Article Text

Download PDFPDF

Associations of screen work with neck and upper extremity symptoms: a systematic review with meta-analysis
Free
  1. Pieter Coenen1,
  2. Henk F van der Molen2,
  3. Alex Burdorf3,
  4. Maaike A Huysmans1,
  5. Leon Straker4,
  6. Monique HW Frings-Dresen2,
  7. Allard J van der Beek1
  1. 1 Department of Public and Occupational Health, Amsterdam UMC, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands
  2. 2 Coronel Institute of Occupational Health, University of Amsterdam, Amsterdam, The Netherlands
  3. 3 Department of Public Health, Erasmus MC, Rotterdam, The Netherlands
  4. 4 School of Physiotherapy and Exercise Science, Curtin University, Perth, Western Australia, Australia
  1. Correspondence to Dr Pieter Coenen, Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Public and Occupational Health, Amsterdam Public Health research institute, Amsterdam, The Netherlands; p.coenen{at}vumc.nl

Abstract

Objectives It has often been suggested that screen work (ie, work on desktop, laptop, notebook or tablet computers) is a risk factor for neck and upper extremity symptoms. However, an up-to-date overview and quantification of evidence are lacking. We aimed to systematically review the association of exposure to screen work with neck and upper extremity symptoms from prospective studies.

Methods An electronic database search (PubMed, Embase, Cinahl and Scopus) for prospective studies on the association of exposure to screen work and musculoskeletal symptoms was conducted. Studies were synthesised regarding extracted data and risk of bias, and meta-analyses were conducted.

Results After screening 3423 unique references, 19 articles from 12 studies (with 18 538 participants) were included for the current review, with the most recent exposure assessment reported in 2005. Studies described duration and input frequency of screen work (ie, computer, keyboard and mouse use, assessed using self-reports or software recordings) and musculoskeletal symptoms (ie, self-reported neck/shoulder and distal upper extremity symptoms and diagnosed carpal tunnel syndrome [CTS]). Although there was overall an increased occurrence of musculoskeletal symptoms with larger exposure to screen work (relative risk: 1.11 [1.03 1.19]), findings were rather inconsistent with weaker (and statistically non-significant) risks when screen work was assessed by software recording (1.05 [0.91 1.21]) compared to with self-report (1.14 [1.03 1.19]).

Conclusions We found an increased risk of musculoskeletal symptoms with screen work. However, the evidence is heterogeneous, and it is striking that it lacks information from contemporary screen work using laptop, notebook or tablet computers.

  • screen work
  • computer use
  • musculoskeletal
  • systematic review
  • meta-analysis

Statistics from Altmetric.com

Key messages

What is already known about this subject?

  • Screen work (ie, work on desktop, laptop, notebook or tablet computers) has suggested to be a risk factor for neck and upper extremity musculoskeletal symptoms.

  • An up-to-date overview and quantification of existing evidence are lacking.

What are the new findings?

  • From data of 12 studies (n=18 538), we showed an 11% increased occurrence of musculoskeletal symptoms with larger exposure to screen work.

  • However, the evidence is heterogeneous, for example, with weak (and statistically non-significant) risks when screen work was assessed by software recording.

  • Also, current evidence lacks information from contemporary (eg, laptop, notebook or tablets) screen work.

How might this impact on policy or clinical practice in the foreseeable future?

  • Occupational health practice should target excessive screen work in order to prevent neck and upper extremity musculoskeletal symptoms.

  • Nevertheless, more evidence on contemporary screen work is needed.

