Strength and multiple types of physical activity predict cognitive function independent of low muscle mass in NHANES 1999–2002

Abstract

Introduction

Multiple domains of cognitive function decline with age, resulting in a significant burden on quality of life and the healthcare system. Recent studies increasingly point to links between muscle mass, particularly low muscle mass, and risk of cognitive decline. However, complex relationships exist between muscle mass, muscle function, physical activity, and overall health.

Methods

Data from 1,424 adults 60+ years old in the 1999‐2000 and 2001‐2002 editions of the National Health and Nutrition Examination Survey (NHANES) were used to investigate the relationship between low muscle mass and cognitive function after accounting for strength, physical activity, and nutritional and metabolic risk factors for cognitive decline.

Results

Muscle strength and physical activity independently predicted performance in the digit symbol substitution test, with muscle mass and muscle strength explaining 0.5% and 5% of the variance in cognitive function, respectively. In graphical network analyses, the association between low muscle mass and cognitive function appeared to be primarily mediated by neuromuscular function. Physical activity was associated with strength but, surprisingly, not muscle mass, which was instead more closely related to total mass.

Conclusions

Low muscle mass is a relatively poor predictor of cognitive function after accounting for physical activity and strength in older individuals from a representative population dataset in the US. Future studies should account for the way in which muscle mass is accrued, which is likely to confound any association between muscle mass and health outcomes.

INTRODUCTION

Cognitive performance tends to decline with age. However, the increasing prevalence of neurodegenerative diseases, such as dementia, seems to suggest pathological neurodegeneration beyond what would be expected of normal aging. Given an increasing adult population over the age of 65, preserving cognition or delaying the onset of cognitive decline is critical due to the existing interrelationship between cognitive function and quality of life.

In addition to cognitive decline, skeletal muscle mass, strength, and function decline with age, beginning in the fourth decade of life in sedentary individuals, with losses in muscle mass and strength ranging between 0.8% and 3% per year. This loss of muscle mass and function may be accelerated in those who are not regularly physically active. Moreover, this loss of muscle mass seems to be coupled with brain atrophy in certain pathologies. Consistent links between sarcopenia—age-related declines in muscle mass, strength, and function—and dementia are apparent, pointing to a shared pathophysiology between the two. In the same vein, it has been established that physical activity and, importantly, resistance training, protects against the loss of skeletal muscle and may reduce the risk of cognitive decline.

While some analyses have found an association between low skeletal muscle mass and declining cognition, this association is not always present. One possible explanation for this observed discrepancy is the mode through which skeletal muscle mass was accrued. For example, gains in total mass in response to overfeeding, even in the absence of additional physical activity, can contribute to gains in muscle mass. Indeed, individuals with higher body mass or body mass indices will have more muscle in general; however, muscle gained as a result of a caloric surplus without changes in physical activity may not be associated with similar increases in muscle function. For example, obesity is associated with reduced muscle function, the relationship of which is likely strengthened as a function of aging and inactivity. It is, therefore, possible that any relationship between lower muscle mass and cognitive function may depend on the stimulus for muscle gain and resulting changes in neuromuscular function or strength. In support of this, a recent longitudinal cohort study similarly identified an inverse association between muscle mass and cognitive function; however, this association was attenuated once adjusted for grip strength. Higher levels of leg strength have also been shown to correlate with better Mini-Mental State Examination scores and executive functions, and whole-body strength is positively associated with scores in cognitive test batteries. This suggests that muscle function may be a more important predictor of cognitive outcomes in older individuals.

It remains challenging to separate the differential effects of muscle mass and strength on cognition, as they often go hand in hand. Nonetheless, it is clear that physical inactivity increases the risk of cognitive decline, and the collective evidence suggests that low muscle mass and function may present a target for prevention and/or treatment. We seek to explore this relationship by examining population data to determine associations between muscle mass, strength, physical activity, and established metabolic health predictors of cognition in American adults aged 60 years and older.

