Workplace pedometer interventions for increasing physical activity

Abstract

Background

The World Health Organization (WHO) recommends undertaking 150 minutes of moderate‐intensity physical activity per week, but most people do not. Workplaces present opportunities to influence behaviour and encourage physical activity, as well as other aspects of a healthy lifestyle. A pedometer is an inexpensive device that encourages physical activity by providing feedback on daily steps, although pedometers are now being largely replaced by more sophisticated devices such as accelerometers and Smartphone apps. For this reason, this is the final update of this review.

Objectives

To assess the effectiveness of pedometer interventions in the workplace for increasing physical activity and improving long‐term health outcomes.

Search methods

We searched the Cochrane Central Register of Controlled Trials, MEDLINE, Embase, the Cumulative Index to Nursing and Allied Health Literature (CINAHL), Occupational Safety and Health (OSH) UPDATE, Web of Science, ClinicalTrials.gov, and the WHO International Clinical Trials Registry Platform from the earliest record to December 2016. We also consulted the reference lists of included studies and contacted study authors to identify additional records. We updated this search in May 2019, but these results have not yet been incorporated. One more study, previously identified as an ongoing study, was placed in 'Studies awaiting classification'.

Selection criteria

We included randomised controlled trials (RCTs) of workplace interventions with a pedometer component for employed adults, compared to no or minimal interventions, or to alternative physical activity interventions. We excluded athletes and interventions using accelerometers. The primary outcome was physical activity. Studies were excluded if physical activity was not measured.

Data collection and analysis

We used standard methodological procedures expected by Cochrane. When studies presented more than one physical activity measure, we used a pre‐specified list of preferred measures to select one measure and up to three time points for analysis. When possible, follow‐up measures were taken after completion of the intervention to identify lasting effects once the intervention had ceased. Given the diversity of measures found, we used ratios of means (RoMs) as standardised effect measures for physical activity.

Main results

We included 14 studies, recruiting a total of 4762 participants. These studies were conducted in various high‐income countries and in diverse workplaces (from offices to physical workplaces). Participants included both healthy populations and those at risk of chronic disease (e.g. through inactivity or overweight), with a mean age of 41 years. All studies used multi‐component health promotion interventions. Eleven studies used minimal intervention controls, and four used alternative physical activity interventions. Intervention duration ranged from one week to two years, and follow‐up after completion of the intervention ranged from three to ten months.

Most studies and outcomes were rated at overall unclear or high risk of bias, and only one study was rated at low risk of bias. The most frequent concerns were absence of blinding and high rates of attrition.

When pedometer interventions are compared to minimal interventions at follow‐up points at least one month after completion of the intervention, pedometers may have no effect on physical activity (6 studies; very low‐certainty evidence; no meta‐analysis due to very high heterogeneity), but the effect is very uncertain. Pedometers may have effects on sedentary behaviour and on quality of life (mental health component), but these effects were very uncertain (1 study; very low‐certainty evidence).

Pedometer interventions may slightly reduce anthropometry (body mass index (BMI) ‐0.64, 95% confidence interval (CI) ‐1.45 to 0.18; 3 studies; low‐certainty evidence). Pedometer interventions probably had little to no effect on blood pressure (systolic: ‐0.08 mmHg, 95% CI ‐3.26 to 3.11; 2 studies; moderate‐certainty evidence) and may have reduced adverse effects (such as injuries; from 24 to 10 per 100 people in populations experiencing relatively frequent events; odds ratio (OR) 0.50, 95% CI 0.30 to 0.84; low‐certainty evidence). No studies compared biochemical measures or disease risk scores at follow‐up after completion of the intervention versus a minimal intervention.

Comparison of pedometer interventions to alternative physical activity interventions at follow‐up points at least one month after completion of the intervention revealed that pedometers may have an effect on physical activity, but the effect is very uncertain (1 study; very low‐certainty evidence). Sedentary behaviour, anthropometry (BMI or waist circumference), blood pressure (systolic or diastolic), biochemistry (low‐density lipoprotein (LDL) cholesterol, total cholesterol, or triglycerides), disease risk scores, quality of life (mental or physical health components), and adverse effects at follow‐up after completion of the intervention were not compared to an alternative physical activity intervention.

Some positive effects were observed immediately at completion of the intervention periods, but these effects were not consistent, and overall certainty of evidence was insufficient to assess the effectiveness of workplace pedometer interventions.

Authors' conclusions

Exercise interventions can have positive effects on employee physical activity and health, although current evidence is insufficient to suggest that a pedometer‐based intervention would be more effective than other options. It is important to note that over the past decade, technological advancement in accelerometers as commercial products, often freely available in Smartphones, has in many ways rendered the use of pedometers outdated. Future studies aiming to test the impact of either pedometers or accelerometers would likely find any control arm highly contaminated. Decision‐makers considering allocating resources to large‐scale programmes of this kind should be cautious about the expected benefits of incorporating a pedometer and should note that these effects may not be sustained over the longer term.

Future studies should be designed to identify the effective components of multi‐component interventions, although pedometers may not be given the highest priority (especially considering the increased availability of accelerometers). Approaches to increase the sustainability of intervention effects and behaviours over a longer term should be considered, as should more consistent measures of physical activity and health outcomes.

Plain language summary

Do health promotion programmes in the workplace increase people's physical activity if they include a pedometer?

Key messages

Overall, there is not enough evidence to show whether workplace health promotion programmes involving a pedometer affect people's physical activity, especially in the long term.

What is a pedometer?

A pedometer is a small, portable electronic device that counts the number of steps a person takes, but unlike an accelerometer, there is no record of intensity. Pedometers aim to encourage people to increase their physical activity by giving them feedback on their daily steps.

Why we did this review

Most people do not do enough physical activity. According to the World Health Organization, doing at least 30 minutes of moderately intense physical activity on most days can reduce a person's risk of developing cardiovascular disease, diabetes, and some cancers. We wanted to find out if workplace health promotion programmes that involve wearing pedometers would motivate people to increase their physical activity.

What did we do?

We searched for studies of workplace programmes that used pedometers to promote health in employees. We looked for randomised controlled studies, where the treatment each person receives is decided randomly.

Search date: We included evidence published up to December 2016.

What we found

We found 14 studies including 4762 people in different workplaces, ranging from offices to construction sites, mostly in high‐income countries. In all studies, pedometers were part of a health programme that included other components, such as walking groups, counselling, or dietary advice. Studies compared the effects of participating in pedometer programmes with:

‐ receiving no or a minimal health programme such as regular advice about the benefits of physical activity; and

‐ participating in other health programmes, that did not include a pedometer.

The programmes lasted from two weeks to two years; assessments continued for three to ten months afterwards.

We were most interested to see whether there pedometers had a lasting effect on physical activity and health. We were also interested in learning about effects on sedentary behaviour (time spent sitting), risk factors for cardiovascular disease and diabetes (blood pressure, body mass index (BMI) and levels of cholesterol in the blood), quality of life (well‐being) and adverse (unwanted) effects.

What are the results of our review?

Compared with no or a minimal health programme:

Pedometer programmes may not affect physical activity at least one month after the programme end (6 studies), but they may reduce sedentary behaviour and may improve people's well‐being (1 study). We have very little certainty about these results.

Pedometer programmes may slightly reduce body mass index (3 studies, low certainty), but probably make little to no difference in blood pressure (2 studies, moderate certainty) and may reduce unwanted effects such as injuries (2 studies, low certainty).

No studies measured cholesterol or disease risk scores at least one month after the programme ended.

Compared with another health programme:

Pedometer programmes may affect physical activity after at least one month since the programme end (1 study), but we have very little certainty about this result.

We could not draw conclusions about unwanted effects. The evidence was not good enough for us to be certain about effects on sedentary behaviour, BMI, blood pressure, cholesterol, cardiovascular disease risk and well‐being. Some effects were seen but findings were not consistent.

Conclusions

Exercise programmes can have positive effects on an employee's physical activity and health, but we did not find enough reliable evidence to be certain whether a pedometer programme is better than other types of health programmes, especially for achieving long‐term changes in behaviour.

Evidence is uncertain because results were reported by a small number of studies — sometimes only one study. In most studies, the people involved knew which study group they were in, which can affect results. Many people dropped out of studies before the studies ended, so not enough results were collected. Some studies did not report any results for some measures we were interested in or did not assess/evaluate if benefits were maintained after the program ended.

As pedometers are largely being replaced by more sophisticated devices such as accelerometers and Smartphone apps, we will not update this review again.

Summary of findings

Summary of findings 1. Pedometer intervention compared to minimal intervention for increasing physical activity.

Pedometer intervention compared to minimal intervention for increasing physical activity
Patient or population: employees Setting: workplace Intervention: pedometer intervention Comparison: minimal intervention
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with minimal intervention	Risk with pedometer intervention
Physical activity assessed with combined measures Follow‐up after completion: range 3 to 10 monthsa	See comment	See comment	‐	724 (6 RCTs)	⊕⊝⊝⊝ Very lowb,c,d	Pedometer interventions may have no effect on physical activity compared to minimal intervention, but the effect is very uncertain. Available studies could not be meta‐analysed due to very high heterogeneity (I² = 94%)
Sedentary behaviour Follow‐up after completion: 6 months	Mean sedentary behaviour was 510 min/de	MD 33 min/d lower (84.28 lower to 18.28 higher)	‐	172 (1 RCT)	⊕⊝⊝⊝ Very lowf,g,h	Pedometer interventions may have an effect on sedentary behaviour compared to minimal intervention, but the effect is very uncertain
CVD risk factor: body mass index (BMI) Follow‐up after completion: range 3 to 6 months	Mean CVD risk factor: BMI was 27.94 kg/m²i	MD 0.64 kg/m² lower (1.45 lower to 0.18 higher)	‐	486 (3 RCTs)	⊕⊕⊝⊝ Lowj,k	Pedometer interventions may reduce BMI slightly compared to minimal intervention, based on low‐certainty evidence
CVD risk factor: systolic blood pressure (SBP) Follow‐up after completion: range 3 to 8 months	Mean CVD risk factor: SBP was 127.6 mmHgi	MD 0.08 mmHg lower (3.26 lower to 3.11 higher)	‐	315 (2 RCTs)	⊕⊕⊕⊝ Moderatel	Pedometer interventions probably result in little to no difference in SBP compared to minimal intervention, based on moderate‐certainty evidence
CVD risk factor: LDL cholesterol At completion of interventionm	Mean CVD risk factor: LDL cholesterol was 120.2 mg/dLn	MD 3.58 mg/dL lower (10.76 lower to 3.59 higher)	‐	127 (2 RCTs)	⊕⊝⊝⊝ Very lowg,o,p	Pedometer interventions may have an effect on LDL cholesterol compared to minimal intervention, but the effect is very uncertain
Quality of life: mental health component (QoL) assessed with: SF‐36v2 Health Survey Scale from 0 to 100 Follow‐up after completion: range 3 to 6 months	Mean QoL: mental health component was 51.8e	MD 1.3 higher (1.8 lower to 4.4 higher)	‐	58 (1 RCT)	⊕⊝⊝⊝ Very lowh,q,r	Pedometer interventions may have an effect on QoL compared to minimal intervention, but the effect is very uncertain
Adverse effects Follow‐up after completion: range 6 to 9 months	24 per 100n	14 per 100 (9 to 21)	OR 0.50 (0.30 to 0.84)	286 (2 RCTs)	⊕⊕⊝⊝ Lowh,o,s	Pedometer interventions may reduce the rate of adverse effects compared to minimal intervention, based on moderate‐certainty evidence. Two additional studies measured adverse effects and found 0 events
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). BMI: body mass index; CI: confidence interval; CVD: cardiovascular disease; LDL: low‐density lipoprotein; MD: mean difference; OR: odds ratio; QoL: quality of life; RR: risk ratio; SBP: systolic blood pressure; SF‐36v2: Short Form 36 version 2.
GRADE Working Group grades of evidence. High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aAll follow‐up periods (unless otherwise stated) measured from completion of the intervention period, to identify lasting effects once the intervention ceases. Longest available follow‐up periods preferred. All study intervention periods were of medium duration (range 10 weeks to 6 months).

^bDowngraded by one level for risk of bias, as a large proportion of included studies were at high or unclear risk for three or more domains (blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other concerns).

^cDowngraded by two levels for heterogeneity: I² was 94%, which was pre‐specified in the protocol to be too high for the meta‐analysis to be useful.

^dDowngraded by one level for imprecision. The 95% CI could not be calculated as the data were not pooled, but the range of effects in included studies included a possible decrease or increase in the outcome.

^eMean of control group in single included study.

^fDowngraded by one level for risk of bias as the single included study was at high or unclear risk for three or more domains (blinding of participants and personnel, blinding of outcome assessment, and incomplete outcome data).

^gDowngraded by two levels for imprecision, as the 95% confidence interval included a meaningful decrease or increase in the outcome and was based on a small sample size.

