The association between sedentary behaviour and indicators of stress: a systematic review

Background Emerging evidence shows sedentary behaviour may be associated with mental health outcomes. Yet, the strength of the evidence linking sedentary behaviour and stress is still unclear. This study aimed to synthesise evidence regarding associations between time spent in sedentary behaviour and stress in adults. Methods A systematic search was conducted (January 1990 – September 2019). Following PRISMA guidelines, an evaluation of methodological quality, and best-evidence synthesis of associations between time in sedentary behaviour (including sitting time, TV viewing, computer use) and stress were presented. Twenty-six studies reporting on data from n = 72,795 people (age 18-98y, 62.7% women) were included. Results Across the studies (n = 2 strong-, n = 10 moderate- and n = 14 weak-quality), there was insufficient evidence that overall time spent in sedentary behaviour and sitting time were associated with stress, particularly when using self-report measures of sedentary behaviour or stress. There was strong evidence of no association between TV viewing, or computer use and stress. Amongst studies using objective measures of sedentary behaviour and/or stress there was also strong evidence of no association. Conclusion Although previous research suggested sedentary behaviour may be linked to mental health outcomes such as depression and anxiety, the evidence for an association between various types of sedentary behaviour and stress is limited in quality, and associations are either inconsistent or null. High-quality longitudinal/interventional research is required to confirm findings and determine the direction of associations between different contexts (i.e. purpose) and domains (i.e. leisure, occupational, transport) of sedentary behaviour and stress.

are linked to reduced risk of psychological stress, as well as other mental health outcomes such as depression and anxiety symptoms [17][18][19]. However, less is known regarding the relationship between sedentary behaviour (i.e. sitting or reclining behaviours requiring minimal energy expenditure [20]) and stress. Given that there is an increasing evidence base to suggest that sedentary behaviour (such as computer use, TV viewing, and overall sitting) is linked to poor mental health, specifically depression [21] and anxiety [22], it is plausible that sedentary behaviour may also be linked to stress.
Theoretically, sedentary behaviour could be associated with stress via a number of potential pathways. For example, screen-based sedentary behaviours such as TV viewing, computer or electronic device use (i.e. smartphones, tablets) can induce feelings of addiction [23], 'brain burnout' [5] and/or sleeping problems [24], potentially leading to heightened levels of psychological stress. Further, engaging in such sedentary behaviours may displace time spent in other important activities such as undertaking household or work-related responsibilities, or physical activity, which may then increase feelings of stress [25]. Alternatively, since TV viewing is a popular strategy used by many adults in developed countries to manage stress (e.g. reported by 85% of respondents from the Stress and Wellbeing in Australia Survey, n = 1521; and 39% in the Stress in America Survey, n = 3361) [5], it is possible that some sedentary behaviours may reduce stress. Yet, there is a lack of clarity regarding associations between different types of sedentary behaviour and stress, and no previous review has summarised the evidence to date. Thus, the aim of this review was to investigate associations between time spent in sedentary behaviour and stress in adults.

Search strategy
A systematic electronic search was undertaken for articles published from January 1990 to September 2019. Findings are reported according to PRISMA reporting guidelines [26]. Articles prior to 1990 were not included as increases in population sedentary behaviour levels did not begin to be reported until after 1990 with the advent of widespread online technology use [27]. Databases included Medline/Medline Complete, CINAHL Complete, PsychINFO, SPORTDiscus and EMBASE. Full search strings are provided in Additional file 1: Table S1, however, principal search terms encompassed: 1) 'Sedentary behaviour' (sedentary behaviour, screen time, screen-based, television, computer, electronic device, video game, smart phone, sitting, passive transport, and tablet); and 2) 'Stress' (stress, cortisol, adrenocortical/glucocorticoid hormones). Search strings were further limited to peer-reviewed articles written in English. The literature search, and each of the following stages was led by LDS. Duplicates were identified by LDS in two stages. The majority of duplicates were first identified using EndNote's automated 'find duplicates' function. Remaining duplicates were then identified in EndNote by sorting article titles alphabetically before manual checking, then sorting author names alphabetically before again checking manually. After removal of duplicates, article titles were initially screened for inclusion by LDS and MT. Abstracts were then assessed, followed by retrieval of full texts which were read to determine suitability. Reference lists from retrieved full-text articles and authors' own bibliographic libraries were also searched, yielding no additional articles. MT and LDS reviewed all final articles to determine their inclusion in the review.

