Total Sedentary Time and Cognitive Function in Middle-Aged and Older Adults: A Systematic Review and Meta-analysis

Abstract

Background

An estimated 47 million people have dementia globally, and around 10 million new cases are diagnosed each year. Many lifestyle factors have been linked to cognitive impairment; one emerging modifiable lifestyle factor is sedentary time.

Objective

To conduct a systematic review and meta-analysis of peer-reviewed literature examining the association between total sedentary time with cognitive function in middle-aged and older adults under the moderating conditions of (a) type of sedentary time measurement; (b) the cognitive domain being assessed; (c) looking at sedentary time using categorical variables (i.e., high versus low sedentary time); and (d) the pattern of sedentary time accumulation (e.g., longer versus shorter bouts). We also aimed to examine the prevalence of sedentary time in healthy versus cognitively impaired populations and to explore how experimental studies reducing or breaking up sedentary time affect cognitive function. Lastly, we aimed to conduct a quantitative pooled analysis of all individual studies through meta-analysis procedures to derive conclusions about these relationships.

Methods

Eight electronic databases (EMBASE; Web of Science; PsycINFO; CINAHL; SciELO; SPORTDiscus; PubMed; and Scopus) were searched from inception to February 2021. Our search included terms related to the exposure (i.e., sedentary time), the population (i.e., middle-aged and older adults), and the outcome of interest (i.e., cognitive function). PICOS framework used middle-aged and older adults where there was an intervention or exposure of any sedentary time compared to any or no comparison, where cognitive function and/or cognitive impairment was measured, and all types of quantitative, empirical, observational data published in any year were included that were published in English. Risk of bias was assessed using QualSyst.

Results

Fifty-three studies including 83,137 participants met the inclusion criteria of which 23 studies had appropriate data for inclusion in the main meta-analysis. The overall meta-analysis suggested that total sedentary time has no association with cognitive function (r = −0.012 [95% CI − 0.035, 0.011], p = 0.296) with marked heterogeneity (I² = 89%). Subgroup analyses demonstrated a significant negative association for studies using a device to capture sedentary time r = −0.035 [95% CI − 0.063, − 0.008], p = 0.012). Specifically, the domains of global cognitive function (r = −0.061 [95% CI − 0.100, − 0.022], p = 0.002) and processing speed (r = −0.067, [95% CI − 0.103, − 0.030], p < 0.001). A significant positive association was found for studies using self-report (r = 0.037 [95% CI − 0.019, 0.054], p < 0.001). Specifically, the domain of processing speed showed a significant positive association (r = 0.057 [95% CI 0.045, 0.069], p < 0.001). For prevalence, populations diagnosed with cognitive impairment spent significantly more time sedentary compared to populations with no known cognitive impairments (standard difference in mean = −0.219 [95% CI − 0.310, − 0.128], p < 0.001).

Conclusions

The association of total sedentary time with cognitive function is weak and varies based on measurement of sedentary time and domain being assessed. Future research is needed to better categorize domains of sedentary behaviour with both a validated self-report and device-based measure in order to improve the strength of this relationship. PROSPERO registration number: CRD42018082384.

Key Points

• 1. The association of total sedentary time with cognitive function varies based on the method of sedentary time measurement and the cognitive domain being assessed

• 2. Populations that have been diagnosed with mild cognitive impairment or dementia spend significantly more time sedentary compared to cognitively healthy populations.

• 3. Future research is needed to investigate associations of individual sedentary behaviours with cognitive function and examine the impact of cognitive load on this association.

Introduction

Cognitive function refers to multiple mental abilities including learning, thinking, reasoning, remembering, problem solving, decision making and attention [1]. Dementia can be defined simply as a significant loss of cognitive function that impacts functions of daily living [2]. An estimated 47 million people have dementia globally and around 10 million new cases are diagnosed each year [3]. Unfortunately, this number will continue to rise at an exponential rate due to a global increase in the number of people living to older age [4]. Dementia has a major impact on the individual, but also has detrimental physical, emotional, psychological, social, and economic effects on caregivers, families, and society as a whole [5]. It is estimated that the total global societal cost of dementia is US$818 billion per year (1.1% of global gross domestic product); making it a public health priority [6, 7].

With the absence of a pharmacological treatment for the disease, current medications can only modestly alleviate symptoms [8]. Thus, other strategies (i.e., lifestyle modifications) are imperative for addressing the forthcoming dementia pandemic. Although cognitive decline tends to materialize later on in life, it is experienced by every individual at different severities and rates [1]. The latest Lancet review published in 2020 on dementia prevention, intervention and care concluded that 40% of worldwide dementia cases can be attributed to 12 modifiable risk factors [9], which is three more risk factors than the original review published in 2017 [10]. While we know a lot about some risk factors, there may be other unexplored factors, for example, targeting the reduction of sedentary time and how it may impact cognitive function. Despite the well-known benefits of moderate to vigorous physical activity, only 15% of adults aged 65–79 achieve at least 150 min of moderate to vigorous physical activity per week [11] and in addition, spend an average of 10.1 h/day being sedentary [12].

The associations between sedentary time and non-communicable diseases such as cardiovascular disease, type 2 diabetes, some forms of cancers, as well as all-cause mortality are now well established [13]. Sedentary behaviour is defined as any waking behaviour in a seated or lying posture while expending less than or equal to 1.5 metabolic equivalents of energy [14]. The evidence on the health risks associated with too much sitting is now informing public health guidelines. For example the Canadian 24-h movement guidelines recommend that in addition to accumulating 150 min of moderate to vigorous physical activity per week, people should keep prolonged sitting time to a minimum, advising no more than eight hours of sitting per day [15].

Sedentary time has become a known modifiable determinant of health and an important predictor of healthy aging [16,17,18,19]. More recently, studies have emerged suggesting that higher levels of sedentary time may also be linked to lower levels of cognitive function and an increased risk of cognitive decline [20,21,22]. Previous reviews have investigated the relationship between sedentary behaviour and cognitive function [23,24,25,26]. From these reviews, we know that the relationship is inconsistent, and observed relationships are rather weak. More specifically, Falck et al. (2016) suggested higher levels of sedentary behaviour were associated with lower cognitive performance [24]. This was concluded from eight observational studies including adults 40 years and over. A later review by Copeland and colleagues (2017) investigated the relationship in adults 60 years and older [23]. They were able to include 14 studies, including five studies featured in Falck’s 2016 review [24]. Of these, only half reported finding associations between increased sedentary behaviour and decreased cognitive function and the results were not differentiated according to the exposure and outcome measures used (i.e., self-report versus device or specific cognitive domains). The review by Loprinzi (2019) included humans of all ages and focused specifically on memory, including 25 studies [25]. Loprinzi observed a conflicting association of sedentary behaviour with cognitive function. Results from the most recent review are not much clearer. Olanrewaju et al. (2020) also found varied and inconclusive evidence on the relationship between the two variables [26]. The main difference of this review from the preceding ones was that it excluded studies involving participants diagnosed with dementia. Out of the 18 studies included, seven of these studies reported associations between increased sedentary behaviour time and decreased cognitive function. Only four of the 18 studies were consistent with those included in the Copeland (2017) review [23]. Despite including 18 studies, there were no interventions identified; and there was too much heterogeneity to perform a meta-analysis. The current systematic review builds on these previous ones, addressing specific conceptual and methodological issues that these reviews could not avoid due to the nature of the literature at that time.