METHODS

Included data

Data for the study were taken from the 1999−2000 and 2001−2002 waves of the National Health and Nutrition Examination Survey (NHANES) as they included data on cognitive function using the digit symbol substitution test (DSST), body composition assessments using dual-energy x-ray absorptiometry (DEXA), and isokinetic strength of the knee extensors (quadriceps) in the same individuals. Additional data included laboratory assessments of overall health that are associated with cognitive function and risk of cognitive decline—systolic blood pressure, homocysteine, glycated hemoglobin (HbA1c), hemoglobin, C-reactive protein (CRP), and waist circumference—as well as demographic data for age, sex, education level, and self-reported physical activity. All adults of 60+ years old with complete data for all the above parameters were included. Within the NHANES DEXA results, some participants had missing data that were imputed (n = 5 imputations). As our exploratory analyses aimed to examine complex network relationships between muscle mass, cognitive function, health, and physical activity, we chose to only perform complete case analyses (n = 1424) excluding those who had imputed DEXA data (n = 299). For completeness, comparisons between the included cohort and those excluded are shown in Supplemental Table . Compared to those who were included in the final analysis, excluded individuals with imputed DEXA data were less likely to have the highest level of daily physical activity and had higher CRP and homocysteine levels as well as lower hemoglobin levels. Groups were similar with respect to age, sex, level of education, strength, and cognitive function scores.

Parameter selection and feature engineering

For the primary outcome of cognitive function, the number of correct digits on the DSST (maximum 133 in 120 s) was Z-scored to 5-year age groupings (60–64, 65−69, 70−74, 75−79, 80+). Two metrics of muscle mass were derived from DEXA data—the appendicular lean mass index (ALMI) and the whole-body fat-free mass index (FFMI). ALMI is the skeletal muscle mass in kilograms of the limbs divided by height in meters squared (kg/m²). FFMI (also in kg/m²) was calculated using the total fat-free mass in kilograms (including bone) divided by height in meters squared. Using Canadian-specific cutoffs (<7.30 kg/m² for males and < 5.42 kg/m² for females), Tessier et al. recently showed that low ALMI was associated with the rate of cognitive decline. This resulted in around 20% of their cohort of adults aged 65−80 having low ALMI. We generated a similar analysis in our cohort by assigning low ALMI to those in the bottom quintile of ALMI by sex— < 7.06 and < 5.43 kg/m² for males and females, respectively. As ALMI requires a full DEXA scan that may not always be available clinically, FFMI may be a more accessible measure of fat-free mass as it requires only body weight, height, and an estimate of body fat percentage. Parallel groups of those below the 20th percentile for FFMI and body fat percentage were also generated—<17.4 and <14.3 kg/m² and < 25.9% and < 37.0% in males and females, respectively. To allow for the combination of data by sex, Z-scores of both FFMI and body fat percentage were calculated within each sex separately.

Daily physical activity was coded on a continuous three-level ordinal scale as reported by the question “Please tell me which of these four sentences best describes your usual daily activities?” (1) sits during the day and does not walk very much, (2) stands or walks about a lot during the day but does not have to carry or lift things very often, (3) lifts light loads or has to climb stairs or hills often, or (4) does heavy work or carries heavy loads. Due to the low level of respondents in level four, this was combined into the third group referred to as “loads or stairs/hills.” Vigorous physical activity was dichotomized by a response of “yes” to the question “Over the past 30 days, did you do any vigorous activities for at least 10 minutes that caused heavy sweating, or large increases in breathing or heart rate? Some examples are running, lap swimming, aerobics classes, or fast bicycling.” Resistance training (approximately one or more times per week) was dichotomized based on those answering with a number >3 to the question “Over the past 30 days, how often did you do physical activities designed to strengthen your muscles such as lifting weights, push-ups or sit-ups.” Muscle strength was quantified as a continuous variable using peak force (in newtons) during isokinetic testing of the quadriceps using the Kin Com MP dynamometer (Chattanooga Group Inc., Chattanooga, TN). Peak force was Z-scored by sex for use in the final analyses. A measure of relative force was also derived as the ratio of lower limb lean mass to peak force, in newton per kilogram.

All analyses were adjusted for several confounders and health-related factors known to be associated with cognitive function. Education was included as (1) less than high school graduate, (2) high school graduate, (3) some college or associate degree, and (4) college graduate or above. Individuals with missing data or who refused to answer were included in the lowest education category. The “gender” variable in NHANES was interpreted as sex assigned at birth. Elevated homocysteine was dichotomized at 13 μmol/L, which is a level previously shown to be associated with accelerated cognitive decline and brain atrophy. Though different studies have used different cutoffs, we defined elevated CRP as >1.0 mg/dL, as the risks of several health-related outcomes appear to increase above this level. Anemia was defined as a hemoglobin level <12.0 g/dL in females or <13.0 g/dL in males. Systolic blood pressure and HbA1c were included as continuous variables as cardiovascular risk factors associated with cognitive decline.