^hDowngraded by one level for indirectness, as the included studies were conducted in people who were either sedentary or overweight, which may substantively affect their outcomes.

ⁱUnweighted average of means of control groups in included studies.

^jDowngraded by one level for imprecision, as the 95% confidence interval included both a negligible increase and a meaningful decrease in the outcome.

^kDowngraded by one level for risk of bias as one of the included studies was at high or unclear risk of bias in three or more domains, and sensitivity analysis indicated that removal of this study would have led to a smaller and less precise effect estimate.

^lDowngraded by one level for imprecision, as the 95% confidence interval included a meaningful increase and decrease in the outcome.

^mMeasured at completion of intervention period. No follow‐up available after completion of the intervention.

ⁿMean of control group in one of the included studies that used endpoint measures, while the other study reported change from baseline.

^oNot downgraded for risk of bias, although one of the included studies was at high or unclear risk of bias in three or more domains, as sensitivity analysis indicated that removal of this study had no important impact on the overall result.

^pDowngraded one level for indirectness, as no post‐intervention follow‐up was available, only end of intervention, and all studies were conducted in people who were either sedentary or overweight, which may substantively affect their outcomes.

^qDowngraded by one level for imprecision, as the 95% confidence interval included both a negligible decrease and a meaningful increase in the outcome and was based on a small sample size.

^rDowngraded by one level for risk of bias, as the only included study was at high or unclear risk or bias for three or more domains (random sequence generation, blinding of participants and personnel, and blinding of outcome assessment).

^sDowngraded by one level for imprecision due to small sample size.

Summary of findings 2. Pedometer intervention compared to alternative physical activity intervention for increasing physical activity.

Pedometer intervention compared to alternative physical activity intervention for increasing physical activity
Patient or population: employees Setting: workplace Intervention: pedometer intervention Comparison: alternative physical activity intervention
Outcomes	Anticipated absolute effects* (95% CI)		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
	Risk with alternative intervention	Risk with pedometer intervention
Physical activity assessed with combined measures Follow‐up after completion: 3 monthsa	Mean steps per day was 10,100b	RoM 707 steps higher (303 steps lower to 1818 steps higher)	Ratio of means 1.07 (0.97 to 1.18)	143 (1 RCT)	⊕⊝⊝⊝ Very lowc,d,e	The effect of pedometer interventions on physical activity compared to an alternative physical activity intervention is very uncertain
Sedentary time assessed with % of accelerometer wear time At completion of interventionf	Mean change in sedentary time was ‐2.05%b	MD 1.48% higher (1.52 lower to 4.49 higher)	‐	62 (1 RCT)	⊕⊝⊝⊝ Very lowd,g,h	The effect of pedometer interventions on sedentary time compared to alternative physical activity interventions is very uncertain
CVD risk factor: body mass index (BMI) At completion of interventionf	Mean CVD risk factor: BMI was 29.14 kg/m²i	MD 1.23 kg/m² lower (2.85 lower to 0.39 higher)	‐	144 (2 RCTs)	⊕⊝⊝⊝ Very lowg,j,k	The effect of pedometer interventions on BMI compared to alternative physical activity interventions is very uncertain
CVD risk factor: systolic blood pressure (SBP) At completion of interventionf	Mean CVD risk factor: SBP was 124.5 mmHgb	MD 4 mmHg lower (10.15 lower to 2.15 higher)	‐	94 (1 RCT)	⊕⊝⊝⊝ Very lowg,j,k	The effect of pedometer interventions on SBP compared to alternative physical activity interventions is very uncertain
CVD risk factor: LDL cholesterol At completion of interventionf	Mean CVD risk factor: LDL cholesterol was 112.6 mg/dLb	MD 0.6 mg/dL higher (15.08 lower to 16.28 higher)	‐	94 (1 RCT)	⊕⊝⊝⊝ Very lowg,j,l	The effect of pedometer interventions on LDL cholesterol compared to alternative physical activity interventions is very uncertain
Quality of life (QoL) ‐ not reported	‐	‐	‐	‐	‐	One study measured QoL but did not report the results.
Adverse effects ‐ not measured	One study reported that no adverse effects were reported		‐	‐	‐	No conclusions could be drawn about the adverse effects of pedometers and other physical activity interventions
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). BMI: body mass index; CI: confidence interval; CVD: cardiovascular disease; LDL: low‐density lipoprotein; MD: mean difference; OR: odds ratio; QoL: quality of life; RCT: randomised controlled trial; RoM: ratio of means; RR: risk ratio; SBP: systolic blood pressure.
GRADE Working Group grades of evidence. High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

^aFollow‐up period measured from completion of the intervention period to identify lasting effects once the intervention ceases. Longest available follow‐up periods preferred. All study interventions periods were of medium duration (range 3 to 6 months).

^bMean value in the control group of the single included study.

^cDowngraded by one level for risk of bias, as the single included study was at high or unclear risk of bias in three or more domains (blinding of participants and personnel, blinding of outcome assessment, and selective outcome reporting).

^dDowngraded by one level for imprecision, as the 95% CI included both a negligible decrease and a meaningful increase in the outcome and was based on a small sample size.

^eDowngraded by one level for indirectness, as the single included study was conducted only in sedentary women. Gender and activity levels were pre‐specified in the protocol for this review as possible effect modifiers.

^fMeasured at completion of the intervention period. No follow‐up available after completion.

^gDowngraded one level for indirectness, as no post‐intervention follow‐up period was measured and lasting impact is therefore unknown.

^hDowngraded by one level for risk of bias, as the single included study was at high or unclear risk of bias in three or more domains (blinding of participants and personnel, incomplete outcome data, and selective outcome reporting).

ⁱUnweighted average of means of control groups in included studies.

^jDowngraded by one level for risk of bias, as the included study or studies were at high or unclear risk of bias in three or more domains (random sequence generation, allocation concealment, blinding of participants and personnel, incomplete outcome data, and selective outcome reporting).

^kDowngraded by one level for imprecision, as the 95% CI included both a negligible increase and a meaningful decrease in the outcome, and was based on a small sample size.

^lDowngraded by two levels for imprecision, as the 95% CI included both a meaningful decrease or increase in the outcome and was based on a small sample size.

Background

Description of the condition

The World Health Organization (WHO) recommends that most adults should undertake at least 150 minutes of moderate‐intensity physical activity per week, or equivalent, as this reduces the risk of cardiovascular disease, diabetes, and some cancers (WHO 2014). Although the health benefits of physical activity are recognised, in 2010, only 23% of the world’s population was undertaking adequate amounts of physical activity (WHO 2014), and rates have been declining (Brownson 2005; Food and Agricultural Organization of the UN 2006; Norman 2003; WHO 2010). This trend is likely to continue, as physical activity is continuously being reduced in all life environments including at home, during school/work, during recreation, and in transport (Brownson 2005; WHO 2004; WHO 2010). Currently, physical inactivity is the fourth leading global risk for mortality and the 11th leading global risk for burden of disease (WHO 2009). Physical inactivity and low physical activity accounted for 3.2 million global deaths and 2.8% of disability‐adjusted life‐years in 2010 (Lim 2014; WHO 2014).

Description of the intervention

Workplace as a setting for health promotion

The workplace has become a key setting for health promotion and disease prevention (Freak‐Poli 2010; Freak‐Poli 2011a; Freak‐Poli 2012; Freak‐Poli 2013a; WHO 2002; WHO 2004; WHO & WEF 2008). The potential to positively influence behaviour in the workplace setting is especially important as occupations have gradually become more sedentary (Ferro‐Luzzi 1996; Puig‐Ribera 2008; WHO 2000; WHO 2009). Health promotion interventions are increasingly conducted at workplaces to access groups of participants in their daily lives, and to provide opportunities for employers to improve worker health, reduce absenteeism, and increase productivity (Marshall 2004; WHO & WEF 2008). Workplace health intervention evaluations have demonstrated improvement in the leading global risk factors for chronic disease (Rongen 2013; WHO 2004); they have also been found to benefit the employer (Batt 2009; Speck 2009; WHO & WEF 2008).

Pedometer use in health promotion

A pedometer, or step counter, is a small, light, portable, easy‐to‐use electronic device that counts the number of steps taken by an individual. Pedometers are usually around the size of a matchbox, and they can be worn clipped to the person's clothing at the hip, or at another convenient place. They are low in cost, usually priced between USD 20 and USD 35, making them an accessible and feasible intervention.

By wearing a pedometer for a period of time, during either ordinary daily activities or a specific period of walking, the individual receives feedback on the number of steps taken and thereby a measure of his or her physical activity. Pedometers have been used as a measurement tool by athletes and in fitness training programmes, and have served as health promotion interventions aimed at increasing physical activity levels. Health promotion interventions usually encourage participants to wear a pedometer during waking hours to record and give feedback on the number of steps taken on a daily or weekly basis (Bravata 2007; Freak‐Poli 2011a; Kang 2009; Ogilvie 2007; Richardson 2008). These interventions encourage individuals to increase their level of walking (a moderate‐intensity activity) or running (a vigorous‐intensity activity) and often provide a target step goal, such as the commonly used 10,000 steps per day (Behrens 2007; Dishman 2009; Low 2007; Maruyama 2010; Rush 2009; Warren 2010).

Pedometers are rarely used alone; health interventions may include a variety of additional components such as a diary or a website for logging steps, dissemination of additional health promotion information, motivational reminders, shared reporting of step counts, counselling sessions, group facilitators, weekly meetings, a website for communication among participants, team competition, participation rewards, or group physical activity sessions (Aittasalo 2004; Behrens 2007; Chan 2004; Croteau 2004; De Cocker 2010; Dishman 2009; Faghri 2008; Farag 2010; Freak‐Poli 2011a; GCC 2010; Gemson 2008; Gilson 2007; Goetzel 2009; Goetzel 2010; Haines 2007; Kwak 2010; Low 2007; Lubans 2009; Maruyama 2010; Naito 2008; Puig‐Ribera 2008; Racette 2009; Rush 2009; Speck 2009; Thomas 2006; Warren 2010). Pedometer use can also be incorporated as a component of broader health promotion interventions incorporating elements such as mass media, community‐based activities, physical health checks, and healthy eating initiatives. A systematic review of workplace physical activity interventions concluded that a pedometer was among the components that assisted with intervention effectiveness (To 2013).

It is important to note that the use of pedometers in health promotion interventions is now being superseded by widespread access to more sophisticated devices such as wearable accelerometers and Smartphone apps.

How the intervention might work

Pedometers provide immediate, specific feedback on levels of physical activity that is intended to motivate individuals to increase their activity over time (Matevey 2006). Health interventions that utilise pedometers are generally based on social‐cognitive theory, identifying self‐efficacy as the main driver of positive physical activity and health behaviour change (Bandura 2001; Culos‐Reed 2001; De Cocker 2010; Lemon 2010; Lubans 2009; Maruyama 2010; Prabhakaran 2009; Prodaniuk 2004; Tudor‐Locke 2009). Pedometer‐based interventions promote self‐efficacy by focusing on walking or running activities, which usually have few barriers to participation. A pedometer can facilitate progressive individual goal‐setting and can allow the individual to be flexible in amount and scheduling of physical activity. In this way, the pedometer acts as both a motivator and a monitor of activity. The use of additional components such as targets, education, and rewards is intended to increase that motivation.

By setting a health promotion intervention incorporating pedometers in a workplace context, social‐cognitive motivation is combined with an ecological approach, addressing the environment in which people interact (Prodaniuk 2004). The workplace is a pre‐existing social setting in which collegiate camaraderie and endorsement of leaders can reinforce participation in interventions. Available facilities can be used to undertake physical activity, and existing communication networks, such as email or a common notice board, can be used to encourage and inform participants (Freak‐Poli 2011a).

This review aims to measure effects of the unique monitoring and motivational role of pedometers used to increase physical activity in workplace settings, including relatively simple interventions in which pedometers are the main intervention (although they may be supported by the components listed above) and broader interventions incorporating pedometers as a component. Although it is more difficult to assess the impact of pedometers in the context of a complex, multi‐component intervention, it is important to consider the evidence for these interventions, as this is often how pedometers are used in health promotion practice.

There is a further particular question to be examined, which is whether any measured impact of interventions incorporating pedometers in the short term can be translated into long‐term increases in physical activity that could feasibly lead to a reduction in risk factors for and incidence of a range of chronic diseases such as cardiovascular disease and diabetes.

Why it is important to do this review

The World Health Organization and the World Economic Forum recommend that further research is needed to strengthen current knowledge of workplace health interventions, particularly regarding effectiveness and use of simple instruments (WHO & WEF 2008).

A number of Cochrane Reviews have assessed the evidence surrounding the effectiveness of different interventions to increase physical activity within certain environments, including community‐wide interventions and school‐based interventions (Baker 2011; Baker 2015; Dobbins 2013).

