randomized control studies are the most informative method in measuring what affects change over time, they are, by no means, infallible—especially when using human subjects. Studies often fall short when it comes to unbiased randomization between conditions; the absence of true randomization inhibits the ability to test causation. In addition, as with cross-sectional and longitudinal studies, experimental studies are also vulnerable to more general methodological issues such as recruitment bias. For example, a clinical intervention trial recruiting from a population diagnosed with ARD faces the possibility that individuals who respond to community recruitment efforts may be a biased representation (do not suffer from mobility issues, are more self-sufficient, have reason to be worried about cognitive decline, are not in hospice care already, etc.). Intervention retention rates can suffer from attrition bias as well; individuals who choose to remain enrolled in and end up finishing the clinical trial may have certain characteristics in common. Finally, intervention studies often have limited follow-up, making it difficult to establish if positive results resulting from an intervention have any long-term efficacy.

General findings

Even more so than in the cross-sectional or longitudinal literature, findings from published intervention studies trend towards regional specificity; cardiovascular fitness effects were often isolated within the hippocampus [19, 23, 37, 104–108]. It is unclear whether such regional specificity is because intervention studies tend to target changes in the hippocampus as a region of interest or if the focus was post hoc—i.e., driven by their results. In other words, the bias towards finding isolated structural changes in the hippocampus in intervention studies could be due to a priori focus on the hippocampus, with few researchers looking elsewhere in the brain. Alternatively, the pattern may suggest that the hippocampal region may have increased sensitivity to changes in cardiovascular fitness.

Many intervention studies investigating fitness-related changes in the hippocampus measure regional gray matter volume as the outcome variable of interest. However, there are some alternative physiological variables within the hippocampus that may offer additional information regarding the neurobiological foundation of changes in overall gray matter volume [37, 105, 106]. For example, measurements of mean diffusivity in hippocampal tissue offer fine-grained information about the microstructural integrity of the hippocampal tissue itself [105, 109]. Mean diffusivity is an estimate of the diffusion of water molecules through tissue, and low diffusivity is indicative of gray matter tissue that is more densely packed, suggesting stronger structural integrity. There is evidence that increases in VO₂Max are associated with decreases in mean diffusivity in the hippocampus [105, 109]. As decreases in mean diffusivity are associated with accompanying increases in overall hippocampal volume, this suggests that increases in total hippocampal gray matter volume due to exercise may be, in part, driven by an increase in integrity in the tissue and in cell membranes, which operate as diffusion barriers [105].

In a similar vein, some research suggests increases in measurements of blood flow, or perfusion, within the hippocampus are associated with increases in aerobic fitness [37, 106, 110, 111]. One hypothesis is that aerobic fitness may influence microcirculation and cerebral vasculature, which can have regional effects on delivery of metabolites, oxygen, and nutrients, leading to increased rates of neurogenesis [106, 110, 111]. Changes in hippocampal perfusion are also closely linked with bilateral volumetric changes in the hippocampus [106]. However, exercise-related improvements in perfusion may be age-dependent as the brain’s capacity for vascular plasticity declines with age. This could have implications on the efficacy of such interventions in the later stages of life such that older individuals who exhibit reductions in hippocampal volume may not see robust increases in perfusion, despite increasing exercise, due to age-related neurobiological limitations [106].

Despite the main focus in intervention research on changes in the hippocampal region, there have been findings in other areas of the brain. For example, studies have found that individuals who engaged in a long-term, low-intensity cardiovascular fitness intervention showed greater cortical thickness or gray matter volume in the prefrontal cortex post intervention than comparative controls [112, 113]. However, rather than any increase in cortical thickness or gray matter volume from baseline, those in the exercise condition had lower rates of atrophy. In other words, all participants saw decreases in cortical thickness in the prefrontal cortex and in overall gray matter volume over the course of the intervention—those in the exercise condition just declined at a slower rate [112, 113]. Thus, long-term, low-intensity exercise may help to preserve cortical thickness in the prefrontal cortex over time; however, it is unclear if cardiovascular fitness can altogether prevent or reverse cortical atrophy.