Statistical analysis

Descriptive statistics were used to describe the cohort based on those who did and did not have low ALMI. Body composition measures were compared after stratification by sex. Groups were compared with respect to the demographic and health variables using t-tests and chi-square tests for continuous and categorical variables, respectively. Using linear regression with robust standard errors (rigr v1.0.4), the effect of low ALMI and low FFMI on Z-scored DSST performance was determined. The independent effects of level of daily physical activity, vigorous physical activity, and resistance training were also determined, adjusting for sex, sex-specific body fat percentage Z-score, systolic blood pressure, HbA1c, elevated homocysteine, elevated CRP, anemia, and level of education. Partial R² values were extracted from the fully adjusted models to explore the proportion of variance in DSST score predicted by knee extensor strength.

To examine the association between physical activity, strength, muscle mass, health parameters, and cognitive function, a graphical network analysis approach was used. Partial covariance matrices were constructed using the maximum likelihood estimation method and Fisher Z-transformed 95% confidence intervals (CI), as described by Williams and Rast. For network analyses only, the four-level education variable and three-level daily physical activity variables were included as continuous predictors. Significant associations in the resulting precision matrices were depicted as interconnected network diagrams using the ggraph (2.1.0), igraph (v1.5.0.1), and qgraph (v1.9.5) libraries in R. All p-values <0.05 were considered statistically significant. Similar network analysis was performed to (1) examine the relationship between FFMI and body fat percentage after Z-scoring by sex and (2) examine the relationship between relative force and DSST Z-score after taking into account all the covariates identified above. p-values from the associated partial correlation matrices were extracted and displayed as heatmaps using the ppcor (v1.1) library. All analyses were conducted using the R statistical package (v4.1.2, Foundation for Statistical Computing, Vienna, Austria, ).

RESULTS

Demographic variables are shown in Table . Overall, compared to those with normal ALMI, those with low ALMI were older had a smaller waist circumference and lower HbA1c but a higher homocysteine level. Body fat percentage was lower in females with low ALMI but not in males. Individuals with low ALMI had lower peak leg force and were less likely to have performed vigorous physical activity in the preceding 30 days. Levels of education, CRP, hemoglobin, systolic blood pressure, daily physical activity, and relative lower limb force were similar between groups. Of those with low ALMI, 80.5% (n = 227) also had low FFMI (<20th percentile). When separated by fat-free mass, individuals with low ALMI and low FFMI had a lower number of correct DSST answers (median 41 correct answers vs. 45 correct answers for both), but neither was statistically significant (Figure ).

TABLE 1 Descriptive statistics of included individuals and variables by ALMI status.