Although the definition of sedentary behaviour was properly defined in previous reviews, the large-scale heterogeneity found within the available studies at the time could stem from the broadly defined study exposure (i.e., sedentary behaviour) needed in order to synthesize the current state of the literature. In other words, a specific domain of sedentary behaviour (i.e., television viewing) needed to be synthesized alongside studies measuring total sedentary time. This is problematic as we know that television viewing is a poor proxy of overall sedentary time [27]. Thus, no reviews have specifically aimed to review the association of total sedentary time with cognitive function. Secondly, the inconsistent and weak relationships between sedentary behaviour and cognitive function reported in previous reviews suggests that this relationship needs to be examined under certain moderator conditions. None of the reviews examined the association between studies that have used self-reported measures versus those that have used a device (i.e., activPAL™). This is an important moderator to investigate as self-report has been shown to underestimate sedentary time when compared to device-based measurements [23]. Although touched upon in the review by Falck (2016), another moderator that warrants attention is which specific domains of cognition are most affected by total sedentary time. Insight here will assist researchers’ focus on cognitions that are more salient to total sitting time. Third, no previous reviews have interpreted the findings based on a threshold or cut-off that compares two categorical groups (i.e., high versus low sedentary time) as opposed to comparing the relationship using continuous variables. It is unknown whether this relationship will be more concrete if more extreme groups or categorical groups are used; this may have implications for how researchers collect and analyze total sitting time data. Fourth, no previous review has interpreted the associations based on how sedentary time was accumulated throughout the day (i.e., short, frequent bouts vs. longer uninterrupted bouts). Recent investigations have highlighted the negative aspects of accumulating prolonged bouts of sedentary time regardless of meeting physical activity guidelines [17]. Thus, it is recommended to incorporate short bouts of activity to break up sedentary periods throughout the day [28]. Fifth, no previous reviews have synthesized the findings based on the prevalence of sedentary time in populations diagnosed with mild cognitive impairment or dementia versus populations deemed as cognitively healthy. Since the older adult population is the most sedentary and inactive population, it makes it hard to tease out a bidirectional relationship (i.e., is more sitting causing cognitive decline or is cognitive decline resulting in more sitting?). Therefore, differentiating populations that may be more susceptible to increased amounts of sitting, (i.e., people with cognitive impairment, transitioning into retirement, etc.) could help tease out any mixed association. This also highlights the importance of including both middle-aged and older adults. Sixth, experimental studies were scarce at the time of these previous reviews, but with the growing interest in this relationship, synthesizing any experimental studies aiming to reduce or break up sedentary time is important as it could illustrate whether we are able to manipulate cognitive function with sitting time. Seventh and finally, due to the heterogeneity of the literature at the time of these previous reviews, they were not able to perform a meta-analysis to quantify the effect size of the relationship. Since these reviews, there have been many more studies published on the relationship of sedentary behaviour with cognitive function, which allows this review to have a narrower view of the exposure variable, sedentary time, alongside being able to evaluate the relationship with the aforementioned moderators.

Addressing the abovementioned issues, the overarching objective of this study was to conduct a systematic review to explore the relationship of total sedentary behaviour time with cognitive function in middle-aged and older adults. A specific objective was to examine the relationship under the following moderator conditions: a) self-reported versus device based sedentary time measurement; b) the cognitive domain being assessed (e.g., working memory, processing speed, etc.); c) looking at sedentary time using categorical variables (i.e., high versus low sedentary time) and d) the pattern of sedentary time accumulation (e.g., longer versus shorter bouts). We also aimed to examine the prevalence of sedentary time in healthy versus cognitively impaired populations. Additionally, we aimed to explore how experimental studies reducing or breaking up sedentary time affects cognitive function. Lastly, we aimed to conduct a quantitative pooled analysis of all individual studies through meta-analysis procedures to derive conclusions about these relationships. In doing so, we intended to determine whether the aforementioned variables on their own, or in conjunction with one another served to change the strength and/or direction of relationships between total sedentary time and cognitive function.

Methods

The protocol for this review is registered on PROSPERO (registration number: CRD42018082384). The review and meta-analysis were also completed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PRISMA). This paper was also conducted in accordance with the PRISMA in Exercise, Rehabilitation, Sports Medicine, and Sports Science guidelines (PERSIST).

Search Strategy

We searched the following electronic databases from inception to February 13th, 2021: Excerpta Medica (EMBASE); Web of Science; PsycINFO; Cumulative Index to Nursing and Allied Health Literature (CINAHL); Scientific Electronic Library Online (SciELO); SPORTDiscus; PubMed; and Scopus. The databases were searched using a combination of controlled vocabulary (MeSH) and free text terms. The review authors along with librarians at The University of Queensland and Western University developed the search strategy. The search included terms related to the exposure (i.e., sedentary behaviour), the population (i.e., middle-aged and older adults), and the outcome of interest (i.e., cognitive function) (See Additional file 1 for an example search strategy for Scopus). All database searches were appropriately revised to suit the specific database. Where the database software had the function, we used "forward" citation searches (showing the sources that cite these articles) on the studies we included in this systematic review. Additionally, reference lists of all relevant reviews on cognitive outcomes and sedentary behaviour from the last 10 years were hand-searched. These reviews were identified in the initial title and abstract screening and through a search of our key terms in Cochrane. All additional articles identified through these other sources were subject to the same eligibility criteria and screening process as those found through the electronic database searches.

Selection Criteria

Types of Participants

The focus of this review was middle-aged and older adults. Therefore, we included studies where the mean age of the population was aged 40 or over [29]. Studies from any country were included. If the data set had been used more than once, the publication most relevant and appropriate for the current review was included.