Introduction

Musculoskeletal symptoms are highly prevalent,1 can have a large impact on people’s quality of life2 and put a major burden on our society, for example, through sickness absence.3 The prevalence of musculoskeletal symptoms can vary considerably between occupational groups.4 Among office workers, symptoms in the neck and upper extremities have shown to be particularly prevalent,5 with a 12-month prevalence of 42% and 20% for the head/neck and wrist/hand areas, respectively. Despite these high prevalences, musculoskeletal symptom incidences have shown to range from 2% to 6% per 12 months in various body areas.6 7 Also, the occurrence of clinically diagnosed musculoskeletal disorders has shown to be relatively low among office workers, for example, a 6% point prevalence of diagnosed epicondylitis.8

It has been suggested that screen work (ie, work on desktop, laptop, notebook or tablet computers) may be associated with neck and upper extremity symptoms in office workers,9–11 possibly contributing to the occurrence of musculoskeletal symptoms among office workers. If such an association would be confirmed, physical work demands, as well as psychosocial and organisational factors associated with screen work may play an important (and possibly interacting) role in the development of neck and upper extremity symptoms.10 12 For example, awkward body postures and repetitive movements during screen work have been suggested to be risk factors for musculoskeletal symptoms,9 as they may cause increased muscle tension in the forearm and neck/shoulder region. Increased muscle tension has also been suggested to arise from psychosocial and organisational work environment factors.13 An alternative explanation may be that psychosocial workplace stressors associated with screen work14 may be associated with a central sensitisation response in the body that can lead to musculoskeletal symptoms.15 To conclude, musculoskeletal risks associated with exposure to screen work can be physical or psychosocial, and may also interact with other aspects of work.

International guidelines and European directives16 advise workers to refrain from working with a computer for uninterrupted periods longer than 2 hours, while daily computer usage should be limited to 5–6 hours. These guidelines are, however, rather outdated. As the exposure–outcome association between screen work and neck and upper extremity symptoms is currently unclear, an evidence base for revising these guidelines, therefore, is lacking. An up-to-date evidence-base is needed to support revised guidelines for screen work given the high and growing amount of screen work in the current occupational setting. For example, it has been shown that while in 2005, only 36% of the working population was exposed to at least moderate amounts of screen work, this proportion had increased to 56% in 201517. In the Netherlands, it has been shown that present-day workers report spending an average of 3.9 hours per day on their computer at work, with 38% of the working population spending >6 hours per day.18

Although there are several systematic reviews on the association of screen work with neck and upper extremity symptoms,9–12 19 20 these reviews are either outdated or have not expressed a pooled risk estimate in quantified units of exposure to screen work. Furthermore, these reviews have not considered the new generation (such as laptop, notebook and tablet) screen work.

We aimed to perform a meta-analysis on evidence for the association of exposure to screen work with the occurrence of neck and upper extremity symptoms from longitudinal cohort studies. To get a broad overview of the evidence on this topic, we aimed to incorporate evidence from studies in which screen work was assessed by either self-reports or software recordings. Also, we aimed to consider self-reported pain or discomfort, or specific and/or diagnosed conditions to assess musculoskeletal symptoms. We aimed to incorporate evidence for traditional (eg, desktop) and more contemporary devices.

Methods

Search for literature

This systematic review was a-priori registered21 and executed according to Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) statement guidelines.22 A literature search for articles was conducted in the electronic databases PubMed, Embase, Cinahl and Scopus from 1981 (the introduction of the IBM personal computer) to 15 January 2019. Search terms included controlled terms (from MeSH in PubMed and EMtree in EMBASE) as well as free-text terms (online supplementary file 1), including terms expressing: (1) the exposure (ie, screen work), (2) the population (ie, workers and students), (3) the outcome (ie, neck and upper extremity symptoms) and (4) the study design (ie, longitudinal cohort studies on humans).

Supplementary file 1

Titles and abstracts of identified records were assessed on eligibility by two reviewers independently (PC and HvdM) using the Covidence online tool (www.covidence.org). Full-text copies of the identified articles were retrieved and assessed for inclusion. In case of any disagreement, consensus was reached during a meeting, if necessary with a third researcher (MH). Reference lists from included articles and personal databases of review authors were cross-checked for additional articles.

