Directionality between cognitive function and daily physical activity patterns

Abstract

INTRODUCTION

Physical activity is a modifiable risk factor for dementia, but cognitive function is also important for physical activity engagement. This study evaluated the directionality of associations between daily physical activity and cognitive function in a sample of cognitively and physically intact older adults.

METHODS

Cognitive factor scores for domains including global cognition, memory, language, executive function/attention, and visuospatial processing, and physical activity patterns from wrist accelerometry were measured at two visits (mean: 1.8 years) among 237 cognitively intact older adults in the Baltimore Longitudinal Study of Aging (BLSA) (mean age: 76.5 years). Bivariate latent change score models estimated directionality of associations between changes in cognitive factor scores and physical activity patterns. Models were adjusted for age, sex, race, education, comorbidities, and body mass index.

RESULTS

Higher total amount of activity, longer activity bouts, less sedentary time, and less activity fragmentation at baseline were associated with less annual cognitive decline across multiple cognitive domains (X ² > 4.11, 1 df for all). In contrast, baseline cognitive factor scores were not associated with changes in any activity pattern (X ² < 3.20, 1 df for all).

DISCUSSION

Increasing movement and/or decreasing sedentary behavior is associated with less prospective cognitive decline. Targeting reductions in sedentary time and lengthening activity bouts may slow cognitive decline among older adults at risk for dementia.

Highlights

• Greater activity engagement is related to less annual cognitive decline.

• Baseline cognition is not associated with short‐term changes in activity patterns.

• Promoting daily movement and lowering sedentary time may have cognitive benefits.

1. INTRODUCTION

Physical activity is an established modifiable risk factor for cognitive impairment and dementia. ¹ Higher self‐reported physical activity is associated with better cognitive function ² , ³ , ⁴ and lower risk of dementia among older adults. ² , ³ Yet, self‐reported measures of physical activity often overlook light intensity everyday activities. This has led to a growing interest and use of accelerometers to assess physical activity in the free‐living environment. ⁵ , ⁶ Indeed, accelerometry studies have shown that more active time and less sedentary time are associated with better executive function, memory, and global cognition among cognitively intact older adults ⁷ and analogously that persons with cognitive impairment exhibit lower total activity during certain times of day, ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ blunted diurnal patterns of activity, ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ and more fragmented ⁸ and less complex ¹¹ activity. As most of these studies are cross‐sectional, the directionality of these associations remains unclear.

Emerging evidence also suggests that cognitive function is important for physical activity. ¹⁴ , ¹⁵ , ¹⁶ Volitional self‐regulation of physical activities requires cognitive processes that stem from neural networks and include functions such as planning and decision‐making, sustaining attention, and inhibiting distractions. ¹⁴ , ¹⁵ , ¹⁷ Indeed, functional connectivity in these networks predicts adherence to exercise interventions ¹⁸ and is strengthened through physical activity and exercise. ¹⁹ Higher executive function is also linked to better self‐efficacy and exercise intervention adherence. ²⁰ In addition, greater brain atrophy is related to declines in objectively measured total physical activity and increased sedentary behavior, ²¹ as well as lower total amount of activity and higher activity fragmentation ²² among cognitively unimpaired older adults. Objective measures of neuropsychological performance also suggest that poorer memory ¹⁶ , ²³ and executive function ¹⁴ significantly predict self‐reported physical activity decline. Despite this evidence, the directionality between objectively measured free‐living physical activity and cognitive function remains understudied.

This study evaluates the directionality of associations between 2‐year changes in daily physical activity patterns and 2‐year changes in cognitive function among cognitively intact older adults. We hypothesize that this association is bidirectional, as defined by significant associations between baseline daily physical activity patterns with changes in cognitive function and baseline cognitive function with changes in daily physical activity patterns. Understanding this directionality among older adults without cognitive impairment has potential implications for both dementia prevention and physical activity engagement in later life.

