Cross-Sectional Relationships of Physical Activity and Sedentary Behavior With Cognitive Function in Older Adults With Probable Mild Cognitive Impairment

Abstract

Background

Mild cognitive impairment (MCI) represents a transition between normal cognitive aging and dementia and may represent a critical time frame for promoting cognitive health through behavioral strategies. Current evidence suggests that physical activity (PA) and sedentary behavior are important for cognition. However, it is unclear whether there are differences in PA and sedentary behavior between people with probable MCI and people without MCI or whether the relationships of PA and sedentary behavior with cognitive function differ by MCI status.

Objective

The aims of this study were to examine differences in PA and sedentary behavior between people with probable MCI and people without MCI and whether associations of PA and sedentary behavior with cognitive function differed by MCI status.

Design

This was a cross-sectional study.

Methods

Physical activity and sedentary behavior in adults dwelling in the community (N = 151; at least 55 years old) were measured using a wrist-worn actigraphy unit. The Montreal Cognitive Assessment was used to categorize participants with probable MCI (scores of <26/30) and participants without MCI (scores of ≥26/30). Cognitive function was indexed using the Alzheimer Disease Assessment Scale-Cognitive-Plus (ADAS-Cog Plus). Physical activity and sedentary behavior were compared based on probable MCI status, and relationships of ADAS-Cog Plus with PA and sedentary behavior were examined by probable MCI status.

Results

Participants with probable MCI (n = 82) had lower PA and higher sedentary behavior than participants without MCI (n = 69). Higher PA and lower sedentary behavior were associated with better ADAS-Cog Plus performance in participants without MCI (β = −.022 and β = .012, respectively) but not in participants with probable MCI (β < .001 for both).

Limitations

This study was cross-sectional and therefore could not establish whether conversion to MCI attenuated the relationships of PA and sedentary behavior with cognitive function. The diagnosis of MCI was not confirmed with a physician; therefore, this study could not conclude how many of the participants categorized as having probable MCI would actually have been diagnosed with MCI by a physician.

Conclusions

Participants with probable MCI were less active and more sedentary. The relationships of these behaviors with cognitive function differed by MCI status; associations were found only in participants without MCI.

By 2030, there will be nearly 1 billion older adults worldwide.¹ Since age is the greatest risk factor for dementia,² the number of dementia cases is expected to increase substantially.³ Pharmaceutical therapies to treat dementia are still in their infancy,^4,5 and thus reducing the risk of dementia—and potentially dementia incidence—requires the development of effective lifestyle-based strategies.

Mild cognitive impairment (MCI) represents a critical phase to intervene, since it is a transitional stage between healthy cognition and dementia.⁶ MCI is defined as cognitive decline greater than expected for age and education level which does not interfere with independence,⁷ and is associated with up to a 30% increased risk of developing dementia within 5 years.⁸ By comparison, older adults without MCI develop dementia at a rate of 1% to 2% within 5 years.⁹ Providing effective strategies to maintain cognitive health during this transition period might slow the conversion to dementia.

One potential strategy is physical activity (PA)—a behavior associated with both cognitive and physical health.¹⁰ High levels of PA are prospectively linked to lower incidence of MCI and dementia,^11–15 and an estimated 17.7% of Alzheimer disease cases could be prevented through PA.¹⁶ Physical activity also has multidimensional health benefits, including reduced risk of mortality and chronic diseases such as type 2 diabetes mellitus and cardiovascular disease.^17–19 Given the importance of PA for health, current recommendations for older adults suggest 30 minutes of PA (in bouts as brief as 10 minutes of moderate intensity) 5 days/week.²⁰ Unfortunately, >95% of older adults are physically inactive (ie, do not engage in ≥150 minutes/week of activity ≥3.0 metabolic equivalents [METs]) and thus fall short of these recommendations.²¹