Types of Interventions/Exposure

This review examined studies where the intervention or exposure was sedentary behaviour, as defined by each individual study. The definition of sedentary behaviour is sitting or lying behaviours during waking time that expend low levels of energy (≤ 1.5 metabolic equivalents) [14]. Studies were included if they reported the total time spent ‘sedentary’ per day or week, or if they reported the percentage of daily waking hours in ‘sedentary behaviour’ per day or week. Thus, any reference to ‘sedentary time’ in the current review can be interpreted as ‘total’ or all-encompassing sedentary time unless stated otherwise. Studies were excluded if they defined and/or measured sedentary behaviour as a lack of physical activity, included sleep time in the reported sedentary behaviour time, only measured leisure sedentary time, or only measured a specific domain of sedentary behaviour. Studies were also excluded if they were a multicomponent lifestyle intervention where sedentary behaviour was only one component of the intervention. In other words, if sedentary behaviour was only a ‘control’ condition of the intervention, it was excluded. Studies specifically investigating bedrest or designed for cognitive rehabilitation or training were excluded.

Types of Comparison

Studies with any comparators or no comparators were eligible for inclusion.

Types of Outcome Measures

Studies measuring cognitive function and/or cognitive impairment/decline (i.e., dementia or mild cognitive impairment [MCI]) using a recognised method or measure were included. Cognitive outcomes from each study were tabulated and categorized based on the authors’ classification of which domain the task measured (i.e., processing speed, episodic memory, etc.). If the task was used in more than one study and the authors reported different domains, discrepancies were solved by the current study’s authors using guidance from an article which reviewed the general structures of cognitive domains, along with assessment strategies for differentiating them [30]. The various outcomes of cognitive function from each study were grouped into one or more of the following domains: (1) processing speed; (2) episodic memory; (3) global cognitive function; (4) motor skills and construction; (5) executive function; (6) cognitive flexibility and (7) working memory. The definition and acceptable cognitive tests for each domain are described in Additional file 2. Studies that measured cognitive function as defined by brain volume or cerebral blood flow using Magnetic Resonance Imaging (MRI) or Positron Emission Tomography (PET) were excluded.

Types of Studies

Quantitative empirical studies published in any year were included, including observational (cross-sectional and/or longitudinal) and intervention/experimental studies. Discussion articles, conference proceedings, book chapters, theses or commentaries not presenting empirical research in a peer-reviewed journal were excluded. Studies published in a language other than English were also excluded.

Selection of Studies

Titles and abstracts of the identified studies were independently checked by two review authors using Covidence systematic review software and those that clearly did not meet the inclusion criteria were excluded. Full texts were then also reviewed by two authors. At both stages of the screening process, the authors discussed any discrepancies in their initial judgements and reached a consensus.

Data Extraction

Data from each of the included studies were extracted independently by one review author and checked by a second author for accuracy. The following data from each of the included studies were extracted and can be found in Tables 1, 2, 3 and 4: (1) author name and year; (2) study design; (3) participant characteristics (country, number of participants in each group, mean age, percent of female participants, and sedentary time reported; (4) measure of exposure (i.e., sedentary behaviour); (5) measure of outcome (i.e., cognitive function); (6) covariates used in the least and fully adjusted models (if applicable) and (7) main findings/numerical results for each outcome of interest (i.e., correlation coefficients, mean [SD], effect sizes, p values, etc.). Additional information that was extracted, but not reported within Tables 1, 2, 3 and 4 consisted of: (1) study objectives; (2) source of recruitment/method; (3) inclusion/exclusion criteria; (4) number of groups and method of group allocation (if application); (5) type of data (i.e., binary or continuous) for both exposure and outcome variables; (6) time frame of intervention/observation; (7) statistical methodology (i.e., subgroup analysis, intention to treat, etc.); (8) conclusion; (9) limitations identified by the authors and 10) any declared conflict of interest. If any study reports were incomplete or if data were missing, corresponding study authors were contacted via email.

Table 1 Summary and characteristics of cross-sectional and longitudinal studies reporting an association of sedentary behaviour with cognitive function

Table 2 Summary and characteristics of cross-sectional studies reporting the associations for pattern of sedentary behaviour accumulation with cognitive function

Table 3 Summary and characteristics of cross-sectional studies reporting prevalence of sedentary behaviour for clinical and non-clinical populations

Table 4 Summary and characteristics of experimental studies reporting an association of sedentary behaviour with cognitive function

Quality Assessment

Two review authors assessed the quality of each paper independently using the study quality checklist QualSyst proposed by Kmet et al. (2004), allowing assessment of both experimental and quasi-experimental studies [31]. Studies were scored (0—No, 1—Partial, 2—Yes) on 14 criteria. Aspects covered included the quality of study design, confounders, blinding, selection bias and misclassification bias. The sum of all scores was then divided by the highest possible score for each study. Any discrepancies between the two authors’ assessment were further discussed until a quality percentage score was agreed upon. A quality score of ≥ 75% indicates strong quality, a score between 55 and 75% moderate quality, and a score < 55% weak quality.

Data Synthesis and Analysis

Systematic Review

First, regardless of the statistical procedure used, all studies that reported an association (cross-sectional, longitudinal and experimental studies) of sedentary behaviour with cognitive function were synthesized into a pinwheel based on the method of sedentary behaviour measurement and the cognitive domain that was assessed. Method of measurement was separated into three categories (1) self-report; (2) activPAL™ or (3) other device (i.e., Actigraph). The two device-based measures were separated due to activPAL being the gold standard for measuring sedentary time, as they can differentiate sitting from standing. Cognitive function was separated into the seven various cognitive domain categories stated previously. Second, studies were synthesized based on associations reported using sedentary behaviour as a categorical variable (i.e., high versus low groupings). Third, studies that reported on the pattern of sedentary time accumulation (i.e., 30 + min bouts or number of interruptions) and its association with cognitive function. Fourth and finally, studies comparing prevalence of sedentary behaviour between healthy versus cognitively impaired populations.