2. METHODS

2.1. Study population

The Baltimore Longitudinal Study of Aging (BLSA) is a human aging study conducted by the National Institute on Aging Intramural Research Program. BLSA participants are community‐dwelling adult volunteers free of cognitive and physical impairments and major chronic conditions upon enrollment. A full sample description is provided elsewhere. ²⁴ The current analytic sample includes cognitively intact older adults (i.e., without diagnosed cognitive impairment or dementia) aged ≥ 60 years. Participants had two visits between August 2015 and February 2020 with both cognitive function and ≥ 3 valid days of wrist accelerometry data. The study protocols were approved by the Internal Review Board of the Intramural Research Program of National Institutes of Health. All participants provided written informed consent.

2.2. Cognitive function

In a prior analysis, cognitive tests administered between June 1987 and July 2019 were used to derive cognitive domain factor scores for global cognition, memory, executive function/attention, language, and visuospatial processing using separate unidimensional confirmatory factor analysis models. The individual cognitive tests are presented in Table A.1. Test scores were moderately‐to‐strongly correlated between the two visits (Table A.2). Separate unidimensional confirmatory factor analysis models with maximum likelihood estimation were used to estimate each factor score. This latent variable method uses all available cognitive test data at each study visit to construct cognitive factors reflected by observed test indicators. This method allows for tests to have unique, empirically derived weights and residual errors. ²⁵ , ²⁶

2.3. Daily physical activity patterns

Daily physical activity was objectively measured using the triaxial ActiGraph GT9X wrist accelerometer (ActiGraph, Pensacola, FL, USA) between August 2015 and February 2020. Participants wore the accelerometer on the nondominant wrist 24 hours per day for 7 days. Raw data were aggregated to the minute‐level. Days with over 10% missing data as identified by the Choi algorithm were excluded. ⁷ , ²⁷ Time in bed was identified using the ActiGraph auto‐sleep detection algorithm based on the 2007 Troiano Wear Time Validation Algorithm.

Minute‐level data were processed using the arctools package ²⁸ in R to generate daily physical activity patterns. A cut‐point of 1853 activity counts was used to define active (≥ 1853) versus sedentary behavior (< 1853). ²⁹ Time in bed was removed. Patterns of interest include: (1) total amount of activity, defined as the sum of total activity counts per day (TAC); (2) minutes spent active per day; (3) minutes spent sedentary per day; (4) intensity of activity (max 5), defined as the sum of total activity in the five most active consecutively occurring minutes of the day; (5) mean active bout length; (6) mean sedentary bout length; (7) physical activity fragmentation (active‐to‐sedentary transition probability [ASTP]), defined as the probability of transitioning from an active‐to‐sedentary state; and (8) sedentary‐to‐active transition probability (SATP), defined as the probability of transitioning from a sedentary‐to‐active state. All patterns were averaged across valid days and then transformed to z‐scores by subtracting the sample mean and dividing by the sample standard deviation.

2.4. Covariates

Age (years) was measured at baseline. Sex was categorized as male or female. Race was categorized as self‐reported White, Black, or other. Due to small sample sizes, participants who self‐reported their race as American Indian or Alaska Native, Asian, Native Hawaiian or Other Pacific Islander, two or more races, do not know, or refused were combined in the other race group category. Years of education were self‐reported. Chronic conditions included cardiovascular disease (myocardial infarction, congestive heart failure, angina, peripheral arterial disease, vascular‐related procedures), stroke or transient ischemic attack, hypertension, diabetes, hyperlipidemia, pulmonary disease (chronic bronchitis, emphysema, chronic obstructive pulmonary disease, asthma), arthritis or osteoarthritis, and cancer (excluding squamous or basal cell cancers). Chronic conditions were identified through provider evaluations and self‐reported medical history and medications. The number of individual conditions was summed. Body mass index (BMI) was measured as body mass (kg) divided by height squared (m²). Work or volunteer status was self‐reported from the following questions: “Do you currently work for pay, either at a regular job, consulting, or doing odd jobs?” and “Do you currently work or volunteer?” Usual gait speed was measured by usual walking pace (meters/second). In a prior analysis, missing values for BMI, usual gait speed, number of comorbidities, and work or volunteer status were imputed using multiple imputation.