That exercise and cardiovascular fitness may preserve rather than increase brain volume supports longitudinal evidence that increases in cardiovascular fitness are associated not with positive changes in brain integrity, but rather with lower rates of atrophy or decline over time ([23, 107, 112, 113]). These findings suggest that cardiovascular fitness may act as a buffer against future decline but may not be an effective mechanism through which to reverse atrophy [23, 107]. This is an important distinction as it would mean that the potential efficacy of cardiovascular fitness interventions is limited to slowing or preventing future brain atrophy progression as opposed to offering a way to reverse atrophy that has already occurred.

As there is yet to be any definitive evidence determining which exercise is best for targeting changes in cardiovascular fitness, some intervention studies look to find answers through comparing the efficacy of different types of exercise. For example, whether a dance intervention consisting of a 90-min dance class twice a week had more of an effect on brain structure than other intervention types consisting of an equivalent amount of time spent in general aerobic fitness classes [114] needs to be determined, comparing groups doing cardiovascular exercise to groups doing balance and coordination training [108], or anaerobic resistance training [104], or even splitting training groups by intensity level [105].

While studies do occasionally find clear differences between different types of exercise groups [104, 114], these findings tend to be inconsistent and inconclusive. Instead, it appears that the most consistent predictor of positive structural outcome is increases in cardiovascular fitness regardless of assigned condition [105, 108, 114]. The lack of differentiation in outcomes between conditions is also seen in intervention studies with a single exercise condition and a stretching/flexibility control [19, 106, 107, 115]. In other words, in many intervention studies, if any participant’s cardiovascular fitness improves, regardless of assigned condition, they were more likely to show positive changes in brain integrity post intervention.

Thus, in line with current research trends, it does not seem to matter which type of exercise contributed to said change in cardiovascular fitness, just that positive change occurred, suggesting that the type of exercise individuals participate in matters only insofar as how effective the exercise is at changing their overall cardiovascular fitness [97, 116]. However, it may also be an indication that exercise might not be the most effective way to achieve changes in cardiovascular fitness. There are a variety of ways to effect cardiovascular fitness aside from increasing physical activity such as a healthier diet (e.g., [117]) or the addition of blood pressure medications (e.g., [118, 119]). In certain clinical populations where exercise is counter indicated, such as individuals with type 2 diabetes (e.g., [119]) or heart failure with preserved ejection fraction (e.g., [120]), such non-exercise interventions may be necessary in place of exercise interventions or to improve physical health to the point that exercise is no longer counter indicated.

Unlike in the cross-sectional and longitudinal fitness literature, null or inconsistent findings in the intervention literature are less rare (e.g., [121]). However, the plethora of null findings could be due to the fact that many of them are not published as such. A lack of difference between outcomes in the experimental condition when compared to the control condition is, in effect, a null finding. Such an outcome can be an indication that the isolated mechanism of interest may not have any effect on the targeted phenotype (e.g., cardiovascular fitness, or changes in cardiovascular fitness, may not independently affect structural brain integrity in older adults). Even if significant findings are reported in a study, if the findings are not specific to the experimental group, the study results still suggest evidence of a null main effect. Some intervention studies fail to explicitly state this.

However, it should be noted that a lack of group differences is rarely simple to interpret as the outcome can be influenced by a variety of confounds, making it difficult to ascertain if a null finding is truly a lack of effect or if it is actually due to methodological issues. Examples of this seen in the intervention studies discussed include issues such as non-random assignment to conditions [113], non-compliance to exercise protocol [104], biased recruitment or attrition [107], small sample sizes [115], or a lack of any control group altogether [105, 115]. Such methodological vulnerability as well as the significant heterogeneity in study parameters across published intervention studies hinder a conclusion in either direction. That is to say, it is as yet unclear if the lack of group effects seen in the cardiovascular fitness intervention literature suggests an actual null effect or if some combination of noise, bias, and insufficient dose (i.e., need to increase length of intervention or intensity of exercise in the intervention, as suggested in a recent 2020 review by Valenzuela et al.) may be concealing the true impact exercise and changes in cardiovascular fitness could potentially have on the brain (Fig. 3).

Fig. 3.