Meta-analysis

Studies that were of high quality (i.e., scoring 75% or greater on the quality score outcome) were considered for inclusion in the meta-analysis. Studies that analysed cognitive function and sedentary behaviour as continuous variables were pooled into a random effects model meta-analysis using Comprehensive Meta-analysis Software (version 3). Random-effects models were chosen as heterogeneity was expected given the differences in study populations and procedures. Heterogeneity was determined by Cochran’s Q statistic and I² values (values of 25, 50 and 75 were considered to indicate low, moderate and high heterogeneity, respectively) [32]. Planned sub-group and meta-regression analyses were conducted to examine the contribution of specific variables to heterogeneity. These were specified a priori and were as follows: measurement type (activPAL™, other device or self-report), outcome assessment (cognitive flexibility, episodic memory, executive function, global cognitive function, motor skills and construction, processing speed and working memory), physical activity (studies controlling for physical activity time), number of covariates (number of covariates used in the analysis), age (mean age of population), percent female (percent of the population reported as females) and sedentary time (sedentary time reported in hours). Sub-group analyses were employed when five or more studies were available that used the same design (i.e., cross-sectional), and reported a correlation, standardized or unstandardized regression coefficient [33]. Where r was not reported, we transformed standardized beta (β) values to r based on the formula described by Peterson et al. (2005) [r = β + 0.05λ [34]. Where unstandardized betas were reported, they were converted to standardized values with the following formula: [standardized beta = unstandardized beta x (SD of independent variable/SD of the dependent variable)]. If a study reported data for more than one model, the least adjusted model was used within the analysis to improve comparability between studies. If a study reported data for more than one outcome, each study was used as the unit of analysis in the overall model. Studies reporting prevalence of sedentary time between populations diagnosed with mild cognitive impairment or dementia and populations considered cognitively healthy were also analyzed in a random effect meta-analysis using means, sample sizes and p values. If a study was deemed unsuitable for the meta-analysis (i.e., units not comparable or missing key information) the authors were contacted in an attempt to obtain the necessary data. We had intended to also calculate the pooled mean effect of the longitudinal, experimental and studies reporting on patterns of sedentary behaviour time accumulation. However, this was not possible due to the low number and heterogeneity between the studies. Publication bias was assessed using a funnel plot of all included studies to allow for visual inspection of publication bias.

Results

Search Results/Study Characteristics

The initial electronic database search identified 5886 papers with one additional paper found through forward searching. After duplicates were removed, 4673 papers remained for title and abstract screening. After the initial screening stage, 124 papers were identified for full text review. Following the full text review, 71 were excluded (see Additional file 3 for detailed reasons of exclusion) leaving 53 to meet the inclusion criteria (see Fig. 1 for the PRISMA flow diagram). Twenty-eight of the 53 studies were secondary data analyses; with 41 of the studies being published in 2017 or later. Tables 1, 2, 3 and 4 summarize the characteristics of the included studies, including participant characteristics, exposure measurement method, outcomes reported (task and corresponding domain), covariates used and the main outcomes of interest. Of the 53 studies included in this review, most were observational in nature. Forty-three studies employed a cross-sectional design; four of those reporting both cross-sectional and longitudinal data [35,36,37,38] and two studies reporting longitudinal data only [39, 40]. Three of the 43 cross-sectional studies also implemented isotemporal substitution models [41,42,43] and the remaining eight consisted of various experimental designs (i.e., randomized crossover). The average quality score was 83%, ranging from 59 to 100% (see Table 5).

Fig. 1

PRISMA flow diagram of the search and screening process in review of sedentary behaviour with cognitive function

Table 5 Quality assessment of included studies

Population

The total sample size from the 53 studies was 83,137 middle-aged and older adults (study population sizes ranged from 12 to 32,715 (median of 150) with mean ages ranging from 40.8 to 88.0 years old. Studies were mainly conducted in the USA (n = 20), but other countries included were Canada (n = 5), Australia (n = 4), Japan (n = 5), Netherlands (n = 6), Sweden (n = 2), Taiwan (n = 2), Brazil (n = 1), Chile (n = 1), China (n = 2), England (n = 1), Finland (n = 1), Germany (n = 1), Ghana (n = 1), Hong Kong (n = 1), India (n = 1), Ireland (n = 1), Italy (n = 1), Mexico (n = 1), Portugal (n = 1), Russia (n = 1), Scotland (n = 1), Singapore (n = 1), South Africa (n = 1) and Switzerland (n = 1).

Exposure (Sedentary Time)

Tables 1, 2, 3 and 4 describe and summarize the method of sedentary behaviour measurement for each study. The majority of the studies (n = 44) used a device (i.e., accelerometer, inclinometer) to measure sedentary behaviour. Eight various devices were used, and only 10 of the 44 device-based studies used an activPAL™. The reported device measured sedentary behaviour time ranged from 405 min per day [44] to 1038 min per day [45]. Three of the studies using a device also used a self-report measure. Eight studies used only a self-reported measure. Five different self-report measures were used, and two studies [46, 47] did not specify their measurement tool. Self-reported sedentary behaviour time ranged from 225 min per day [48] to 803 min per day [49].

Outcome (Cognitive Function)

Tables 1, 2, 3 and 4 include all measures used within each study and its corresponding cognitive domain. Processing speed and episodic memory were assessed with 16 and 15 different measures respectively. Nine different measures were used to assess global cognitive function. Ten different measures were used to assess executive function, seven for cognitive flexibility and eleven for working memory. Lastly, five different measures were used to assess motor skills and construction.

Association of Total Sedentary Time with Cognitive Function

Studies assessing the association of sedentary time with cognitive function are summarized in Tables 1, 2, 3 and 4 and illustrated in Fig. 2. Overall, most studies report no association, with some reporting that more sedentary time was associated with worse cognitive function and the fewest studies reporting the more sedentary time was associated with better cognitive function. We can also observe that experimental studies in the area are lacking, with mixed associations found from longitudinal studies.

Fig. 2

Forest plot of all studies reporting on an association of sedentary behaviour with cognitive function based on method of measurement (device or self-report) and cognitive domain

Association of Total Sedentary Time with Cognitive Function Based on Measurement Type

Figure 2 illustrates that six studies showed positive associations [38, 41, 43, 47, 50, 51] (i.e., more sitting resulting in better cognitive function), with most of these studies (n = 4) using self-reported measures of total sedentary behaviour time [38, 43, 47, 51]. Figure 2 also shows that 20 studies found negative associations [35,36,37, 39, 40, 43, 47, 48, 52,53,54,55,56,57,58,59,60,61,62,63] (i.e., more sitting resulting in worse cognitive function), with most of these studies (n = 16) using a device to measure total sedentary behaviour time [35,36,37, 40, 47, 53,54,55,56,57,58,59,60,61,62,63].