Only one other review, published elsewhere, has examined pedometers in a workplace context (Bravata 2007), reporting inconclusive results.

To understand whether workplace health interventions incorporating pedometers offer an avenue for improving physical activity and consequent health risk factors, a systematic review incorporating the current literature is required. This review is an update of a previous version of the review (Freak‐Poli 2013). In the previous version, we identified four relevant studies providing data for 1809 employees. Although we observed that independent studies reported health benefits associated with the pedometer programmes, we determined that there was "a need for more high‐quality randomised controlled trials to assess the effectiveness of pedometer interventions in the workplace for increasing physical activity and improving subsequent health outcomes". We concluded that "there [were] limited and low‐quality data providing insufficient evidence to assess the effectiveness of pedometer interventions in the workplace for increasing physical activity and improving subsequent health outcomes". A number of new studies on this topic have now been published, warranting an updated review.

Although it is useful to bring this review up‐to‐date, we note that growing use of accelerometers and Smartphone apps to provide feedback on physical activity is superseding the use of pedometers. For this reason, this will be the final update of this review.

The original protocol for this review was published (Freak‐Poli 2011). Changes made between the protocol or the last published version and this version are detailed in the Differences between protocol and review section.

Objectives

To assess the effectiveness of pedometer interventions in the workplace for increasing physical activity and improving long‐term health outcomes.

Methods

Criteria for considering studies for this review

Types of studies

We included individual and cluster‐randomised controlled trials (RCTs).

Types of participants

We included studies conducted with employed adults. Adults were defined as 16 years of age or older. Mixed‐age populations were eligible if a separate analysis of adult participants was available; however, none were available for inclusion. Studies conducted with trained athletes would have been excluded; however, none were available.

Types of interventions

We included any workplace health intervention that incorporated the use of a pedometer. To be eligible, the pedometer had to be incorporated into the health intervention for the entire duration of the intervention, and participants had to be able to view their step count. We included studies in which the pedometer was the sole component of the intervention; studies in which the pedometer was the main focus of the intervention but was supported by other intervention components like step goals, diaries, teams, or rewards to increase motivation; and broader health promotion interventions that incorporated pedometers as one of many components. We aimed to explore the modifying effects of additional intervention components through subgroup analysis.

We did not include health interventions incorporating accelerometers rather than pedometers. Although both accelerometers and pedometers are unobtrusive, accurate motion sensors, there are four main distinctions. First, the mechanics of an accelerometer function differently from a pedometer, allowing detection of three‐dimensional movement in addition to simple step counting (Tudor‐Locke 2002). Second, an accelerometer allows more complex analysis, with the capacity to segregate recorded movements into subsets of time and analyse their frequency or intensity (Tudor‐Locke 2002). Additional information such as speed, distance, caloric expenditure, and total physical activity time may be available, dependent on the brand, and could be an extra motivator for the user. Third, an accelerometer unit is at least four times more expensive than a pedometer unit; the usual price ranges between USD 120 and USD 299, but it can cost up to USD 450 per unit (Tudor‐Locke 2002; Tudor‐Locke 2004b). Fourth, to access the information that an accelerometer provides, it often must be plugged in to a computer with a specific software or synchronised with a mobile device. The cost of the accelerometer, the use of a computer or device, the costs of specific software, and the time required to set up and monitor these devices increase barriers to accelerometer use in health promotion, particularly for populations less comfortable with the use of technology (Tudor‐Locke 2002; Tudor‐Locke 2004b). Due to differing mechanical function, additional information types, and increased costs, we did not view accelerometers as a low‐cost, easy‐to‐use device; therefore, we have not assessed them in this review. However, over time, with technological advances and the introduction of companies such as Fitbit® (www.fitbit.com) and Bellabeat (webshop.bellabeat.com), accelerometers are becoming less expensive and may render pedometers as obsolete. Note that we did include studies that used accelerometers solely to measure physical activity as an outcome, but only if the accelerometers were available during limited data collection rounds for both the pedometer intervention group and the control group, and were not available throughout the study.

We included all comparator groups in the review, including any intervention without a pedometer, or no intervention.

Types of outcome measures

We aimed to report the following outcomes, but only the primary outcome was required as part of the eligibility criteria of studies for inclusion in the review.

Primary outcomes

The primary outcome was physical activity.

• If multiple physical activity measurements were reported, preference was first given to activity assessed over the whole day. For example (in order of preference), metabolic equivalents (METs), step count, METs for moderate and vigorous activity combined, incidental activity (incorporated into work or leisure time (Prodaniuk 2004)), duration of physical activity.

• Second, if one of these was not available, preference was given to one specific type of activity assessed over the whole day. For example (in order of preference), moderate and vigorous activity, moderate activity, vigorous activity, light activity.

• Third, preference was given to activity assessed within work time rather than within leisure time (Godin 1985).

• When more than one measure (e.g. objective, self‐report) was available within one of these three categories, preference was given to measures with less risk of bias. For example (in order of preference), physical activity that is objectively measured, followed by observed and then self‐reported activity (e.g. the Stanford Usual Activity Questionnaire (Sallis 1985), the Dutch short questionnaire to assess health‐enhancing physical activity (Wendel‐Vos 2003), the International Physical Activity Questionnaire (IPAQ 2011)).

We categorised time points into two categories: those occurring at completion of the intervention (labelled intervention "duration"), and those occurring at follow‐up after completion of the intervention (labelled follow‐up "term"). The primary measurement time points of interest were those at follow‐up after completion of the intervention period, as these may identify whether any sustainable change in behaviour has occurred. Within these categories, duration of the intervention and follow‐up time were grouped as short term (less than one month), medium term (at least a month but less than one year), and long term (at least one year). Thereby the duration of an intervention and the follow‐up term after completion of the intervention were categorised independently of each other.

We excluded outcomes that were measured only in subsets of the study population.

Secondary outcomes

Health risk factors of interest included the following (in order of preference).

• Sedentary behaviour (e.g. time sitting, time spent under 1.5 metabolic equivalent of task units (METs) ‐ a measure of energy consumption, time watching television or other media).

• Cardiovascular disease and type 2 diabetes risk factors.

• • Anthropometric measures (e.g. body mass index (BMI), waist circumference (WC), waist‐to‐hip ratio, body fat, body weight, hip circumference).

  • Blood pressure (e.g. systolic blood pressure (SBP), diastolic blood pressure (DBP), hypertension, resting heart rate (for comparison, as heart rate should not change due to the health intervention)).

  • Biochemical measures (e.g. blood glucose (fasting, not fasting), blood cholesterol (high‐density lipoprotein (HDL), high‐ to low‐density lipoprotein ratio, low‐density lipoprotein (LDL), blood triglycerides).

  • Disease risk scores (e.g. cardiovascular disease risk (D'Agostino 2008), type 2 diabetes risk (Baker IDI Heart and Diabetes Institute 2008)).

• Quality of life (e.g. Short Form 36 or 12 Health Survey (SF‐36 2011), the Social Support Inventory (Dunkel‐Schetter 1986), the Satisfaction With Life Scale (Diener 1985)).

• Adverse effects including injury.

Search methods for identification of studies

Electronic searches

We searched the following sources from the earliest record. The latest date on which the search was fully up‐to‐date across all sources was 1 December 2016 (see Appendices for complete dates, including all updates and specific search strategies).

• CENTRAL (the Cochrane Central Register of Controlled Trials, in the Cochrane Library).

• CINAHL (Cumulative Index to Nursing and Allied Health Literature).

• MEDLINE.

• Embase.

• OSH (Occupational Safety and Health) UPDATE (CISDOC, HSELINE, NIOSHTIC, NIOSHTIC‐2, RILOSH, IRSST, and INTERNATIONAL BIBLIOGRAPHIC databases) (this source was excluded from the search after 1 December 2014 due to lack of relevant results).

• Web of Science (SCI‐Expanded, SSCI, and A&HCI databases).

• ClinicalTrials.gov.

• WHO International Clinical Trials Registry Platform (ICTRP).

Before publication of this review, to assess the currency of the search, we conducted a provisional update of the search across all electronic databases in May 2019, although the results of this provisional search are not fully incorporated into the review and are reported separately in the Results of the search section.

We developed our systematic search strategy with help from information specialists at Cochrane Occupational Safety and Health (see Acknowledgements). We tested the strategy against a set of 13 known relevant studies from across the globe before running final searches. We did not limit the search by language.

Searching other resources

We searched the websites of organisations actively involved in workplace physical activity interventions, for example, the World Health Organization, including Global Strategy on Diet, Physical Activity and Health (WHO 2004), and Preventing Noncommunicable Diseases in the Workplace Through Diet and Physical Activity (WHO & WEF 2008). These sources did not reveal any included studies in the initial search, and we did not search them again when we updated the review.

We scanned the references of included studies for additional studies. For the first version of this review, we sent a comprehensive list of relevant articles together with the inclusion criteria for the review to the first author of each paper that met the inclusion criteria, asking if they knew of any additional published or unpublished studies that might be relevant. No additional studies were found, and this process was not repeated when the review was updated.

Data collection and analysis

Selection of studies

Two review authors or other contributors undertook an initial screening of titles and abstracts independently, to remove those that were obviously outside the scope of the review. We sought full‐text translations or evaluations of all relevant non‐English articles. We rejected articles at the initial screening stage if both review authors agreed based on title and abstract that it met one of these specified criteria.

• The article was not a report of an RCT.

• The intervention was not tested on employed adults.

• The trial did not address a pedometer‐based physical activity intervention.

We were over‐inclusive at this stage, and, if in doubt, we did not exclude the record. For the 2016 update of the search, one review author screened titles and abstracts as above. For the duplicate screen, search results were initially screened by Cochrane Crowd (http://crowd.cochrane.org), Cochrane's crowd‐sourcing platform, to identify records referring to randomised trials. This reduced set was then forwarded for screening against the remaining criteria by a second review author as normal.

We obtained the full text of all papers potentially meeting the inclusion criteria. We linked together multiple publications and reports on the same study. Two review authors (of RFP, MC, and LAB) screened all full‐text papers independently. We rejected articles at this stage if both review authors agreed, based on the full text, that the study did not meet all eligibility criteria. When two review authors were uncertain about the classification of a study, a third review author was consulted and/or the study authors were contacted via email to request further information.

For the provisional updated search conducted in May 2019, results were initially screened to identify randomised trials using Cochrane Crowd, and the remaining studies were provisionally screened by only one review author (MC or LAB).

Data extraction and management

Two review authors (of RFP, MC, and LAB) independently completed a data extraction form for each included study. We tailored the data extraction form to the requirements of this review, and we piloted it to assess its ability to capture study data. We incorporated items for assessing risk of bias into the data extraction form. In addition, we assembled and compared multiple reports and publications of the same study to ensure completeness and to identify possible contradictions.

We collected data on the study population, the study environment, intervention specifics, study methods, and outcomes for each study. We recorded all measures identified as primary or secondary outcomes, regardless of how the information was reported (e.g. categorical cut‐offs, continuous mean ± standard deviation data). When studies reported more than one time point, we collected all outcomes at all time points.

Assessment of risk of bias in included studies

Two review authors (of RFP, MC, and LAB) independently assessed the risk of bias of each included study using a descriptive approach, as outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b). We assessed the following key criteria: random sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessment; incomplete outcome data; selective outcome reporting; and other sources of bias (e.g. baseline imbalance, risks associated with cluster‐randomised designs such as differences in recruitment and comparability of clusters) (Higgins 2011b). We assessed each study as having low, high, or unclear risk of bias. A judgement of unclear risk of bias indicated either lack of information or uncertainty over the potential for bias. There was no persisting difference of opinion, so a third review author was not required to review additional papers.

We considered trials that were not assessed as low risk in three or more of the above criteria to be at high risk of bias overall.

Measures of treatment effect

When possible, we aimed to express effect sizes for dichotomous outcomes as risk ratios (RRs). For continuous outcomes, we aimed to use mean differences (MDs) between post‐intervention values of intervention and control groups to express effect sizes when possible.

When studies used different scales to measure the same outcome, we aimed to use standardised mean differences (SMDs). For the primary outcome of physical activity, the included studies were very heterogeneous across a number of domains and used diverse definitions and measures of physical activity. These factors would almost certainly influence the standard deviations (SDs) in each study, undermining the required assumption for the calculation of an SMD that the SDs reflect only variation in measurement scales, and not other important differences (Deeks 2011). We therefore made the decision to express this outcome as a ratio of means (RoM), using the methods outlined in Friedrich 2008.

For the outcome of adverse effects, results were reported in one study as dichotomous data (Aittasalo 2012), and in another as a continuous measure of the number of events over two 12‐month periods (Morgan 2011). We converted both to log(odds ratio) to enable meta‐analysis of these two studies using the methods described in Deeks 2011.

We reported all effect measures with a 95% confidence interval (CI).