Although the importance of PA for cognitive health is well established, less is known about the impact of sedentary behavior on both physical and cognitive health. Sedentary behavior is any behavior which incurs ≤1.5 METs and includes activities such as sitting, television watching, and lying down; PA is any behavior incurring ≥3.0 METs.²² Recent evidence suggests sedentary behavior may be associated with poorer cognitive function and increased risk of cognitive impairment, although epidemiological data are needed to confirm this association.^23,24 Accumulating evidence also suggests sedentary behavior is associated with numerous chronic diseases, including type 2 diabetes mellitus and cardiovascular disease.^25,26 These chronic diseases are concomitantly linked with increased risk of cognitive decline and dementia,^27,28 providing further evidence that high sedentary behavior is a risk factor for cognitive health. Due to the increasing evidence suggesting PA and sedentary behavior are associated with cognitive health, older adults are recommended to limit discretionary sedentary behavior to <2 hours/day; avoid sitting for >30 minutes without standing; and concomitantly increase PA to ≥150 minutes/week.²⁹

Physical therapists are in a unique position to help influence both PA and sedentary behavior. The potent position of clinicians to maximize patient compliance by influencing behavior is why the US Preventive Task Force has recommended clinicians provide PA counseling since 1989.^30,31 In addition, randomized controlled trials using activity counseling among older adults in a primary care setting have been highly successful at increasing patient PA.^32,33 As such, these data further illustrate the importance of activity counseling in the clinical setting and suggest even brief questions about patient activity levels can have demonstrable improvements on patient health outcomes, such as cardiovascular disease and mortality risk.³⁴ Thus, as an important step to promote older adult cognitive health, clinicians should consult their older adult patients about their PA and sedentary behavior.

Since the current evidence suggests PA and sedentary behavior are important for healthy cognitive aging, a next step—given the window of opportunity to intervene for people with MCI—is an analysis of PA and sedentary behavior differences between people with MCI and those without MCI. Importantly, it is unclear whether people with MCI engage in different amounts of PA and sedentary behavior than their peers without MCI and whether the associations of PA and sedentary behavior with cognitive function are the same or different between older adults with MCI and those without MCI. Indeed, because of underlying neurobiological differences between older adults with MCI and those without MCI,³⁵ a functional weakening in the relationships of health behaviors with cognitive function may occur in MCI.³⁶ However, data are still needed showing the associations of PA and sedentary behavior with cognitive function attenuate based on MCI status in order to confirm this hypothesis. Answering these questions may help inform clinicians concerned about the cognitive health of their older adult patients, and help determine which of their older adult patients may see the greatest benefits to cognitive health from PA and sedentary behavior counseling.

To address these gaps in knowledge, we first investigated differences in PA and sedentary behavior between older adults with probable MCI and those without. In addition, we determined whether the relationships of PA and sedentary behavior with cognitive function differed based upon MCI status.

Methods

All participants provided written informed consent. Ethics approval for this study was obtained from the Vancouver Coastal Health Research Institute and the University of British Columbia's Clinical Research Ethics Board (H14-01301).

Protocol

For this cross-sectional study, we recruited and collected data between August 2014 and June 2016. At study entry, we ascertained general health, demographics, socioeconomic status, and education by questionnaire. Subsequently, we screened participants for MCI using the Montreal Cognitive Assessment (MoCA), with a score of <26 indicating probable MCI status.³⁷ Participants’ PA and sedentary behavior were then observed for ≥ 4 days using the MotionWatch 8 wrist-worn actigraphy unit (CamNtech Ltd; Cambridge, United Kingdom). Following MotionWatch 8 observation, we measured cognitive function for all participants using the Alzheimer Disease Assessment Scale-Cognitive-Plus (ADAS-Cog Plus).³⁸

Participants

Participants were recruited from Vancouver, British Columbia, by advertisements placed in local community centers, newspapers, and word-of-mouth referrals. Potential participants were considered eligible if they met the following 3 criteria: men and women at least 55 years old and living in the greater Vancouver area; scores of >24/30 on the Mini-Mental State Examination (MMSE)³⁹; and ability to read, write, and speak English with acceptable visual and auditory acuity. Participants were ineligible if they were diagnosed with dementia of any type; diagnosed with another type of neurodegenerative or neurological condition; taking medications that may negatively affect cognitive function; planning to participate in or currently enrolled in a clinical drug trial; or unable to speak as judged by an inability to communicate by phone. Of the 152 total potential participants who were recruited for this study, only 1 dropped out because of a transient ischemic attack unrelated to the study. Thus, our obtained sample size was 151 participants.