	Mean (SD) / N (% by group)
	Normal ALMI (Quintiles 2−5)	Lowest Quintile of ALMI	p-value
Number included	1142	282	–
Female sex	571 (50.0)	141 (50.0)	1.00
DSST (number correct)	44.8 (18.9)	43.0 (17.9)	0.14
Age (years)	69.3 (7.0)	73.0 (7.7)	<0.0001
Level of education
Less than high school/Unknown	433 (37.9)	108 (38.3)	0.48
High school graduate	261 (22.9)	75 (26.6)
Some college/AA degree	239 (20.9)	55 (19.5)
College graduate or higher	209 (18.3)	44 (15.6)
Fat-free mass (kg): Males	59.0 (7.0)	49.0 (5.0)	<0.0001
Fat-free mass (kg): Females	41.8 (5.2)	34.0 (3.6)	<0.0001
Body fat (%): Males	30.2 (4.8)	29.7 (5.2)	0.22
Body fat (%): Females	42.1 (4.8)	38.1 (5.8)	<0.0001
ALMI (mg/kg2): Males	8.3 (0.8)	6.6 (0.4)	<0.0001
ALMI (mg/kg2): Females	6.6 (0.9)	5.0 (0.3)	<0.0001
FFMI (mg/kg2): Males	19.8 (1.6)	16.6 (1.1)	<0.0001
FFMI (mg/kg2): Females	16.6 (1.7)	13.6 (0.9)	<0.0001
Waist circumference (cm): Males	103.5 (9.8)	94.5 (9.4)	<0.0001
Waist circumference (cm): Females	97.1 (1.08)	82.4 (9.7)	<0.0001
Low FFMI (< 20th percentile)	57 (5.0)	227 (80.5)	<0.0001
Low body fat (<20th percentile)	191 (16.7)	88 (31.2)	<0.0001
HbA1c (%)	5.9 (1.2)	5.6 (0.9)	<0.0001
CRP (mg/dL)	0.4 (0.6)	0.5 (0.9)	0.50
Elevated CRP (>1.0 mg/dL)	107 (9.4)	26 (9.2)	1.0
Homocysteine	9.8 (4.8)	10.3 (4.0)	0.037
Elevated homocysteine	137 (12.0)	55 (19.5)	0.001
Hemoglobin (g/dL)	14.3 (1.3)	14.3 (1.3)	0.82
Anemia (< 12 or <13 g/dL)	70 (6.1)	17 (6.0)	1.00
Systolic blood pressure (mmHg)	139 (21)	141 (23)	0.079
Peak leg force (N)	341 (108)	252 (83)	<0.0001
Lower limb relative force (N/kg fat-free mass)	22.8 (5.5)	22.7 (5.5)	0.77
Average physical activity
Mainly sitting	230 (20.1)	64 (22.7)	0.63
Standing and walking	708 (62.0)	170 (60.3)
Loads or stairs/hills	204 (17.9)	48 (17.0)
Resistance training 1+ times/week	173 (15.1)	39 (13.8)	0.64
Vigorous exercise last 30 days	253 (22.2)	29 (10.3)	<0.0001

[IMAGE OMITTED. SEE PDF]

In a fully adjusted linear model accounting for sex, systolic blood pressure, HbA1c, elevated homocysteine, elevated CRP, anemia, and level of education, no independent effect of low ALMI was seen (Figure ). The DSST Z-score (95% CI) associated with low ALMI relative to normal ALMI was 0.03 (−0.08, 0.15; p = 0.54). By comparison, independent significant associations of leg strength, resistance training, vigorous activities, and daily level of physical activity were all seen. For each standard deviation (SD) increase in peak leg force, DSST Z-score (95% CI) increased by 0.10 (0.05, 0.14), with peak leg force Z-score explaining 5% of the variance in DSST Z-score (R² = 0.049, p < 0.0001). Resistance training at least once per week and vigorous physical activities in the past 30 days were associated with a DSST Z-score increase of 0.15 (0.04, 0.26; p = 0.008) and 0.12 (0.01, 0.22; p = 0.027), respectively. Compared to those who reported being sedentary, those who regularly lifted loads, performed heavy work, or regularly climbed hills or stairs had a DSST Z-score that was 0.20 (0.07, 0.34; p = 0.003) higher. In a similar fully adjusted model, no independent effect of low FFMI was seen (Figure ). The DSST Z-score (95% CI) associated with low ALMI relative to normal ALMI was 0.03 (−0.08, 0.15; p = 0.60). The magnitude and significance of strength and physical activity variables on DSST performance were almost identical to those seen in the ALMI model.

[IMAGE OMITTED. SEE PDF]

[IMAGE OMITTED. SEE PDF]

As low ALMI and low FFMI appeared to similarly predict DSST scores, FFMI was chosen for the primary graphical network analysis as it is a potentially more accessible and interpretable variable (Figure ). In these analyses, significant relationships are depicted after considering how all the variables in the network are associated with one another, which then allows for potential interpretation of mediation and pathways of action. Level of education was the strongest predictor of DSST performance Z-scored to age. Female sex was positively associated with DSST Z-score, while systolic blood pressure, HbA1c, and anemia, were all negatively associated with DSST Z-score. Of the physical parameters of interest, peak leg force Z-score, vigorous activities, and level of daily physical activity were all significantly associated with DSST Z-score. Resistance training appeared to act on DSST Z-score by contributing to vigorous activity and daily physical activity level. After taking into account physical activity and strength relationships, low FFMI was weakly positively associated with DSST Z-score but strongly negatively associated with lower peak leg force, suggesting that any beneficial effect of higher FFMI on cognitive function is primarily mediated by lower body strength as a measure of neuromuscular function. A similar analysis using low ALMI as the indicator of lean muscle mass showed the same network of relationships (Supplemental Figure ). Heatmaps of the associated p-values from the partial correlation matrices are displayed in Supplemental Figures (low FFMI model) and S3 (low ALMI model).