Unit of analysis issues

If a study had more than two arms, we considered the interventions in each arm and did not analyse any arms not relevant to the review. When two intervention arms were considered comparable for the purposes of this review, we combined the two and treated them as a single arm for the purposes of our analysis (either combining the data ourselves or obtaining them from the study authors). When it was not appropriate to combine groups, we would have conducted separate comparisons of each arm of interest (e.g. one intervention arm versus control, then the second intervention arm versus control), taking care not to include the same participants twice within a meta‐analysis and preventing unit of analysis error, but this situation did not arise.

The review included a number of cluster‐randomised controlled trials, in which participants were allocated to intervention or control arms in groups (e.g. workplaces). Unit of analysis errors may occur when studies allocate participants in clusters but analyse the results by the total number of individuals. This can result in overestimation of the statistical significance of the results by not accounting for clustering of individuals in the data. Correcting the error by analysing results by the unit of randomisation (the cluster) can underestimate the statistical significance of the results, particularly when clusters are very large. In our meta‐analysis, we assessed the included cluster‐randomised trials for unit of analysis errors. When studies correctly accounted for clustering, we extracted and used these data in our analyses (usually presented as overall effect estimates, e.g. odds or risk ratios based on a multi‐level model and analysed through the generic inverse variance method).

When clustering was not accounted for by an included study (or when the separate intervention and control group data required for calculation of the RoM were unadjusted), we adjusted the data according to the methods outlined in Higgins 2011c. We adjusted for clustering by calculating robust standard errors using the svy linearized commands for control and intervention groups in Stata version 15 (StataCorp 2015).

For Linde 2012 (for which individual participant data were available), to indicate whether a variable correlated more strongly within workplaces than between workplaces, we calculated two statistics: the design effect and the intracluster correlation coefficient. The design effect (deff) assessed the loss of statistical efficiency arising from the study’s cluster‐sampling method compared with the hypothetical use of a simple random sample (Campbell 2004; Carlin 1999). Intracluster correlation coefficients (ICCs) were calculated as deff minus one, divided by the sum of minus one and division of the average cluster size by the number of clusters (Campbell 2004; Carlin 1999). For physical activity (total METs per week), these were calculated as deff = 16.02 and ICC = 0.07, and for sedentary behaviour, deff = 44.87, ICC = 0.20, indicating a low clustering effect. For BMI, these were calculated as deff = 2.20, ICC = 0, indicating no correlation within clusters.

For Audrey 2015 and Morgan 2011 (physical activity only as other outcomes were already adjusted) and Parry 2013, data were adjusted by calculating an effective sample size according to the methods outlined in Higgins 2011c, using an ICC of 0.12 (95% CI 0 to 0.3) calculated by Audrey 2015, for physical activity. This ICC was used in preference to those calculated for Linde 2012, as it assumed a higher level of correlation and therefore produced more conservative estimates with wider standard errors.

Dealing with missing data

When information was missing from the included studies, we contacted the study authors to request additional information. Linde 2012 provided individual patient data for physical activity, sedentary behaviour, and BMI via personal communication. We assessed the risk of bias arising from incomplete outcome data as part of the overall risk of bias assessment.

When a standard deviation (SD) was not available for a continuous outcome, we used the methods demonstrated by Higgins 2011a to obtain one. In one case (Ribeiro 2014), SDs were not available for the outcome of physical activity and could not be calculated, and so we imputed appropriate SDs from the most closely comparable study in group mean values and outcome definitions for each comparison (Maruyama 2010 for Comparison 1, Talbot 2011 for Comparison 2).

Assessment of heterogeneity

We considered the clinical heterogeneity of included studies before conducting any analyses. We considered the following PICO elements too different to be combined.

For interventions:

• studies of pedometers alone;

• pedometer‐focused interventions with supporting components to increase motivation (e.g. step goals, diaries, teams, rewards); and

• broader health promotion interventions that incorporated pedometers as one of many components.

For comparison groups:

• studies comparing pedometer interventions to no intervention;

• studies comparing pedometer interventions to similar components without a pedometer; and

• studies comparing pedometer interventions to larger‐scale health promotion interventions and other active interventions without a pedometer.

For follow‐up periods:

• measures taken at completion of the intervention period;

• follow‐up measures taken after completion of the intervention period; and

• in either category, short‐, medium‐, and long‐term periods.

We quantified and evaluated the amount of statistical heterogeneity to determine whether the observed variation in study results was compatible with the variation expected by chance alone (Higgins 2003). We assessed heterogeneity through examination of forest plots and quantified it using the I² statistic. When we observed an I² statistic greater than 90%, we considered heterogeneity to be too high to conduct a meta‐analysis.

Assessment of reporting biases

If ten or more studies were included in a meta‐analysis, we aimed to assess the possibility of publication bias by using funnel plots (Higgins 2011a). We also aimed to investigate alternative explanations for funnel plot asymmetry (such as clinical or methodological heterogeneity, statistical artefacts, or chance) (Egger 1998). We aimed to assess the potential impact of any suspected small‐study effects by using a comparison between fixed‐effect and random‐effects meta‐analysis models. However, no meta‐analysis included more than nine studies, and these methods were not required.

We also aimed to assess the risk of bias arising from selective outcome reporting within studies as part of the overall risk of bias assessment.

Data synthesis

For data synthesis, we followed the meta‐analysis methods outlined in Deeks 2011. When we considered studies to be clinically and methodologically homogeneous, and when comparable data were available from at least two studies measuring the same outcome, we performed meta‐analyses using Review Manager 5 software (RevMan 2011). We used a random‐effects model as the default to incorporate the assumption of heterogeneity between studies.

Subgroup analysis and investigation of heterogeneity

If more than two trials that reported data in each category were available, we aimed to explore the following participant characteristics using subgroup analyses.

• Gender: men compared to women.

• Age (as the probability of maintaining good health diminishes as an individual gets older (AIHW 2008), there may be differing motivations for participation in pedometer‐based workplace health interventions depending on age): younger (< 40 years) compared to older (≥ 40 years).

• Educational status: tertiary education completed compared to not completed.

If more than two trials that reported data in each category were available, we aimed to explore the following intervention characteristics.

• Eligibility of participants.

  • Are interventions targeting high‐risk employees more effective than interventions targeting all employees?

  • Are interventions targeting sedentary or office‐based employees more successful than interventions targeting active or manual employees?

• Step goal: are interventions that utilise a daily step goal (e.g. 10,000 steps per day) more effective than non‐step goal‐defined interventions?

• Step diary: are interventions that utilise a step diary (e.g. daily or weekly record of steps) more effective in changing physical activity than non‐diary interventions?

• Duration of intervention period: are short‐duration interventions (less than one month), medium‐duration interventions (more than a month but less than one year), or long‐duration interventions (equal to or more than one year) more effective?

• Provider: are interventions with an external intervention provider more effective than interventions undertaken internally within the workplace?

Sensitivity analysis

We aimed to carry out a sensitivity analysis including only studies at overall low risk of bias, defined as being assessed at low risk of bias in at least four of the seven domains assessed.

We aimed to use sensitivity analysis to assess the impact of any suspected publication bias by comparing fixed‐effect and random‐effects meta‐analysis models.

'Summary of findings' and assessment of the certainty of evidence

The following outcomes were selected as most important for summarising findings of the review: physical activity, sedentary behaviour, BMI, SBP, LDL cholesterol, quality of life (mental health component), and adverse effects. For each of these outcomes, the longest available period of follow‐up after completion of the intervention was used as the main measure. When no follow‐up was conducted after the intervention period, the measure at completion of the longest available duration of interventions was used.

For these measures, two review authors (MC, LAB) assessed the certainty of evidence using the GRADE approach (Schünemann 2011). We summarised the level of certainty as ‘high’, ‘moderate’, ‘low’, or ‘very low’, based on risk of bias, directness of evidence, heterogeneity and its causes, precision, and publication bias. We incorporated these assessments into our descriptions of these outcomes by using standardised language recommended by Cochrane Consumers and Communication (Ryan 2016). Accordingly, throughout the text, we indicated whether the observed effect represents a meaningful change, a slight change, or little to no change. When the certainty of evidence is moderate, we use the word 'probably'; when it is low, we use the word 'may'; and when it is very low, we use the word 'may' but indicate that the result is too uncertain to estimate. Details of the GRADE assessments are included in Table 1 and Table 2.

Summary of findings and assessment of the certainty of the evidence

The following outcomes were selected as the most important to summarise the findings of the review: physical activity, sedentary behaviour, BMI, SBP, LDL cholesterol, quality of life (mental health component), and adverse effects. For each of these outcomes. the longest available period of follow‐up after completion of the intervention was used as the main measure. Where no follow‐up after the intervention period was conducted, the measure at completion of the longest available duration of interventions was used.

For these measures, two authors (MC, LAB) assessed the certainty of evidence using the GRADE approach (Schünemann 2011). We summarised the level of certainty as ‘High’, ‘Moderate’, ‘Low’, or ‘Very Low’, based on risk of bias, directness of the evidence, heterogeneity and its causes, precision and publication bias. We incorporated these assessments into our descriptions of these outcomes using standardised language recommended by Cochrane Consumers and Communication (Ryan 2016). Accordingly, throughout the text, we indicate whether the observed effect represents a meaningful change, a slight change or little or no change. Where the certainty of the evidence is moderate, we use the word 'probably'; where it is low, we use the word 'may', and where it is very low, we used the word 'may' but indicate that the result is too uncertain to estimate. Details of the GRADE assessments are included in Table 1 and Table 2.

Results

Description of studies

Results of the search

In the first published version of this review (Freak‐Poli 2013), we identified and screened 5947 records, and we included four studies. Since that time, the search has been updated three times: in March 2013, December 2014, and December 2016 (see Figure 1). In total, based on both original and updated searches, we identified 8785 records (6197 unique records after removal of duplicates). Of these, we considered 321 potentially eligible based on the title and abstract, and we assessed them in full text. Of these, we excluded 302 (see Excluded studies section and Characteristics of excluded studies table). An additional two met our inclusion criteria, but these studies were ongoing at the time of the search (Audrey 2019; Pillay 2012; see Characteristics of ongoing studies table). We included 14 studies (see Included studies section and Characteristics of included studies table).

1.

PRISMA flow chart.
* Note that additional records associated with included studies were identified during the course of the review, including through author correspondence and targeted searching for study protocols.

We contacted the authors of 24 studies to clarify eligibility, identify additional related publications, and obtain additional results. We received responses regarding 19 studies, providing additional information (Aittasalo 2012; Audrey 2015; Bort Roig 2012; Dishman 2009; Finkelstein 2015; Kazi 2012; Mansi 2013; Martin 2013; Maruyama 2010; Mattila 2013; Morgan 2011; Parry 2013; Pillay 2014; Puhkala 2011; Ribeiro 2014; Swartz 2014; Talbot 2011; Thøgersen‐Ntoumani 2010; Viester 2012). One study was eligible based on the study author's response, but published data were available for only one of the three randomised groups, and no response was received from the larger study's custodians; this study is awaiting classification (Mattila 2013; Studies awaiting classification).

Note that no additional included studies were identified between December 2014 and December 2016, when the most recent search was completed.

Before publication of this review, we conducted a provisional updated search to identify the extent to which additional studies that could be eligible for the review might be available. The provisional search was conducted in May 2019, and we identified 4617 records (2824 unique records after removal of duplicates). Provisional screening of 1162 records based on titles and abstracts was performed by one review author (MC or LA) via Cochrane Crowd to remove non‐randomised trials. Seventy‐nine records were screened in full text by one review author (MC), of which only one appeared to be potentially eligible. This was the final publication of Audrey 2019, which was previously identified as an ongoing study in the main search. This study and one other that could not be excluded pending translation are described in the Characteristics of studies awaiting classification table. We determined that the findings of this provisional search were unlikely to substantively impact the conclusions of the review, and the results of this search have not been fully incorporated into the review. We identified four additional studies in progress that may be eligible for inclusion when completed (see Characteristics of ongoing studies).

Included studies

We included 14 studies in this review (Aittasalo 2012; Audrey 2015; Carr 2013; Dishman 2009; Linde 2012; Mansi 2013; Maruyama 2010; Morgan 2011; Parry 2013; Pillay 2014; Ribeiro 2014; Swartz 2014; Talbot 2011; Viester 2012). In total, the included studies had recruited 4762 employees, with the greatest contribution of 1747 employees coming from Linde 2012, followed by 1442 employees from Dishman 2009; each of the remaining studies contributed fewer than 320 participants. Key features of these studies are summarised below, and more detailed descriptions are given in the Characteristics of included studies table.

Intervention

Duration

One study had a short duration of one week (Swartz 2014); 12 studies were of medium duration (between ten weeks and six months), and one had a long duration of two years (Linde 2012).

Theoretical basis

Several studies mentioned that their interventions had a theoretical basis. These included theory‐based behaviour modification principles built around goal‐setting theory (Dishman 2009), social‐cognitive theory (Carr 2013; Morgan 2011), self‐regulation theory (Mansi 2013), and stages of change theory (Maruyama 2010).