RESEARCH IN CONTEXT

1. Systematic review: The authors reviewed the literature using established sources (i.e., PubMed). While publications have found that physical activity is a modifiable risk factor for cognitive decline/dementia and that cognitive function is important for self‐reported physical activity, the directionality of associations between cognitive function and daily physical activity among cognitively intact older adults is unclear.

2. Interpretation: Consistent with findings currently in the public domain, our findings suggest that higher total amount of activity, longer activity bouts, lower sedentary time, and lower physical activity fragmentation are associated with less cognitive decline across multiple domains among cognitively intact older adults. However, baseline cognition was not related to changes in any activity pattern over a mean of 1.8 years.

3. Future directions: Future studies should examine whether promoting or intervening upon daily movement bouts and/or sedentary behavior could help to slow cognitive decline among older adults at risk for dementia.

2.5. Statistical analysis

Baseline was defined as the first of the two visits in this analysis. Sample characteristics at baseline were compared overall and by tertiles of TAC.

Separate bivariate latent change score models were fit for each cognitive domain and each activity pattern using the laavan package ³⁰ in R. This model is a latent variable method that estimates a change score representing change in daily physical activity and/or cognitive factor scores between baseline and follow‐up using cross‐lagged autoregressions. ³¹ , ³² For each activity pattern‐cognitive domain combination, four models with different assumptions about directionality were tested: (1) a baseline model estimated the change scores for activity patterns or cognitive domains but modeled no relationship between baseline levels of activity patterns and cognitive domains with changes in the other outcome; (2) extension of the baseline model by adding a coupling parameter linking baseline activity patterns with changes in cognitive domains (i.e., baseline level of physical activity → change in cognitive domain); (3) extension of the baseline model by adding a different coupling parameter linking baseline cognitive domains with changes in activity patterns (i.e., baseline level of cognitive domain → change in physical activity); and (4) the dual coupling model by including both coupling parameters to evaluate both directions (i.e., baseline level of physical activity → change in cognitive domain AND baseline level of cognitive domain → change in physical activity). The four models were compared using likelihood ratio tests and were adjusted for age, sex, race, years of education, number of comorbidities, and BMI. A maximum likelihood estimator was used to account for missing data. ²⁵ , ³¹ Model fit parameters were evaluated. ³¹ The structural equation model is presented in Figure 1. ³¹

FIGURE 1.

Structural equation model of latent change score model estimating directionality of associations between changes in daily physical activity patterns and changes in cognitive factor scores.

We also conducted several sensitivity analyses. First, we stratified by whether participants self‐reported currently working or volunteering. Second, we stratified by age 80 (60–79 vs. 80–96) years at baseline. Finally, we additionally adjusted for usual gait speed. Analyses were conducted using R Version 4.3.2 (http://CRAN.R‐project.org).

3. RESULTS

3.1. Sample characteristics

The analytic sample included 237 cognitively intact older adults who had two visits with both cognitive function and wrist accelerometry data. Sample characteristics are described in Table 1. Participants were on average aged 76.5 years, 48% female, and 22% Black. Those who were more physically active at baseline (i.e., highest tertile of total amount of physical activity [TAC]) were younger, more likely to be female, had higher usual gait speed and physical function, and had lower prevalence of cardiovascular disease, diabetes, or cancer than those who were less physically active. There were no statistically significant differences in apolipoprotein E (APOE) ‐ε4 status by tertiles of total amount of physical activity. Sample characteristics of participants in the combined other race group category are presented in Table A.3.

TABLE 1.

Sample characteristics overall and by tertiles of amount of activity (total activity counts) (n = 237).