Measurement of PA and Sedentary Behavior

We measured PA and sedentary behavior using a valid and reliable measure among older adults, the MotionWatch 8.⁴⁰ Briefly, the MotionWatch 8 is a uniaxial, wrist-worn accelerometer designed to observe acceleration ranging in magnitude from 0.01g to 8g with a frequency of 3–11 Hz. The filtered acceleration signal is digitized, and the magnitude is summed over a user-specified time interval. At the end of each interval, the summed value or activity “count” is stored in memory and the integrator is reset. For the current study, we used 60-second epochs.⁴¹

At study entry, participants were fitted with the MotionWatch 8. Details of the measurement protocol used for the MotionWatch 8 can be found elsewhere.⁴² Consistent with established protocol for MotionWatch 8, participants wore the device on the nondominant wrist for a period of ≥4 days,⁴² which is enough to provide reliable estimates of PA (ICC = .90, 95% CI = .86–.93) and sedentary behavior (ICC = .91, 95% CI = .87–.94). After collection, stored activity counts were downloaded and saved to an IBM-compatible computer (IBM; SPSS, Chicago, Illinois) for subsequent data reduction and analysis.

Data Reduction

Data were analyzed using MotionWare 1.0.27 (CamNtech Ltd; Cambridge, United Kingdom). Data prior to recorded wake time on the first full day of recording were manually removed in order to only investigate full 24-hour recordings of activity. Thus, a participant with 6 nights of sleep recorded had 5 full days of activity recorded. Each day of activity consisted of when the participant self-reported being awake and out of bed. Participant self-report was confirmed via event marker time stamps from MotionWatch 8 or a consensus sleep diary that participants completed during the observation period.

A Microsoft Excel (Microsoft Corporation; Redmond, Washington) macro written by R.S.F. was used to reduce data to the following 4 components: wake time for the participant by day; daily calculations of time the participant spent in sedentary behavior (<1.5 METs) and PA (>3.0 METs), as determined from the established cutoff points;⁴⁰ average daily amounts of time spent in PA and sedentary behavior; and average percentage of the day spent in PA and sedentary behavior. Nonwear time of MotionWatch 8 was assessed as a period of consecutive 0 counts ≥120 minutes in length.⁴³ In total, only 2 participants had periods of non-wear time (X̅ = 202 minutes/14 days; SD = 107 minutes/14 days) according to this criterion. We therefore assumed participants did not remove the MotionWatch 8 during observation.

Measurement of Sedentary Behavior and PA Bouts Per Day

We also examined the average 10+ min bouts/day of PA and 30+ min bouts/day of sedentary behavior. A Python 2.7 code written by B.K.C. was used to analyze the data. The script loaded each participant's data from the Microsoft Excel spreadsheet into a Python data table. The script then cleaned and separated the data into tables, 1 for each type of activity level (ie, PA and sedentary behavior). The data loaded into the tables was a sequence of 1’s and 0’s, wherein a “1” on an epoch meant an activity level of the corresponding threshold was detected, and a “0” meant the activity level corresponding to the threshold was not detected. The script then iterated through each day's epochs with an incrementing counter, such that when a “1” was detected, the counter incremented. Once a break (ie, “0”) was detected, the counter was then saved to another data table that kept track of the length of the bouts (ie, streaks of activity at the corresponding activity threshold). The counter was then reset and began counting again when the next bout of activity was detected. Once the bout lengths were counted, we determined the average 10+ min bouts/day of PA and the average 30+ min bouts/day of sedentary behavior for each participant.

Cognitive Function

We used the ADAS-Cog Plus to examine global cognitive function.³⁸ The ADAS-Cog Plus uses a multidimensional item response theory model that can flexibly utilize item scores from multiple cognitive assessment instruments to generate a global cognitive function score and standard error of measurement for that score.⁴⁴ Scores are defined by the Alzheimer Disease Neuroimaging Initiative sample,⁴⁵ wherein the mean score for cognitively healthy older adults is about −1.0, the mean for MCI is about 0.0, and the mean for dementia cases is about 1.0. Thus, higher scores indicate poorer cognitive performance. The ADAS-Cog Plus score was computed using the following 4 methods: the 13-item Alzheimer Disease Assessment Scale (validity: ICC = .80; test-retest: r = .93);⁴⁶ Trail Making Tests A and B (validity: r = .36–.93; test-retest: r = .67);⁴⁷ Digit Span Forward and Backward (validity: r = .48–.85; test-retest: r = .62–.82);⁴⁸ and verbal fluency, consisting of animal fluency and vegetable fluency (validity: r = .44–.87; test-retest: r = .74).⁴⁸ Detailed descriptions of the procedures used for these tests can be found elsewhere.^46–48