Intervention components

None of the studies used pedometers alone. All included studies used broad health promotion interventions that incorporated pedometers as one of many components. Interventions were heterogeneous in terms of the other components that they incorporated. No studies used pedometers alone, nor were co‐interventions limited to pedometer‐focused supporting components to increase motivation (e.g. step goals, diaries, teams, rewards), although all included some components of this type.

Three used organisational action to engage local staff in development and implementation of the programme, such as employees nominated or volunteered to be champions (Audrey 2015), management and employees involved in project objectives, implementation and encouragement via joint employee–management steering committees (Dishman 2009), or employees on advisory panels to provide feedback on planning, implementation, and acceptability of all intervention activities (Linde 2012). Five used professional individualised contact with a dietician and a physical trainer (Maruyama 2010), a researcher (Aittasalo 2012; Morgan 2011), or a counsellor (Maruyama 2010; Morgan 2011; Ribeiro 2014; Talbot 2011). Three used personalised websites (Carr 2013; Maruyama 2010; Morgan 2011). Two used group‐based incentives for reaching designated goals such as lunch bags, intervention t‐shirts, recognition plaques, free catered lunch, and local sporting equipment store gift vouchers (Dishman 2009; Morgan 2011). Three used environmental prompts such as signage (Dishman 2009); motivational signs, decorative posters, and music (Linde 2012); and motivational postcards (Talbot 2011).

All included studies used personal goal‐setting. Two specifically had the goal of 10,000 or more pedometer steps each day (Maruyama 2010; Talbot 2011). Four used participants' personal baseline average number of daily steps as the basis for further step goals: to increase by a minimum of 2000 steps per day over the three months (Ribeiro 2014), to gradually increase the level of activities by 5% to reach at least 10,000 steps per day at the end of the three‐month period (Mansi 2013), or to gradually increase the level of activities by 10% to reach the recommended 150 minutes of moderate‐intensity physical activity per week (Pillay 2014); one did not specify the individual goals (Aittasalo 2012). Two used the 10,000 steps goal if individually chosen with or without an alternative physical activity goal (Dishman 2009), or amongst other personalised strategies to address body weight loss, to reduce energy intake, and to increase energy expenditure (Morgan 2011); three had more general goals of (1) incorporating walking into the journey to and from work (Audrey 2015), and (2) providing advice on how to set goals and suggestions for daily active time (Carr 2013), or personal goals for walking at work (Linde 2012), or walking 100 steps for every 60 minutes of sedentary time (Swartz 2014). Seven used step diaries (Audrey 2015; Dishman 2009; Mansi 2013; Maruyama 2010; Morgan 2011; Ribeiro 2014; Talbot 2011).

Eleven studies reported the pedometer brand: seven studies used a Yamax pedometer, model SW 200 (Dishman 2009; Mansi 2013; Morgan 2011; Swartz 2014; Talbot 2011), SW 700 (Parry 2013), or PW 610 (Ribeiro 2014); and four used an Omron pedometer, model HJ‐7101T (Maruyama 2010), HJ‐150 (Carr 2013), HJ‐750 ITC (Pillay 2014), or Walking Style II (Aittasalo 2012).

Control

Eleven studies compared the pedometer intervention to a 'no intervention' or 'minimal intervention' control condition (Aittasalo 2012; Audrey 2015; Carr 2013; Dishman 2009; Linde 2012; Mansi 2013; Maruyama 2010; Morgan 2011; Pillay 2014; Ribeiro 2014; Viester 2012).

Four studies compared the pedometer intervention to a 'minimal intervention' control condition, which included completing the Centers for Disease Control and Prevention (CDC) health‐risk appraisal and receiving monthly newsletters describing the health benefits of physical activity (Dishman 2009), receiving bi‐weekly general motivational messages (Pillay 2014), receiving three 15‐minute individual sessions and a booklet for general advice on physical activity benefits (Ribeiro 2014), or receiving 'usual care', which included a non‐compulsory periodic health screening (Viester 2012).

Although seven studies compared the pedometer intervention to a 'no intervention' control condition, only two did not intervene at the end of the pedometer intervention (Audrey 2015; Mansi 2013). One study used a cross‐over design, and follow‐up data were not eligible for this review (Maruyama 2010). Two studies stated that they were wait‐list controls and offered intervention material at the end of the three‐month intervention (Dishman 2009), or following the 3.5‐month intervention (Morgan 2011). Another two studies offered the control group a one‐hour seminar and intervention material at the end of the three‐month intervention (Aittasalo 2012), or they provided a DVD containing intervention materials at the end of the two‐year intervention (Linde 2012; two of the three sites obtained).

Four studies compared the pedometer intervention to an alternative physical activity intervention (Parry 2013; Ribeiro 2014; Swartz 2014; Talbot 2011, noting that Ribeiro 2014 included multiple intervention groups and contributed to both comparisons). The alternative physical activity interventions included an active workstation or office ergonomics (Parry 2013); aerobic exercise (Ribeiro 2014); computer prompts to undertake steps (and as a wait‐list control, they also received a one‐hour seminar and intervention material at the end of three months; Swartz 2014); and the US National Guard's usual fitness improvement programme (Talbot 2011).

Eligibility and recruitment

Pedometer interventions versus 'no intervention' control

All but three studies recruited participants based on health status (Audrey 2015; Linde 2012; Viester 2012). Two studies recruited participants based on being healthy (Dishman 2009 ‐ without overt cardiovascular, pulmonary, or metabolic disease; Pillay 2014 ‐ without cancer and with physical ability). Two studies recruited participants based on being unhealthy (Aittasalo 2012 ‐ insufficiently physically active; Maruyama 2010 ‐ risk factors for developing metabolic syndrome). The remaining four recruited participants based on a mix of healthy and unhealthy criteria (Carr 2013 ‐ sedentary, overweight, and physically inactive but without major medical problems; Mansi 2013 ‐ physically inactive but physically able; Morgan 2011 ‐ overweight or obese but without a history of major medical problems; Ribeiro 2014 ‐ physically inactive during leisure time but without a number of health conditions).

All studies first recruited the participating work sites, and then the participants within each site, or they undertook recruitment at only one company.

Studies used a variety of methods to recruit participants but could generally be categorised as individualised or passive. Nine studies used individualised techniques: Aittasalo 2012 circulated information to all employees after the occupational healthcare units recruited the workplaces; Audrey 2015 contacted participants by email or letter after recruitment of the workplace through a mailed leaflet to all Bristol Chamber of Commerce employers; Carr 2013 emailed all employees; Linde 2012 sent emails to all employees deemed eligible by the human resources department of businesses (selected from the business directory) that responded to phone calls and a letter; Maruyama 2010 recruited face‐to‐face at regular medical check‐ups; Morgan 2011 emailed all staff; Pillay 2014 emailed all employees; Viester 2012 recruited face‐to‐face at non‐compulsory periodic health screening; and Dishman 2009 used both individualised (employees sent e‐messages) and passive (onsite flyers, interoffice mail, face‐to‐face meetings, and posters) techniques after the work sites agreed to participate. Two studies used passive techniques: Mansi 2013 put up advertisements (posters) at different work sites including the health clinic, plant administration, cafeterias, and all department notice boards; Ribeiro 2014 put up pamphlets and posters.

Pedometer interventions versus alternative intervention without pedometer

All recruited based on health status. One study was based on being healthy (Talbot 2011 ‐ without a history of major medical problems), two on sedentary occupation (considered unhealthy ‐ Parry 2013; Swartz 2014), and one on a mix (Ribeiro 2014 ‐ physically inactive during leisure time but without a number of health conditions).

Three studies were undertaken at only one company (Ribeiro 2014; Swartz 2014; Talbot 2011), and it is unclear how the three large government organisations were recruited for one study (Parry 2013).

One study did not report participant recruitment (Swartz 2014), two studies used passive techniques (Parry 2013 ‐ at regular monthly staff meetings; Ribeiro 2014 ‐ put up pamphlets and posters), and one study used individualised techniques (Talbot 2011 ‐ automatic after failed physical activity test).

Employee demographics

Pedometer interventions versus 'no intervention' control

Workplaces in this group included 20 office‐based work sites (Aittasalo 2012 ‐ specifics not described); 17 workplaces from professional, scientific and technical, manufacturing, transportation, education, accommodation and food services, public administration, and financial and insurance activities (Audrey 2015); two universities (Carr 2013; Pillay 2014); a home improvement store chain (Dishman 2009); two community colleges; a regional insurance office; a beauty industry corporate headquarters with an attached manufacturing and distribution centre; a utility company office and a national headquarters for a health‐related nonprofit organisation (Linde 2012); a meat processing plant (Mansi 2013); a health insurance association (Maruyama 2010); an aluminium factory (Morgan 2011); a university hospital (Ribeiro 2014); and a construction company (Maruyama 2010; Viester 2012).

The proportion of male participants ranged from 0% in Ribeiro 2014 to 100% in Maruyama 2010, Morgan 2011, and Viester 2012. Among studies that recruited males and females, the proportion of males ranged from 10% in Carr 2013 to 74% in Talbot 2011 (mean 34% ± SD 18%; median 34%). For studies that reported gender by intervention/control group, males were more likely to be randomised to the control group (mean 41% ± SD 24%; median 44%) than to the intervention group (mean 32% ± SD 20%; median 29%) (Aittasalo 2012; Audrey 2015; Carr 2013; Mansi 2013; Pillay 2014; Talbot 2011). Studies recruited adults aged between 17.3 years in Audrey 2015 and 75 years in Linde 2012. The mean age of participants ranged from 32.8 years in Talbot 2011 to 46.64 years in Viester 2012 (overall mean 41.31 ± SD 4.14 years; median 42.9 years). For studies that reported age by intervention/control group, the mean age of participants ranged from 32.7 years (± SD 10.1; Talbot 2011 intervention group) to 47.6 years (± SD 9.9; Carr 2013 control group). For intervention groups, the overall mean of age means was 41.68 ± SD 4.51 years (median 43 years). For control groups, the overall mean of age means was 40.59 ± SD 5.23 years (median 40 years).

Pedometer interventions versus alternative intervention without pedometer

Workplaces in this group included three large government organisations (Parry 2013): a university hospital (Ribeiro 2014); a university (Swartz 2014); and the US National Guard (an army reserve organisation).

The proportion of male participants was 0% (Ribeiro 2014), 19.4% (Parry 2013), 32% (Swartz 2014), and 74.15% (Talbot 2011). The mean age of participants ranged from 33 years in Talbot 2011 to 46 years (± SD 10.5) in the Swartz 2014 pedometer intervention group.

Excluded studies

Of the 321 papers assessed in full text, we determined that 302 did not meet the pre‐specified eligibility criteria for this review, and we excluded them (i.e. because they were RCTs, recruited participants who were not employed, were not undertaken in a workplace setting, did not use a pedometer, did not use a pedometer throughout the intervention period, used accelerometers, also provided pedometers to the control group, did not allow participants to view their step count, or did not measure a physical activity outcome). Those studies most likely to have been considered eligible for the review are listed in the Characteristics of excluded studies table, along with specific reasons for their exclusion. We also excluded two studies because only one workplace cluster was allocated to each of the intervention and control arms (Baghianimoghaddam 2016; Racette 2009). In our opinion, this was not adequate to reduce the risk of imbalance of confounders between the two study arms. This was an additional criterion that was not originally planned (at the protocol stage).

Risk of bias in included studies

For details of risk of bias in included studies, see the Characteristics of included studies table. A brief visual summary is given in Figure 2. As we had hypothesised, given the nature of physical activity as an intervention and the public workplace setting, none of the studies was able to successfully blind all participants, personnel, or outcome measurements, and most studies were rated at unclear or high risk for a number of other domains. Only one study was assessed as overall at low risk of bias for all included outcomes (Carr 2013). A summary of assessments against each risk of bias domain is outlined below. Overall risk of bias was assessed separately for each outcome, incorporated into GRADE assessments of the certainty of evidence, and reported accordingly throughout the review.

2.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Allocation

We judged random sequence generation to be at unclear risk of bias for four studies due to insufficient detail provided about methods of generating the random sequence (Mansi 2013; Pillay 2014; Swartz 2014; Talbot 2011). We judged allocation concealment to be at unclear risk for five studies (Aittasalo 2012; Dishman 2009; Pillay 2014; Ribeiro 2014; Talbot 2011), again due to insufficient information provided on this element of the study methods. All other studies were at low risk of bias in these two domains.

Blinding

Only one study was judged at low risk for blinding of participants and personnel (performance bias). Morgan 2011 was able to achieve blinding of participants through allocation of clusters of shift workers in an industrial setting: workers receiving the intervention were not on the same roster as workers in the control group. All other studies were considered at high risk in this domain.