Study selection criteria
For the purpose of this review, stress was operationalised as both self-reported (e.g. perceived psychological stress) and objective (e.g. changes in stress hormones such as cortisol) measures of stress. Articles were eligible for inclusion if they: 1) were published in a peer-reviewed journal in English between January 1990 to September 2019; 2) examined apparently healthy adults aged ≥18 years (i.e. those not specifically recruited among populations with underlying chronic physical conditions (e.g. diabetes) or mental disorders (e.g. depression)); 3) examined self-reported or objective measures of screen-based sedentary behaviour or other forms of sitting time; 4) assessed self-reported or objective measures of stress; and 5) employed a cross-sectional, longitudinal, direct observation (e.g. where time in sedentary behaviour was collected for one year or one month and averaged), or controlled experimental study design. Qualitative studies and review articles were excluded. Intervention studies that primarily investigated the direct relationship between sedentary behaviour and a measure of stress were eligible to be included; however, studies that reported the effect of an intervention on stress, or sedentary behaviour, independent of one another, were not eligible. Due to potential confounding, articles relating to yoga or meditation for stress relief, or that measured stress in relation to engagement in violent screen-based activities (e.g. viewing violent TV shows, playing violent computer games) were excluded from the review, as were conference abstracts, dissertations, theses, and articles published in non-peer-reviewed journals.

Data extraction
Data extraction was performed by LDS using a data extraction form that was pilot-tested initially by MT, SAC and LDS on three included articles. Key study characteristics were extracted from identified studies included country in which the study took place, study population characteristics (sample size, age, and sex of participants), study design, sedentary behaviour type (e.g. screen-time, sitting time) and measures (e.g. self-reported or objective [i.e. direct observation or device assessed]) of sedentary behaviour, measures of stress (e.g. self-reported or objective measures) and study results in terms of the association between sedentary behaviour and stress.

Methodological quality
Methodological quality of each study was evaluated using the Effective Public Health Practice Project Quality Assessment Tool for Quantitative Studiesrecommended by the Cochrane Public Health Review Group [28]. The tool assesses research studies on selection bias (e.g. representativeness, response rate), study design (e.g. longitudinal, randomized controlled trial), confounders (e.g. controlling for confounders such as sociodemographic characteristics), blinding (e.g. researcher/participant awareness of group allocation), data collection methods (e.g. validity and reliability of measures), withdrawals and dropouts (e.g. reasons, proportion of sample with complete data), intervention integrity (e.g. percent receiving intervention) and analyses (e.g. appropriate statistical analyses for study design). For all studies, each component was given an overall quality score of weak, moderate, or strong, following the established protocol [28]. An overall study rating was assigned to each study as follows: weak (if they received≥2 weak ratings); moderate (if they received one weak rating); or strong (if they received no weak ratings). The methodological quality of studies was independently assessed by two reviewers (LDS and SAC). Any discrepancies were resolved via discussion.

Best-evidence synthesis
To enable conclusions on associations between sedentary behaviour and stress on the basis of the methodological quality of studies, a best-evidence synthesis [29] was conducted by two authors (MT and LDS). Due to the heterogeneity in exposure and outcomes, a bestevidence synthesis was selected over a meta-analysis given it would not be meaningful to calculate the average effect as per a meta-analysis [29,30]. This method has been used previously in systematic reviews in the area of sedentary behaviour and health outcomes [22,31,32]. Adapted from guidelines outlined in previous reviews that applied best-evidence synthesis [22,31], the evidence was graded as strong, moderate, or insufficient. Consistency was defined on two levels: 1) within a study (i.e. ≥75% of results in same direction within a study), to account for multiple modelling; and 2) between studies (i.e. ≥75% of results in same direction across studies examined). Strong evidence was defined as consistent results in ≥2 strong/ moderate quality studies. Moderate evidence was defined as consistent results in one strong/moderate quality study and at least one weak-quality study; or consistent results in ≥2 weak-quality studies. Insufficient evidence was defined as having only one available study or inconsistent results in ≥2 studies. When ≥2 studies were of strong/ moderate methodological quality, those with weak-quality were disregarded in the evidence synthesis [31]. In this manner, evidence was weighted in terms of study design/ methodological quality.
To determine whether associations between time in sedentary behaviour and stress could be explained by the nature of the sedentary behaviour and stress measures, studies were grouped and results analysed firstly on the basis of utilising objective versus self-report measures of stress, and secondly, on the use of objective (i.e. device assessed or direct observation) and self-report measures of sedentary behaviour.