Data Analyses

We performed all of our statistical analyses using SPSS 22.0. One participant was removed from our analyses as an outlier due to an ADAS-Cog Plus score which was >3 standard deviations above the mean score, suggesting the participant had possible severe cognitive impairment (MMSE = 25; MoCA = 17). Thus, our final sample size was 150. Given the nonnormal distribution of average 10+ min bouts/day of PA, we transformed it by means of a natural log transformation which was used for all subsequent analyses as it most closely approximated a normal distribution. Because this was an exploratory analysis, we did not use a Bonferroni adjustment to account for multiple comparisons.

Participant characteristics based on probable MCI status

Means and standard deviations were calculated for all variables of interest based upon probable MCI status. To determine demographic differences based on probable MCI status, we performed independent sample t tests for continuous variables and chi-square tests for categorical variables, using probable MCI status (yes/no) as the grouping variable. In addition, we performed analyses of covariance (ANCOVA) to determine differences in percentage of the day spent in PA; 10+ min bouts/day of PA; percentage of the day spent in sedentary behavior; 30+ min bouts/day of sedentary behavior; and cognitive function based on probable MCI status. These models controlled for age and sex differences while using probable MCI status as the grouping variable.

Relationship of cognitive function with PA and sedentary behavior based on probable MCI status

We then examined the relationship of cognitive function with PA and sedentary behavior for all participants. Multiple linear regression models were generated using ADAS-Cog Plus score as the dependent variable, while controlling for age, sex, and education. We generated 4 models, where the major independent variable of interest was percentage of the day spent in PA; 10+ min bouts/day of PA; percentage of the day spent in sedentary behavior; or 30+ min bouts/day of sedentary behavior. Beta estimates, model R², and P values are presented for each model. We plotted the predicted relationship between each major independent variable and ADAS-Cog Plus score.

We then examined if the relationship of cognitive function with PA and sedentary behavior differed based upon probable MCI status. Participants were stratified based upon probable MCI status and the same 4 multiple linear regression models were generated using ADAS-Cog Plus as the dependent variable, while controlling for age, sex, and education. We then performed z tests to determine if beta estimates for PA and sedentary behavior differed significantly based on probable MCI status. Beta estimates, model R², and P values are presented for each model based upon probable MCI status. The z scores and P values for comparing major independent variable estimates based on probable MCI status are also presented. We then illustrated the differences in the predicted relationship between each independent variable and ADAS-Cog Plus based upon cognitive status.

Role of the Funding Source

This work was supported by funding from the Jack Brown and Family Alzheimer Research Foundation Society. The sponsors played no role in the design, methods, participant recruitment, data collection, data analysis, or preparation of this article.

Results

Participant Characteristics Based on Probable MCI Status

Participant characteristics are described in Table 1. The mean age was 71.11 years (SD = 7.22 years) and 67.10% were female. Older adults without MCI had a mean MMSE score of 29.22 ± 0.10 (range = 27–30), and a mean MoCA score of 27.19 ± 0.13 (range = 26–30); older adults with probable MCI had a mean MMSE score of 28.65 ± 0.13 (range = 25–30), and a mean MoCA score of 22.84 ± 0.23 (range = 14–25). Older adults categorized with probable MCI were significantly older (t = 2.70, df = 149, P = .008) and more likely to be male (χ² = 5.22, P = .022). While controlling for age and sex differences, participants with probable MCI also engaged in significantly less percentage of daily PA (F = 4.81; df = 1,150; P = .030); fewer 10+ min bouts/day of PA (F = 7.94; df = 1,150; P = .005); and more 30+ min bouts/day of sedentary behavior (F = 4.04; df = 1,150; P = .046); they also had poorer cognitive performance on the MMSE (F = 7.36; df = 1,150; P = .007), MoCA (F = 215.78; df = 1,150; P < .001), and ADAS-Cog Plus (F = 25.22; df = 1,150; P < .001).

Table 1.