Given the challenges of blinding in this field, the level of risk for blinding of outcome assessment (detection bias) largely depended on whether the outcome measure was robust enough to be unaffected by the absence of blinding. Six studies were rated as low risk in this domain for physical activity and sedentary behaviours, largely when these were electronically collected by a pedometer or accelerometer and were not self‐reported (Audrey 2015; Carr 2013; Maruyama 2010; Parry 2013; Pillay 2014; Swartz 2014). No studies were assessed at low risk for quality of life measures. All studies were assessed at low risk for objectively measured risk factors for chronic disease, including BMI, weight, blood pressure, cholesterol, etc. Two studies were assessed at low risk for reporting of adverse effects (Audrey 2015; Morgan 2011), two at unclear risk (Linde 2012; Parry 2013), and the remainder at high risk.

Incomplete outcome data

A number of participants withdrew or were lost to follow‐up in most studies, but the level of risk in this domain was assessed using sensitivity analysis to identify whether the observed effect measure was robust to plausible high and low risk assumptions about missing data. As such, this domain also varied by outcome. Of those studies reporting adverse effects, only one was judged at low risk of bias (Morgan 2011). For all other outcomes, five studies were assessed at low risk of bias (Carr 2013; Mansi 2013; Pillay 2014; Ribeiro 2014; Viester 2012).

Selective reporting

We attempted to locate study protocols or trials register records for all included studies, to compare available outcomes and analyses with those planned ahead of time. Four studies had published protocols (Audrey 2015; Mansi 2013; Parry 2013; Ribeiro 2014), and four were registered at a clinical trials register (Carr 2013; Linde 2012; Morgan 2011; Ribeiro 2014). Five studies were judged at low risk of bias on this basis (Audrey 2015; Carr 2013; Linde 2012; Mansi 2013; Pillay 2014). Three were judged at high risk of bias, with identifiable unreported or selectively reported outcome data (Parry 2013; Ribeiro 2014; Viester 2012). The remainder were judged unclear.

Other potential sources of bias

Three studies were judged to be at high risk of bias for reasons arising from their cluster‐randomised design. In Audrey 2015, recruitment of participants occurred after randomisation, and it is possible that knowledge of the intervention affected the types of participants recruited (noting that there was a substantive difference in numbers recruited to each group). In Dishman 2009, although work sites were allocated at random, the nature of the intervention offered at each specific site would have been clear to individual participants before they voluntarily enrolled. Following cluster‐randomisation, recruitment of individuals at each work site was voluntary. Volunteers were permitted to join the study after allocation and baseline measurement, at which point the nature of the intervention would have been clear, leading to possible differences in the types of participants recruited (and again, there were differences in recruitment rates between groups). In Morgan 2011, the small number of clusters may have led to an increased chance of an important baseline imbalance between randomised groups in terms of the clusters or the individuals.

Effects of interventions

See: Table 1; Table 2

Pedometer intervention versus 'no intervention' control

Findings for our main outcomes for this comparison and the detailed assessment of certainty using GRADE are summarised in Table 1. We found no studies that assessed long‐term follow‐up after completion of the intervention (> 12 months ‐ our optimal outcome time point) or short‐term follow‐up after completion of the intervention (< 1 month). We found no studies with a short‐duration intervention period (< 1 month).

Primary outcome: physical activity

Follow‐up after completion of the intervention

Six studies assessed physical activity at medium‐term follow‐up after completion of the intervention (range 3 to 10 months; see Analysis 1.1) (Aittasalo 2012; Audrey 2015; Mansi 2013; Maruyama 2010; Ribeiro 2014; Viester 2012). Due to very high statistical heterogeneity (I² = 94%), we were unable to undertake meta‐analysis. We assessed the evidence to be at very low certainty, meaning that pedometer interventions may have no effect on physical activity but the effect is very uncertain. Individual studies reported a range of effects on physical activity from little to no effect (ratio of means (RoM) 0.91, 95% confidence interval (CI) 0.75 to 1.11, equivalent to the intervention group undertaking 46 fewer METs minutes/week than the control group, or 971 fewer steps/d; Maruyama 2010) to likely improvement (RoM 2.66, 95% CI 2.17 to 3.26, equivalent to the intervention group undertaking 863 more total METs minutes/week than the control group, or 3700 more steps/d; Mansi 2013). This last study appeared to be an outlier, reporting a substantially larger effect than the other studies and accounting for the high heterogeneity. However, no specific characteristic could be readily identified to cause this difference based on our pre‐specified categories considered likely to modify the effect (see Subgroup analysis and investigation of heterogeneity).

1.1. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 1: Physical activity: combined (follow‐up after completion)

At completion of the intervention

One study assessed physical activity immediately at completion of a long‐duration (two year) intervention (see Analysis 1.2). Linde 2012 observed that participants in the intervention group on average appeared to have lower levels of physical activity (RoM 0.91, 95% CI 0.63 to 1.31; 1299 participants) but with a wide confidence interval consistent with the possibility of a large reduction or a large increase in physical activity.

1.2. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 2: Physical activity: combined (at completion of intervention)

Nine studies assessed physical activity immediately at completion of interventions of medium duration (range 10 weeks to 6 months) (Aittasalo 2012; Carr 2013; Dishman 2009; Mansi 2013; Maruyama 2010; Morgan 2011; Pillay 2014; Ribeiro 2014; Viester 2012). Participation in the intervention appears to improve physical activity (RoM 1.37, 95% CI 1.11 to 1.71; 1672 participants), which is roughly equivalent to the intervention group taking 1900 more steps/d or 470 more METs minutes/week compared to the control group, although heterogeneity between studies was very high (I² = 86%).

Measurements at completion of the intervention were not our main outcomes of interest, and we performed no GRADE assessment.

In summary, although pedometer interventions appear to improve physical activity during the course of medium‐ or long‐duration interventions, it is uncertain whether effects are sustained in the medium term after completion of the intervention.

Secondary outcome: sedentary behaviour

Follow‐up after completion of the intervention

One study assessed sedentary behaviour at medium‐term (six months) follow‐up after completion of the intervention (Aittasalo 2012; see Analysis 1.3). Pedometers may have an effect on sedentary behaviour, but the effect is very uncertain (mean difference (MD) ‐33.00 minutes/d, 95% CI ‐84.28 to 18.28; 172 participants).

1.3. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 3: Sedentary behaviour (min/d)

At completion of the intervention

Two studies measured sedentary behaviour immediately at completion of medium‐duration interventions (range 3 to 6 months; see Analysis 1.3) (Aittasalo 2012Carr 2013). Researchers observed a large reduction in sedentary time of MD ‐44.09 minutes/d (95% CI ‐99.51 to 11.33; 227 participants), although the confidence intervals were also consistent with the possibility of an increase in sedentary time, and heterogeneity between studies was moderate (I² = 57%).

One study assessed sedentary behaviour at completion of a long‐duration intervention (two years) and observed on average a slight increase in sedentary behaviour, but with wide confidence intervals consistent with a large decrease or increase (MD 16.60 minutes/d, 95% CI ‐61.11 to 93.18; 1354 participants) (Linde 2012).

Measurements at completion of the intervention were not our main outcomes of interest, and we performed no GRADE assessment.

In addition, one study did not report its findings (Audrey 2015), and another study collected data on sedentary time for only a subset of participants (Viester 2012).

In summary, although pedometer interventions appear to improve sedentary behaviour during the course of medium‐duration interventions, it is uncertain whether effects are sustained over the course of a long‐term intervention or after the intervention is finished.

Secondary outcome: cardiovascular disease and type 2 diabetes risk factors

Anthropometric measures

Follow‐up after completion of the intervention

Body mass index (BMI) was assessed at medium‐term (range 3 to 8 months) follow‐up after completion of the intervention by three studies (Aittasalo 2012; Mansi 2013; Viester 2012; see Analysis 1.4). Low‐certainty evidence suggests that pedometer intervention may improve (reduce) BMI slightly in the medium term (MD ‐0.64 kg/m², 95% CI ‐1.45 to 0.18; 486 participants).

1.4. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 4: CVD risk factor: body mass index (BMI; kg/m²)

Three studies measured waist circumference (WC) at medium‐term follow‐up after completion of the intervention (range 3 to 8 months; see Analysis 1.5) (Mansi 2013; Ribeiro 2014; Viester 2012), finding an uncertain effect consistent with the possibility of a slight improvement in WC (MD ‐0.77 cm, 95% CI ‐1.91 to 0.37; 439 participants), although with low heterogeneity (I² = 16%).

1.5. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 5: CVD risk factor: waist circumference (cm)

At completion of the intervention

Seven studies (see Analysis 1.4) measured BMI immediately at completion of medium‐duration interventions (range 10 weeks to 6 months) and observed a slight improvement (reduction) (MD ‐0.67 kg/m², 95% CI ‐0.1.28 to ‐0.06; 751 participants) (Aittasalo 2012; Carr 2013; Mansi 2013; Maruyama 2010; Morgan 2011; Pillay 2014; Viester 2012). Heterogeneity was moderate (I² = 60%). Linde 2012 assessed BMI immediately at completion of a long‐duration intervention (two years) and observed a negligible increase (worsening) in BMI on average (MD 0.67 kg/m², 95% CI ‐0.17 to 1.51; 1405 participants).

Seven studies (see Analysis 1.5) measured WC at completion of medium‐duration interventions (range 10 weeks to 6 months; MD ‐1.67 cm, 95% CI ‐3.85 to 0.51; 694 participants) (Carr 2013; Mansi 2013; Maruyama 2010; Morgan 2011; Pillay 2014; Ribeiro 2014; Viester 2012). This represents an uncertain effect consistent with a possible increase or decrease, and heterogeneity for medium‐term measures was high (I² = 85%). WC was not a main outcome of interest; therefore no GRADE assessment was performed. No study reported WC at completion of a long‐term intervention.

In addition, Ribeiro 2014 reported weight through the time of publication and did not report BMI.

In summary, pedometer interventions may slightly improve BMI during the course of medium‐duration interventions and at medium‐term follow‐up after completion of the intervention, but it is uncertain if any effect is sustained over the course of a long‐duration intervention. There is no clear evidence of an effect on WC.

Blood pressure

Follow‐up after completion of the intervention

Two studies assessed systolic (SBP) and diastolic blood pressure (DBP) following medium‐term (three to eight months) follow‐up after completion of the intervention (see Analysis 1.6 and Analysis 1.7) (Mansi 2013; Viester 2012). Moderate‐certainty evidence shows that pedometer interventions probably make little or no difference to SBP (MD ‐0.08 mmHg, 95% CI ‐3.26 to 3.11; 315 participants). The DBP finding was consistent with the systolic finding (MD 0.70 mmHg, 95% CI ‐1.34 to 2.73; 315 participants), although DBP was not considered a main outcome for this review, and no GRADE assessment was performed. No statistical heterogeneity was observed for either measure (I² = 0).

1.6. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 6: CVD risk factor: systolic blood pressure (mmHg)

1.7. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 7: CVD risk factor: diastolic blood pressure (mmHg)

At completion of the intervention

Six studies measured blood pressure at completion of medium‐duration interventions (range 10 weeks to 6 months) and observed no clear evidence of an effect on SBP (MD ‐1.56 mmHg, 95% CI ‐3.66 to 0.53; 571 participants) or DBP (MD ‐0.71 mmHg, 95% CI 2.57 to 1.14; 571 participants) (Carr 2013; Mansi 2013; Maruyama 2010; Morgan 2011; Pillay 2014; Viester 2012). Heterogeneity was very low for both measures (I² = 0%). No GRADE assessment was performed. No studies assessed blood pressure after long (> 12 months) pedometer interventions.

One study assessed resting heart rate and reported that on average, those allocated to the medium‐length (three‐month) pedometer intervention had a lower resting heart rate at completion of the intervention than those allocated to the control group (MD ‐7.90 beats per minute, 95% CI ‐11.59 to ‐4.21; 110 participants) (Morgan 2011). No GRADE assessment was performed. We hypothesised that resting heart rate would not change due to a low‐impact health intervention. It is important to note that this benefit may be due to other factors such as increased participant calmness at the second round of data collection. Hence, the significant benefits for resting heart rate in this one small study should be interpreted with caution.

In summary, there is no clear evidence that pedometer interventions affect blood pressure during the course of medium‐duration interventions, and there is moderate certainty that they make little to no difference to blood pressure at medium‐term follow‐up after completion of the intervention.

Biochemical measures

Follow‐up after completion of the intervention

We found no studies that assessed biochemical measures at follow‐up after completion of the intervention.

At completion of the intervention

Two studies assessed biochemical measures at completion of medium‐duration (three to four months) interventions (Carr 2013; Maruyama 2010). Low‐density lipoprotein (LDL) cholesterol was considered a main outcome of interest for this review (see Analysis 1.8). Pedometer interventions may have an effect on LDL cholesterol (MD ‐3.58 mg/dL, 95% CI ‐10.76 to 3.59; 2 studies; 127 participants), but the effect is very uncertain.