Articles describing studies that examined exposures to occupational screen work (ie, work conducted on desktop, laptop, notebook or tablet computers) and its association with neck and upper extremity symptoms were included. Here, occupational screen work refers to any screen activities undertaken during paid work or student activities, as we believed that there is no substantial difference in mechanisms for symptom development between these two contexts. All potential definitions for the quantification of exposure to screen work were considered, for example, duration of uninterrupted bouts as well as the total duration of screen work, but also type/click frequency of a keyboard and/or mouse.

Studies on visual display terminals (VDT) were only included if such terminals consisted of input devices comparable to those of desktop, laptop, notebook or tablet computers, that is, having a mouse and keyboard rather than a joystick. Studies that assessed screen work using self-reports or other methods, such as software recordings or structured observations, were included. Studies on symptoms of the upper extremity (including the hand, wrist, elbow and shoulder) and neck, including self-reported (non-specific) pain or discomfort, or specific and/or diagnosed conditions (eg, carpal tunnel syndrome [CTS] or epicondylitis) in neck or upper extremity regions of the musculoskeletal system were included. Here, only studies assessing new occurrences of musculoskeletal symptoms, either by excluding participants with baseline symptoms or by the original study authors adjusting for baseline symptoms, were included. We excluded studies that did not quantify the specified association (with the exposure quantified in input frequency and/or duration using effect sizes such as risk ratios [RR], HRs or ORs) or that did not enable differentiation of exposure to screen work from exposure to other non-screen (office) tasks. Only articles published in peer-reviewed journals, written in English and describing original research from prospective epidemiological studies with a follow-up period of at least 1 week were included. Only articles from studies on adult (aged between 18 and 65 years) participants conducting occupational or educational screen work were included.

Data extraction and risk of bias assessment

The following information from the included studies was extracted by two reviewers: first author and year of publication, study name, study design (and follow-up period), sample description (number of participants, relevant inclusion/exclusion criteria, %female, age, country and occupation), adjustment (confounders and/or effect modifiers), description of screen work (type of screen work or input devices), way of assessment and exposure categories, description of outcome (way of assessment and occurrence during the follow-up period) and effect measures (eg, RR, OR and HR). Corresponding authors were contacted and asked for additional information in cases where data provided in the published articles were insufficient.

For risk of bias assessment, we used the ‘Quality in Prognosis study Tool’, a previously published scoring system23 with six criteria related to the study participation, attrition, prognostic factor (ie, exposure) measurements, outcomes, confounding and statistical analysis (online supplementary file 2). Sum scores were calculated by averaging individual criteria scores (high risk=0; medium risk=0.5; low risk=1), with sum scores ranging from 0 (high risk of bias across all criteria) to 1 (low risk of bias across all criteria).

Supplementary file 2

Data analysis

Included studies were described according to their extracted data and assessed risk of bias. A meta-analysis was conducted, if possible, for which reasonable differences in the definition of exposure and musculoskeletal symptoms were accepted. In the case of multiple articles reporting on the same study, only complementary information from various articles was used for further analyses.

Regarding the meta-analysis, for all exposure metrics and outcome definitions in each individual study, an exposure–outcome association (expressed as beta) was calculated using linear regression analysis. The exposure metric was standardised for the duration of screen work by 10 hours/week and input frequency by 10 clicks or strokes/minute. In order to do so, information was retrieved from the original studies. As some studies used categories of exposure to screen work, the midpoint of an exposure category was used as average exposure, for example, 5 hours/week for a 0–10 hour/week exposure category. For the highest exposure category, with an unbounded upper value, a measure was chosen to reflect a meaningful estimate. For example, computer use >40 hours/week was set at 40 hours/week. It was assumed that several risk estimates could be interpreted equally. In the regression analysis, exposure–response associations were forced through the origin under the assumption that workers without exposure to screen work do not have an increased occurrence of musculoskeletal symptoms attributable to screen work.