Mean (SD) or N (%)	Overall (N = 237)	Low activity a (N = 79)	Middle activity a (N = 79)	High activity a (N = 79)
Demographics
Age (years), mean (SD)	76.5 (8.3)	78.0 (8.7)	77.3 (8.0)	74.2 (7.8)
Female, N (%)	114 (48)	32 (41)	33 (42)	49 (62)
Black, N (%)	53 (22)	12 (15)	17 (22)	24 (30)
Education (years), mean (SD)	17.4 (2.3)	17.3 (2.6)	17.5 (2.1)	17.3 (2.2)
Physical function
BMI (kg/m2), mean (SD)	27.0 (4.9)	27.0 (5.1)	27.1 (4.8)	26.8 (4.9)
Gait speed (m/s), mean (SD)	1.10 (0.23)	1.04 (0.25)	1.11 (0.22)	1.14 (0.21)
SPPB, mean (SD)	11.4 (1.4)	11.2 (1.6)	11.4 (1.6)	11.8 (0.6)
Dif. walking ¼ mile, N (%)	12 (5)	6 (8)	5 (7)	1 (1)
Mental health
CES‐D, mean (SD)	5.9 (5.9)	6.8 (6.7)	5.6 (6.1)	5.4 (4.8)
Work/volunteering, N (%)	160 (68)	52 (66)	50 (63)	58 (73)
Comorbidities
# conditions, mean (SD)	2.2 (1.3)	2.0 (1.2)	2.3 (1.5)	2.3 (1.3)
CVD, N (%)	23 (10)	8 (10)	13 (17)	2 (3)
Stroke/TIA, N (%)	16 (7)	7 (9)	3 (4)	6 (8)
Hypertension, N (%)	118 (50)	32 (41)	45 (58)	41 (52)
Diabetes, N (%)	43 (18)	12 (15)	9 (12)	22 (28)
High cholesterol, N (%)	141 (60)	46 (59)	50 (64)	45 (57)
COPD, N (%)	32 (14)	11 (14)	9 (12)	12 (15)
Osteoarthritis, N (%)	134 (58)	42 (54)	39 (51)	53 (69)
Cancer, N (%)	81 (35)	29 (38)	33 (43)	19 (24)
APOE‐ε4 carrier, N (%)	54 (23)	17 (22)	16 (20)	21 (27)
Years between visits, mean (SD)	1.8 (0.6)	1.7 (0.7)	1.8 (0.6)	1.8 (0.5)

Abbreviations: APOE, apolipoprotein; BMI, body mass index; CES‐D, Center for Epidemiologic Studies Depression scale; COPD, chronic obstructive pulmonary disease; CVD, cardiovascular disease; SPPB, Short Physical Performance Battery; TAC, total activity counts; TIA, transient ischemic attack.

^a Categories of total amount of physical activity are defined by tertiles of total activity counts.

3.2. Baseline level of physical activity and changes in cognitive domains

Over an average of 1.8 (range: 0.9–4.0) years, there was a mean decline in memory of 0.021 SD (standard error [SE] = 0.055), executive function/attention of 0.072 SD (SE = 0.032), language of 0.048 SD (SE = 0.041), visuospatial processing of 0.14 SD (SE = 0.053), and global cognition of 0.064 SD (SE = 0.025) (Figure 2).

FIGURE 2.

Model‐estimated annual rates of change in cognitive factor scores and daily physical activity patterns (N = 237). Model estimates come from separate latent change score models of each outcome. Estimates are adjusted for: age, sex, race, years of education, number of comorbidities, and body mass index. The top panel displays annualized standard deviation changes in cognitive factor scores over follow‐up. The bottom panel shows annualized standard deviation changes in daily physical activity patterns over follow‐up. Follow‐up was on average 1.8 years (range 0.9 to 4.0 years).

