Background

The burden of musculoskeletal pain (MSP) conditions has increased in recent decades, contributing to substantial health care costs [1]. According to 2019 Global Burden of Disease (GBD) estimates, age-standardised disability-adjusted life years attributable to MSP conditions excluding low back pain (LBP) increased from 1990 to 2019 by some 30.7 percentage points [2]; whereas the 2017 GDB report ranked LBP as the second-highest contributor to years lived with disability [3]. The prevalence of MSP conditions has increased in parallel with the rising burden of chronic disease and is most pronounced in those with multi-morbidities [3, 4]. Also, MSP can substantially limit mobility and engagement in regular physical activity, thereby predisposing to increased risk of other chronic conditions [3].

The biological mechanisms contributing to MSP conditions are heterogeneous; nonetheless, obesity, static working postures, physical inactivity, smoking, and aging, as well as cardiometabolic and systemic inflammation, are some factors identified to increase the prevalence of MSP [5, 6]. While there is convincing evidence of beneficial associations of physical activity with outcomes related to MSP conditions [7, 8] there is an additional element to consider in this nexus – sedentary behaviour (SB). Defined as time spent in sitting and/or reclining postures during waking hours, with energy expenditure less than 1.5 metabolic equivalents (METs) [9] – SB is associated with increased risk and unfavourable outcomes of chronic diseases, including cardiovascular disease, metabolic disorders, musculoskeletal diseases, and some cancers, as well as all-cause mortality [10, 11]. Intervention trials have shown that reducing sitting time can result in modest improvements in some biomarkers of health risk [12, 13]. From a population health perspective, excessive time spent sitting is common among older adults, especially in those with co-morbidities such as cardiovascular and metabolic disorders [14, 15].

Epidemiological evidence indicates higher volumes of SB are associated with several MSP conditions, including osteoarthritis, back pain, and neck/shoulder pain [16, 17]. Some of these findings are from low-level evidence cross-sectional studies and there could be potential reverse causality bias [16]; inferring a causal relationship between SB and MSP may therefore be problematic as pain and chronic disease could predispose to engagement in excessive SB [18]. There is, however, an inconsistent body of evidence of associations of SB with MSP conditions and related outcomes from high-level evidence-based studies [19, 20]. Some previous systematic reviews of studies including higher-level study designs have reported no associations of SB with the prevalence of some MSP conditions [19,20,21,22,23,24], whereas others have reported either positive [20, 25] or negative [26] associations with some MSP-related outcomes such as pain intensity. Methodological differences and limitations within the individual studies reviewed in these systematic reviews could impact the quality of evidence and comparability of these reviews as some of the studies were based on self-reported and surrogate estimates of SB which increases the risk of bias [19, 21, 22, 24, 27]. The emergence of evidence on device-measured SB, especially from studies using the ActiGraph and activPAL devices has improved the quality of SB evidence in recent research outputs [25,26,27].

There could be other reasons for the equivocal associations, including factors related to the influence of the specific domains of SB (e.g., work, transport, domestic) and the relative exposure of the studied population. This perspective suggests potential contributions of different domains of SB to the risk of adverse health outcomes, which may differ from the effects of total full-day SB [28,29,30]. Moreover, evidence on differences in health effects of different SB domains has been identified as a key knowledge gap by the 2020 World Health Organisation (WHO) physical activity and SB guidelines development group [31]. Existing systematic reviews have not identified differences according to domains in the associations of SB with MSP conditions.

This distinction is important, partly because, most working adults accumulate SB in both occupational and non-occupational settings. That said, SB could predispose to MSP conditions in certain occupational groups such as desk-based workers who commonly engage in a prolonged sitting [32, 33]. In this context, interventions to reduce prolonged workplace sitting time by breaking up sitting with standing and/or light walking have shown beneficial associations with a reduction in MSP or musculoskeletal system discomfort among desk-based workers [34, 35]. Thus, SB associations may also reflect plausible biomechanical or biological pathways explaining MSP conditions in those exposed to prolonged static sitting postures [36,37,38]. Paradoxically, however, in occupational groups such as tradespeople who engage in more labour-intensive manual work, SB may be a protective behaviour against MSP conditions and other chronic diseases [39,40,41].

We conducted a systematic review to examine evidence on the associations of SB with MSP conditions in observational and experimental/intervention studies of adults. Specifically, we examined and synthesised evidence separately for associations of SB with MSP conditions in the occupational and non-occupational SB domains.

Methods

Review design

We used a standard Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines-based pre-designed protocol (PROSPERO ID: CRD42020166412 – amended to limit the scope of the review) to ensure a transparent review [42, 43]. The a priori research question and search strategy were formulated according to the Population, Intervention, Control/Comparison, and Outcome (PICO) framework [44] to enhance search precision and ensure extensive data extraction to be representative and unbiased [45]. The research question was: What are the associations of occupational and non-occupational SB with MSP conditions in adults?

Search strategy

Using a comprehensive search strategy, search terms were identified and combined using Boolean operators to search the following electronic databases: MEDLINE Complete, CINAHL Complete, PsycINFO, Web of Science, Scopus, Cochrane Library, SPORTDiscus, and AMED. Additionally, three online grey literature databases, including Google Scholar, WorldChat, and Trove, were searched to also identify non-peer-reviewed studies to help to minimise publication bias [46]. The search was conducted by one reviewer, for consistency, with the guidance of a librarian (Australian Catholic University, Melbourne) initially on January 5, 2020; and, further updated on November 1, 2020, and March 17, 2021. The search filter was set to limit search results to studies published from January 1, 2000, onwards. This timeframe was chosen because the field of SB is relatively new, the early definitive papers were published at the beginning of this period, and SB research output has grown significantly over the past two decades [9].

The search terms format, guided by the PICO framework, included keywords, terms, and phrases related to SB (Exposure/Intervention); MSP conditions (Outcome); and adults (Population). The search was optimized by adding to the search string, newly identified key terms that consistently appear in titles and abstracts of retrieved studies during the search [47]. A supplementary file (Supplementary Table 1: Search key terms and strings strategy – A sample Medline database search syntax) describing the comprehensive search term framework is attached.

Study eligibility and selection

Inclusion and exclusion criteria

The selection of eligible studies was based on pre-determined inclusion and exclusion criteria. The reviewed studies satisfied all the criteria below:

  1. a.

    An original quantitative study involving either an observational or intervention/experimental design. This included cross-sectional, case–control studies, and prospective studies, as well as randomized controlled trials (RCTs) and non-randomized experimental study designs.

  2. b.

    The study was conducted in adults aged 18 years or older and examined relationships between SB (the exposure of interest) and MSP conditions (the outcome of interest).

  3. c.

    The study included a measure of any kind of MSP condition, including inflammatory and non-inflammatory MSP conditions such as back pain, joint/osteoarthritis, and pain in extremities (except for pain attributable, acutely or recently, to trauma). Autoimmune-related MSP conditions, for example, rheumatoid arthritis and fibromyalgia were not included in this review because the pathophysiology of these conditions is mainly attributable to the processes and progression of specific clinical disease entities with autoimmune causations. Some studies did not measure a specific type of MSP condition but produced a composite measure of MSP conditions. Those that measured arthritis but excluded fibromyalgia were considered for inclusion because the majority of reported cases of arthritis are likely to be osteoarthritis rather than rheumatoid arthritis. There is no universally accepted measure for MSP conditions; therefore, any acceptable measures described in studies provided the basis for considering studies to be appropriately inclusive of MSP conditions.

  4. d.