Exposure–outcome associations of individual studies (expressed in RR) were pooled using Review Manager (RevMan) V.5.3. Pooled exposure–outcome associations were presented in forest plots. Heterogeneity of the findings was assessed using I 2 statistics and visual inspection of the forest plots. A random-effects model was applied as there was evidence of heterogeneity with I 2 >50%.24

In accordance with our a-priori registered protocol,21 results were stratified based on (1) exposure device: computer versus mouse versus keyboard (with no evidence on other devices), (2) exposure assessment method: self-reported versus software recorded, (3) exposure quantification: that is, duration versus input frequency of screen work and (4) musculoskeletal symptom area (with three categories: self-reported neck/shoulder, self-reported distal upper extremity symptoms and clinically/physically diagnosed CTS). We used both univariate and multivariate meta-regression analyses to investigate how the above-mentioned four variables affected pooled risk estimates.25 The variable exposure device was dichotomised (computer vs mouse/keyboard), and so was the variable musculoskeletal symptom area (self-reported symptoms vs clinically/physically examined CTS). This dichotomisation was chosen after exploring the data, showing comparable findings for mouse and keyboard use, as well as for neck/shoulder and distal upper extremity symptoms. Due to the low number of studies with a low risk of bias, no subgroup analysis regarding the risk of bias was performed (which would have been in accordance with our a-priori registered protocol).

Results

Selection process

The flow chart of the search and selection process is shown in figure 1. The literature search generated a total of 5326 references. After removing duplicates, 3423 unique references were screened by their title and abstract, and 54 articles by full text. Of the latter, 37 were excluded for various reasons (online supplementary file 3), yielding 17 eligible articles. Another 2 articles were retrieved from screening reference lists, resulting in 19 articles from 12 studies being included in this review6 7 26–42 (with 7 articles reporting on study data that were already considered in another article). In these studies, 18 538 participants were measured in total. Two articles reported on duplicate data already reported in another article and were not used for qualitative or quantitative analysis.32 40 Moreover, one article reported on two different studies.43

Supplementary file 3

Figure 1

Flowchart depicting the literature search.

General descriptives

Extracted data from the included articles are summarised in the online supplementary file 4. The vast majority of studies was from North/West-European origin with two studies on a US sample.26 27 Participants in the studies were, in general, sampled from office-based populations with one study among university students28 and two studies in a sample of the general working population.27 29 Articles, in general, described studies on participants from working age of both male and female participants.

Supplementary file 4

Regarding the risk of bias (figure 2; online supplementary file 5), included articles scored relatively well, with an average (SD) methodological quality of 0.78 (0.08), ranging from 0.67 to 0.92. All studies scored a low risk of bias on the statistical analysis, whereas not all studies scored that well on study attrition, prognostic factor (exposure to screen work) and outcome (musculoskeletal symptom) assessment. A reason for the risk of bias due to these assessments was the use of not well described and/or non-validated assessment methods.

Supplementary file 5

Figure 2

Risk of bias of included articles. The risk of bias (ie, + is low risk of bias, − is high risk of bias and ? is unclear) in six domains (ie, study participation, study attrition, prognostic factor, outcome, study confounding and statistical analysis) is depicted, while also sum scores are shown.

Exposure to screen work

The assessment of exposure to screen work in the included studies was executed in periods from 1984 to 2005. We found 10 studies assessing computer use,26–35 3 on keyboard use,7 32 35–38 4 on mouse use6 7 32 33 35–39 and 1 on VDT use.40 In included studies, both duration and input frequency of keyboard and mouse were assessed, whereas for computer and VDT, only the duration of use was assessed. We found nine studies in which the exposure was assessed using self-reports,26–31 33 35 40 while in two studies (the NUDATA and the PROMO study, respectively) both self-reports and software recordings (ie, using dedicated software to record the duration and input frequency of use) of screen work were reported6 7 32 36–39; in one study, the exposure was assessed using a job–exposure matrix.27 No studies on laptop, notebook and/or tablet computer use were identified, despite our aim and search to include such studies in this particular review.