Compared to models with no coupling, baseline levels of daily physical activity were associated with change in cognitive factor scores for 8 of 40 variable combinations we tested (Table 2, Figure 3). Greater baseline total amount of physical activity (TAC) was associated with less annual memory decline (estimate = 0.099, SE = 0.046). More baseline sedentary time was associated with steeper declines in global cognition (estimate = −0.057, SE = 0.024), language (estimate = −0.076, SE = 0.036), and visuospatial processing (estimate = −0.073, SE = 0.035). Greater baseline physical activity fragmentation was associated with steeper declines in memory (estimate = −0.11, SE = 0.048) and visuospatial processing (estimate = −0.072, SE = 0.034), whereas longer mean active bout length at baseline was associated with less decline in these domains (memory: estimate = 0.13, SE = 0.047; visuospatial: estimate = 0.078, SE = 0.034). As expected, models with a dual coupling showed better fits than their respective baseline models, although this improvement in fit was driven by the couplings from baseline level of physical activity to changes in cognitive domains.

TABLE 2.

Likelihood ratio tests of latent change score model of associations between changes in daily physical activity patterns and changes in cognitive factor scores.

Activity pattern	Model a	Cognitive domain b
		Mem	EF/Att	Lang	Vis	Global
Amount	No coupling	REF	REF	REF	REF	REF
	PA → ∆ Cog	4.11 b	0.014	0.40	0.90	2.39
	Cog → ∆ PA	0.98	0.014	2.18	0.52	0.18
	PA ↔ Cog	5.60	0.026	2.78	1.59	2.59
Active time	No coupling	REF	REF	REF	REF	REF
	PA → ∆ Cog	3.11	0.000	0.094	0.57	1.72
	Cog → ∆ PA	0.21	0.014	1.31	0.078	0.052
	PA ↔ Cog	3.55	0.016	1.49	0.70	1.81
Sedentary time	No coupling	REF	REF	REF	REF	REF
	PA → ∆ Cog	0.27	0.15	4.68 b	4.14 b	5.43 b
	Cog → ∆ PA	2.42	0.18	0.35	0.088	0.004
	PA ↔ Cog	2.53	0.37	4.82	4.43	5.43
Intensity	No coupling	REF	REF	REF	REF	REF
	PA → ∆ Cog	0.25	0.13	0.29	1.36	0.50
	Cog → ∆ PA	1.002	0.20	0.12	3.20	1.24
	PA ↔ Cog	1.44	0.37	0.46	4.64	1.89
Physical activity fragmentation	No coupling	REF	REF	REF	REF	REF
	PA → ∆ Cog	6.35 b	0.050	2.06	4.16 b	2.87
	Cog → ∆ PA	0.90	0.020	0.02	0.12	0.004
	PA ↔ Cog	6.50	0.064	2.14	4.52	2.90
Mean active bout length	No coupling	REF	REF	REF	REF	REF
	PA → ∆ Cog	7.27 b	0.098	1.94	4.96 b	2.73
	Cog → ∆ PA	0.12	0.052	0.52	0.24	0.15
	PA ↔ Cog	7.27	0.14	2.69	5.54	2.98
SATP	No coupling	REF	REF	REF	REF	REF
	PA → ∆ Cog	0.000	0.27	0.29	0.27	2.35
	Cog → ∆ PA	0.08	0.13	0.054	0.016	0.028
	PA ↔ Cog	0.08	0.43	0.37	0.30	2.42
Mean sedentary bout length	No coupling	REF	REF	REF	REF	REF
	PA → ∆ Cog	0.016	0.12	0.57	0.042	1.77
	Cog → ∆ PA	0.008	0.002	0.04	0.032	0.04
	PA ↔ Cog	0.02	0.12	0.65	0.066	1.78

Note: Model: (1) No coupling (reference); (2) Baseline physical activity associated with changes in cognition (PA → Cog); (3) Baseline cognition associated with changes in physical activity (Cog → PA); 4) Dual coupling (PA ↔ Cog).

Abbreviations: Cog, cognition; EF/Att, executive function/attention; Lang, language; Mem, memory; PA, physical activity; REF, reference; SATP, sedentary‐to‐active transition probability; Vis, visuospatial processing.