Association of Total Sedentary Time with Cognitive Function Based on Cognitive Domain

Associations of sedentary time with cognitive function, based on cognitive domain are summarized in Tables 1, 2, 3 and 4 and illustrated in Fig. 2. For studies using a device for measurement of sedentary behaviour time, there were 12 studies assessing cognitive flexibility [40,41,42, 44, 50, 51, 53, 54, 58, 64,65,66], 16 studies for episodic memory [47, 50, 53, 54, 58, 59, 61, 62, 65,66,67,68,69,70,71,72], 11 studies for executive function [40, 50, 53, 54, 58, 59, 62, 66, 71,72,73], 16 studies for global cognitive function [35,36,37, 47, 53,54,55, 57, 60, 63, 67, 74,75,76,77,78], three studies reporting on the domain of motor skills and construction [35, 56, 72], 20 studies for processing speed [35, 40, 41, 41, 47, 50, 51, 53, 54, 56, 58, 59, 62, 65,66,67,68,69, 71, 72, 79] and 15 for working memory [40,41,42, 50, 51, 53, 54, 58, 62, 64,65,66, 70, 72, 73]. For studies using self-report to measure sedentary behaviour time, there were two studies for cognitive flexibility [49, 51], five studies for episodic memory [43, 47, 49, 80, 81], seven for global cognitive function [38, 39, 47, 48, 52, 80, 82], six for processing speed [43, 47, 49, 51, 80, 81] and two for working memory [49, 51]. Overall, results of the studies were mixed, with no consistent associations found for any of the domains. Figure 2 demonstrates that the relationship of sedentary time with some cognitive domains remains unexplored (i.e., domain of motor skills and construction and executive function for studies using self-report to measure sedentary behaviour time).

High Versus Low Sedentary Time

More detailed results of the studies that used categories of sedentary time can be found in Table 1. One study grouped sedentary time into two groups (i.e., high versus low) [52]. They report that those in the high sedentary time group had significantly higher odds for mild cognitive impairment. Three studies [36, 53, 82] separated sedentary time into tertiles. One study [82] reported the people in the highest tertile of sedentary time had the highest cognitive function while the other two [36, 53] found the opposite (i.e., participants in the highest tertile of sedentary time had the worst cognitive function). One study used quartiles to group sedentary time, based on level of physical activity. They reported no associations for time spent in sedentary behaviour with any of their cognitive measures [67].

Associations Based on Pattern of Accumulation

Table 2 describes the 11 studies that assessed the association of various patterns of sedentary time accumulation with cognitive function [35, 53,54,55, 64, 68, 74,75,76, 83, 84]. Six studies investigated whether more prolonged bouts of sedentary time (30 + min) were associated with cognitive function [53, 55, 64, 68, 74, 76]. Four studies reported no association [53, 68, 74, 76] and two studies reported a significant association [55, 64] for more prolonged sedentary bouts with worse cognitive function. Five studies reported on sedentary behaviour pattern of accumulation (i.e., interruptions in sitting time via sit to stand transitions, duration of sedentary bouts or number of sedentary behaviour bouts) and cognitive function, and all five observed no association [35, 64, 68, 75, 76].

Prevalence (Cognitively Impaired vs. Healthy)

Table 3 describes the 12 studies that assessed the prevalence of sedentary time between populations with cognitive impairment or dementia and cognitively healthy populations [45, 46, 52, 54,55,56,57, 68, 77, 83,84,85,86]. Eight studies reported the cognitively impaired population spending more time sedentary [45, 46, 52, 56, 68, 83, 84, 86] while five studies reported no significant difference [54, 55, 57, 77, 85]. Five studies reported if prolonged bouts of sedentary time differed between cognitively impaired and non-impaired populations [54, 55, 68, 83, 84]. Three studies reported the number of prolonged bouts did not differ between cognitively impaired and non-impaired populations [54, 83, 84] and two studies reported the number of prolonged bouts did significantly differ [55, 68]. For pattern of sedentary behaviour time accumulation, three studies reported no difference between the two groups [54, 68, 83].

Experimental Studies

Figure 2 illustrates the eight studies that utilized an experimental design and Table 4 describes them in detail. The experimental periods ranged from 3 h to 12 months. Overall, the shorter experimental protocols (i.e., 3 h to four days) showed less consistent findings than the longer-term protocols (i.e., eight weeks to 12 months). For example, two of the three studies that were eight weeks or longer showed a positive effect of reducing sedentary behaviour on cognitive functioning while only one of the five shorter-term studies (i.e., four days or less) showed any benefit.

Meta-analysis Results

Total Sedentary Time and Cognitive Function

Twenty-three studies including 41,334 participants were pooled and showed a non-significant association of higher sedentary time with worse cognitive function (Fig. 3: r = −0.012 [95% CI − 0.035, 0.011], p = 0.296). Heterogeneity between the studies was statistically significant (Q = 219.694, df = 25, p < 0.001) and large in magnitude (I² = 89%). Visual assessment of the funnel plot (Fig. 4) suggests no notable asymmetry.

Fig. 3

Forest plot of all eligible studies for the meta-analysis reporting on the association of sedentary behaviour with cognitive function using a random-effects model. ACE-III Addenbrooke’s Cognitive Examination, ADAS-cog Alzheimer’s Disease Assessment Scale—Cognitive, MMSE Mini Mental State Examination, 3MS Modified Mini Mental State Examination, MoCA Montreal Cognitive Assessment. Note a is the SALSA cohort, b is the PATH cohort, c is the SGS cohort, and d is the SLAS2 cohort reported within one paper

Fig. 4

Funnel plot for all eligible studies in the meta-analysis reporting on the association of sedentary behaviour with cognitive function using a random-effects model

Meta-regression

The results of the a priori subgroup meta-regression analyses to explore heterogeneity are shown in Table 6. Measurement type and sedentary time were significant (p < 0.05).

Table 6 Meta-regression main results for random effects

Subgroup Analyses of Device-Based Studies

Subgroup analyses of studies using a device to measure sedentary time suggested that more sedentary time was significantly associated with worse cognitive function (Fig. 5; r = −0.035 [95% CI − 0.063, − 0.008], p = 0.012). Heterogeneity between the studies was statistically significant (Q = 87.629, df = 19, p < 0.001) and large in magnitude (I² = 78%).

Fig. 5

Forest plot of all eligible studies for the meta-analysis that used a device as the method a measurement for sedentary behaviour reporting on the association of sedentary behaviour with cognitive function using a random-effects model. c is the SGS cohort. ADAS-cog Alzheimer’s Disease Assessment Scale—Cognitive, MMSE Mini Mental State Examination, MoCA Montreal Cognitive Assessment

Subgroup Analyses of Cognitive Domains from Device-Based Studies

There were sufficient studies to meta-analyse the relationship of device-based sedentary time with five of the seven domains of cognitive function. Global cognitive function and processing speed showed significant negative associations for sedentary time with cognitive function (Fig. 6; r = −0.061 [95% CI − 0.100, − 0.022], p = 0.002; Q = 27.597, df = 9, p < 0.01, I² = 67% and Fig. 7; r = −0.067, [95% CI − 0.103, − 0.030], p < 0.001; Q = 15.858, df = 11, p = 0.146, I² = 31% respectively). No associations were found for the domains of working memory (Fig. 8; r = 0.000 [95% CI − 0.039, 0.039], p = 0.995; Q = 9.065, df = 7, p = 0.248, I² = 23%), episodic memory (Fig. 9; r = 0.027 [95% CI − 0.064, 0.117], p = 0.558; Q = 46.583, df = 7, p < 0.001, I² = 85%) or cognitive flexibility (Fig. 10; r = −0.007 [95% CI − 0.043, 0.029], p = 0.698; Q = 7.929, df = 7, p = 0.339, I² = 12%).