Participant Characteristics

Characteristic	All Participants (N = 150)	Older Adults Without MCI (n = 69)	Older Adults With Probable MCI (n = 81)	P Value
Age (y)	71.11 (7.22)	69.42 (6.37)	72.54 (7.62)	.008
% Women	67.10	77.90	59.80	.022
Education (%)
High school diploma or less	18.40	14.70	20.70	.303
Trade school	11.20	7.40	13.40
Some university	15.80	14.70	17.10
University diploma or higher	54.60	63.20	48.80
% Retired	77.60	83.80	81.70	.802
Smoking status (%)
Current smoker	2.00	1.50	2.40	.877
Former smoker	50	50.00	48.80
Never smoked	50.00	48.50	48.80
Physical activity and sedentary behavior
%PA-Timea	10.25 (6.51)	12.07 (7.19)	8.69 (5.45)	.030b
LN(average 10+ min bouts/day of PA)c	0.52 (0.44)	1.20 (1.45)	0.40 (0.36)	.005b
%Sedentary behavior-Timed	59.62 (12.00)	57.24 (12.38)	61.65 (11.35)	.161b
Average 30+ min bouts/day of sedentary behavior	3.72 (1.83)	3.30 (1.73)	4.07 (1.85)	.046b
Cognitive function
Mini-Mental State Examination	28.91 (1.07)	29.22 (0.87)	28.65 (1.15)	.007b
Montreal Cognitive Assessment	24.84 (2.77)	27.19 (1.10)	22.84 (2.11)	<.001b
ADAS-Cog Pluse	−0.79 (0.65)	−1.11 (0.57)	−0.52 (0.59)	<.001b

^aAverage percentage of the day spent in physical activity (PA).

^b  F test while controlling for age and sex.

^cNatural log transformation for average 10+ minute bouts/day of PA.

^dAverage percentage of the day spent in sedentary behavior.

^eAlzheimer Disease Assessment Scale Plus.

Association of Cognitive Function With PA and Sedentary Behavior

Our models describing the relationship of cognitive function with PA and sedentary behavior are described in Table 2. We found a significant relationship between higher percentage of the day spent in PA and better cognitive performance (β = −.017, P = .024), and there was a marginal relationship between greater 10+ min bouts/day of PA and better cognitive performance (β = −.203, P = .070). There was also a marginal relationship between higher percentage of the day spent in sedentary behavior and poorer cognitive performance (β = .007, P = .089). Finally, we found a significant association between greater 30+ min bouts/day of sedentary behavior and poorer cognitive performance (β = .061, P = .016). These relationships are illustrated in Figure 1.

Figure 1.

Association of sedentary behavior and physical activity (PA) with cognitive function. (A) Association of percentage of day spent in PA (%PA-Time) with ADAS-Cog Plus score. (B) Association of number of 10+ minute bouts/day of PA with ADAS-Cog Plus score. (C) Association of percentage of day spent in sedentary behavior (%sedentary behavior-Time) with Alzheimer Disease Assessment Scale Cognitive Subscale Plus score (ADAS-Cog Plus score). (D) Association of number of 30+ minute bouts/day of sedentary behavior with ADAS-Cog Plus score. Models controlled for age, sex, and education.

Table 2.

Association of Sedentary Behavior and Physical Activity With Alzheimer Disease Assessment Scale-Cognitive-Plus (ADAS-Cog Plus) Score^a

Independent Variable (IV)	ModelbR2	ΔR2 Due to IV	IV β	IV P
Physical activity (PA)
%PA-Timec	.321	.025	−0.017	.024
LN(average 10+ min bouts/day of PA)d	.312	.016	−0.203	.070
Sedentary behavior
%Sedentary behavior-Timee	.311	.014	0.007	.089
Average 30+ min bouts/day of sedentary behavior	.324	.027	0.061	.016

^aThe dependent variable was the ADAS-Cog Plus score.

^bModels controlling for age, sex, and education.

^cAverage percentage of the day spent in PA.

^dLog-transformed average 10+ min bouts/day of PA.

^eAverage percentage of the day spent in sedentary behavior.