1.8. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 8: CVD risk factor: cholesterol: LDL

In the same studies, little to no effect on high‐density lipoprotein (HDL) cholesterol was observed (MD ‐0.94 mg/dL, 95% CI ‐3.77 to 1.89; 2 studies; 127 participants; see Analysis 1.9), and the effect on triglycerides was uncertain (MD ‐10.83 mg/dL, 95% CI ‐33.84 to 12.18; 2 studies; 127 participants; see Analysis 1.10). Maruyama 2010 assessed blood glucose at completion of a four‐month intervention and observed moderate improvement (reduction) (MD ‐4.80 mg/dL, 95% CI ‐9.14 to ‐0.46; 87 participants; 4‐month intervention). No statistical heterogeneity was observed for these outcomes (I² = 0%). These outcomes were not considered main outcomes for this review; therefore no GRADE assessment was performed.

1.9. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 9: CVD risk factor: cholesterol: HDL

1.10. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 10: CVD risk factor: triglycerides

In summary, it is uncertain whether pedometer interventions improve LDL cholesterol during the course of medium‐duration interventions. We found no studies that assessed biochemical measures during follow‐up after completion of the intervention. Effects on other biochemical measures were also uncertain, and no clear evidence was observed.

Disease risk scores

Viester 2012 listed the European Systematic Coronary Risk Evaluation (SCORE) in the study protocol but did not report the results. None of the other included studies reported disease risk scores as an outcome in this comparison.

Secondary outcome: quality of life

Follow‐up after completion of the intervention

Mansi 2013 assessed the mental health component of quality of life using the Short Form 36 version 2 scale (SF‐36v2) at medium‐term (three months) follow‐up after completion of the intervention. Pedometer interventions may have an effect on the mental health component of quality of life, but the effect is very uncertain (MD 1.30 units, 95% CI ‐1.80 to ‐4.40; 58 participants; see Analysis 1.11).

1.11. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 11: Quality of life: mental health component

Two studies assessed the physical health component of quality of life using the SF‐36v2 and RAND‐36 scales at medium‐term (range 3 to 6 months) follow‐up after completion of the intervention (Mansi 2013; Viester 2012). The intervention appeared to make little to no difference (standardised mean difference (SMD) 0.08 units, 95% CI ‐0.14 to 0.30; 310 participants; see Analysis 1.12), with no statistical heterogeneity (I² = 0%). This was not a main outcome of interest for the review, and no GRADE assessment was performed.

1.12. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 12: Quality of life: physical health component

At completion of the intervention

Two studies measured the mental health component using the SF‐36v2 or SF‐12 scale (which can be converted to the same units) at completion of medium‐duration interventions (range 3 to 3.5 months) and observed a very slight improvement (MD 3.18 units, 95% CI ‐1.32 to 07.68; 168 participants; see Analysis 1.11) (Mansi 2013; Morgan 2011), although heterogeneity was high (I² = 70%).

Two studies measured the physical health component using the SF‐36v2 and RAND‐36 scales at completion of medium‐duration interventions (six months) and observed little to no difference (MD 0.08 units, 95% CI ‐0.14 to 0.30; 210 participants; see Analysis 1.12), with moderate heterogeneity (I² = 53%) (Mansi 2013; Viester 2012). One additional study reported the physical health component at completion of a medium‐duration intervention (three months) but reported the results as change from baseline using the SF‐12 scale and could not be standardised alongside the other two studies in a single meta‐analysis (Morgan 2011). This study reported a consistent finding of little to no difference (MD 2.8 units, 95% CI ‐0.30 to 5.9; 110 participants). These time points were not considered main outcomes for the review, and no GRADE assessment was performed.

We found no studies that assessed long‐duration (> 12 months) pedometer interventions.

Aittasalo 2012 planned to but did not assess quality of life. Ribeiro 2014 measured quality of life but did not report the results.

In summary, the effect of pedometer interventions on quality of life is very uncertain. A slight improvement in mental health quality of life was observed at completion of medium‐duration interventions, but all observed effects were small to negligible.

Secondary outcome: adverse effects

Follow‐up after completion of the intervention

Two studies provided low‐certainty evidence suggesting that pedometer interventions probably reduce the rate of adverse effects at medium‐term follow‐up (range 6 to 9 months) after completion of the intervention (odds ratio (OR) 0.50 95% CI 0.30 to 0.84; 286 participants; see Analysis 1.13), with no statistical heterogeneity (I² = 0%) (Aittasalo 2012; Morgan 2011). Aittasalo 2012 measured exercise‐related adverse events six months after the intervention, and Morgan 2011 compared reported work site injuries in the 12 months before to the 12 months including and following intervention implementation (8.5 months' follow‐up). The absolute impact of this reduction would depend on the definitions and rates of adverse effects in the population, given that some populations observed no adverse effects. In a population comparable to Aittasalo 2012, where a relatively high rate of adverse effects of 24 people experiencing an adverse effect per hundred was observed in the control group, this would translate to a reduction to 14 people per hundred (95% CI 9 to 21 people).

1.13. Analysis.

Comparison 1: Pedometer intervention vs minimal intervention, Outcome 13: Adverse effects

In addition, Audrey 2015 reported zero adverse effects due to walking to work (including physical and social adverse effects) at eight months' follow‐up after completion of the intervention, but this information could not be incorporated into the meta‐analysis.

At completion of the intervention

Linde 2012 reported zero adverse effects (not defined) at completion of a long‐duration intervention (two years), but these data could not be included in a meta‐analysis. Aittasalo 2012 reported exercise‐related adverse effects at completion of a medium‐duration intervention (six months) and observed uncertain evidence of effect (OR 0.87, 95% CI 0.39 to 1.94; 180 participants; see Analysis 1.13). These time points were not of main interest to this review, and no GRADE assessment was performed.

In addition, Mansi 2013 collected adverse events only in the pedometer intervention group as part of a feasibility trial (reported no adverse events), which could not be analysed in this review.

In summary, moderate‐certainty evidence shows that pedometer interventions probably reduce the rate of adverse effects in the medium term after completion of the intervention, although the reason for this is uncertain. Aittasalo 2012 measured an increase in physical activity following completion of the intervention, and so in theory, increased exercise could lead to improved health and fitness. Morgan 2011 did not report physical activity at a comparable time point. The effects remain at completion of pedometer interventions.

Pedometer intervention versus alternative intervention without pedometer

Findings for our main outcomes for this comparison and detailed assessment of certainty using GRADE are summarised in Table 2. We found no studies that assessed long‐term follow‐up after completion of the intervention (> 12 months ‐ our optimal outcome time point) or short‐term follow‐up after completion of the intervention (< 1 month). We found no studies that measured outcomes immediately at completion of a long‐duration intervention (> 12 months).

Primary outcome: physical activity

Follow‐up after completion of the intervention

Ribeiro 2014 assessed physical activity at medium‐term follow‐up (three months) after completion of the intervention. Pedometer interventions may improve physical activity, but the effect is very uncertain (RoM 1.07, 95% CI 0.97 to 1.18; 143 participants; 3 months' follow‐up after completion of the intervention; see Analysis 2.1). To assist interpretation, an RoM of 1.07 is roughly equivalent to the intervention group taking 681 steps/d more than the comparison group, but with a confidence interval also consistent with the possibility of a small decrease or a large increase.

2.1. Analysis.

Comparison 2: Pedometer intervention vs alternative physical activity intervention, Outcome 1: Physical activity: combined

At completion of the intervention

Three studies assessed physical activity at completion of medium‐duration interventions (three to six months), reporting an uncertain effect on physical activity (RoM 1.04, 95% CI 0.92 to 1.17; 273 participants; see Analysis 2.1) with moderate heterogeneity (I² = 48%) (Parry 2013; Ribeiro 2014; Talbot 2011). To assist interpretation, an RoM of 1.04 is roughly equivalent to the intervention group taking 398 steps/d or 10 METs minutes/workday more than the comparison group, but with a confidence interval also consistent with an important decrease or increase.

Swartz 2014 assessed physical activity at completion of a short‐duration intervention (two weeks) and appeared to observe an increase in physical activity (RoM 1.26, 95% CI 0.96 to 1.66; 60 participants), but the result was uncertain, with a confidence interval also consistent with a small decrease or a very large increase in physical activity. For Swartz 2014, the RoM of 1.26 was equivalent to the intervention group taking 1032 steps/workday more than the alternative intervention group. These time points were not of main interest to this review, and no GRADE assessment was observed.

In summary, it is uncertain whether pedometer interventions increase physical activity in comparison to alternative physical activity interventions.

Secondary outcome: sedentary behaviour

Follow‐up after completion of the intervention

We found no studies that assessed sedentary behaviour at follow‐up after completion of the intervention.

At completion of the intervention

One study measured sedentary behaviour as a percentage of pedometer wear time at completion of a medium‐duration intervention (three months) (Pillay 2014). Pedometer interventions may improve sedentary behaviour, but the effect is very uncertain (MD 1.48% of accelerometer wear time, 95% CI ‐1.52 to 4.49; 62 participants).

One study measured sedentary behaviour in minutes per workday after completion of a short‐duration intervention (two weeks) and appeared to observe an increase (worsening) in sedentary time in the pedometer group compared to the alternative exercise intervention group (MD 25.00 minutes/workday, 95% CI 0.61 to 49.39; 60 participants), although no GRADE assessment was performed on this time point.

In summary, it is uncertain whether pedometer interventions improve sedentary behaviour during the course of short‐ to medium‐duration interventions in comparison to alternative exercise interventions.

Secondary outcome: cardiovascular disease and type 2 diabetes risk factors

Anthropometric measures

Follow‐up after completion of the intervention

We found no studies that assessed BMI at follow‐up after completion of the intervention.

Ribeiro 2014 assessed WC (cm) at medium‐term follow‐up (three months) after completion of the intervention. No clear evidence of effect was observed (MD 0.75 cm, 95% CI ‐0.49 to 1.99; 148 participants).

At completion of the intervention

Two studies (see Analysis 2.2) assessed BMI at completion of medium‐duration interventions (three to six months) (Parry 2013; Talbot 2011). Pedometer interventions may improve BMI at completion of medium‐duration pedometer interventions, but the effect is very uncertain (MD ‐1.23 kg/m², 95% CI ‐2.85 to 0.39; 144 participants). No statistical heterogeneity was observed (I² = 0%).

2.2. Analysis.

Comparison 2: Pedometer intervention vs alternative physical activity intervention, Outcome 2: CVD risk factor: body mass index (BMI; kg/m²)

In addition, Ribeiro 2014 reported BMI as a planned measure in the study protocol but did not report the results.

Ribeiro 2014 measured WC at completion of its medium‐duration intervention (three months) and observed no clear evidence of effect (MD 0.76, 95% CI ‐0.48 to 2.00; 148 participants). We found no studies that assessed WC at completion of a short‐duration intervention. This outcome was not of main interest to the review, and no GRADE assessment was performed. .

In summary, it is uncertain whether pedometer interventions affect adiposity in comparison to alternative exercise interventions.

Blood pressure

Follow‐up after completion of the intervention

No studies reported blood pressure at follow‐up after completion of the intervention.

At completion of the intervention

Talbot 2011 reported blood pressure at completion of a medium‐duration intervention (six months). Pedometers may improve SBP at completion of medium‐duration interventions, but the effect is very uncertain (MD ‐4.00 mmHg, 95% CI ‐10.15 to 2.15; 94 participants). The result for DBP was similar (MD ‐2.20 mmHg, 95% CI ‐6.74 to 2.34; 94 participants), although no GRADE assessment was performed on this result.

In addition, Ribeiro 2014 measured blood pressure but did not report the results.

In summary, it is uncertain whether pedometer interventions improve blood pressure in comparison to alternative exercise interventions.

Biochemical measures

Follow‐up after completion of the intervention

No studies reported LDL cholesterol at follow‐up after completion of the intervention.

At completion of the intervention

Talbot 2011 measured LDL cholesterol at completion of a medium‐duration intervention. Pedometer interventions may have an effect on LDL cholesterol, but the effect is very uncertain (MD 0.60 mg/dL, 95% CI ‐15.08 to 16.28; 94 participants). The same study appeared to observe an increase in HDL cholesterol with the pedometer intervention (7.50 mg/dL, 95% CI 1.55 to 13.45), although no GRADE assessment was performed on this outcome. This result should be interpreted with caution.

No studies reported cholesterol at completion of a short‐term intervention.

In addition, triglycerides were included in the protocol for Talbot 2011, but study authors did not include these results in their study report. Ribeiro 2014 listed blood glucose in the study protocol but did not report the results.

In summary, it is uncertain whether pedometer interventions improve cholesterol in comparison to alternative physical activity interventions. Pedometer interventions may be associated with a comparative increase in HDL cholesterol during the course of a medium‐duration intervention, but this should be considered uncertain.