[IMAGE OMITTED. SEE PDF]

We further examined the association between body composition metrics and health as well as the associations between relative strength, relative body composition, and cognitive function. FFMI and fat mass were strongly correlated (Figure ). In a linear model including all the variables presented in Figure , the body fat percentage Z-score explained 12% of the variance in a muscle mass Z-score (p < 0.0001). A higher FFMI Z-score was strongly associated with greater peak force but also with higher HbA1c. A higher body fat Z-score was associated with higher likelihood of having elevated CRP as well as a lower peak leg force. After taking these relationships into account, physical activity measures were associated with greater peak force, but FFMI was not associated with any measures of physical activity. The associated heatmap of p-values is displayed in Supplemental Figure .

[IMAGE OMITTED. SEE PDF]

We then performed a final network analysis including relative peak force (newton per kg lower limb fat-free mass) (Figure ). Greater relative force was positively correlated with cognitive function, with daily activity and vigorous exercise still independently associated with cognitive function. The same biochemical risk factors for cognitive decline remained significantly associated with cognitive function. The associated heatmap of p-values is displayed in Supplemental Figure .

[IMAGE OMITTED. SEE PDF]

DISCUSSION

In a cohort of nearly 1500 older US adults, we explored the complex relationships between muscle mass, muscle strength, physical activity, and cognitive function, while accounting for other important health-related predictors of cognitive function and cognitive decline. We show that, in our primary linear models, quadricep strength and physical activity, but not low muscle mass, independently predicted cognitive function as assessed by the DSST. In graphical models, a weak association between low muscle mass and cognition was seen, but the main effect of low muscle mass on cognitive function appeared to be mediated by other health-related factors, especially muscle strength. In addition, when examined as a continuous variable, muscle mass (as FFMI) was not associated with self-reported physical activity levels. This suggests that the majority of muscle mass in this cohort was accrued as a part of greater total body mass. In line with this, higher muscle mass was associated with higher fat mass as well as higher HbA1c, indicating worse glycemic control with increasing total body mass. Associations between muscle mass and health outcomes in datasets such as NHANES are therefore likely to be confounded by the way in which muscle mass is accrued. Importantly, however, both relative strength (force per kg of lower fat-free limb mass as a proxy of neuromuscular function) and multiple types of self-reported physical activity were independently associated with higher DSST scores, underscoring benefits of exercise and physical function on cognitive function that go beyond changes in body composition.

While muscle strength often parallels muscle mass, this relationship is not always observed, and measures of strength and muscle mass are not interchangeable. Care should be taken to separate strength and muscle mass when considering older populations, as age-related loss of muscle strength is disproportionately greater than observed losses in muscle mass. Furthermore, Mesinovic and colleagues reported a positive relationship between waist circumference and greater muscle size, but poorer muscle strength and quality. Therefore, we further examined the association between body composition metrics and health as well as the associations between relative strength, relative body composition, and cognitive function. This allowed us to isolate the effects of muscle function, where force per kilogram of lower fat-free limb mass would be expected to be associated with muscle health and neuromuscular function. Together, our results support the idea that more muscle mass alone may not be a predictor of cognitive function in older adults, as higher muscle mass does not necessarily reflect greater neuromuscular function.

When considering metrics of body composition and relative strength, body composition appeared to be associated with cognitive function through its effect on relative force production. We therefore hypothesize that muscle mass accrued through physical activity, which should result in greater lean mass gain relative to fat mass gain, is likely more strongly associated with higher cognitive function due to both local and global improvement in specific strength and neuromuscular function. We also expect that greater muscle mass may independently benefit metabolic and overall health, but only when paired with greater physical function as a result of higher physical activity levels. However, as muscle mass and vigorous physical activity were not associated with one another in our dataset, it is perhaps not surprising that low muscle mass and cognitive function were weakly positively related. Indeed, the opposite of some of the expected relationships were seen, with those in the low ALMI/FFMI groups also having lower HbA1c levels. This may be explained by broader associations between muscle mass, fat mass, and overall health. For instance, the FFMI Z-score was positively associated with HbA1c. In males in particular, those with normal ALMI had higher waist circumferences despite similar levels of body fat. This may suggest that males in the normal ALMI group had higher levels of visceral fat, which is known to be associated with higher levels of systemic inflammation and worse metabolic health.