    The study clearly defined or stated the measure of SB. Specifically, the study reported a self-report measure or device-based measure of occupational or non-occupational SB. This included population-based or occupational/workgroup cohort studies that measured SB exposures that aligned with the focus of our review.

Studies were excluded if they met any of the criteria described below:

  1. a.

    all qualitative studies and those quantitative studies involving children and adolescent populations aged below 18 years;

  2. b.

    studies that did not appropriately define SB; those that used proxy estimates, such as “less active”, “inactive” or “does not engage in physical activities”; those that did not make a clear distinction between SB and physical inactivity and included these as overlapping behaviours or used these terms interchangeably;

  3. c.

    studies that focused on SB as an outcome but did not explicitly examine the relationship of SB with MSP conditions; studies that focused only on the relationship between physical activity and MSP conditions;

  4. d.

    studies conducted exclusively in clinical groups with existing clinically diagnosed MSP conditions, e.g., knee osteoarthritis patients that focused on symptom severity as outcome measures;

  5. e.

    opinion or perspective articles, conference papers, editorials, newsletters, and review studies, however, the reference lists of some literature reviews on a similar topic were hand-searched for relevant studies;

  6. f.

    studies published in languages other than English.

Screening and selection process

A two-stage approach was used to process all identified studies before arriving at the final set of studies for inclusion in this review. First, the reviewer (FD), exported all the retrieved studies into Endnote reference manager software [48], checked and removed duplicate studies. The refined list of studies was exported into collaboration-supported Rayyan systematic review software [49] for screening. One reviewer (FD) initially screened and removed irrelevant studies by title and abstract according to our inclusion and exclusion criteria, but where there was uncertainty regarding inclusion, such studies were considered in stage two screening. The second stage involved retrieval of full-text articles of retained studies, and two reviewers (FD and CB), independently read and assessed the full-text articles for inclusion. Disparities were discussed and resolved among the two reviewers; however, when uncertainty remained, they consulted with three senior reviewers (AC, NO and DD). Records of retained studies as well as reasons for exclusion (at stage two) were documented using a PRISMA flowchart (Fig. 1).

Data extraction

A pre-designed data extraction form was used to organise relevant information from the studies reviewed, to ensure data quality, and to minimise errors [50]. Reviewer FD extracted data from all the studies, and this was verified independently by CB. The verification process involved the comparison of data extracted by CB from randomly selected studies (not less than 20%) with the extracts of FD [51]. Disagreements were resolved harmoniously. Extracted data included:

  • Descriptive details – study title, author name, year of publication, place of study, study aim

  • Study design – cross-sectional, case–control, prospective, experiment/RCT/non-RCT

  • Study population – population-based, occupational/workgroup cohort

  • Sample size

  • Demographic information of study participants – e.g., gender, mean age or age range, and BMI.

  • SB and measures – occupational SB, non-occupational SB, self-report and objective measures.

  • Outcome variables and measures – MSP conditions, e.g., back pain, neck/shoulder pain, osteoarthritis, and extremities pain.

  • Intervention/experiment detail (when applicable) – type, duration, assessment point(s), effect size, etc.

  • Other relevant data relating to the MSP condition outcomes and their measures – e.g., pain intensity and disability.

Study quality assessment

Quality assessment for the included studies was undertaken (independently by two reviewers) using the quantitative checklist of QualSyst (Standard Quality Assessment Criteria for Evaluating Primary Research Papers from a Variety of Fields) [52]. Briefly, the quantitative QualSyst checklist is scored on 14 criteria as either “YES = 2”, “PARTIAL = 1”, “NO = 0” or “NOT APPLICABLE” (N/A) depending on the extent to which each criterion item is satisfied by the study report. Items marked ‘N/A’ were excluded from the computation of the QualSyst summary score. For each paper, a summary score was computed by summing scores across items and dividing this by the maximum possible score for all relevant items [i.e., 28 – (number of ‘N/A’ items × 2)] [52]. Disparities in the assessments were discussed and resolved between the assessors, and if required, the three senior reviewers arbitrated. Note, however, that the quality assessment score was not a criterion for study selection but was to be considered in the determination of the robustness of our data synthesis.

Data synthesis

The extracted data were first categorised broadly as either general population or occupational cohort studies. Thereafter, they were summarised as either observational or experimental/intervention studies. The observational studies were then further organised according to study design (cross-sectional/case–control and prospective), and experimental/intervention studies were categorised as RCTs and non-RCTs to simplify the evidence synthesis. Within the categories, the SB domain measured was organised into occupational and non-occupational SB, and the measuring instrument into device-measured and self-reported SB. Further, grouping was completed according to measured SB [full-day, leisure-time, workplace sitting, computer time, vehicle time (time spent sitting in a vehicle), and sedentary behaviours (SBs) – time spent watching television, on computer/video gaming, reading or talking on the phone], as well as the type of MSP condition outcomes. The MSP conditions included back pain (low back pain – LBP and upper back pain – UBP); neck/shoulder pain; knee osteoarthritis (pain); extremities pain (upper and lower); and other MSP conditions (included MSP conditions reported no more than three in the reviewed studies; a general MSP/discomfort or collectively measured MSP conditions; and arthritis).

Descriptive tables and narrative text provide a general overview of the studies reviewed. MSP condition outcomes (e.g., back pain, neck/shoulder pain, and knee osteoarthritis) reported in three studies or more with permissible variations in the study designs and measures were quantitatively synthesised. Otherwise, the MSP condition is presented in a narrative review.

Narrative review

In the case whereby meta-analysis was not feasible, individual study findings were systematically described and integrated using the best-evidence synthesis in a narrative text [53, 54]. This commonly used synthesis approach takes into account the quality and the consistency of reported findings of the studies in three levels – strong evidence (≥ 75% of the studies show consistent significant findings in the same direction of ≥ 2 high-quality studies; moderate evidence (consistent significant findings in the same direction of a high-quality and at least a low-quality studies or ≥ 2 low-quality studies; and insufficient evidence (inconsistent findings in ≥ 2 studies or just a single available study). When there were ≥ 2 studies of high quality in a category, our conclusion on the evidence of associations was based on the within- and between-relationships of the high-quality studies.

Quantitative synthesis

Pooled meta-analysis was performed on homogenous data for SB and MSP condition outcomes when permissible. The RevMan5 (Review Manager 5.4.1) inverse-variance approach was used to estimate the pooled effect size (in odds ratio) based on random effect due to the heterogeneity of the data [55]. When there were sufficient studies, subgroup analysis was performed based on self-reported and device-measured SB. To gain insight on how occupation type could mask the association of workplace sitting with MSP conditions, a subgroup analysis by occupation type was performed. Further, subgroup analysis was conducted for studies that reported neck, shoulder, and neck/shoulder pain, and for a subgroup that reported extremities pain. Pooled effect relationships were illustrated by forest plots, and data heterogeneity was estimated by I2, Tau2, and Cochran’s Chi-square. The robustness of our estimated pooled effect sizes was examined in a sensitivity analysis by excluding studies of low quality from the estimate; we used a funnel plot to illustrate potential publication bias.

In general, evidence synthesised by narrative review (the best-evidence synthesis) or quantitative synthesis (meta-analysis) from observational studies was regarded as either of low quality for cross-sectional/case–control studies-based evidence or high quality for prospective studies-based evidence. Evidence synthesised from experimental/intervention studies was regarded as of moderate/high quality depending on the relative contribution of non-RCT and RCT studies in the evidence.