A hypothetical scenario illustrating the likely impact of two major methodological components of fitness interventions—type of physical activity and prescribed dose—that vary considerably between studies. Physical activities are used in interventions to increase participants’ cardiovascular fitness (VO₂Max). These varying activities differ in their efficacy in increasing VO₂Max and thus may make it more or less difficult to surpass the amount of change in VO₂Max necessary to see corresponding changes in the brain. However, depending on which physical activity is being utilized, the dose prescribed by interventions (e.g., low, medium, or high as illustrated above) may fall short of the dose required for sufficient change in VO₂Max to see changes in brain structure. Thus, interventions may fail to result in positive change because participants are falling short of this hypothetical threshold due to some combination of dose and exercise type, making it difficult to discern whether an intervention failing to find positive results is actually a true null finding or can be instead contributable to some combination of chosen activity and insufficient dose. Dose is defined here as a combination of the total minutes of exercise done per day, how many days per week exercise occurs, and the effort participants put into each session. Example types of recreational physical activities are, in order from lowest efficacy to highest efficacy, yoga/stretching, walking, biking, and running

As evidence of this, studies that seem to be more methodologically sound do seem to find differences between conditions. Perhaps the best example illustrating the importance of methodology is Erickson et al. [23]—one of the few published intervention studies showing significant post-intervention differences between their control and exercise groups potentially due to their relatively more rigorous methodology in comparison to other interventions. For example, the number of participants enrolled in the intervention was comparatively large relative to most other interventions (N = 156), participants were randomly assigned to either conditions, the intervention lasted for a full year, and all exercise and control stretching sessions were standardized and supervised by a trained exercise leader. Unlike many intervention studies, the exercise group in Erikson et al. (2011) saw greater increases in VO₂Max compared to the stretching control group, validating the exercise component of the intervention. Accordingly, only the exercise group showed increases in hippocampal gray matter volume. Unfortunately, while this study offers coherent evidence supporting exercise as a way to target cardiovascular fitness and brain structure in older adults, a single study requires replication.

A more recently published meta-analysis of 14 fitness-related interventions that targeted changes in hippocampal volume did find that exercise interventions decreased rates of gray matter atrophy over time in the left hippocampus, although the meta-analysis found no overall effect of exercise on increasing total hippocampal volume [15]. However, as with the interventions discussed here, the intensity, frequency, and length of intervention in each exercise program, as well as each study’s overall method and parameters, differed considerably across the studies, limiting the effectiveness of meta-analytic statistical analyses. Hence, the results of the meta-analysis cannot support any definitive conclusions.

Limitations

As mentioned above, there was extreme heterogeneity in methods across published intervention studies. Interventions were as short as 12 weeks [115] or longer than 2 years [113]. Sample sizes also differed considerably, with some only having 20–30 adults in total within the intervention [115] and others recruiting hundreds ([23, 113]). The majority of samples consisted of healthy, cognitively normal individuals; however, some studies targeted populations with MCI or probable AD diagnosis [104, 107, 112]. In addition, the physical activities implemented within experimental conditions varied greatly. Some studies had participants engage in supervised use of a cardio machine, such as treadmill [106] or stationary cycle (e.g., [105]). Others merely assigned a certain amount of time (e.g., 30–60 min, three times per week; e.g., [19]) with minimal supervision or specification regarding the type of exercise. While some studies included ways to measure the intensity and amount of effort put in by the participant during prescribed activities, many did not, increasing within-group noise.

High levels of variability across study procedures are in line with an older meta-analysis from 2005 surveying 41 different exercise interventions in older adults to compare their ability to change cardiovascular fitness. Huang et al. [49] found that the characteristics of the training programs utilized were hard to compare due to differences in the length of each intervention, the type of recreational exercise utilized, and the duration and frequency of individual training sessions. As a consequence, each aerobic intervention’s effectiveness in increasing participants’ VO₂Max fluctuated, although the majority did show significant positive effect sizes (with the biggest gains seen in the lengthier interventions). Due to the disparity among the studies included in the meta-analysis, Huang et al. [49]B52 were hesitant to summarize any conclusive findings and encouraged caution in interpreting their results [49].