Musculoskeletal symptoms

The average (with SD) duration of the follow-up period across studies was 2.4 (1.4) years, ranging from 0.8 to 5.0 years. Nine studies reported neck or neck/shoulder symptoms,6 9 26 28 30 31 33 35 37 40 four studies shoulder pain only,6 30 34 35 37 three elbow symptoms,7 31 34 39 five studies forearm/wrist/hand symptoms,7 26 30 32 33 35 39 one study on all body areas38 and four studies CTS.27 29 36 Studies reported on complaints, discomfort, pain, symptoms and disorders, which were self-reported in 10 studies6 7 26–35 37–40 and clinically/physically diagnosed (ie, only CTS) in 3 studies.27 29 For further qualitative analyses (see below), above-mentioned outcomes were categorised into self-reported neck/shoulder symptoms, self-reported distal upper extremity symptoms and clinically/physically diagnosed CTS.

New occurrences of symptoms were assessed in all studies, either by excluding persons with baseline symptoms6 7 26–29 32–40 or by adjusting for baseline symptoms.30 31 41 42 The recall period of the symptoms studied varied from current status26–29 to 1 week6, 3 months32 34–36 41 42 and 12 months.7 30 31 33 37–40 Apart from this variation in recall period, symptom definition was heterogeneous with definitions such as ‘pain severity >4 (on a 0–7 scale)’,6 severity of >6/10 (yes/no),26 ‘regular or prolonged pain (yes/no)’32 and ‘≥8 days of pain’.40

Quantitative analyses

Data from a total of 13 articles (reporting 58 different exposure-outcome associations) were used for quantitative analysis, assessing exposure–outcomes for both duration and input frequency of screen work. According to the findings of our meta-regression (table 1), exposure to screen work was associated with an increased occurrence of musculoskeletal symptoms, showing RR of 1.11 (1.03 to 1.19). Associations were stronger (in a statistically significant manner) for exposure to keyboard/mouse use (RR=1.16 [1.06 to 1.27]) compared with computer use (RR=1.00 [0.90 to 1.11]). Associations were stronger (although not statistically significant) when the exposure was quantified by self-reports (RR=1.14 [1.06 to 1.24]) compared with software recordings (RR=1.05 [0.91 to 1.21]). Associations were also stronger (although not statistically significant) when the exposure was expressed in duration (RR=1.13 [1.05 to 1.23]) compared with input frequency (RR=0.96 [0.90 to 1.03]). Moreover, associations were stronger (although not statistically significant) when the musculoskeletal outcome was self-reported neck/shoulder or distal upper extremity symptoms (RR=1.13 [1.05 to 1.21]) compared with clinically/physically diagnosed CTS (RR=0.94 [0.69 to 1.29]).

Table 1

Univariate and multivariate meta-regression models, with exposure device, exposure assessment, exposure quantification and outcome modelled as independent variables and effect size (expressed in beta) as a dependent variable

More specific findings are tabulated in table 2 and the online supplementary files 6−13. Computer use, as assessed by self-reports, showed a (statistically non-significant) increased risk of neck/shoulder and distal upper extremity symptoms, and a reduced risk of CTS (online supplementary file 6). When computer use duration was assessed by software recordings (from one high-quality study), a lower risk of symptoms with more hours of computer use was seen (online supplementary file 7). Keyboard use, assessed by self-report, showed a higher risk of symptoms with more hours of keyboard use (online supplementary file 8). However, when keyboard use was assessed by software recordings, results were inconsistent and non-significant (online supplementary file 9). A higher risk of symptoms with more hours of mouse use was observed when mouse use was assessed by self-reports (online supplementary file 10) as well as by software recordings (online supplementary file 11), although not statistically significant for all associations. Regarding input frequency of keyboard and mouse use, one study per outcome was incorporated showing a (non-statistically significant) tendency for a higher input frequency of screen work being associated with a reduced risk of musculoskeletal symptoms (online supplementary files 12 and 13).