Self-report sitting
Week day sitting: sitting time [49]. Reliability and validity of all the above measures of sedentary behaviour and stress are reported in Additional file 2: Table S2.  Table S2).

Methodological quality
Results are presented as a whole sample (i.e. combined findings) first, then for specific sedentary behaviours (i.e. overall sitting time, TV viewing, computer use), and finally for objectively versus subjectively assessed sedentary behaviour and stress.
Both positive and inverse associations between sedentary behaviour and self-reported stress were found in one weak-quality study [37], with the direction of these associations dependent on the source of stress assessed (e.g. total sitting was positively associated with 'argument-related' stress, but negatively associated with 'running late' stress) [37]. Two weak-quality studies which examined self-reported stress [36] and salivary cortisol [52] respectively, found inverse associations between time spent in sedentary behaviour and stress, whereby increased time in sedentary behaviour was associated with lower levels of stress. Eight remaining studies (n = 6 moderate- [39,43,48,50,53,54]; n = 2 weak-quality [33, 47]) found no association between sedentary behaviour and stress. Stress was self-reported in four of those studies [33,47,48,53]; three used objective measures of stress (hair cortisol [43,54]; systolic and diastolic blood pressure, heart rate, and salivary cortisol [39]); whilst one study used both a self-report and objective (hair cortisol) measure [50].

Overall sitting time
Unadjusted results showed that across all studies, there were 60 models that assessed the association between sitting time and stress. This resulted in 23 positive associations (38%); seven inverse associations (12%) and 30 null results (50%).
One weak-quality study showed an inverse association between sitting time and objectively measured stress (i.e. salivary cortisol) [52]. Four studies (n = 1 moderate- [41]; n = 3 weak-quality [38,44,51]) showed mixed findings. That is, they all included both positive and null results. For example, Lee & Kim (2018) showed a positive association between total and weekday sitting time and stress, but no association between weekend sitting time Fig. 2 Harvest plot: Evidence for association between indicators of sedentary behaviour and stress. Columns represent individual studies with reference numbers above. Column height represents methodological quality of the study (3 = strong, 2 = moderate, 1 = weak). Shading represents objective measure of stress used (not shaded = subjective measure used). Horizontal lines represent objective measure of sedentary behaviour used (no lines = subjective measure used). Notefor studies that included > 1 model, overall association for those studies was calculated on consistency (≥75%) of findings (see also Table 1) and stress [44]. Gubelmann et al. (2018) showed a positive association between sitting and diurnal cortisol slope, but no association with mean or awakening cortisol [41]. Nine studies (n = 1 strong- [45]; n = 5 moderate- [39,[48][49][50]54]; n = 3 weak-quality [37,47,57]) showed predominately no association between sitting time and stress. Of note, however, are the findings from Rebar et al. (2016) [49]. That study compared sitting in three domains (transport-related, leisure-time, occupational) as well as overall sitting. Transport-related sitting time was positively associated with stress, whilst all other models (three of the four [i.e. 75%]) showed no association (and hence the study was labelled as a null overall finding).

Best-evidence synthesis
Based on the inconsistent findings across studies (i.e. among the 10 moderate-strong quality studies included, three showed positive associations (30%), one showed mixed (10%), six showed null results (60%)), the bestevidence synthesis resulted in insufficient evidence for a relationship between overall sitting time and stress.