Relationship of Cognitive Function With PA and Sedentary Behavior Based on Probable MCI Status

Table 3 describes the relationship of cognitive function with PA and sedentary behavior based on probable MCI status. For older adults without MCI, higher percentage of the day spent in PA and greater 10+ min bouts/day of PA were associated with better cognitive performance (β = −.022 [P = .024] and β = −.286 [P = .046], respectively). However, neither PA characteristic was associated with cognitive performance for older adults categorized with probable MCI. In addition, the beta estimates for percentage of the day spent in PA (z = 2.412, P = .016) and 10+ min bouts/day of PA (z = 1.986, P = .047) differed significantly based on probable MCI status.

Table 3.

Association of Sedentary Behavior and Physical Activity With Alzheimer Disease Assessment Scale-Cognitive-Plus (ADAS-Cog Plus) Score Based on Probable Mild Cognitive Impairment (MCI) Status^a

	Non-MCI (n = 69)				Probable MCI (n = 81)				IV β Differences Based on Probable-MCI Status
Independent Variable (IV)	ModelbR2	ΔR2 Due to IV	IV β	IV P	ModelbR2	ΔR2 Due to IV	IV β	IV P	z Score	P Value
Physical activity (PA)
%PA-Timec	.209	.066	−0.022	.024	.322	<.001	<0.001	0.993	2.412	.016
LN(average 10+ min bouts/day of PA)d	.195	.052	−0.286	.046	.329	.007	0.148	.378	1.986	.047
Sedentary behavior
%Sedentary behavior-Timee	.199	.056	0.012	.038	.322	<.001	<0.001	.948	1.536	.125
Average 30+ min bouts/day of sedentary behavior	.188	.045	0.075	.064	.353	.010	0.033	.282	0.830	.407

^aThe dependent variable was the ADAS-Cog Plus score.

^bModels controlling for age, sex, and education.

^cAverage percentage of the day spent in PA.

^dLog-transformed average 10+ min bouts/day of PA.

^eAverage percentage of the day spent in sedentary behavior.

We also found higher percentage of the day spent in sedentary behavior was associated with poorer cognitive performance for participants without MCI (β = .012, P = .038); there was a marginal relationship between greater +30 min bouts/day sedentary behavior and poorer cognitive performance for participants without MCI (β = .075, P = .064). By comparison, there was no relationship for either sedentary behavior characteristic with cognitive performance for older adults categorized with probable MCI. Finally, the relationship of percentage of the day spent in sedentary behavior with cognitive performance was marginally different based on probable MCI status (z = 1.536, P = .125). These relationships are illustrated in Figure 2.

Figure 2.

Association of sedentary behavior (SB) and physical activity (PA) with cognitive function based on the presence of mild cognitive impairment (MCI). (A) Association of percentage of day spent in PA (%PA-Time) with ADAS-Cog Plus score. (B) Association of number of 10+ minute bouts/day of PA with ADAS-Cog Plus score. Models controlled for age, sex, and education. (C) Association of percentage of day spent in sedentary behavior (%sedentary behavior-Time) with Alzheimer Disease Assessment Scale Cognitive Subscale Plus score (ADAS-Cog Plus score). (D) Association of number of 30+ minute bouts/day of sedentary behavior with ADAS-Cog Plus score). Models controlled for age, sex, and education. Solid lines = older adults without MCI; dashed lines = older adults with probable MCI.

Discussion

Our results suggest older adults with probable MCI engage in less PA and more sedentary behavior compared with adults without MCI. In addition, the relationship of PA and sedentary behavior (to a lesser degree) with cognitive function differs by cognitive status such that PA and sedentary behavior are only associated with cognitive function among participants without MCI. We now provide potential explanations for these findings, as well as discuss how these findings can be applied to clinical practice.

Differences in PA and Sedentary Behavior by Cognitive Status: Does Activity Level Change Because of MCI Conversion?

Previous studies suggest low PA and high sedentary behavior are both risk factors for cognitive impairment in later life;^15,29 however, our study is the first to show that older adults with probable MCI are less active and more sedentary than their cognitively healthy peers. One potential explanation is younger adults at high risk for MCI in later life—due to lower PA and higher sedentary behavior—continue to be less active and more sedentary into older adulthood, often becoming cognitively impaired as they age. Indeed, the current evidence suggests that PA and sedentary behavior patterns worsen from childhood into young adulthood and then stabilize from middle adulthood onward.⁴⁹ It is therefore plausible that older adults with MCI are less active and more sedentary throughout their adult lives, and this behavior has continued into older adulthood.