Supplementary file 6

Supplementary file 7

Supplementary file 8

Supplementary file 9

Supplementary file 10

Supplementary file 11

Table 2

Overview of study findings

Supplementary file 12

Supplementary file 13

Of the studies not used in the quantitative analysis (four in total), one study reported a significantly increased risk of those having had >5 years of exposure to computer use, for hand/arm symptoms (RR=2.0 [1.2 to 3.3]), but not for neck/shoulder symptoms (RR=0.9 [0.6 to 1.4]).26 From one study, although the exposure was quantified, no exposure–outcome associations were reported,31 just that ‘very often’ use of the computer (compared with seldom or never) led to an increased (and not statistically significant) risk of elbow/wrist hand symptoms OR=1.42 [0.70 to 2.86], but not neck/shoulder symptoms OR=0.94 [0.60 to 1.48]. There was only one study on VDT use that could not be included in the pooled analysis,40 showing no difference in neck pain risk among those who spend <50% compared with those who spend >50% of their work time on using the VDT (OR=1.0 [0.6 to 2.9]). We identified one study on musculoskeletal symptoms in general (not in one particular body area), showing a strong association of increased occurrence of symptoms among those exposed to mouse use >30 hours/week compared with <9 hours/week (OR=8.4 [2.5 to 29]).38 In the same study, however, such an association was absent for exposure to keyboard use.

Discussion

In this study, we explored the association of exposure to screen work using different devices with the occurrence of neck and upper extremity symptoms in a systematic review of prospective studies. Overall, we found an 11% increased risk of musculoskeletal symptoms with screen work (with an increment in every 10 hours/week or 10 clicks or strokes/minute). Based on these numbers, the increased risk of musculoskeletal symptoms in workers who spend a lot of time on their computer (eg, office workers having 40 hours of screen work per week) compared with those who do not use the computer at all, might be substantial (ie, up to 44%). It has to be acknowledged, however, that risk of musculoskeletal symptoms with screen work differs considerably between studies in which screen work was measured by software recordings (5% increased risk) compared with self-reports (14% increased risk), with the former risk not being statistically significant.

Our findings are in the line with earlier body of evidence,20 while we add results from our meta-regression to be able to differentiate different sources of heterogeneity. As such, we have quantified associations of different modalities of screen work (ie, computer, keyboard and mouse use, expressed in duration and input frequency of the use of these devices and when measured with self-reports and by software recordings) and musculoskeletal symptoms (ie, self-reported neck/shoulder and distal upper extremity symptoms and clinically/physical diagnosed CTS). This information may be important in the development and evaluation of future interventions to prevent musculoskeletal symptoms due to screen work.43

Exposure device, assessment and quantification

Stronger (and statistically significant) associations with musculoskeletal symptoms were found for keyboard and mouse use than for computer use in general. This is in line with an earlier review by IJmker and colleagues reporting stronger associations for mouse use,9 with the authors attributing this effect to a lack of variation in working postures during mouse use,44 potentially leading to longer duration of continuous muscle loading.45

We found weaker (and not statistically significant) associations with musculoskeletal symptoms when exposure to screen work was software recorded. It has been shown before that self-report measurements generally overestimate the duration of computer use, resulting in misclassification.46 47 If this misclassification would be non-differential, it would lead to an underestimation of the true exposure–outcome association.48 This is, however, not in line with findings from our review. Namely, our results suggest that people who are prone to musculoskeletal symptom development may perceive a higher exposure to screen work than people who do not develop musculoskeletal symptoms. These findings support the notion that while exposure to screen work measured by software recordings is likely to be capturing ‘true’ exposure, self-reported screen work may reflect a combination of duration/frequency of exposure and experienced muscle tension or sustained stress. This confirms earlier research showing that perceived muscular tension in office workers is a strong predictor of musculoskeletal symptoms.49 50 An additional implication of these findings is that (self-reported) perceived exposure may be important in identifying those at risk for developing musculoskeletal symptoms. As perceived exposure is more sensitive than (software recorded) ‘true’ exposure in associations with musculoskeletal symptoms, it could be an important focus of musculoskeletal symptom prevention in office workers.