^a Models adjusted for: age, sex, race, education, number of comorbidities, and body mass index.

^b Indicates coupling parameter significantly contributes to the model as determined by likelihood ratio tests, where X ² with 1 df = 3.841.

FIGURE 3.

Latent change score model estimates of change in cognitive factor scores per one standard deviation higher in baseline daily physical activity patterns. Statistically significant associations are indicated by **. Bivariate latent change score models are adjusted for: age, sex, race, years of education, number of comorbidities, and body mass index.

3.3. Baseline level of cognitive domains and changes in physical activity

There was no evidence of change in daily physical activity patterns over follow‐up (Figure 2). Likely due to the small changes in physical activity, no baseline cognitive factor scores were associated with changes in daily physical activity patterns (Table 2). Model fit statistics for all models are included in Tables A.4–8.

3.4. Differences by work/volunteer status

Participants currently working or volunteering (n = 160, 68%) had more difficulty walking one‐quarter mile, higher prevalence of diabetes, high cholesterol, or cancer, and lower prevalence of depressive symptoms, cardiovascular disease, or stroke than non‐workers or non‐volunteers (Table A.9).

Among workers/volunteers, findings for TAC with memory, mean active bout length with visuospatial processing, and time spent sedentary with language remained robust. All other primary findings lost statistical significance. However, the magnitude of the coefficients was similar (Table A.10). Additionally, one SD higher in baseline visuospatial processing and global cognition were separately associated with greater annual decline in activity intensity [max 5]) (Figure A.1). Among the n = 77 not working or volunteering, findings for physical activity fragmentation (ASTP) and mean active bout length with memory remained robust, though all other findings lost statistical significance (Figure A.1). The magnitude of the coefficients was similar for TAC with memory and sedentary time with global cognition (Table A.10).

3.5. Differences by age

Participants aged ≥ 80 years at baseline (n = 100, 42%) were less likely to be female or Black individuals, had lower BMI and usual gait speed, and had higher prevalence of certain comorbidities than those < 80 years (Table A.9).

All primary findings lost statistical significance among participants aged ≥ 80 years. The magnitude of the coefficients was similar for sedentary time with language or global cognition (Table A.10). Additionally, one SD higher in baseline global cognition was associated with less annual decline in mean active bout length (Figure A.2). Among those aged < 80 years (n = 137), findings for physical activity fragmentation (ASTP) and mean active bout length with memory and visuospatial processing remained robust. All other primary findings lost statistical significance, but the magnitude of the coefficients was similar (Table A.10). Additionally, one SD higher in baseline visuospatial processing was associated with greater annual decline in activity intensity (max 5) (Figure A.2).

3.6. Adjustment for usual gait speed

All findings remained robust after additionally adjusting for usual gait speed. Baseline cognitive function remained unassociated with changes in daily physical activity patterns (data not shown).

4. DISCUSSION

Among this sample of cognitively intact older adults, certain daily physical activity patterns at baseline were related to less cognitive decline across multiple domains over a mean of 1.8 years. However, inconsistent with our hypothesis, baseline cognitive function was not associated with short‐term changes in any daily activity pattern. This study extends the large body of work suggesting that increasing engagement in daily movement and/or decreasing sedentary behavior might be protective against cognitive decline among older adults without cognitive impairment. Future studies are needed to evaluate whether intervening on these activity patterns can help prevent dementia.