It is also plausible that a reciprocal association is present, such that conversion to MCI influences PA and sedentary behavior. In particular, this effect might occur through diminished executive function.^8,50,51 Briefly, executive function is a broad term used to define planning and problem-solving and is known to significantly decline with age.^52,53 Loss of executive function capability has been shown to negatively impact older adult independence and functionality.^54–56 Increasing evidence also suggests PA and sedentary behavior have a bidirectional relationship with executive function such that changes in executive function can predict changes in activity levels, and vice versa.^57–59 Thus, older adults living with MCI may have impaired decision making about engaging in PA or sedentary behavior, due to impaired executive function capabilities.

Differences in the Relationships of PA and Sedentary Behavior With Cognitive Function by MCI Status: Are There Underlying Differences in the MCI Brain?

To our knowledge, this is the first study to report differences in the relationships of PA and sedentary behavior with cognitive function based upon cognitive status. One explanation is that there might be a minimum threshold of PA and/or a maximum threshold of sedentary behavior required to elicit a relationship with cognitive function. As discussed previously, older adults with probable MCI engaged in less PA and more sedentary behavior than adults without MCI. As such, older adults with MCI may not meet a minimum threshold level of PA—or may exceed a maximum threshold of sedentary behavior—which may lead to nonsignificant associations between these health behaviors and cognitive function. This interpretation of our findings may help explain why exercise and PA interventions for older adults with MCI can lead to significant improvements in cognitive function.^60,61 Potentially, exercise and PA interventions for older adults with MCI help ameliorate cognitive function by providing a necessary threshold level of PA and concomitant reduction in sedentary behavior. By comparison, the effects of PA and exercise interventions on older adults without MCI appear to be less substantial,⁶² perhaps due to higher levels of basal activity in these adults.

An alternative explanation for our findings is that the relationships of PA and sedentary behavior with cognitive performance differ by MCI status because of underlying neurobiological differences between older adults with MCI and those without.⁶³ For example, compared with older adults without MCI, those with MCI have greater amounts of beta-amyloid accumulation,⁶⁴ accelerated atrophy in the medial temporal lobe,⁶⁵ and decreased connectivity of the posterior cingulate gyrus and hippocampus with the whole brain.⁶⁶ These underlying changes in the MCI brain may potentially alter the relationships of health behaviors with cognitive function,³⁵ leading to an attenuation—or functional weakening—of the relationships of PA and sedentary behavior with cognitive function.³⁶

Clinical Applications

The findings of our study are also applicable toward improving clinical practice. First, given that PA and sedentary behavior can have important implications on older adult physical and cognitive health,^10,29 all clinicians should make a serious effort to counsel their patients on PA and sedentary behavior.³⁴ Although the health care system has serious potential as a tool for promoting changes in older adult PA and sedentary behavior, its potential is not being fully realized in clinical practice.⁶⁷ A first step to utilizing this untapped potential to promote behavior change is for clinicians to track their older adult patients’ PA and sedentary behavior using objective monitors, such as a pedometer or a FitBit. Although activity trackers may not be feasible for some clinicians to use in their practice, clinicians should at the very least ask brief questions about activity during their consultations with older adult patients since it can have important implications on both the physical and cognitive health of older adults.⁶⁸

Second, our data suggest that PA appears to have a stronger relationship with cognitive function than does sedentary behavior. A recently published systematic review examining the relationship of sedentary behavior with later life cognitive decline found that in order to best promote healthy cognitive aging, older adults should limit their sedentary behavior and concomitantly increase their PA to ≥150 minutes/week.²⁹ In order to best promote healthy cognitive aging, it may thus be prudent at this time for clinicians to focus on ensuring their older adult patients obtain this threshold level of PA, and to advise their patients to limit sedentary behavior. Over 95% of older adults are underactive, and thus the greatest benefits to older adult physical and cognitive health may occur by increasing PA.

Although the main findings of our study are complex and require future investigation, the recommendations highlighted above are practical for both older adults with MCI and those without. We therefore strongly suggest that clinicians who are concerned about their older adult patients’ cognitive health should track patient activity levels with objective monitors, if possible, and at least ask simple questions about activity levels during patient visits; and encourage their older adult patients to engage in ≥150 minutes/week of PA and limit their sedentary behavior as much as possible.