Results

The search identified 5060 studies (Fig. 1) and 3690 remained after removing duplicates. These studies were screened by title and abstract according to the review’s inclusion and exclusion criteria. A total of 178 studies were retained for full-text screening. Of these, we excluded 99 studies (Supplementary Table 2: Studies excluded after full-text screening) after the full-text screening, leaving 79 studies published from 2000 to 2021 for the evidence synthesis, including 26 studies for meta-analysis. The included studies had representation from 36 different countries. Several of these countries were the settings for five or more studies: Australia (10), Denmark (8), Brazil (8), South Korea (5), the USA (5), and the UK (5).

Fig. 1
figure 1

PRISMA flow diagram of the studies record

Full size image

Characteristics of the included studies

The characteristics of the studies are detailed in Tables 1, 2, and 3 for the general population cohorts, observational occupational cohorts, and experimental/intervention occupational cohorts, respectively. Overall, 24 observational studies were categorised as general population cohort studies; 55 studies as occupational cohort studies, which included 40 observational studies and 15 experimental/intervention studies. The occupational category comprised studies of office workers (21); professionals – physicians, specialists, nurses, university staff, teachers, students, and police duty officers (20); tradespeople and manual workers – construction, factory, manufacturing, cleaning, transport, handicraft, sewing machine operators, steel plant workers and beauticians (14); and bus drivers (3), included a study [56] that recruited office workers, professionals, and tradespeople; and another study [57] was also of professionals and tradespeople. Cross-sectional designs and a case–control design accounted for 75% and prospective designs 25% in the general population category, whereas 85% of the observational studies in the occupational category were cross-sectional and 15% had prospective designs. Among the experimental/intervention studies, however, there were six randomised controlled trials (RCTs), two randomised cross-over trials, and two non-randomised experiment without control; one study each of non-RCT, randomised trial (RT) without control, non-RT without control (a pilot study), non-randomised cross-over trial, and a cross-sectional analysis of a dataset from an RCT.

Table 1 Characteristics of the general population studies
Full size table
Table 2 Characteristics of the observational occupational cohort studies
Full size table
Table 3 Characteristics of the experimental/intervention of occupational cohort studies
Full size table

In the general population category, SB was most frequently measured (79%) in the non-occupational domain. In contrast, in the occupational category, SB was most frequently measured (85%) in the occupational domain. Most (i.e., 54 out of 79) of the studies measured self-reported SB. In total, 19 studies investigated device-measured SB, including ActiGraph (general population category, four studies; occupational category, eight studies), activPAL (five – all in the intervention studies of occupational category), and both ActiGraph and activPAL (one intervention study of occupational category). Four studies in the experimental/intervention category, however, were based on pre-determined or usual workplace sitting conditions.

Among the studies that examined full-day SB or sitting, more than twice as many were in the general population category (15 studies) as were in the occupational category (seven studies). More studies in the occupational category examined workplace sitting (21 studies) and leisure-time sitting (seven studies) than in the general population category (workplace sitting time, two studies and leisure time, zero studies). Time spent watching television and/or other SBs were investigated in seven studies (six in the general population and one in the occupational cohort categories). Also, computer time (five studies) and vehicle time (five studies) were examined only in the occupational category. In addition to SB or sitting time, five studies examined SB/sitting bout duration, four of these studies were in the occupational category. Finally, 11 experimental/intervention studies examined changes in self-reported or device-measured sitting time.

Regarding MSP condition outcomes, 38 studies investigated a single MSP condition, 30 studies investigated multiple MSP conditions and 11 studies investigated general MSP. In general, LBP (50 studies) and neck/shoulder (28 studies) were the most frequently investigated. Except for two studies in the general population category that examined either medical record data or clinical examination data, all the studies investigated self-reported MSP conditions. In total, 22 studies investigated MSP-related pain intensity (19 studies) or MSP-related disability, and only three of these studies were in the general population category.

Regarding the population, 10 of 24 general population studies were of adults ≥ 45 years, including three studies of older adults (≥ 65 years). Also, one study in this category which was conducted in 2013 was of a 1946 birth cohort. In the occupational category, the studies were of adults ≥ 18 years; among these, five studies specifically recruited young or middle-aged adults, and one study was of a cohort of 21-year olds.

Inter-rater reliability and quality assessment

There was 83.9% agreement between the two reviewers for including or excluding studies. Decisions on seven studies were made after consultation with the three senior reviewers.

Quality assessment scores for the studies are presented in Tables 1, 2, and 3 for the general population, observational-occupational, and experimental/intervention studies, respectively. On average, the studies in each of the categories were of high quality with mean scores of 0.83, 0.80, and 0.76 for the general population, observational-occupational, and experimental/intervention studies, respectively. The lowest scores in these categories were 0.41 for Aweto et al. [58], a cross-sectional study in the general population category; 0.36 for Omokhodion et al. [94], a cross-sectional study in the observational-occupational category; and 0.42 for Engelen et al. [122], a non-RT without control design pilot study in the experimental/intervention category. The highest score among the general population category was 0.95 scored in six studies [16, 64, 66, 67, 75, 76]. In the occupational category, the highest score in observational studies was 0.95 scored by six studies [39, 57, 100, 109, 114, 129], and in experimental/intervention studies was 0.96 for one study, Brakenridge et al. [121].

The low-quality studies mostly scored low for QualSyst checklist item-11, “Some estimate of variance is reported for the main results?”. Most of the experimental/intervention studies scored low on item 9, “Sample size appropriate?”. In general, most of the studies scored average on item 8, “Outcome and (if applicable) exposure measure(s) well defined and robust to measurement/misclassification bias? Means of assessment reported?”. Overall, based on a relatively liberal cut-off threshold of 0.55 put forward by Kmet & Lee [52], six studies scored ≤ 0.55 (general population two, observational-occupational three, and experimental/intervention occupational one); when based on a relatively conservative 0.75 cut-off threshold, 56 studies scored > 0.75 (general population 18, observational occupational 28 and experimental/intervention occupational 10). Studies that scored above 0.75 were considered high-quality, and those that scored below were considered low-quality studies.

Associations of non-occupational sedentary behaviour with musculoskeletal pain conditions

Table 4 shows the key associations of non-occupational SB with MSP conditions and Table 5 summarises the findings.

Table 4 Summary of key associations of sedentary behaviour with musculoskeletal pain conditions by studies quality
Full size table
Table 5 Summary of findings synthesised by meta-analysis and the best-evidence synthesis
Full size table