We also showed that the association of exposure with musculoskeletal symptoms was generally stronger (although not statistically significant) when the exposure was expressed in duration than when it was expressed in input frequency (ie, in clicks or keystrokes per minute) of screen work. However, we identified little evidence on the input frequency of screen work, as a result of which we recommend to incorporate software recordings of screen work in future work to gain more evidence on the input frequency of screen work as well as on variation in exposure. Future research should also examine other potential features, such as kinematics, muscle activity and click or keystroke force, in prospective studies to identify potential mechanisms and thus targets for intervention.

Musculoskeletal symptom definition

We reported the strongest (although not statistically significant) risk estimates for exposure to screen work with self-reported neck/shoulder or distal upper extremity symptoms. Associations for computer use with clinically/physically diagnosed CTS were generally weaker and even showed a protective effect of computer use with CTS for some of the associations. An explanation for the latter may be that two of the studies reporting on CTS27 29 addressed a sample of non-office workers, in which a lack of exposure to screen work may constitute an exposure to other physical work demands (eg, manual handling). Nevertheless, our findings are in line with those by Waersted and colleagues who concluded that there was only limited evidence for an association of computer use and clinically diagnosed musculoskeletal symptoms.11 Notably, in our attempt to include studies on clinically/physically diagnosed conditions, we only identified research on CTS. Other reviews indicated more consistent associations of screen work with hand-arm symptoms than with neck-shoulder symptoms,9 which we have not been able to replicate in the current review. Namely, we found associations with outcomes of neck/shoulder symptoms to be fairly comparable to those of more distal upper extremity symptoms.

We identified studies with a large variation in recall periods for the musculoskeletal symptom assessment, which makes it difficult to differentiate acute from chronic symptoms. In our review, we aimed to investigate the effect of screen work on new occurrences of musculoskeletal symptoms, by addressing studies that either excluded participants with baseline symptoms or adjusted for baseline symptoms. However, due to the recurrent nature of musculoskeletal symptoms and as musculoskeletal symptoms develop as early as during adolescence,51 it is likely that many persons in working age may have experienced symptoms before enrolment in the study population.

Implications for ergonomic practice and prevention research

Interventions for the prevention of musculoskeletal symptoms in office workers targeting screen work (eg, by trying to modify the workplace or altering the work content) have had limited effect thus far,52 53 with moderate evidence at best for interventions, such as mouse use feedback, forearm support and office workstation adjustments. A reason for this lack of evidence may be that the risk of screen work for musculoskeletal symptoms is modest at best and, thus, interventions targeting screen work may only have limited effect. This is reflected in the lack of consistent and conclusive empirical evidence in systematic reviews.9 11 19 54 An alternative reason for the lack of success in musculoskeletal symptoms’  above-mentioned prevention in office workers could be the limited translation of knowledge on the mechanisms underlying the potential association of screen work with musculoskeletal symptoms into well-designed intervention programmes.43 55 For example, our study findings suggest that mainly the duration of exposure to screen work drives the association of screen work with musculoskeletal symptoms. However, most intervention studies seem to focus on making modifications to the physical work setting, thereby aiming to reduce the internal biomechanical loading during screen work, instead of targeting the exposure duration per se,52 53 which would impact on work productivity. For successful musculoskeletal symptom prevention, underlying mechanisms should be revisited (considering all evidence) and should be incorporated into intervention development.43

Methodological strengths and limitations

A strength of the current study is that only articles from prospective studies on the association of quantified exposure to screen work with musculoskeletal symptoms were included. These studies, in which, the exposure precedes the musculoskeletal symptoms, provide the opportunity to infer on the causality of this association.56 Moreover, studies included in this meta-analysis were, in general, of high methodological quality, providing a limited risk of bias.