The lack of a significant positive association between muscle mass and cognitive function in our dataset is interesting because it contrasts with similar previous population-based reports, such as those using UK Biobank data, which found that muscle mass positively correlated with cognition. Besides the possibility for other discrepancies between muscle mass and function not considered here, a possible explanation for this could be the different cognitive tests used and the brain regions that these models assess. For example, in the UK Biobank study, cognitive function was interrogated using the fluid intelligence test, whereas in NHANES the DSST was employed. While both tests are used to detect the presence of cognitive dysfunction and do overlap in some regard (e.g., processing speed), the DSST captures general cognition with low specificity for different regions of the brain, whereas the fluid intelligence test is said to be more related to the mass of the anterior prefrontal cortex and the integrity of frontal white matter.

A shortcoming of broad and unstratified population health data is the invariably linear correlation that exists between FFMI and body fat percentage. Previous studies using NHANES data identified that those with a higher body mass index have a decreased risk of all-cause mortality; however, when excluding those with sarcopenia, this benefit disappears. As such, it appears that higher muscle mass in the context of large population datasets such as NHANES typically implies more body weight overall rather than an elevated degree of physical strength. These differentiating factors address the mode in which muscle is gained and maintained, that is, muscle mass gained through greater gains in total body mass versus exercise and resistance training, the latter of which would support greater muscle quality. Importantly, muscle gained through resistance training is protective against age-related declines in mobility, metabolic health, and cognitive function, all of which are not mutually exclusive. This discrepancy between how muscle is accrued, and the associated differences in muscle function may also explain recent findings showing that higher muscle mass in men was associated with higher risk of heart disease and mortality, while higher grip strength was associated with lower risk in the same cohort. Overall, our data and that of others suggest that muscle mass is only protective of health and cognitive function if it is proportionately associated with muscle function (e.g. strength and power), which is more likely to be the case in the setting of exercise-induced hypertrophy.

Increasing and maintaining muscular strength necessitates a regular muscle hypertrophic stimulus. Resistance training is the primary mode of exercise for increasing both muscle mass and strength and preventing their age-related decline in elderly populations. Compared to endurance exercise, significant morphological adaptations to resistance training occur in type-II muscle fibers, the fiber type preferentially lost with age. Improvements in muscle strength and quality can be achieved in elderly adults who engage in resistance training, and strength appears to increase more rapidly than muscle mass. Critically, however, muscle gain is possible at any age, and age only predicts 10% of the variance in muscle gain in response to resistance training. Moreover, resistance training in older adults results in improved balance and stability. This can occur in as little as 6 weeks training 2 days/week when specifically designed to incorporate a balance component. Relevant to both our own and other analyses, as power and strength are generally lost faster than muscle mass, loss of muscle mass may be a lagging indicator of poor neuromuscular function. Measures of relative and absolute strength will, therefore, be more important when examining associations between muscle mass and cognitive function at the population level. Indeed, emerging research gives greater recognition for muscle strength as an indicator of cognitive function, with grip strength as a suggested tool to detect cognitive decline in aged adults. Further, while increased physical activity of any type or intensity is associated with greater cognitive function, several systematic reviews and meta-analyses now provide a strong evidence-base that resistance training can be employed as an intervention to improve cognition, and that the addition of resistance training to aerobic training interventions may hold greater benefits on cognitive outcomes than aerobic training alone. Resistance exercise interventions have also been shown to increase frontal lobe connectivity and executive function while reducing white matter atrophy and improving memory, attention, and information processing.