Fig. 6

Forest plot of all eligible studies for the meta-analysis that used a device as the method a measurement for sedentary behaviour reporting on the association of sedentary behaviour with the cognitive domain of global cognitive function using a random-effects model. c is the SGS cohort. ADAS-cog Alzheimer’s Disease Assessment Scale—Cognitive, MMSE Mini Mental State Examination, MoCA Montreal Cognitive Assessment

Fig. 7

Forest plot of all eligible studies for the meta-analysis that used a device as the method a measurement for sedentary behaviour reporting on the association of sedentary behaviour with the cognitive domain of processing speed using a random-effects model. RT reaction time; TMT A Trail Making Test A

Fig. 8

Forest plot of all eligible studies for the meta-analysis that used a device as the method a measurement for sedentary behaviour reporting on the association of sedentary behaviour with the cognitive domain of working memory using a random-effects model. TMT B Trail Making Test B

Fig. 9

Forest plot of all eligible studies for the meta-analysis that used a device as the method a measurement for sedentary behaviour reporting on the association of sedentary behaviour with the cognitive domain of episodic memory using a random-effects model. HVLT-R Hopkins Verbal Learning Test-Revised

Fig. 10

Forest plot of all eligible studies for the meta-analysis that used a device as the method a measurement for sedentary behaviour reporting on the association of sedentary behaviour with the cognitive domain of cognitive flexibility using a random-effects model. TMT B Trail Making Test B

Subgroup Analyses of Self-Report-Based Studies

Subgroup analyses of studies using a self-report measure of sedentary time suggested that more sedentary time was significantly associated with better cognitive function (Fig. 11; r = 0.037 [95% CI − 0.019, 0.054], p < 0.001). Heterogeneity between the studies was statistically significant (Q = 28.994, df = 7, p < 0.001) and large in magnitude (I² = 76%) (Fig. 12).

Fig. 11

Forest plot of all eligible studies for the meta-analysis that used self-report as the method a measurement for sedentary behaviour reporting on the association of sedentary behaviour with cognitive function using a random-effects model. MMSE Mini Mental State Examination, 3MS Modified Mini Mental State Examination. Note a is the SALSA cohort, b is the PATH cohort, and d is the SLAS2 cohort reported within one paper

Fig. 12

Forest plot of all eligible studies for the meta-analysis that used self-report as the method a measurement for sedentary behaviour reporting on the association of sedentary behaviour with the cognitive domain of global. MMSE Mini Mental State Examination, 3MS Modified Mini Mental State Examination, SR self-report. Note a is the SALSA cohort, b is the PATH cohort, and d is the SLAS2 cohort reported within one paper

Subgroup Analyses of Cognitive Domains from Self-Report Studies

There were sufficient studies to meta-analyse the relationship of self-reported sedentary time with two of the seven domains of cognitive function. Global cognitive function showed no significant association (Fig. 13; r = 0.024 [95% CI − 0.005, 0.053], p = 0.106; Q = 20.827, df = 5, p < 0.05, I² = 76%) while processing speed showed a significant positive association (Fig. 13; r = 0.057 [95% CI 0.045, 0.069], p < 0.001; Q = 1.014, df = 4, p = 0.908, I² = 0%).

Fig. 13

Forest plot of all eligible studies for the meta-analysis that used self-report as the method a measurement for sedentary behaviour reporting on the association of sedentary behaviour with the cognitive domain of processing speed using a random-effects model. TMT A Trail Making Test A, DSST Digit Symbol Substitution Test. Note a is a population with over 7 h of sleep, and b is a population with under 7 h of sleep

Prevalence of Sedentary Behaviour and Cognitive Function

Eight studies provided sufficient data to be included in the meta-analysis examining the prevalence of sedentary behaviour in cognitively impaired versus cognitively healthy populations. The random effect model showed a significant standard difference in mean (SDM) hours spent sedentary (Fig. 14; SDM = −0.219 [95% CI − 0.310, − 0.128], p < 0.001; Q = 1.858, df = 7, p = 0.967, I² = 0%) for cognitively impaired populations spending more time sedentary when compared to cognitively healthy populations.

Fig. 14

Forest plot of all eligible studies for the meta-analysis that reported on the prevalence of sedentary time for a clinical (i.e., diagnosed with cognitive impairment) versus non-clinical (i.e., cognitively healthy) population. a are females, and b are males. Cog impaired, cognitively impaired, CD cognitive decline, MCI mild cognitive impairment, AD Alzheimer’s disease

Discussion

The purpose of this paper was to systematically review and quantify the size and direction of the relationship of total sedentary time with cognitive function and/or cognitive impairment under selected moderator conditions (i.e., method of sedentary time measurement, the cognitive domain being examined, categorical (i.e., high versus low) sedentary time cut-offs and pattern of sedentary time accumulation). We also aimed to explore the prevalence of sedentary time in populations diagnosed with mild cognitive impairment or dementia versus populations deemed as being cognitively healthy in middle-aged and older adults. Furthermore, we aimed to investigate whether experimental studies aiming to reduce or break up sedentary time affect cognitive function. Lastly, we aimed to conduct a quantitative pooled analysis of all individual studies through meta-analysis procedures to derive conclusions about these relationships. While other reviews have been conducted on the relationship of sedentary behaviour with cognitive function, to our knowledge, this is the first to quantify the association using a meta-analysis. Based on the pooled co-efficient estimates, we found a relationship of sedentary time with cognitive function, but the direction differed depending on the moderator being assessed. More specifically, there was a detrimental relationship for studies using a device to measure sedentary time while studies using self-report tended to find a beneficial relationship. For the device-based studies, more sedentary time was shown to be associated with worse cognitive function when assessing global cognitive function or processing speed. For the self-report-based studies, more sedentary time was shown to be associated with better cognitive function when assessing processing speed. We were unable to perform a meta-analysis on the studies assessing sedentary time using categorical cut-offs, pattern of sedentary time accumulation, or the experimental studies due to large heterogeneity. We were able to perform a meta-analysis for the prevalence studies and found that cognitively impaired populations spend significantly more time sedentary compared to non-impaired populations. Beyond these general findings, the following specific findings warrant commentary.