TV viewing
Unadjusted results showed that across all studies, there were 15 models that assessed the association between TV viewing and stress. This resulted in six positive associations (40%); one inverse associations (7%) and eight null associations (53%).
Specifically, ten studies (n = 1 strong-quality [45]; n = 4 moderate-quality [43,49,53,54]; n = 5 weak-quality [33,36,46]) investigated associations between TV viewing time and stress. One weak-quality study found a positive association between TV viewing and self-reported stress (i.e. increased TV viewing was associated with greater stress) [46]. Conversely, one weak-quality study found an inverse relationship between TV viewing and self-reported stress (i.e. increased TV viewing was associated with reduced stress) [36]. Mixed results, in this case positive and null associations, were found in two weak- [33] and one strong- [45] quality study. For example, one longitudinal study found no cross-sectional association between TV viewing and perceived stress at baseline, but stress at baseline predicted increased TV viewing at follow-up [45]. Five (n = 1 weak-quality [33] and n = 4 moderate-quality [43,49,53,54]) studies found no association between TV viewing and stress.

Best-evidence synthesis
Among the five moderate-strong quality studies that examined TV viewing, four (80%) showed null associations and therefore based on the best-evidence synthesis there was strong evidence for no association between TV viewing and stress.

Computer use
Unadjusted results showed that across all studies, there were two models that assessed the association between computer use and stress, both of which found null associations (100%). Specifically, two moderate-quality studies examined the association between computer use [49,54] and stress. No association was found between computer use and self-reported stress [49], or computer use and objectively measured stress (hair cortisol) [54] in these studies (100%). Based on the best-evidence synthesis there was strong evidence for no association between computer use and stress.

Objectively measured versus self-reported stress
Unadjusted results showed that across all studies, there were 18 models that assessed the association between sedentary behaviour and objectively measured stress. This resulted in two positive associations (11%); six inverse associations (33%) and ten null associations (56%).
Specifically, of the seven studies that used an objective measure of stress (n = 2 weak- [40,52] and n = 5 moderatequality [39,41,43,50,54]), two reported positive [40] or mixed (in this case positive and null) associations between sedentary behaviour and stress [41] (weak-and moderate quality studies, respectively). Four moderate-quality studies found null associations [39,43,50,54], while one weakquality study showed inverse associations between sedentary behaviour and stress [52]. Based on the best-evidence synthesis (i.e. four of five moderate-quality studies showed null association (80%), there was strong evidence for no association between any type of sedentary behaviour and objectively measured stress.
Unadjusted results showed that across all studies, there were 67 models that assessed the association between sedentary behaviour and self-reported stress. This resulted in 32 positive associations (48%); eight inverse associations (12%) and 27 null results (40%).

Objectively measured versus self-reported sedentary behaviour
Unadjusted results showed that across all studies, there were 39 models that assessed the association between objectively measured sedentary behaviour (i.e. device, direct observation) and stress. This resulted in 13 positive associations (33%); seven inverse associations (18%) and 19 null results (49%).