Full-day sedentary behaviour or sitting time

Low back pain

Fourteen studies in total (10 general population [59,60,61, 63, 66, 68,69,70, 75, 77] and four occupational [57, 108, 109, 129]) examined the association of full-day SB/sitting time with LBP [59,60,61, 63, 66, 68,69,70, 75, 77] or LBP-intensity [57, 109, 129], including two studies [69, 129] that also examined full-day SB bout. Among these studies, 11 were cross-sectional [57, 59,60,61, 63, 66, 68,69,70, 108, 109, 129] and three applied a prospective [57, 75, 77] design; one study [57] reported both cross-sectional and prospective analyses. In the cross-sectional studies, six reported a positive association [60, 66, 68,69,70, 109] and four reported no association [59, 61, 63, 108, 129]. Five of the positive association studies [60, 66, 69, 70, 109] and three with no associations [57, 59, 129] were of high quality. Further, one of the two high-quality cross-sectional studies that investigated full-day SB/sitting bout reported a positive association in obese individuals [69]; whereas the other study [129] reported a positive association in non-overweight individuals (BMI < 25kgm−2), and a negative association in overweight/obese individuals (BMI ≥ 25kgm−2). This suggests probable evidence of cross-sectional association of full-day SB/sitting bout with LBP-intensity which is moderated by BMI. Eight of these cross-sectional studies were considered in a meta-analysis, including five studies [60, 61, 63, 66, 70] that investigated self-reported full-day SB and LBP and three studies [57, 109, 129] that analysed device-measured full-day SB/sitting and LBP-intensity (Fig. 2). The overall pooled effect size indicated full-day SB is positively associated with LBP [OR = 1.19(1.03 – 1.38), p = 0.02], though a significantly moderate-high heterogeneity was observed (I2 = 77%, p < 0.00001). A subgroup analysis by self-reported and device-measured full-day SB showed a cross-sectional association of self-reported full-day SB with LBP [OR = 1.33(1.13 – 1.57), p = 0.007; I2 = 62%, p = 0.03], but no association of device-measured full-day SB/sitting with LBP-intensity in mostly tradespeople [OR = 1.05(0.86 – 1.29), p = 0.65; I2 = 75%, p = 0.008]. The robustness of the analysis was tested in a sensitivity analysis (Supplementary Figure 1A) by excluding two studies [61, 63] with low-quality; the overall and the self-reported full-day SB subgroup associations remained significant.

Fig. 2
figure 2

A forest plot for inverse-variance meta-analysis using a random effect of cross-sectional studies that investigated non-occupational sedentary behaviour (SB) showing the pooled effect size of the association of full-day SB with low back pain (LBP); subgroup analysis by self-reported SB and LBP and device-measured SB and LBP-intensity

Full size image

For the prospective studies, the evidence was inconsistent with a positive association of full-day SB with LBP reported in one low-quality study [77], and two high-quality studies reported no association of self-reported full-day SB [75] and device-measured [57] full-day sitting with LBP [75] and LBP-intensity [57] respectively.

Neck/shoulder pain

There were two high-quality cross-sectional studies [64, 110] that investigated the association of device-measured full-day SB with neck/shoulder pain-intensity [110] and shoulder pain [64]. One study [110] of tradespeople reported a positive association of high full-day SB with neck/shoulder pain-intensity. The other study [64] of severely obese individuals reported no association of low full-day SB with shoulder pain, which may imply a high full-day SB could be associated with shoulder pain. Thus, there is inconclusive evidence of a cross-sectional association of full-day total SB with neck/shoulder pain.

Knee/hip pain/arthritis

Four high-quality cross-sectional studies, three of adults ≥ 45 years in the general population cohorts [16, 59, 60] and one study of adults < 40 years in the occupational cohorts [105] reported a positive association of full-day SB with knee pain (osteoarthritis), including one study that reported the association only in men [59]. There was one prospective study [76] that reported no association of extensive full-day SB with knee pain. According to the best-evidence synthesis, we concluded there is strong evidence of cross-sectional association of full-day SB with knee pain in middle-aged to older adults, however, there is insufficient evidence whether the association is gender-dependent. Also, there is insufficient evidence of a prospective association of full-day SB with knee pain. Also, of the two high-quality cross-sectional studies [60, 64], one reported a positive association of self-reported full-day SB with hip pain [60], and the other a positive association of device-measured low full-day SB with hip pain, indicating a protective association of high full-day SB with hip pain [64]. Therefore, there is insufficient evidence of a cross-sectional association of full-day SB with hip pain. Furthermore, two high-quality cross-sectional studies [62, 70] in the general population category reported a positive association of full-day SB with arthritis of adults ≥ 50 years [70] or ≥ 65 years [62] old. Another high-quality cross-sectional study [60], however, reported no association of full-day SB with osteoarthritis of adults ≥ 65 years old. Thus, there is evidence of a cross-sectional association of full-day SB with arthritis in adults ≥ 50 years.

Extremities pain

One high-quality cross-sectional study in the general population cohort reported an association of wrist/hand pain with a high volume of full-day SB, but no association with a low volume of full-day SB [64]. However, evidence in one study is insufficient to conclude.

General musculoskeletal pain

Four cross-sectional studies investigated full-day SB and general MSP. Two high-quality studies of the general population category reported a positive association [65, 67] and two studies (one high-quality [106] and one low-quality [107]) study of the occupational category reported no association. Based on the high-quality studies, there is strong evidence of a cross-sectional association of full-day SB with general MSP. However, the evidence of a prospective association is inconclusive with only one low-quality study in the general population category reporting a positive association [17].

Time spent in sedentary behaviours – sitting watching TV, video games, reading, listening to music

Five cross-sectional [58, 61, 72, 73, 86] and two prospective [78, 114] studies – five of general population [58, 61, 72, 73, 78], two of occupational [86, 114] – investigated time spent in SBs and MSP conditions [58, 61, 72, 73, 86] or MSP-related outcomes [78]. Three were of high-quality [72, 73, 86] and two low-quality [58, 61]. There were variations in the MSP condition outcomes, hence meta-analysis was not performed for these studies. Among the cross-sectional studies, only one study [58] (low-quality) reported positive associations of SBs > 3 h/day with LBP, UBP, knee pain, and ankle pain, and no associations with neck/shoulder pain and elbow pain. Another study [86] (high-quality) also reported a positive association of TV-viewing time (TV time) > 2 h/day with clinically diagnosed MSP condition of > 50 year old adults. Most of the cross-sectional studies reported no associations of TV time (≥ 2 or 3 h/day) with LBP [61, 72], neck/shoulder pain [73], back/neck pain [86], or limb pain [86]. Based on the best evidence, there is insufficient evidence of cross-sectional associations of SBs or TV time with MSP conditions.

For the two prospective studies, both of high quality, one reported no association of TV time > 2 h/day with LBP-intensity, but a positive association with LBP-disability only in women [78]. The other study [114], however, reported a positive association of TV time with general MSP. Herein also, prospective evidence of associations of TV time with MSP conditions and MSP-related outcomes are insufficient.

Leisure-time sedentary behaviour

Five cross-sectional studies (four high-quality [104, 109,110,111] and one low-quality [56]) of occupational category examined the associations of self-reported [56, 104] and device-measured leisure-time SB [109,110,111] or SB bout [111] with LBP [56, 104], LBP-intensity [109], neck/shoulder pain [56], neck/shoulder pain-intensity [110, 111] and lower extremities pain [56]. All these studies except one [104] were of tradespeople, and two were from a single large study – “Danish PHysical ACTivity cohort with Objective measurements (DPHACTO) [110, 111]. Three of the studies reported a positive association of leisure-time SB with LBP [56], LBP-intensity [56, 109], and neck/shoulder pain-intensity [110], whereas three studies reported no association of SB [56, 104] or SB bout [111] with LBP [104], neck/shoulder pain [56], neck/shoulder pain-intensity [111] or lower extremities pain [56]. Based on the best-evidence synthesis, there is insufficient evidence of cross-sectional associations of leisure-time SB or SB bout with LBP, LBP-intensity, neck/shoulder pain, neck/shoulder pain-intensity, or lower extremities pain.

Associations of occupational sedentary behaviour with musculoskeletal pain conditions

Table 4 (above) shows the key associations of occupational SB with MSP conditions and Table 5 summarises the findings.