The variability within the various methods of sedentary time measurement (i.e., self-reported versus device-based) has been highlighted in previous literature [23] and our findings further demonstrate how it continues to be an obstacle. It is important to note that recall bias from self-reporting sedentary time is an issue for all age-groups and populations, but especially in older adults who may be more susceptible to cognitive impairment; as older adults have more memory complaints which may be related to poorer understanding of the questions due to cognitive impairment [87]. This could be a reason for the discrepancy in our findings and is supported by our meta-regression which showed that measurement type (i.e., device versus self-report) was a significant predictor in the overall model (~ 68%). However, this is not to suggest that we should abandon self-report methods, as previous studies have also indicated the importance of self-report measures for capturing the context of behaviour (i.e., television viewing versus reading or writing) [23]. Furthermore, due to the definition of sedentary behaviour being distinct from a lack of physical activity, it is imperative to differentiate between the two when assessing each behaviour. Many of the studies included in the current review that did measure sedentary time with a device were unable to distinguish physical inactivity (i.e., lack of movement) from sedentary time (i.e., low energy expenditure in a seated or lying posture). For example, data from a hip-worn ActiGraph GT3X + is shown to consistently underestimate time spent sitting compared to the thigh worn activPAL™ monitor [54] due to its lack of ability to distinguish sitting from standing still. Future studies need to ensure that they are not reporting physical inactivity (i.e., not meeting the physical activity guidelines) as sedentary time in order to properly advance this field of research. Additionally, future work in this field should perform a meta-analysis on studies only using an activPAL™ and compare the results to studies using other devices when there is more available research. Bias could also be due to different measures of sedentary time (i.e., total sedentary time, percent sedentary in waking hours, 24-h wear period, etc.). However, total sedentary time and percent sedentary during waking hours should be equivalent as sedentary behaviour does not include sleep, and for that reason we excluded studies that reported sedentary behaviour that included sleep time.

We also found the relationship of sedentary time with cognitive function may differ depending on which cognitive domain was being assessed (e.g., working memory, processing speed, etc.). Amongst the various cognitive domains that were assessed, when using a device, more sedentary time was shown to be associated with worse cognitive function when assessing global cognitive function or processing speed. When using a self-report measure of sedentary time, more time spent sedentary was associated with better cognitive performance when assessing processing speed. While our findings highlight variability in the domain of processing speed, it is premature to conclude that it is the most important cognitive domain at this time. However, it is noted as the domain most strongly correlated with impairments in everyday functions [30]. The differences in the associations found for processing speed between device-based and self-reported measured sedentary time studies illustrates how important the measure of sedentary time is (as discussed in the previous paragraph), since the same task can show different associations. For example, the Digit Symbol Substitution task showed a negative association in the device-based forest plot (Fig. 7) and a positive association in the self-report forest plot (Fig. 13). These findings are important as certain parts of the brain may be affected by sedentary behaviour more so than others. For example, one study found that sedentary behaviour is negatively associated with white matter volume, but not grey matter volume [88]. Additionally, cognitive processing speed has been closely related to the structural integrity of white matter [89]. We hypothesize that high levels of sedentary behaviour might be affecting regions of the brain involved in specific domains (i.e., processing speed), more so than others (i.e., working memory). While the specific regions of the brain involved in such tasks are beyond the scope of this review, we believe it to be an important point to bring up for future research to consider. Compared to a previous review on sedentary behaviour and cognitive function, our meta-analysis findings are only consistent for global cognitive function. Falck et al. (2016) indicated associations with memory, executive function and global cognitive functioning. The numerous measures used to assess cognition, even when assessing the same domain is problematic as it makes it challenging to compare studies [24]. For example, Bojsen-Moller et al. (2019) use two different tests to assess processing speed (Trail Making Test A and Digit Symbol) and four different tests to assess working memory (Digit Span Backwards, N-Back, Automated Operation Span, Trail Making Test B) [50] while Kojima et al. (2019) use one test for processing speed that Bojsen-Moller et al. (2019) did not use (Symbol Digit Modality Test) and similarly, two different tests to assess working memory (Symbol Trails and Design Memory) [40]. Falck et al. (2016) intended to amend this problem by recommending exact instruments to be used for each domain going forward, however, this still remained an issue in the current review. Without analogous measures going forward, conclusions about how many and which specific domains are affected by this behaviour are not possible. Furthermore, there needs to be agreement throughout the literature as to which specific domain each cognitive task is assessing. For example, one study used the symbol digit modalities test and deemed it as assessing executive attention [90], while another deemed it as a measure of visual/spatial processing speed and working memory [79]. Without consensus moving forward, we will not be able to better understand and advance the understanding of this relationship.

There were not enough homogenous studies to meta-analyse the relationship of sedentary time as a categorical variable (i.e., high versus low) with cognitive function. For example, one study classified ‘high’ sedentary time as eight or more hours per day, comparing it to those with ‘low’ sedentary time (i.e., less than 8 h) [52] whereas another study used tertiles, separating their participants into three groups (< 180 min/day, > 180 < 308.61 min/day, > 308.61 min/day) [82]. Future studies need to explore if a dose–response relationship exists, not just investigate linear relationships. This is supported by the findings of our meta-regression, in that sedentary time (in hours) was shown to have a significant moderating effect on the overall model. However, there are currently no universally set cut points, as each author determines the cut points for each independent study. Therefore, exploring sedentary time and its association with cognitive function should be investigated using both continuous and categorical variables to investigate any underlying relationship or threshold cut-off.

The current study was not able to provide substantial insight on how the association of sedentary time with cognitive function may differ based on the pattern of accumulation. However, previous studies have shown that there is a difference between long-uninterrupted bouts of sedentary time versus short bouts throughout the day for various health outcomes such as postprandial glucose and insulin responses [91, 92], the low quality and lack of evidence inhibited the ability for us to observe any association. Therefore, more studies are needed before any inferences can be made about how the pattern of sedentary behaviour accumulation is associated with cognitive function.

Interestingly, when looking at the studies reporting on the prevalence of sedentary time for impaired versus healthy populations, we observed a consistent difference. We found that more cognitively impaired people spend significantly more time sedentary compared to their control counterparts. This has ramifications for populations in assisted living environments, as reducing sitting may be a way to reduce or mitigate cognitive decline.

Only eight experimental studies were able to be included in the current review. Due to the large heterogeneity in the populations and designs, results were inconclusive, and we were not able to perform a meta-analysis. Four studies used single day protocol [85, 86, 90, 91] one study used a four-day protocol [58], and three studies used a longer study design (i.e., 8 weeks to 12 months) [79,80,81]. Of those, two of the longer studies [79, 80] and one of the single day studies [91] showed positive effects of reducing sedentary behaviour on cognitive function. More short and long-term experimental research is needed to explore how reducing or breaking up sedentary behaviour affects cognitive function. Future research should also investigate whether increasing sedentary behaviour (i.e., bed rest studies) affects cognitive function.