Discussion
To the best of our knowledge, this is the first review to systematically synthesise evidence of associations between time spent in sedentary behaviour and stress. Findings showed insufficient evidence for the association between overall sedentary behaviour, or sitting time and stress, particularly when using self-report measures of sedentary behaviour or stress. Further, we found strong evidence of no association between TV viewing, or computer use and stress; and strong evidence of no association between sedentary behaviour and stress when using objective measures of these outcomes.
Given that the findings between overall sitting time and stress were inconsistent, but TV viewing and computer use were strongly not associated with stress, this may indicate that the type of sedentary behaviour could potentially play a role in the link with stress. It is possible that time spent watching TV and using computers may have confounded the association between sitting time (which includes, but is not limited to, both of these behaviours) and stress. It may be that other sedentary behaviour indicators (e.g. smartphone and tablet use) that were not captured in these studies, yet contribute to overall sitting time, may be subsequently linked to greater stress. However, no studies utilised measures of these more modern electronic devices (i.e. smartphones/ tablets) and therefore this is a key area identified as requiring further research. It has been hypothesised that engaging in high levels of social media use (predominately undertaken whilst using modern electronic devices such as smartphones or tablets) may adversely impact mental health outcomes (e.g. anxiety, stress, depressive symptoms, including suicidal behaviours [23]), particularly since social media use has been associated with feelings of "addiction" and loss of sleep [58]. Further, it may be that watching TV or using computers (which are likely to be used for both leisure and occupational purposes) are not detrimental to mental health, given that some studies have suggested TV viewing may be used by adults experiencing depressive symptoms as a time to relax and 'switch off' [59].
Another finding of this review was that although there was insufficient evidence for the association between sedentary behaviour and stress when using self-report measures, there was strong evidence for no association between sedentary behaviour and stress when using objective measures of these outcomes. There is mixed evidence regarding the relationship between self-report and objectively-assessed stress. Whilst some studies have shown that self-report retrospective measures of stress are associated with objective measures (e.g. heart rate variability) of stress [60], other studies have shown that self-report measures of stress, assessed only contemporaneously but not retrospectively, are associated with objective measures (e.g. salivary cortisol) [61]. Although self-report measures can be subject to bias and recall error, objective measures also present limitations. For example, salivary cortisol levels fluctuate diurnally and can be influenced by other factors such as age or sleep [62]. Similarly, current objective, specifically device-assessed, measures of sedentary behaviour (e.g. inclinometers) are limited by the lack of distinction between the type (e.g. computer use, electronic device use, TV viewing) and context (e.g. work-related emails, social media use, online shopping) of activity being performed. Given that the type and/or context of sedentary behaviour could potentially be a key factor in the relationship with stress, rather than the sitting posture itself (as indicated by previous research examining associations between sedentary behaviour and other mental health outcomes [21,22]), further studies utilising objective behaviour-specific measures of sedentary behaviour, which capture type and context, are required. This could include utilising existing technology such as smartphone apps (e.g. Moment app) that track the duration of time spent using electronic devices, which currently has not been done in studies examining the link between sedentary behaviour and mental health.
Only one study [49] examined associations between different domains of sitting time (i.e. transport, leisure, occupational) and stress, and results indicated some domains of sitting (in this case transport-related sitting) were more strongly associated with stress than other domains (i.e. leisure, occupational). These findings contrast those of another study that investigated the link between occupational sitting time and stress [40], which found a positive association. Although the domain of sedentary behaviour could potentially play a role in the relationship with stress, currently little research distinguishing domains of sedentary behaviour exists. This is another consideration when selecting measures for future studies, given that sitting at the computer for work purposes could potentially elicit differing stress responses compared to sitting at the computer for leisure purposes. To further add to the complexity of this research area, it has been suggested that standing (as opposed to sitting) whilst engaging in such screen-based behaviours (using a sit-stand workstation for example) may reduce the risk of poor mental health [63]. However, as previously described, the type and context of the sedentary behaviour is likely to be a key determinant in the relationship with stress, and is likely a more important determinant than the actual sitting/standing posture itself [49].
Studies included in this review were limited by other factors including: more than half (54%) were rated as weak-quality and 62% utilised cross-sectional study designs in which temporal associations or cause and effect were unable to be determined. Studies used varied measures of sedentary behaviour and stress, which limits the ability to directly compare findings, and it was not possible to conduct a meta-analysis [29,30]. A number of studies did not report on the reliability or validity of measures of sedentary behaviour and/or stress used. Further, there is a lack of evidence to suggest whether prolonged, uninterrupted sitting has a different effect or association on stress, compared to if one regularly breaks up their sitting time.
In addition, no studies examined interactions between specific sedentary behaviours (e.g. TV viewing, computer use, electronic device use) and total sedentary time. Thus, it could be possible that the effects of, for example, TV viewing would be dependent on the total amount of sedentary behaviour (e.g., with a low total amount of sedentary behaviour the effects of TV viewing could be beneficial but with high total amount of sedentary behaviour the effects of TV viewing could be harmful). This is worthy of further examination.

Conclusions
Although limited (i.e. predominately weak-quality, crosssectional) research has explored associations between sedentary behaviour and stress, this review consolidates the existing evidence and found insufficient (due to conflicting results) evidence for the association between sedentary behaviour and overall sitting time and stress. Further there was strong evidence for no association between other specific types of sedentary behaviour (e.g. TV viewing, computer use) and stress. This review and the results presented provides preliminary information to question why we should investigate the link between sedentary behaviour and stress and highlights the importance of the activity that is being undertaken whilst the sitting is occurring (rather than the sitting itself). High-quality longitudinal/interventional research is required to confirm findings and determine the direction of associations between different types, contexts and domains of sedentary behaviour and stress.