Device-measured workplace sitting time

Low back pain

Three high-quality cross-sectional [33, 109, 129] and two high-quality prospective [39, 57] studies investigated device-measured workplace sitting [39, 57, 109, 129] or sitting bout [129] and LBP [33] or LBP-intensity [39, 57, 109, 129], including a study with both a baseline cross-sectional and a prospective analysis [57]. Two of these studies [39, 129] were from a single large study. One study was of office workers [33] and four studies were of tradespeople [39, 57, 109, 129], which included one study also with healthcare workers [57]. No association was reported in any of the cross-sectional studies, except one that reported a marginally significant positive association with LBP-intensity [109]. One cross-sectional study [129], nonetheless, reported a negative association of total workplace sitting or a moderate sitting bout with LBP-intensity in overweight/obese individuals (BMI ≥ 25kgm−2), and a positive association of brief bout workplace sitting with LBP-intensity in non-overweight individuals (BMI < 25kgm−2). The baseline cross-sectional analysis of one prospective study [57] reported a negative association with LBP-intensity in healthcare workers but no association in construction workers (tradespeople). Meta-analysis was not feasible, hence, the best-evidence synthesis indicates there is insufficient evidence of cross-sectional associations of device-measured workplace sitting with LBP and LBP-intensity in tradespeople and non-tradespeople. For the prospective studies, there were two high-quality studies [39, 57]; the association was inconsistent in one study with a reported negative association with LBP-intensity in healthcare workers but no association in construction workers [57]. The other study of tradespeople, however, reported a negative association of both total workplace SB and SB bout with LBP-intensity [39]. There is, therefore, an indication that sitting at the workplace may have a protective effect which is dependent on occupation type.

Neck/shoulder pain

Two cross-sectional studies [110, 111] and one prospective [112] study all from a single large study (all high-quality) examined the association of device-measured total workplace sitting or sitting bout with neck/shoulder pain-intensity of tradespeople. No association of high total workplace sitting with neck/shoulder pain-intensity was reported in the cross-sectional studies [110, 111]. One cross-sectional study [110], however, reported a negative association of low total workplace sitting with neck/shoulder pain-intensity in males but no association in females. Also, the other cross-sectional study [111] reported equivocal associations of workplace sitting bouts with neck/shoulder pain-intensity; a positive association for a moderate bout, and a negative association for a brief bout. A negative association was reported in the prospective study [112]. The cross-sectional association is inconsistent [110, 111], however, a negative association in a prospective analysis [112] of the same DPHACTO study dataset suggests there is a probable protective association of workplace sitting exposure with neck/shoulder pain-intensity in tradespeople.

Self-reported workplace sitting time

There were 19 cross-sectional [71, 82,83,84,85,86,87,88,89, 92,93,94, 96,97,98, 100,101,102,103], one case–control [74] and three prospective [79, 113, 115] studies that investigated self-reported workplace sitting and MSP conditions – LBP [71, 83, 87,88,89, 92, 94, 96,97,98, 102, 103, 113], neck/shoulder pain [82,83,84,85, 87, 88, 92, 93, 100, 103, 115], knee/hip pain [74, 79, 93] and extremities pain [83,84,85,86, 88, 92, 101]. All but three of these studies [71, 74, 79] were in the occupational category. The durations of the workplace sitting examined varied across the studies, included 20 min continuous [93], > 4.2 h/week [92], ≥ 2 h/day [74, 79, 87, 88, 113], ≥ 3 h/day [94], ≥ 4 h/day [82,83,84,85, 92, 97, 100, 103], ≥ 6 h/day [98], 51.9(11.8)hrs per total weekdays [115], or unspecified durations (prolonged sitting) [71, 86, 89, 96, 101, 102].

For the cross-sectional studies, of the 11 studies (two of office workers, five of professionals, and three of tradespeople, as well as one general population study) that examined associations with LBP, seven reported positive associations [92, 94, 96,97,98, 102, 103] and four reported no association [71, 87,88,89]. All these studies except two [94, 96] were of high-quality. Eight studies (all high-quality) were meta-analysed with a subgroup analysis according to non-tradespeople (office workers [98], professionals [89, 92, 97, 103], and general population [71]) and tradespeople [87, 102] as indicated in Fig. 3. Overall, there is a significant cross-sectional association of workplace sitting with LBP (OR = 1.47(1.12 – 1.92), p = 0.005; however, there is non-significant moderate heterogeneity (I2 = 44%, p = 0.08). The subgroup analysis indicates the association is significant in the non-tradespeople [OR = 1.56(1.18 – 2.05), p = 0.002] with moderate but non-significant heterogeneity (I2 = 31%, p = 0.20), and non-significant association in the tradespeople [OR = 1.40(0.61 – 3.20), p = 0.43] with substantial non-significant heterogeneity (I2 = 70%, p = 0.07). Sensitivity analysis (Supplementary Figure 2A) excluded two studies [71, 102] with lower quality score and the overall association as well as the association for non-tradespeople were still significant, and zero heterogeneity in the non-tradespeople (I2 = 0%).

Fig. 3
figure 3

A forest plot for inverse-variance meta-analysis using a random effect of cross-sectional studies that investigated occupational SB showing the pooled effect sizes for the association of self-reported workplace sitting time with LBP; subgroup analysis by non-tradespeople and tradespeople

Full size image

With neck/shoulder pain, a positive association was reported in eight studies (one of office workers [82], three of professionals [87, 88, 93, 100, 103], and four of tradespeople [84, 85]). Only one study [92] of professionals reported no association. Also, one study [83] reported a negative association only in females. Seven of these studies [84, 87, 88, 92, 93, 100, 103] were of high-quality. A meta-analysis (Fig. 4) of pooled effect sizes of nine studies [82, 84, 85, 87, 88, 92, 93, 100, 103] indicates workplace sitting is associated with increased odds of neck/shoulder pain [Overall OR = 1.73(1.46 – 2.03), p < 0.00001]. Subgroup analysis also shows there is increased odds of neck pain [OR = 1.90(1.35 – 2.68), p = 0.0002], shoulder pain [OR = 1.71(1.31 – 2.22), p < 0.0001] and neck/shoulder pain [OR = 1.62(1.34 – 1.96), p < 0.00001]. The overall heterogeneity was, however, significantly substantial (I2 = 51%, p = 0.02), mainly due to heterogeneity in studies on neck pain (I2 = 74%), as studies on shoulder and neck/shoulder pain were homogeneous (I2 = 0%). Sensitivity analysis (Supplementary Figure 3A) after excluding two studies [82, 85] with low-quality shows the estimate is robust and the association remained significant.

Fig. 4
figure 4

A forest plot for inverse-variance meta-analysis using a random effect of cross-sectional studies that investigated occupational SB showing the pooled effect size for the association of self-reported workplace sitting time with neck/shoulder pain; subgroup analysis by studies that reported neck pain, shoulder pain, and neck/shoulder pain

Full size image

For extremities pain, a positive association with hand/wrist pain was reported in two studies [88, 101]; three studies [83, 84, 86] reported a negative association, including one study [83] with the association only in females; and another study [86] with lower limb disability; no association was reported in four studies [85, 87, 88, 92]. Five of the studies were of high quality. A pooled analysis (Fig. 5) of five studies [84, 85, 88, 92, 101] with considerable heterogeneity (I2 = 88%, p = 0.00001) indicated no association of workplace sitting with extremities pain [OR = 1.17(0.65 – 2.11), p = 0.60]; however, a subgroup analysis of three studies [84, 85, 92] with low and non-significant heterogeneity (I2 = 28%) indicated an inverse association of workplace sitting with lower limbs pain [OR = 0.61(0.46 – 0.80), p = 0.0004]. Sensitivity analysis shows the overall effect size remained non-significant (Supplementary Figure 4A).