Cross-sectionally, our device-based findings are consistent with the previous review by Falck et al. (2016) that suggested higher levels of sedentary time were associated with lower cognitive performance in adults 40 years and over [24]. Although it is difficult to make direct comparisons as this review consisted of mostly studies assessing the relationship of television viewing time as ‘sedentary behaviour’ with cognitive function. Our findings also support the previous review by Copeland and colleagues (2017) in adult 60 years and older in which only half of the studies reported finding associations between increased sedentary time and decreased cognitive function [23]. The findings of our overall meta-analysis can relate to this as without examining the moderators of sedentary time measurement or specific cognitive domains, we would be left with mixed associations. We were able to build upon these previous reviews by differentiating according to the exposure and outcome measures used (i.e., self-report versus device or specific cognitive domains). Similarly, reviews by Loprinzi (2019) [25] and Olanrewaju and colleagues (2020) [26] observed conflicting associations of sedentary time with cognitive function, differing based on sedentary behaviour type (i.e., television viewing, computer use, etc.). This again supports our review and our findings in highlighting how specific moderators vary the direction and strength of the association of sedentary time with cognitive function. It is evident that this field of research is growing at a rapid rate, as 60% of studies included in the current review were published in the year 2017 or later. The heterogeneity in the previous reviews may be the result of differing exposure variables. For example, sedentary behaviour in the current review consisted of ‘total sitting time’, so there is no way to know what specific activities were taking pace during its time; while studies within the aforementioned reviews included domain specific sedentary behaviours (i.e., television watching or computer time only). Although previous research has illustrated that there may be a difference between cognitively demanding sedentary activities (i.e., puzzles) versus passive sedentary activities (i.e., television viewing) [93,94,95]; we believed the first step was to investigate whether sitting, irrespective of domain, was associated with worse cognitive function, which is why only studies reporting ‘total sedentary time’ were included in the current review. Now that there is an established association of sedentary time and cognitive function, the next step would be to investigate specific domains (i.e., leisure time) or specific sedentary behaviours (i.e., television viewing) to investigate how the association may differ.

There are some important limitations to consider with the current review. First, there were many studies that relied solely on self-reported data which as stated earlier, could impose recall bias. Second, the majority of studies were secondary analyses of a study designed to test a different primary hypothesis. In other words, many studies included in the present review did not have cognitive function as their primary outcome, and thus, may have been underpowered to detect changes in cognitive function. Third, our inclusion criterion for middle-aged and older adults was a mean age of 40 years and older. Therefore, studies could have included participants under 40 years of age, which may reduce the strength of the results. Fourth, the heterogeneity found in the review, both statistically generated and through the extraction of study characteristics. Fifth, the use of studies only published in English. Sixth, bed-rest studies, or studies aiming to increase sedentary behaviour were not included in the systematic review. Future studies should investigate whether reduced sedentary time versus forced extended sedentary time impacts the results. Seventh, we only included high quality studies in the meta-analysis and only performed sub-group analyses when five or more studies were available. Eighth, the bulk of the evidence was from cross-sectional studies. Thus, causality cannot be inferred, and reverse causality remains a possibility. Future studies should aim to investigate the association of sedentary behaviour and cognitive function over different stages of the lifespan. It is important to note that these variables may not show accurate associations from one static point in time. As we age, people sit more, while cognitive function declines [96]. Due to these variables naturally going in opposite directions, it may be more appropriate to use evidence from longitudinal studies assessing change over time. Seventh and lastly, we were not able to examine the association while considering other important factors such as physical activity time or sleep. For example, one study investigated the association of sedentary behaviour and cognitive function with participants who had greater or less than 7 h of sleep and found that sleep did in fact have an impact on the association [43]. That said, while these factors are beyond the scope of the current review, we do believe these are important considerations for future research. The main strength of the current review is that by systematically identifying these various moderators and limitations, future research can be improved. Another strength is the use of large population-based datasets from a range of countries and the subsequent methods used to analyze the data. However, there were limited studies from low-income countries or data on ethnicity. Lastly, by systematically comparing self-report versus device-based studies we were able to identify how the various measurement methods affect the relationship and give recommendations for future research in order to improve upon this.

This systematic review and meta-analysis highlights the need for future studies to use standardized measures of sedentary time and cognitive function. The appropriate device needs to be used to ensure we are capturing sedentary behaviour as opposed to physical inactivity. Furthermore, in order to better understand the association of sedentary time with cognitive function, future research needs to establish categories or cut points of sedentary time that represent ‘high’, ‘medium’ and ‘low’ levels of sedentary time. More studies are needed to investigate change in sitting behaviour over time and how this relates to cognitive function. It is unknown whether reducing sedentary time affects cognitive function in an acute or more long-term nature, so future longitudinal and experimental studies are needed to provide confidence in the findings. There are key transition periods throughout one’s life that can drastically affect sedentary behaviour time. For example, there is evidence showing that physical activity levels tend to decrease while sedentary time increases when transitioning into retirement [97]. Assessing change over time for various populations will allow for a better understanding of the relationship. Lastly, total sedentary time is just one piece of the puzzle, and more work is needed to investigate whether certain domains of sedentary behaviour (e.g.., television viewing, reading, etc.) changes the direction of the association. More research is needed to establish how the relationship of sedentary behaviour with cognitive function may differ when assessing cognitively stimulating sedentary activities (i.e., crossword puzzles) in comparison to non-stimulating domains of sitting (i.e., television viewing). It is hypothesized that the former domains will likely be unrelated to poor cognitive function whereas the latter domains will likely be related to poor cognitive function. Determining the domains of sitting and what is happening while people are sitting, is perhaps the most important consideration for future research.

Conclusion

In conclusion, within the high heterogeneity of the studies reviewed, our findings suggest that the association of total sedentary time with cognitive function is weak and varies based on the method of sedentary behaviour measurement and cognitive domain being assessed. Specifically, there was a negative association when using a device for global cognitive function and processing speed and a positive association when using self-report for processing speed. Furthermore, our findings suggest that cognitively impaired populations (i.e., diagnosed with mild cognitive impairment or dementia) spend more time sedentary than non-impaired (i.e., cognitively healthy) populations. Overall, these findings suggest that now that we have established that total sedentary time is associated with cognitive function, future research needs to explore how the sedentary time domain (i.e., occupational, leisure, transportation, etc.) and cognitive load associated with each domain (i.e., cognitively stimulating tasks versus non stimulating) may be affecting the association.