A limitation of this study is the heterogeneity in study data as a result of which we had to make assumptions regarding exposure, outcome and statistical method used. Sources of heterogeneity include the categories used for exposure, the wording used to question about symptoms and recall period used to assess outcome. Moreover, we assumed that several effect estimates (eg, OR, RR and HR) could be interpreted equally, which may be questioned especially when the outcome of interest is more prevalent, such as musculoskeletal symptoms.57 All the above may have caused some inaccuracy in the pooled estimation of the effect size, potentially influencing the accuracy of our inferences.

Another issue is that there is still limited evidence from studies using measurement methods other than self-reports to assess exposure to screen work. Only two studies (although with nine articles), the NUDATA and PROMO studies, used software recording measurements. This measurement method has, however, the limitation that it only records the time spent using the mouse or keyboard, while it uses an algorithm to estimate the time looking at the screen. Other potential measurements methods that would take this issue into account, such as eye tracking, wearable cameras and systematic observations, were not used in the evidence that we have identified and can be considered in future research. The use of such measures would potentially allow researchers to look at quantifications of exposure to screen work beyond total duration and frequency. For example, distinguishing exposure to screen work in one uninterrupted bout from a similar amount of screen work accumulated from smaller bouts can help to get a better understanding of the mechanisms underlying the association of screen work and musculoskeletal symptoms.

Although the use of laptop, notebook and tablet computers is common these days, we did not identify any prospective studies assessing the association of the use of these devices with musculoskeletal symptoms. The most recent publication dated from 2015 and the most recent exposure assessment was from 2005 (ie, before the introduction of the first tablet computer). Given the changes in screens over the past decade (touch screens, larger flat screens and mobile devices), increased use of screen work and evolution of work systems involving screens, there is a clear need for prospective studies assessing the risks associated with contemporary screen work. It is imaginable that the usage of contemporary devices has differential associations with musculoskeletal symptoms. For example, contemporary devices are more portable and are thus used in more ‘unconventional’ work positions, such as while travelling, on a couch or in bed.58 Even when used on a desk, contemporary devices can have more strenuous physical demands (eg, awkward postures and increased muscle activity) than traditional devices such as desktop computers.59 However, studies focussing on contemporary screen work are mostly lab-based, while evidence from longitudinal epidemiological studies is lacking.

Another limitation of our study is that we assumed the association of screen work and musculoskeletal symptoms to be linear, while research on the association of other physical demands with musculoskeletal symptoms suggests this association to be non-linear.60 However, the available evidence in this systematic review was based on too few studies to investigate alternatives for the currently chosen linear association between exposure and response.

Conclusion

An overall association was found of an 11% increased risk of musculoskeletal symptoms with screen work (for each 10 hours/week or 10 clicks or strokes/minute). However, risks were weaker (and not statistically significant) when screen work was assessed by software recording compared with self-report. This review suggests that, although there is evidence for an association of screen work with musculoskeletal symptoms, the evidence is inconsistent for different types of exposure to screen work and symptom areas and may not reflect the risks associated with contemporary screen work using laptop, notebook or tablet computers. The different effect sizes when screen work was measured by software recordings compared with self-reports suggests differences between these two ways of exposure assessment.

Acknowledgments

The authors wish to thank Linda Schoonmade with her help in running the electronic database searches.

References

Footnotes

  • Contributors All authors were involved in study conception and design, drafting the article and/or critically revising it for important intellectual content, and all authors approved the final version to be submitted for publication. PC and HFvdM: study selection, data extraction and methodological quality assessment. PC: conducted the analyses (with the help of HFvdM and AJvdB) and prepared a draft manuscript. AJvdB: study guarantor.

  • Funding The study has partly been funded by the Amsterdam Public Health Research Institute.

  • Competing interests None declared.

  • Ethics approval As this is a systematic review, no ethical approval is required.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Additional data are provided in the supplementary files.

  • Patient consent for publication Not required.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.