Fig. 5
figure 5

A forest plot for inverse-variance meta-analysis using a random effect of cross-sectional studies that investigated occupational SB showing the pooled effect size for the association of self-reported workplace sitting time with extremities pain; subgroup analysis by upper limbs, lower limbs, and hand/wrist pains

Full size image

The only case–control study [74] of the general population reported a positive association of workplace sitting with hip pain, insufficient evidence of association from a single study.

For the prospective studies, one of low-quality reported no association of workplace sitting with LBP [113]; another one of high-quality reported a positive with neck pain [115]; the third study of high-quality reported a negative association with knee pain [79]. Therefore, prospective evidence of association of workplace sitting is insufficient with LBP, neck pain, and knee pain.

Computer time

Five cross-sectional studies of the occupational category (office workers [80, 82] and professionals [86, 97, 103]), including three high-quality investigated computer time and LBP [97, 103], neck/shoulder pain [82, 103] or general MSP [80, 86]. A positive association of computer time ≥ 4 h/day was reported with LBP [97], neck/shoulder pain [82, 103], and general MSP [80], and a negative association reported with LBP in another study [103]. Also, one study reported no association of computer time ≥ 2 h/day with general MSP [86]. There is moderate evidence of a cross-sectional association of computer time with neck/shoulder pain, however, the evidence is restricted to a small number of studies. The evidence with LBP and general MSP is insufficient with limited studies.

Vehicle time

Five occupational category cross-sectional studies of bus drivers [90, 95, 99] and professionals (patrol duty police officers) [81, 91] reported vehicle time and LBP [81, 90, 99] or general MSP [91, 95], including three of high-quality [81, 91, 99]. There is an inconsistent association with general MSP; of the two studies [91, 95], one reported no association [91] and the other a positive association [95]. A similar inconsistent association was reported with LBP; two studies [90, 99] reported a positive association and one study [81] reported no association. In a meta-analysis (Fig. 6), the pooled effect size of the three studies [81, 90, 99] showed considerable heterogeneity (I2 = 95%) but increased odds of LBP with prolonged sitting in a vehicle, although this was not statistically significant [OR = 2.16(0.79 – 5.93), p = 0.13]. After excluding the low-quality study [90] in a sensitivity analysis the association was still non-significant (Supplementary Figure 5A).

Fig. 6
figure 6

A forest plot for inverse-variance meta-analysis using a random effect of cross-sectional studies that investigated occupational SB showing the pooled effect size for the association of vehicle time with LBP

Full size image

Changes in workplace sitting time

Fourteen experimental/intervention studies investigated changes in sitting time and MSP symptoms, including LBP, neck/shoulder pain, extremities pain, and general MSP/discomfort of office workers [35, 116,117,118,119,120,121,122,123,124, 126,127,128] and students [125]. Designs included six RCTs [35, 116,117,118,119,120], two randomised controlled cross-over trial [125, 127], two non-randomised experiment without control [126, 128], one study each of non-RCT [124], RT without control [121], non-randomised cross-over trial [123], and non-RT pilot study.

Duration of experiments/interventions ranged from 65 min [125] to 12 months [121]. Sample sizes ranged from 12 participants [126] to 317 participants [35]. Nine of the studies were of high quality [35, 116, 118,119,120,121, 123, 125, 127] and four of low quality [117, 122, 124, 128]. Of the studies, nine measured sitting time change and reported a reduction in sitting time after the period (device-measured – ActiGraph [35, 123] and activPAL [118, 120, 121, 123, 127]; self-report [119, 122,123,124]) while three studies were based on fixed sitting duration (65 min [125] and 4 h [116, 128]), over 2-h continuous sitting [126] or usual work sitting condition [117].

There were methodological and analytical variations among the studies, therefore, the data were not meta-synthesised. A positive correlation of sitting reduction with a reduction in LBP was reported in six studies [118, 121,122,123,124, 127] (including four high-quality studies with one RCT [118]); reduction in neck/shoulder pain two RCT studies [35, 119] (both high-quality). No study reported a correlation or association of sitting reduction with a reduction in extremities pain. Two high-quality RCT studies [35, 120], however, reported no significant correlation with LBP; three studies [120, 121, 127] with neck/shoulder pain, all high-quality with one RCT [120]. Furthermore, of six studies, two high-quality studies [35, 125] reported sitting reduction correlates with a reduction in general MSP/discomfort; one RCT study [117] of low-quality reported reduced workplace sitting time does not increase the risk of general MSP/discomfort; and three studies [116, 126, 128], one of high-quality [116], reported a positive association of continuous uninterrupted sitting with increased general MSP/discomfort [116, 128] and LBP/discomfort [126]. Also, one high-quality study [118], however, reported a protective association of prolonged workplace sitting bout with extremities pain.

Generally, the best evidence suggests workplace sitting reduction is correlated with reduced LBP and general MSP symptoms. For neck/shoulder pain reduction, the evidence from RCT suggests there is a positive correlation with reduced workplace sitting. Also, there is moderate evidence of association of prolonged uninterrupted sitting with general MSP/discomfort. There is, nevertheless, no evidence of correlation of reduced workplace sitting with a reduction in extremities pain.

Risk of bias

Three studies had lower quality scores detected by the QualSyst checklist, one of which was a pilot study and had a potential risk of bias; however, most of the studies did not show any major risk of bias. The funnel plots (Supplementary Figures 1B, 2, 3, 4 and 5B) of the meta-synthesised studies were mostly asymmetrical; this could be because of the small number of studies available and not likely due to publication bias. Also, the significant heterogeneity observed may have risen from the studies’ methodological heterogeneity in the variables measured and study sample.

Discussion

Key findings

This is the first review to examine separately the associations of occupational and non-occupational SB with MSP conditions in adults. We found in the non-occupational SB domain, strong evidence of cross-sectional associations for full-day SB with MSP conditions, including LBP, knee pain, arthritis, and general MSP. For the occupational SB domain, there is strong evidence of cross-sectional associations of self-reported workplace sitting with MSP conditions, including LBP and neck/shoulder pain. Also, we found moderate evidence of a cross-sectional association of computer time with neck/shoulder pain. Furthermore, we identified from experimental/intervention studies that reduced occupational sitting time was associated with a reduction in LBP, neck/shoulder pain, and general MSP. However, there was insufficient evidence on cross-sectional associations of leisure-time SB and TV time with MSP conditions. Likewise, the evidence on prospective associations of occupational and non-occupational SB with MSP conditions was insufficient, nonetheless, there is an indication that device-measured total workplace sitting could be negatively associated with LBP-intensity in tradespeople.

Non-occupational sedentary behaviour and musculoskeletal pain conditions

We observed in our meta-analysis of cross-sectional studies that full-day SB or sitting time is positively associated with the risk of LBP. However, subgroup analysis by self-reported and device-measured SB indicated the association exists between self-reported full-day SB and LBP, but not for device-measured full-day SB and LBP-intensity, which included studies of mostly tradespeople. This finding is, nonetheless, limited by a small number of studies. The cross-sectional design and self-reported data downgrade the quality in this evidence with the association only present in the case of self-reported SB, but not device-measured SB, with LBP. Our narrative synthesis based on the best-evidence synthesis found that there are cross-sectional associations for full-day SB with knee pain, arthritis, and general MSP, but an inconclusive association with neck/shoulder pain. We found inconsistent cross-sectional associations of full-day SB with hip and extremities pains. Also, limited by the number of studies, there was insufficient evidence of prospective associations of full-day SB with MSP conditions. Furthermore, we observed inconsistent evidence of cross-sectional and prospective associations of SBs, TV time, and leisure-time SB with MSP conditions. These findings were, however, constrained by the limited number of studies available, especially evidence from prospective studies.