Another issue left unaddressed within the current intervention literature is how long effects of interventions last. Studies rarely follow-up with their participants past the conclusion of the intervention, or if there are follow-ups, they rarely go on for more than 6 months post intervention. There is little information indicating whether positive results are enduring or if they require ongoing activity to maintain them. Therefore, the temporal stability of cardiovascular fitness effects remains unclear. For example, it could be that exercise-induced changes in brain volume are permanent and do not require further intervention. However, it is also possible that these changes are transient, and without continued exercise to maintain post-intervention levels of cardiovascular fitness, any structural changes seen will degrade over time.

Conclusion and future directions

Summary: what do we know?

In summary, animal models seem to have established that it is possible to alter the brain with exercise. Further, there does seem to be initial, tentative cross-sectional, longitudinal, and intervention-based evidence of an association between cardiovascular fitness and human brain structure. However, whether these associations extend to other aspects of fitness (i.e., exercise or daily activity) remains unsubstantiated. It is important to note that the fact that cross-sectional studies have found associations between higher cardiovascular fitness and better brain structure but within-subject longitudinal and experimental studies (intervention studies) have suggested higher cardiovascular fitness is only associated with less atrophy which may be an indication that some published cross-sectional findings are artifacts of reverse causation.

Despite the fact that many of the observational cross-sectional and longitudinal studies reviewed do provide tentative evidence for an association between cardiovascular fitness, exercise, and brain structure, they have yet to sufficiently rule out other possible confounding mechanisms—a trend that becomes especially apparent when considering outcomes of recent fitness-related intervention studies. A well-executed exercise intervention study with sufficient randomization (e.g., [23]) can offer causal evidence of a relationship between fitness and brain structure. However, as the majority of intervention studies recently published, discussed in the section “Stage 3: longitudinal studies,” found no group effect due to intervention, the current evidence remains provisional. Instead, intervention results suggest that general increases in cardiovascular fitness—regardless of exercise—across groups are associated with subsequent changes in brain structure. Thus, it is possible that the field has overestimated the extent to which currently published evidence supports a causal relationship between cardiovascular fitness, exercise, and brain structure. Until exercise-specific intervention studies can reliably find and replicate post-intervention group differences in both brain and cardiovascular fitness levels between study members who were randomly assigned to exercise or control groups, causal evidence remains elusive.

Open questions and future directions

To establish causal proof that changes in cardiovascular fitness have neuroprotective effects on age-related rates of structural decline in humans and that exercise is a viable way to induce these changes, several goals should be accomplished. Of highest importance are consistent randomized control trials showing reliable and replicable differences between exercise groups and control groups. Random assignment to conditions is imperative. In addition, observational studies can look to systematically rule out potential neuroselective variables. Changes in cardiovascular fitness must be proven to have efficacy above and beyond any mediating influence of neuroselection.

Future cardiovascular fitness research can build towards these goals in a number of ways. For example, by defining more standardized measurement procedures when assessing cardiovascular fitness and exercise. More rigorous and standardized measurement procedures will allow more cross-study comparison and increased cardiovascular fitness measurement reliability. Increasing the consistency of the type of physical activities and specific doses chosen for exercise interventions will aid in discriminating between activities that are most efficacious at changing cardiovascular fitness and those that are ineffective. It can also help inform our understanding of the threshold of physical activity needed to cause sufficient changes in cardiovascular fitness necessary for consequent changes in brain structure, as well as help establish a set of general parameters that may increase positive outcomes in future clinical trials.

It is important to keep in mind that exercise may not be the recommended way to change cardiovascular fitness for some individuals, particularly those with concurrent health complications (e.g., [117, 118, 120]). In such cases, other interventions such as a healthier diet, addition of blood pressure medication, or controlled weight loss, might be as effective in changing an individual’s overall cardiovascular fitness and cardiovascular health. There may also be potential gender differences in the efficacy of cardiovascular fitness interventions—both in the general population and in specific clinical populations. While I do not discuss gender-specific findings, it is a pertinent consideration and should be addressed by further research.