Higher TAC, longer activity bouts, lower sedentary time, and lower physical activity fragmentation were associated with less annual decline in memory, visuospatial processing, language, and global cognition over a mean of 1.8 years. This aligns with prior work finding positive associations between physical activity and cognitive function. ⁷ , ³³ , ³⁴ , ³⁵ Physical activity exerts positive effects on cognitive function through mechanisms including increasing neurogenesis, ² , ³ , ⁴ which likely occurs due to increased brain neurotrophic factors ² , ³ , ⁴ and cerebral blood flow. ⁴ , ³⁶ This might then improve brain connectivity ² , ⁴ and reduce brain atrophy, ² , ³ , ⁴ , ³⁶ which then relates to behavioral changes such as improved sleep. ² Additionally, physical activity reduces risk factors for cognitive impairment including stress, ² , ³ inflammation, ³ and cardiovascular disease. ² , ⁴

Our finding that baseline activity is related to acute changes in cognitive function aligns with results from exercise interventions. Short‐term exercise interventions (i.e., < 3 months) that target moderate‐to‐vigorous intensity physical activity are associated with improvements in cognitive function, ³⁷ , ³⁸ increased gray matter volume and thickness, ³⁸ and attenuation of hippocampal volume decline ³⁸ among older adults. The present findings were also robust after adjusting for usual gait speed, which is linked with cognitive decline ³⁹ and neuropathology. ⁴⁰ , ⁴¹ This study extends the exercise intervention literature to suggest that free‐living movement including light intensity physical activity, and not just exercise or moderate‐to‐vigorous intensity activity, are positively related to memory, visuospatial processing, language, and global cognition above‐and‐beyond gait speed. This might have potential implications for dementia prevention, as there is controversy over whether current treatments, which have serious side effects, modify the natural disease course. ⁴² The present findings extend work concentrating on moderate‐to‐vigorous intensity activity and suggest interventions to increase free‐living movement or decrease sedentary behavior may have utility for short‐term improvements in cognitive function among older adults at risk for dementia. This work also suggests that different activity patterns relate to different domains of cognitive function, thereby providing possible opportunities to target specific domains (particularly memory and visuospatial processing). Such interventions may also have enhanced effectiveness when conducted in combination with medications targeting cardiovascular risk factors (i.e., high blood pressure ⁴³ , ⁴⁴ and high cholesterol ⁴⁴ ) or other health behaviors.

Executive function is the cognitive domain most commonly associated with physical activity, as it is involved in planning and execution of goal‐oriented behaviors. ¹⁴ , ¹⁵ Yet, we did not find any associations between daily physical activity and executive function/attention in this study. One potential explanation for this inconsistent result is that executive function is often linked with performance of health behaviors. ¹⁵ As such, it is possible that structured, planned activities (i.e., exercise) are more strongly related to executive function than daily movement. It might also be that the factor score for executive function was too broad, as attention and executive function tests were combined due to model convergence issues. Future work should examine these associations using individual cognitive tests and information on the types of activities being performed.

In this study, we did not find cognitive function to be associated with subsequent changes in daily physical activity patterns. This is inconsistent with recent work finding associations between baseline cognitive function and changes in physical activity ²¹ , ²³ or bidirectional associations between cognitive function and physical activity. ¹⁴ , ¹⁶ One potential explanation is the length of follow‐up in this study, as there were no significant changes in daily physical activity patterns over the 1.8‐year study period (Figure 2). Although physical activity declines with age, ⁵ it is possible that patterns of activity engagement are more established and remain relatively stable by later life among some individuals. ⁴⁵ , ⁴⁶ This finding highlights the need for multifaceted interventions to target behavior change in addition to promoting physical activity. It might also be that cognitive decline is associated with changes in activity at certain times in the life course. Indeed, a longitudinal study of adults ≥ 50 years found lower delayed recall or verbal fluency scores were associated with less self‐reported physical activity engagement. This study also found that lower baseline cognition was linked to steeper declines in physical activity over 12 years, suggesting a stronger influence of midlife cognitive function on late life physical activity. ²³ An alternative explanation is the use of cognitive domain factor scores, where individual tests might be more strongly related to daily activity. Although we did not find support for a bidirectional relationship between physical activity and cognitive function in this study, we maintain the importance of cognitive function for engaging in daily movement. Future work is warranted to examine the directionality of these associations in more diverse samples with longer follow‐up.