Considering the established acute and chronic effects of exercise, and the protective role that muscle plays against the hallmarks of aging, the loss of skeletal muscle mass and strength may contribute to cognitive decline. Increasing physical activity also necessarily protects against the negative effects of sedentary lifestyles on mortality. There are several potential direct and indirect mechanisms linking muscle quality to cognitive function that include the physiological response and adaptations to exercise, metabolic health factors, as well as the effects on functional capacities that impact independence and quality of life. These outcomes are modulated by potent antiaging effects of exercise, including but not limited to mitochondrial and vascular adaptations to training, increased blood flow, secretion and signaling of crucial growth factors, cytokines, and the induction of autophagy brought on by muscle contraction, some of which act by chronically lowering systemic inflammatory signaling. The exact mechanisms underlying the cognitive enhancements of resistance training are not fully understood yet, likely due to the pluripotent effects of exercise on both a local (working muscle) and systemic (whole-body) level. Different groups have looked to uncover the mechanistic bases for this crosstalk, resulting in the identification of a variety of metabolites, peptides, and proteins that are implicated in this axis. For example, the MOTS-c (mitochondrial ORF of the 12s rRNA type-c) peptide, secreted in skeletal muscle and circulation during bouts of exercise, was recently shown to play a role in age-dependent physical and cognitive decline. Brain-derived neurotrophic factor (BDNF), shown to exert positive effects on cognition due to its important role in neurogenesis and neuroplasticity, is also secreted during exercise, including in older adults. These are important findings as resting levels of BDNF decline with age, are associated with cognitive function and can be elevated with exercise independent of age. When taken together, the secretion of these factors hints at a positive muscle–brain crosstalk and role for physical activity in the maintenance and preservation of cognitive capacity into advanced age, so much so that exercise mimetics are posited as a potential therapeutic intervention, especially in the context of central nervous system diseases. The molecular bases for muscle–brain crosstalk are beyond the scope of this discussion and have been reviewed extensively in recent excellent reviews looking to synthesize how these molecules act in concert to modulate positive cognitive outcomes.

Another lens through which to approach this issue is that of metabolic disease, and the role it plays in promoting and accelerating cognitive decline. Similarly to muscle mass and cognition, glucose tolerance declines with age, and this decline is associated with an increased risk of neurodegenerative diseases such as dementia and Alzheimer's disease. The English Longitudinal Study of Ageing identified that individuals with type 2 diabetes (T2D) and prediabetes experience more rapid declines in cognitive function with age as compared to those with normal glycemic profiles, even after adjusting for age, sex, education, and several other variables associated health status. Furthermore, data obtained using structural magnetic resonance imaging and the “BrainAGE” algorithm (a biomarker of brain structure) identified fasting blood glucose as the strongest predictor of having a higher BrainAGE relative to biological age in a sample of participants with T2D. Specifically, those with the highest fasting blood glucose values had a BrainAGE 5.5 years older than those with the lowest values. Skeletal muscle represents the primary site of postprandial glucose disposal, responsible for up to 90% of insulin-stimulated glucose uptake, and exercise results in profound improvements in insulin sensitivity and glucose uptake both acutely and chronically. Dela et al. demonstrated that exercise training can result in 2−3 times greater glucose uptake per kilogram of muscle mass even when at rest, suggesting that the more muscle mass gained through training, the greater the glucose uptake by skeletal muscle. Additionally, low muscle mass is associated with the pathogenesis of insulin resistance and T2D, and it is well known that an acute bout of exercise improves insulin sensitivity. This is important as glycemic control and insulin sensitivity are linked to cognitive performance. In addition to more direct effects, muscle mass, and physical activity may, therefore, also indirectly contribute to better cognitive health through its effects on metabolic health.

Lastly, a point that is often overlooked is how the disproportionate age-related decline in muscle strength when compared to a decline in muscle mass can result in reduced overall mobility and physical function (e.g., reduced walking speed and balance). Existing links between cognitive impairment, frailty, fall-related injuries, and reduced fitness suggest that this aspect of functional capacity is at least somewhat related to accelerated brain aging. Losing functional capacity can lead to a loss of independence and profoundly influence potential mediators of cognitive decline such as social isolation, which itself is a risk factor of cognitive decline. Indeed, reduced motor function in elderly populations is associated with self-perceived isolation, reduced social activities in late life (i.e., social isolation), and increased risk of dementia. Additionally, loss of mobility reduces the ability to carry out acts of daily living and overall quality of life, which is highly dependent on the maintenance of an independent lifestyle in advanced age. Improving and preserving strength in the elderly appears important with respect to lifestyle-based behaviors that may blunt declines in cognition, while also increasing an individual's ability to perform exercises that include coordination (e.g., a balance component), which appear to result in additional benefits on cognitive outcomes.