Our cross-sectional findings for LBP are in contrast to a previous review of observational prospective and case–control studies by Chen and colleagues, that showed no associations of a sedentary lifestyle with the risk of LBP [19]. Unlike our review which included only adults, Chen and colleagues’ review included both children and adults [19]. Another review of prospective studies has also reported some inconsistent associations of SB with LBP [20]. A meta-analysis by Alzahrani and colleagues reported no association of SB with the prevalence of LBP but reported positive associations with LBP intensity and disability [20]. Notwithstanding the methodological limitations that might be present in the above-mentioned reviews, a specifically clear distinction was not made between SB and physical inactivity in the inclusion criteria [19], the possibility of reverse causation within cross-sectional designs limits the comparability of our findings with these previous reviews of prospective studies. Adults, especially those with multi-morbidities including MSP conditions may often be less active and resort to SB which may have a pain modulation effect [130]. A review, for instance, had previously found that SB is much common in people with knee osteoarthritis [131]. We found that there is a positive cross-sectional association of SB with knee pain, but of limited strength due to a small number of reviewed studies; however, causal relation cannot be inferred from a cross-sectional finding with a potential reverse causation bias.

Occupational sedentary behaviour and musculoskeletal pain conditions

Our meta-analysis of cross-sectional studies found a positive association of self-reported total workplace sitting with the risk of LBP and neck/shoulder pain. A subgroup analysis by non-tradespeople and tradespeople for the risk of LBP shows the association is significant only in the non-tradespeople. Although limited in terms of the number of studies available, our best-evidence synthesis indicates the association of device-measured workplace sitting with LBP or LBP-intensity was inconsistent in cross-sectional studies of both non-tradespeople and tradespeople but suggests a potential protective association in prospective studies which could be moderated by occupational demand. Also, there is an indication from three studies (including a prospective study) from the same dataset of a negative association of workplace sitting with neck/shoulder pain-intensity in tradespeople. Furthermore, our meta-analysis showed no association of self-reported workplace sitting with the risk of pain in extremities. Nevertheless, a subgroup analysis indicates self-reported workplace sitting may have a protective association for pain in lower limbs. Evidence of prospective associations of self-reported workplace sitting with MSP conditions is insufficient due to a limited number of reviewed prospective studies. Additionally, we observed in a meta-analysis of three cross-sectional studies on vehicle time and LBP that excessive time spent sitting in a vehicle increases the odds of LBP, yet this cross-sectional association is non-significant.

Additionally, though limited by the number of studies, computer time was found to be cross-sectionally but not prospectively associated with neck/shoulder pain in the positive direction, and there was inconclusive evidence on the direction with LBP and general MSP. Also, from the reviewed experimental/intervention studies, we observed evidence of positive associations of reduced workplace sitting with a reduction in LBP, neck/shoulder pain, and general MSP/discomfort; nevertheless, no evidence on whether reduced workplace sitting is associated with a reduction in extremities pain.

A recent review of prospective studies has reported that device-measured workplace sitting among tradespeople to be associated with a reduced risk of LBP and neck pain [26]. Compared to our review, there are some similarities in the findings even though we were limited by the volume of studies reviewed in this context. For example, there was an indication from our reviewed prospective studies that device-measured workplace sitting could have a negative association with LBP-intensity which may be dependent on the physical demand of the occupation. Similarly, there is a likelihood of a negative cross-sectional association of device-measured workplace sitting bout with LBP-intensity which is potentially moderated by overweight/obesity in tradespeople. Additionally, our reviewed studies on device-measured workplace sitting in tradespeople suggest a probable negative association with neck/shoulder pain-intensity. A possible explanation of the observed tendency of protective associations of workplace sitting with some MSP conditions in tradespeople could be the physically intensive nature of some of these occupations compared to desk-based occupations. For instance, we also observed in our meta-analysis that self-reported workplace sitting of cross-sectional studies be positively associated with LBP in non-tradespeople but not in tradespeople, albeit in a limited number of studies. Some proponents of the “physical activity paradox” assert that sitting could be of health benefit in individuals who regularly engage in high occupational physical activity as sitting may allow some form of rest and recovery [40, 41]. These indications in our review are, however, inconclusive and warrant further investigations in diverse occupational settings to ascertain these findings.

Generally, our meta-analysis of cross-sectional studies indicated that self-reported workplace sitting significantly increases the odds of LBP by 1.47 times; and was marginally higher, by 1.56 times, in a subgroup of non-tradespeople (Fig. 3). In contrast, previous reviews have reported no evidence of association of workplace sitting with LBP [22, 23]. These previous reviews included both cross-sectional and prospective studies; in contrast, our evidence was synthesised from only cross-sectional studies, therefore, limiting any interpretation of a causal relationship of workplace sitting with LBP. The possibility of reverse causation along with bias in self-reported data in the cross-sectional studies reviewed may adversely affect the quality of evidence in the observed positive association. Similarly, this may have affected the interpretation of the association between SB and neck/shoulder pain.

Also, our best evidence synthesised indicates there is moderate cross-sectional evidence that computer time (≥ 4 h/day) increases the risk of neck/shoulder pain; two previous systematic reviews of prospective studies [21, 24] and RCT studies [21], however, have reported no association of computer time with the risk of neck pain. Furthermore, there is informative evidence of a probable association between vehicle time and LBP. A pooled meta-analysis of three cross-sectional studies indicates prolonged hours of sitting in a vehicle increase the odds of LBP, but the association is not statistically significant. No published review studies, to our knowledge, have specifically investigated vehicle time and MSP conditions, nonetheless, a recent review has reported that MSP conditions are highly prevalent in vehicle drivers [132]. The cross-sectional evidence of computer and vehicle times is, however, of low quality and limited by a small volume of reviewed studies precluding the possibility of causal relationships.

Evidence on the effects of changes in workplace sitting on MSP conditions is scarce. In contrast, workplace interventions to reduce MSP conditions have provided some insight into the benefit of increased workplace physical activity on musculoskeletal health for comparison [133,134,135,136]. For instance, increased occupational physical activity is reported to be associated with reduced general MSP symptoms [133, 134, 136]. Also, a review of experimental studies has reported that device-measured continuous uninterrupted sitting is associated with the increased immediate report of LBP in adults [25]. The evidence from our review also suggests experiments/interventions that reduce total workplace sitting time or sitting bout duration potentially reduce general MSP/discomfort, especially in the lower back and the neck/shoulder. This is consistent with a review that found that workplace interventions potentially reduce LBP and neck/shoulder pain among workers [133, 134]. These findings should be treated with caution due to the limited number and variations in the reviewed experimental/intervention studies.

This review did not specifically investigate the potential mechanisms that underpin the association of occupational and non-occupational SB with MSP conditions. Nevertheless, some previous studies have speculated the potential mechanisms of the association between SB and MSP conditions such as LBP [37, 137]. For instance, studies that have investigated biomechanical and physiological mechanisms of LBP suggest occupational sitting increases spinal load and accumulation of metabolites that accelerate degenerative changes in vertebral discs [36, 37]. The available systematic review literature on the association between SB and MSP conditions is yet to address potential biological mechanisms. Nonetheless, there is an observation in this current review that indicates the association of occupational SB with, for example, LBP may be modulated by overweight/obesity. Increasingly, higher volumes of SB are linked with adiposity [38]; adipose tissue is metabolically active, releasing pro-inflammatory cytokines and adipokines that may potentiate inflammatory changes in the musculoskeletal systems leading to pain [138]. There is, therefore, a need for further studies on the potential biological mechanisms that explain the associations.