When stratifying by work status or age, we found that baseline cognitive function was related to changes in certain daily activity patterns. Higher global cognition was associated with less annual decline in mean active bout length only among those ≥ 80 years. It is possible that higher levels of global cognition allowed participants to sustain active bouts through a greater ability to plan, maintain attention, and inhibit distractions. ¹⁴ , ¹⁵ , ¹⁷ Cognitive function might also be more important for maintaining activity among this group, as they tended to have lower usual gait speed and higher prevalence of depressive symptoms and cardiovascular disease than those < 80 years. An alternative explanation is selection bias, as older BLSA participants must have still been robust enough to complete the 3‐day, in‐person study visits. We also found a contradictory result that higher baseline visuospatial processing and global cognition were associated with greater annual decline in activity intensity among those who were currently working or volunteering or aged < 80 years compared to their respective counterparts. These groups tended to have lower depressive symptoms and cardiovascular disease, which are thought to mediate the cognition‐activity association. ¹ , ⁴⁷ , ⁴⁸ It is possible that cognitive function is related to activity intensity only among those with greater comorbidity burden. However, we would not expect negative results in the groups with less chronic disease. There is also a small but unlikely chance that some movement accrued through employment activities was not captured if participants had an irregular work schedule (i.e., not working/volunteering every week). This may have underestimated physical activity in this group. Caution is needed when interpreting these results, as the sensitivity analyses were limited by small sample sizes. Future work is needed to examine whether cognitive function might be more strongly related to short‐term changes in daily activity among certain subgroups of older adults.

4.1. Limitations and strengths

This study has some limitations. First, BLSA participants are healthy older adults with high educational attainment, which potentially limits generalizability to the broader population of older adults. Second, there was a relatively short time between baseline and follow‐up (mean: 1.8 [range: 0.9–4.0] years). This suggests that, even at these older ages, large changes in cognitive function or daily activity in relatively healthy older adults may not occur in such a short follow‐up period. Future studies with longer follow‐up in more cognitively diverse populations are warranted to better test whether cognitive function is associated with changes in physical activity. Reverse causality is also a potential concern; however, we directly tested the directionality of associations. Third, participants with cognitive decline may not have returned for a follow‐up visit. We expect this would result in a conservative bias. Fourth, the sensitivity analyses had small sample sizes and should be interpreted as hypothesis generating. Finally, cognitive domains for executive function and attention were combined, which limits our ability to make claims about executive function‐ or attention‐specific associations.

This study also has several strengths. First, the BLSA is a well‐characterized cohort study. Second, cognitive domain factor scores were derived from a neuropsychological test battery using confirmatory factor analysis. Third, wrist accelerometry data were utilized to characterize patterns of free‐living movement. Finally, we employed latent change score models to test the directionality of associations between cognitive domains and physical activity patterns and to evaluate potential reverse causality.

5. CONCLUSIONS

Among cognitively and physically intact older adults, higher levels of total amount of daily activity, longer activity bouts, and lower levels of sedentary time and physical activity fragmentation were associated with reduced decline across multiple cognitive domains. Yet, baseline level of cognitive function was not related to short‐term changes in daily physical activity in this sample. This suggests interventions to improve engagement in daily movement and reduce sedentary behavior might help attenuate short‐term cognitive decline among older adults at risk for dementia, where certain patterns may target different cognitive domains.

CONFLICT OF INTEREST STATEMENT

J.A.S. is a consultant with Edwards Life Sciences and is on the advisory board of BellSant, Inc. All other authors have no conflicts to disclose.

CONSENT STATEMENT

The study protocols were approved by the Internal Review Board of the Intramural Research Program of National Institutes of Health. All participants provided written informed consent.

Supporting information

Supporting Information

Supporting Information

Marino FR, Deal JA, Gross AL, et al. Directionality between cognitive function and daily physical activity patterns. Alzheimer's Dement. 2025;11:e70068. 10.1002/trc2.70068