Limitations and Future Research

This study was cross-sectional and therefore cannot establish whether conversion to MCI attenuates the relationship of PA and sedentary behavior with cognitive function. Although MoCA is a standard measure for determining the cognitive status of older adults, MCI diagnosis is confirmed through a physician based on several criteria, including medical history; assessment of independence and activities of daily living; subjective memory complaints; neurological assessment; and laboratory tests and neuroimaging.^8,50,51 Given that we did not confirm the diagnosis of MCI with a physician, we cannot conclude how many of the participants we classified as living with probable MCI would actually be diagnosed with MCI by a physician.

This was a secondary analysis of a study which did not obtain information on participant comorbidities, and thus we did not account for these potential confounders within our analyses. Future studies will need to determine whether the associations of PA and sedentary behavior with cognitive performance do indeed differ by cognitive status when accounting for comorbid conditions.

Our findings may have been influenced by an overlap of the constructs of PA and sedentary behavior. Specifically, our PA and sedentary behavior data had a medium-sized negative correlation (r = −.515); however, a recent systematic review has also concluded that total PA and total sedentary behavior have a modest to medium-sized correlation.⁶⁹

Although we used a measure of PA and sedentary behavior that has good evidence of validity and reliability within this population,^40,42 given that we previously calibrated this device for PA and sedentary behavior using this sample of older adults, the MotionWatch 8 may still have over- or under-estimated participant PA and sedentary behavior. We therefore suggest that future investigations examine the concurrent validity of the MotionWatch 8 against another type of objective measurement tool (preferably a hip accelerometer)^70,71 and examine whether the findings reported in this manuscript can be replicated in other samples using other types of objective measures.

Future intervention research will also be needed to determine the minimum effective dose of PA—or sedentary behavior—necessary for maintaining or improving cognitive function, and how this might differ by MCI status. In addition, future research will need to examine whether the relationships of PA and sedentary behavior with cognitive function differ by MCI status due to underlying neurobiological changes.

Conclusions

This study found that older adults with probable MCI engage in less PA and more sedentary behavior than older adults without MCI. Our results also suggest the relationships of PA and sedentary behavior with cognitive function differ by probable MCI status. Future investigations should examine if a threshold amount of PA and sedentary behavior is required to maintain cognitive health in older adults, or if underlying neurobiological changes associated with MCI alter the relationships of PA and sedentary behavior with cognitive function.

Author Contributions

Concept/idea/research design: R.S. Falck, J.R. Best, G.J. Landry, J.C. Davis, T. Liu-Ambrose

Writing: R.S. Falck, J.R. Best, J.C. Davis, B.K. Chiu, T. Liu-Ambrose

Data collection: G.J. Landry

Data analysis: R.S. Falck, G.J. Landry, J.R. Best, J.C. Davis, B.K. Chiu, T. Liu-Ambrose

Project management: G.J. Landry, T. Liu-Ambrose

Fund procurement: T. Liu-Ambrose

Providing participants: G.J. Landry, T. Liu-Ambrose

Providing facilities/equipment: T. Liu-Ambrose

Providing institutional liaisons: T. Liu-Ambrose

Clerical/secretarial support: T. Liu-Ambrose

Consultation (including review of manuscript before submitting): R.S. Falck, G.J. Landry, J.R. Best, J.C. Davis, B.K. Chiu, T. Liu-Ambrose

R.S. Falck and G.J. Landry are co-first authors.

Ethics Approval

Ethical approval for this study was obtained from the Vancouver Coastal Health Research Institute and the University of British Columbia's Clinical Research Ethics Board (H14-01301). All participants provided written informed consent.

Funding

This work was supported by funding from the Jack Brown and Family Alzheimer Research Foundation Society. R.S. Falck was funded by the University of British Columbia Rehabilitation Science's Doctoral Award. J.C. Davis is a Canadian Institutes of Health Research Post-Doctoral Fellow. G.J. Landry is a Canadian Institutes of Health Research Post-Doctoral Fellow. T. Liu-Ambrose is a Canada Research Chair (Tier 2) in Physical Activity, Mobility, and Cognitive Neuroscience.

Disclosures/Presentations

The authors completed the ICJME Form for Disclosure of Potential Conflicts of Interest and reported no conflicts of interest.