Implications for practice and research

Despite the methodological challenges within the reviewed studies in this current systematic review, the overall observation which is supported by the evidence from experimental/intervention studies is that SB may have a detrimental association with musculoskeletal health. Theoretically, replacing a portion of time spent in SB with physical activity could beneficially impact MSP conditions. For instance, one of our reviewed studies [67] reported that substituting 30 min of a full day’s total sedentary time with 30 min of moderate-to-vigorous physical activity (MVPA) may reduce general MSP by 29%. Further, evidence from some of the reviewed experimental/intervention studies also indicates that reduced workplace sitting, and increased standing or walking did not worsen general MSP symptoms [116, 121, 123]. Current WHO physical activity and sedentary behaviour guidelines, in part, recommend reducing and interrupting prolonged SB or sitting with physical activity of any intensity for improved health outcomes [139]. This practice guideline could be encouraged in adults, especially in occupational settings to minimise the risk of MSP conditions.

Our review has identified some knowledge gaps for potential further studies. For instance, inconsistent associations were observed for self-reported and device-measured SB. The evidence of positive cross-sectional associations of SB with MSP conditions was mainly based on self-reported SB. The evidence synthesised from the few studies that investigated device-measured SB was inconsistent with MSP conditions. There is evidence of disparities in device-measured and self-reported SB in adults, with increased potential of self-reported tools to either underestimate or overestimate SB [27]. Furthermore, there were some variations in the measures of MSP conditions; some studies investigated single MSP conditions and some multiple MSP conditions, which could impact the studies' quality and their comparability. Also, the review identified insufficient evidence of prospective associations of SB with MSP conditions and could not make definite conclusions regarding possible causal relationships due to the limited number of prospective studies. Hence, future attention on the application of device-measured SB will be relevant in this context to minimise bias in the probable associations, taking into consideration the outcome measure. Specifically, future research focus could explore the use of posture-based activPAL, the gold standard instrument for measuring sitting time, in prospective study designs. Additionally, some contemporary analytical approaches in the field, such as compositional data analysis could be applied to investigate SB associations relative to other 24-h movement behaviours such as physical activity and sleep with MSP conditions [140]. This review mainly examined the associations of SB with different types of MSP conditions and did not consider the underlying pathophysiology of the MSP conditions. Future studies could also examine the direction of the associations in subgroups of particular MSP conditions. For instance, the direction of association of SB with LBP secondary to lumbar disc degeneration may contrast with the association of SB with LBP due to facet joint inflammation.

This review and previous reviews have not investigated the probable interaction of chronic diseases in the association of SB with MSP conditions. Importantly, MSP conditions are highly prevalent in the presence of multi-morbidities [3, 4], and also emerging as common comorbidities in some chronic diseases, especially type 2 diabetes (T2D) [141,142,143]. Evidence from an observational study, for example, suggests there is a potential interaction of SB with the association of T2D with MSP conditions in adults [141]. Therefore, it will be of great interest for potential future studies, including cross-sectional, prospective, and RCTs study designs to also focus on the interaction of some chronic diseases such as obesity, T2D, cardiovascular diseases, etc. with the association of SB with MSP conditions. Research in this direction will also provide insight into the understanding of the potential biological mechanisms of SB/MSP conditions associations.

Strengths and limitations

A key strength of this review is its distinct consideration of occupational and non-occupational SB, as well as a wide range of MSP conditions. Also, the evidence synthesis was organised into SB domains and measures, likewise the type of MSP outcomes. For a better insight into the risk associations, studies conducted exclusively in clinical groups diagnosed with MSP conditions and those of autoimmune disease-related MSP conditions were not reviewed.

However, we acknowledge that there are some limitations, and caution should be applied when interpreting the findings. First, a single reviewer initially excluded irrelevant studies by title and abstract screening in stage one of two-phase screening; this might have contributed to exclusion of some relevant studies [144]; however, where there was uncertainty regarding inclusion, such studies were considered for second-stage screening by two independent reviewers. Second, most of the studies reviewed were cross-sectional in design, hence, causality cannot be inferred. Third, there were a limited number of studies, especially prospective and experimental/intervention studies, as well as high methodological and analytical variations in the reviewed studies. The limited number of experimental/intervention studies, especially RCTs, may be because we used the term “sitting” to search for “sitting reduction interventions” and “sitting experimental studies” instead of searching for specific interventions (e.g., sit-stand workstations, stand-up desk, etc.). Also, the limited number of prospective studies might be a result of publication bias as some prospective studies on risk factors for MSP conditions may have examined sitting as a risk factor or have accounted for SB as a confounder but found no association and did not report in the Abstract; therefore, these studies would not be identified by the search.

Fourth, a small number of studies were included in the meta-analyses to estimate the pooled effect sizes, resulting in moderate-to-high heterogeneity in some of the outputs. It is important, however, to note that the inverse-variance meta-analysis approach has a limitation of estimating a false high heterogeneity [145]. Therefore, the observed heterogeneity may be potentially due to variations within the studies but not bias in the results. Fifth, we did not consider the covariates adjusted for in the individual studies in our evidence synthesis. For instance, evidence synthesised from studies that accounted for physical activity might be different from those that did not control for physical activity in analyses. Similarly, studies that accounted for sitting positions assumed (e.g., leaning forward or backward) and occupational activities may influence the evidence synthesised from those that did not account for these factors. Also, specific sources of potential bias and specific limitations that were commented upon by the authors of the reviewed studies, or which potentially could be identified in the studies might impact the findings but were not considered in the evidence synthesis.

Sixth, strict selection criteria were adapted to enhance the efficiency of the review, however, this might consequently lead to studies with relevant information being excluded. Furthermore, we adapted the PICO format in constructing our search terms which included search terms for the outcome to maximise the search output. There is the possibility that the outcome may not be well described in the title and abstract of potential studies and therefore not indexed in databases with controlled vocabulary terms leading to missing potential studies [146]. Finally, only articles published in the English language were reviewed; this could bias our finding as informative evidence in studies published in other languages may have been missed. To minimise this shortcoming, however, we also searched grey literature to identify more relevant studies.

Conclusions

Our systematic review identified evidence of cross-sectional associations of SB (occupational and non-occupational) with MSP conditions. The direction of the association of occupational SB with some MSP conditions, nonetheless, may be dependent on the type and physical demand of the occupation involved. The possibility of reverse causation could not, however, be discounted from the observed cross-sectional associations. Further, evidence from intervention studies shows that reducing prolonged sitting at work reduces MSP conditions and discomforts. There was, however, limited evidence of prospective associations of SB with MSP conditions. Importantly though, the review highlighted some knowledge gaps, including a limited number of studies using device-measured SB and MSP conditions, as well as limited prospective and RCT study designs. Considering the inconsistencies of the review’s findings, as well as the highlighted knowledge gaps, further research, especially prospective and RCT studies, is required to better understand the association of SB in occupational and non-occupational settings with MSP conditions. Furthermore, as studies of clinical groups with existing MSP conditions were not reviewed in this current study, future review studies could consider exclusively reviewing this study population. Such studies could also consider examining the contribution of the presence of MSP conditions to the engagement in SB. Also, there is the need for tailored studies to understand the potential interactions of chronic diseases such as obesity, T2D, and cardiovascular diseases in the association of SB with MSP conditions.