Going forward, the field would also greatly benefit from longer, more inclusive clinical trials that continue to follow up with study participants post intervention. Larger randomized trials would increase the likelihood that studies are sufficiently powered to detect any real effects of cardiovascular fitness on brain structure between groups. In addition, longer follow-ups would help to delineate any long-term effects of fitness on brain structure, as well as help to define exact thresholds of change in cardiovascular fitness necessary to see stable or permanent change in the brain. It is as yet unclear if any potential buffering of or reversal in brain atrophy that may occur in response to a fitness intervention is permanent—in which case, a succinct cardiovascular fitness intervention may be sufficient—or if the stability of these changes requires maintaining raised levels of cardiovascular fitness—in which case, interventions may need to be targeting lifestyle changes rather than short-term behavior.

Understanding the long-term effects of interventions is of particular relevance as age-related atrophy rates in the brain increase over time which may override any small, temporary benefits caused by increases in cardiovascular fitness. By recruiting larger sample sizes and increasing standardization of the type/dose of exercise, intervention studies would also be more equipped to isolate cardiovascular fitness within and across analyses as any possible sample bias and measurement noise would be better accounted for. Finally, more inclusive or diverse subject pools would help further our understanding of any differences that may exist between particular at-risk populations, informing future clinicians as to which individuals may benefit most from targeted cardiovascular fitness interventions.

The field should also begin to methodically rule out possible neuroselective confounds in order to further claims of a causal link. Confounds can be ruled out simply through more vigorous randomized clinical trials. Intervention studies can take steps to establish stronger proof of a causal relationship above and beyond mediating confounds through ensuring true randomization between groups. However, observational cross-sectional and longitudinal studies can also take steps to help with this process by methodically ruling out neuroselective variables that may be potentially mediating or confounding the relationship between cardiovascular fitness and brain structure. One difficulty here is that the most relevant neuroselective variable to current aging researchers studying structural brain changes in older populations is a measure of childhood brain structure. Without childhood measures of brain structure, there is no way to fully account for baseline brain integrity in observational studies.

Unfortunately, due to the relative recency of modern MRI technology, current populations of older adults cannot have been scanned as children. Thus, no current dataset relevant in aging research (i.e., currently accessible longitudinal data from midlife to older adults) has childhood MRI data. Eventually, more recent ongoing prospective studies with access to such data from childhood will begin to reach middle age and onward. These studies will then offer the unique ability to look at how early life factors may contribute to the relationship between cardiovascular fitness and brain structure. Until then, there are more accessible neuroselective variables, such as childhood IQ, that researchers can consider in future analyses. Cross-sectional and longitudinal studies with access to these alternative early-life variables should look to introduce them into analyses as potentially relevant neuroselective mediators, with the goal towards ruling them out.

One final note to consider is the distinction between testing if increasing cardiovascular fitness can positively affect brain structure and testing if these improvements to brain structure will actually bring about positive changes in brain function. The field has so far focused on proof of this first query, but in order to demonstrate clinical relevance, the latter step is critical. Even if cardiovascular fitness influences brain structure, if this does not translate to viable increases in daily function or in cognitive ability, it holds no relevant efficacy as an intervention for ARD. There is also a difference between participant compliance during clinical trials and translation from a lab setting to everyday life. Depending on how large an increase in cardiovascular fitness is required to reach the threshold necessary for clinically meaningful change in brain structure, it may be unrealistic for the general population or for those already affected by ARD who may have limited physical mobility to reach. These are all things that should be taken into consideration moving forward. However, as exercise is linked to improvements in other aspects of physical and mental health aside from neurostructural changes, there seems to be little drawback in increasing general activity levels in older populations even if it may not lead to targeted changes in the brain.

Conclusion

The field of cardiovascular fitness and brain is relatively young. There are significant obstacles to overcome before definitive conclusions can be drawn about any translatable clinical utility of exercise in mitigating age-related decline. Future work can help address these obstacles at each stage of research. However, while cardiovascular fitness may not be, as proven, a mechanism as previously reported, it has, by no means, been ruled out and merits further investigation.

Funding information

This research was supported by the National Institute on Aging grants AG032282, AG049789, and AG028716 and by the UK Medical Research Council grant MR/P005918.

Compliance with ethical standards

Conflict of interest

The author declares that she has no conflict of interest.