While our analysis is revealing, it does have limitations. To perform our graphical network analyses, participants with imputed DEXA data were excluded from the cohort. This skewed our final set of individuals towards being slightly healthier and more physically active. However, our data still retained a broad range of physical activity levels and health parameters that reflected the cohort as a whole. Nonetheless, we assume our results to be relevant to US adults 60 years and older of a similar demographic to those included here. As this cohort showed little association between physical activity and muscle mass, the results are also unlikely to translate to an athletic or other population whose muscle mass is more closely related to the level of resistance training or other physical activity. Due to the nature of the network analyses, some variables were assumed to have linear relationships with cognitive function which may not be truly linear. Unfortunately, sleep data were also unavailable in this cohort of NHANES, which we believe this would have strengthened this analysis given known links between physical activity, sleep, and cognition, with improved sleep likely to mediate some of the effects of exercise on cognitive function. Moreover, compared to similar studies such as Tessier et al. and Klinedinst et al., our cohort was smaller and did not include longitudinal assessments. This may have precluded us from isolating or identifying any true independent effect of low ALMI or FFMI due to lower statistical power. As the independent effect of lean mass is unlikely to differ markedly between the United States and Canada, differences between our results and those of Tessier et al. may also be explained by differences in the populations. For instance, Tessier et al. defined low ALMI as <1.5 SD below the sex-specific mean, which represents the lowest 6−7% of their population. We used similar numeric cutoffs but defined low ALMI as being in the lowest 20% of our population, and examination of the two cohorts suggests that our population had lower muscle mass overall despite being a similar age. We also acknowledge that deriving ALMI and FFMI from DEXA results provides only proxies of muscle mass and not a true quantitative measure of skeletal muscle mass; however, we do not believe this would alter our findings. Lastly, since this is an observational cohort study, we are unable to infer causation. We also assumed that no interactions existed between our included predictors and their associations with DSST scores as these are difficult to model within graphical networks, though we are not aware of any interactions relevant to the predictors we included in our analysis. Despite these limitations, we hope that these findings galvanize further research into the underlying drivers of the crosstalk behind muscle mass, muscle strength, and cognition across their different axes.

CONCLUSION

Our work dives into findings presented in past reports about underlying correlations between muscle mass and cognition. Network analyses suggest the critical predictor of cognitive function is overall relative neuromuscular function, which may be impacted by body composition and the route in which muscle mass is accrued. In this case, we show that muscle strength, both overall and relative to muscle mass, is a more potent predictor of cognitive outcomes rather than the amount of fat-free mass. Practically, it follows that—from a lifestyle medicine perspective—the integration of an exercise regimen including resistance training to improve strength and neuromuscular function may be an important intervention for those looking to improve long-term cognitive health. Through both direct (e.g., neuromuscular stimulus and improved physical functioning) and indirect (e.g., glucose disposal and myokine release) mechanisms, resistance training in older adults can translate to greater mobility, reduced risk of falls, maintenance of independence, and consequently, cognition.

ACKNOWLEDGEMENT

The authors have nothing to report.

CONFLICTS OF INTEREST STATEMENT

TRW is a paid scientific advisor for Hintsa Performance, Sidekick Health, Thriva LLC, and Rewire Fitness, and is a founding trustee of the British Society of Lifestyle Medicine. AJG is a paid scientific advisor for Momentous and XPT Life, is a founder of RAPID Health Optimization and Absolute Rest, and is on the board of directors for the Health and Human Performance Foundation. KLS is a paid science communicator for Zero Longevity Science, Inc., and is a founder of The Femme Formula. BTH is the founder of Deconstruct Nutrition. GDMP is a founder of Resilient Nutrition Ltd and is a paid scientific advisor for Wellics Ltd and Becoming Co.

DATA AVAILABILITY STATEMENT

All data used in this manuscript are available from the CDC website:

References

Harada CN, Natelson Love MC, Triebel KL. Normal cognitive aging. Clin Geriatr Med. 2013;29(4):737. doi: [DOI: https://dx.doi.org/10.1016/J.CGER.2013.07.002]