Disease risk scores

Follow‐up after completion of the intervention

No studies measured disease risk scores at follow‐up after completion of the intervention.

At completion of the intervention

Talbot 2011 measured 10‐year hard coronary heart disease (CHD) risk levels (recognised and unrecognised myocardial infarction, coronary insufficiency, CHD death) calculated via the Framingham study risk equations, at completion of a medium‐duration intervention (six months). No important difference was observed between pedometers and the alternative exercise intervention (MD 0.00%, 95% CI ‐0.11 to 0.11; 94 participants), although this outcome was not of main interest to the review, and no GRADE assessment was performed.

No studies measured disease risk scores at completion of a short‐duration intervention.

In summary, there appears to be no evidence of a difference between pedometer interventions and alternative exercise interventions in disease risk scores, but this should be considered uncertain.

Secondary outcome: quality of life

Ribeiro 2014 measured quality of life (SF‐36) at completion of a medium‐duration (three‐month) intervention and at medium‐term (three‐month) follow‐up after completion of the intervention but did not report the results. None of the other included studies reported quality of life as an outcome.

Secondary outcome: adverse effects

Parry 2013 reported, "No adverse outcomes were reported for any participants" at completion of a medium‐duration (three‐month) intervention. None of the other included studies reported adverse effects as an outcome.

Subgroup analysis

We conducted subgroup analyses by participant risk profile, that is, comparing studies that specifically recruited participants at high risk of cardiovascular disease or diabetes, or who led lifestyles that were more sedentary than average. These subgroup analyses were conducted for physical activity at medium‐term follow‐up after completion of the intervention (see Analysis 3.1) and for measures at completion of medium‐duration interventions (see Analysis 3.2). No subgroup effect was observed in either case, and the heterogeneity of these studies was not explained by this factor.

3.1. Analysis.

Comparison 3: Subgroup analyses, Outcome 1: Phyical activity: subgroup analysis by participant risk profile of Analysis 1.1 (follow‐up after completion)

3.2. Analysis.

Comparison 3: Subgroup analyses, Outcome 2: Physical activity: subgroup analysis by participant risk profile of Analysis 1.2.1

It was not possible to conduct the remaining planned subgroup analyses in either Comparison 1 or Comparison 2 (see Subgroup analysis and investigation of heterogeneity section). Three included studies included only male participants (Maruyama 2010; Morgan 2011; Viester 2012), but only one study included only women (Ribeiro 2014), and the remaining studies did not report effects by gender to provide sufficient data for this subgroup analysis. No studies reported data broken down by age or educational status. All studies used a step or exercise time goal and a means of recording progress against the goal, such as a step diary.

Subgroup analysis was not possible based on duration of the intervention. In Comparison 1, all intervention periods were of medium duration, with the exception of Linde 2012, which provided an intervention of long duration. In Comparison 2, all studies were of medium duration, with the exception of Swartz 2014, which provided an intervention of short duration.

It was not possible in most cases to classify groups as using an external intervention provider compared to interventions undertaken in the workplace. In most cases, interventions were conducted within the workplace or through existing workplace health service programmes but were initiated by external research groups.

Discussion

Summary of main results

This review aimed to investigate whether introducing interventions that incorporate a pedometer in a workplace setting would be effective for increasing physical activity and improving subsequent health outcomes. We included 14 studies that were heterogeneous in the age of recruited participants, the duration of the intervention, the range of additional components included in the intervention, and the timing of outcome assessment.

Eleven studies compared programmes including a pedometer to what we considered a no‐intervention control group. Our primary interest was the effects sustained after completion of the active intervention. In this comparison, the effects observed on physical activity, sedentary behaviour, and quality of life (mental health component) were very uncertain, and no studies measured biochemical measures or disease risk scores. Low‐certainty evidence suggests that body mass index (BMI) was slightly improved (decreased) at three to eight months' follow‐up after the intervention. Moderate‐certainty evidence shows that pedometer interventions had little or no effect on systolic blood pressure, and low‐certainty evidence suggests that the rate of adverse effects was lower following the pedometer intervention at 6 to 10 months' follow‐up.

Measured at completion of interventions of medium duration (between one month and one year), and indicating changes that may have occurred during the course of the active intervention, pedometer interventions appeared to improve physical activity (average increase approximately 1900 steps/d) and sedentary behaviour (average decrease 44 minutes/d) and appeared to slightly improve BMI (average decrease 0.67 kg/m²) and quality of life (mental health component; average improvement 3.18 units), but these outcomes were not our primary time points of interest, and no GRADE assessment of uncertainty was performed. Effects on systolic blood pressure, low‐density lipoprotein (LDL) cholesterol, and adverse effects were uncertain. No studies measured disease risk scores. One study looked at effects during a long‐term (two‐year) intervention and did not observe any clear effects, throwing further uncertainty on the sustainability of effects of these interventions.

Four studies compared programmes including a pedometer with alternative physical activity interventions. Again, looking first at sustained effects after completion of the active intervention period, the effects observed on physical activity were very uncertain. Sedentary behaviour, BMI, systolic blood pressure, LDL cholesterol, disease risk scores, quality of life (mental health component), and adverse effects were not measured.

Measured at completion of interventions of medium duration, effects on physical activity, sedentary behaviour, BMI, systolic blood pressure, and LDL cholesterol were very uncertain. Quality of life (mental health component) was not reported. One study reported that no adverse effects were observed in either group.

Overall completeness and applicability of evidence

Several limitations are noted for the studies included in this review, which contributed to lack of certainty around the review's conclusions. Although studies included broad working populations and used intervention designs that are likely to be generally applicable to the working population, there were limitations in the completeness of available evidence.

Most included studies used the pedometer as part of a multi‐component intervention compared to no intervention. This makes it difficult to draw conclusions on the effectiveness of pedometers as a specific component to increase physical activity and improve subsequent health outcomes due to the potential confounding effects of other intervention components. Sufficient data were not available to explore the interventions' multiple components, nor other possible sources of heterogeneity, including age, gender, and educational status of participants, or the impact of risk of bias on the results.

Although pedometers formed only one component of the intervention, it was observed that participants did not uniformly find the pedometers useful. As described in relation to Audrey 2015, "this study used pedometers as an optional self‐monitoring tool and participants had mixed views about using them. Although some found them important for goal setting, over half did not use them for practical reasons ... the current study suggests that pedometers were useful for some participants and not for others. It would seem that to develop an intervention that specifically requires their use may restrict the involvement of some potential participants". Although the pedometer is an objective measurer of physical activity, the individual must choose to wear the device. This subjective choice to wear the device may be influenced by forgetfulness, especially after being in contact with water (e.g. when showering or swimming), when the device must be removed; or for appearance reasons, as it may be difficult to wear hip devices with dresses or during certain activities; or deliberately during an intended low activity day.

It is important to note that over the past decade, technological advancement in accelerometers as commercial products has in many ways rendered the use of pedometers outdated. Accelerometers have become cheaper, if not freely available, through Smartphone applications. Accelerometers have been integrated within other devices such as watches and jewelry. They are commonly waterproof, overcoming difficulties with use during water sports and the need to remove during showering. They are now widespread in the community, in contrast to the niche technology of pedometers. Furthermore, '10,000 steps per day' and other public health messages encouraging breaks in sedentary time have been successfully distributed through many avenues, including the workplace. Any future study aiming to test the impact of access to this technology (as distinct from more complex support programmes) would likely find any control arm highly contaminated.

This may be one reason that recent searching (including the final search for this review between 2014 and 2016, and our provisional search update to May 2019) yielded no additional eligible studies (with the exception of one extension of the Audrey 2015 pilot study included in this review), and why further updates of this review may not be useful.

Quality of the evidence

Using the GRADE approach (Higgins 2011a), we assessed the overall certainty of evidence for our main outcomes of interest, and we found most to be of low or very low certainty (see Table 1; Table 2). The certainty of the evidence was commonly downgraded due to high risk of bias in the included studies, which often arose from high levels of attrition and lack of blinding, both of which are likely to be intractable issues in this kind of pragmatic community intervention. Many of the observed effects were imprecise, encompassing a wide range of possible effects, and many outcomes were based on small sample sizes. There were high levels of heterogeneity among observed results, for which adequate evidence was not available to conduct targeted subgroup analyses to identify the causes, among the many variations in populations, interventions, and comparisons in the included studies.

Potential biases in the review process

Throughout the process of conducting this review, we endeavoured to use methods that would minimise risk of bias in our work.

Our literature searches were comprehensive, including a range of electronic databases, trials registers, and contact with identified authors in the field. However, the most recent search (conducted in 2016) is now considerably out of date. Given observed trends in the literature (no eligible studies identified after 2014, and a preliminary update of the search to 2019 identifying only one eligible study that has not been incorporated) and in the community (given the widespread use of accelerometers), it is in our view very unlikely that enough additional studies have been completed to resolve the uncertainties identified in this review. It is more likely that current research in this field will encompass new technology and will focus more broadly on multi‐component support interventions rather than testing of individual technological components.

Data collection for this review was conducted thoroughly and in duplicate. Sufficient detail on the included studies was available for the purposes of analysis, although the level of detail describing the interventions was variable and was dependent on the authors of those studies. Some readers of this review may not find sufficient information available to replicate or assess in detail all the interventions included. Efforts were made to obtain as much additional information as possible, although our focus on the use of pedometers may not have covered all aspects of interest to all readers.

We went to some lengths to ensure that as many data as possible were included in the analysis, including standardised assessment of different measures of physical activity using ratios of means, and imputed values for missing values such as standard deviations. These efforts reduced the level of potential bias arising from exclusion of studies from the analysis, although there is always a level of risk in making the assumptions required in this approach. We were largely unable to test the robustness of data using planned sensitivity analyses, subgroup analyses, and funnel plots, given limitations of the data obtained.

Agreements and disagreements with other studies or reviews

We identified eight systematic reviews related to the broader topics of workplace physical activity interventions, workplace health promotion interventions to increase physical activity, or pedometer interventions in adults. We did not identify any relevant systematic review that assessed sustained changes after completion of interventions.

In relation to the immediate impact of these interventions on physical activity, four reviews found positive effects on physical activity outcomes, broadly consistent with our review. Bravata 2007 reviewed pedometer‐based interventions (27 studies, randomised and non‐randomised; five in a workplace setting). Foster 2013 compared interventions with a pedometer to alternative physical activity interventions (three studies). Commissaris 2016 reviewed interventions to reduce sedentary behaviour and increase physical activity during productive work (six studies of self‐monitoring techniques including pedometers). Jahangiry 2017 reviewed Web‐based physical activity interventions (six studies using pedometer walking interventions). In contrast, Malik 2014 reviewed workplace health promotion interventions for increasing physical activity and identified six studies (five randomised), of which two demonstrated a statistically significant increase against a control group. Reed 2017 reviewed workplace physical activity interventions and did not find a positive effect on metabolic equivalents per week (three studies).

The systematic reviews Bravata 2007, Reed 2017, and Verweij 2011 (11 studies) also found improvements in BMI immediately following an intervention, again consistent with our review.

Workplace pedometer interventions may not increase physical activity enough for blood pressure changes to occur. Igarashi 2018, a systematic review, calculated that although there is no set number of steps (e.g. 10,000 per day) that would cause a reduction in blood pressure, increasing one's step count could do so, which is in contrast to our finding that despite an increase in physical activity, the effect on blood pressure was uncertain. Also in contrast to our uncertain findings on LDL cholesterol is the finding by Reed 2017 that workplace physical activity interventions decrease LDL cholesterol (four studies).

The findings of these reviews reinforce that there may be an immediate impact on physical activity immediately following workplace interventions using pedometers, but these effects may not be greater than those obtained when other physical activity interventions are used, and may not be sustained in the long term. There may be positive effects on other outcomes, including disease risk factors, but uncertainty remains. Considerable heterogeneity in the focus and components of physical activity interventions is a challenge across the literature.

Authors' conclusions

Implications for practice.

Evidence was insufficient to justify clear conclusions about the effectiveness of pedometer interventions in a workplace setting for increasing physical activity and improving subsequent health outcomes.

When compared to minimal intervention, some immediate benefits are seen to occur during a pedometer‐based intervention, but evidence is insufficient to show whether these effects are sustained following completion of the active intervention period, and sustainability during very long interventions (longer than one year) is not assured. This comparison does not demonstrate whether the pedometer is a critical component of physical activity interventions, or whether any similar multi‐component intervention would achieve similar outcomes. When compared to alternative physical activity interventions, the effects overall are very uncertain.

Exercise interventions can have positive effects on employee physical activity and health, although there is currently no reason to suggest that a pedometer‐based intervention would be more effective than other options. Decision‐makers considering allocating resources to large‐scale programmes of this kind should balance expected benefits with realistic expectations about participation (including high levels of withdrawals), and should note that effects may not be sustained over the longer term.

Implications for research.

Several factors could be considered for planning future research in this field. In terms of participants, the existing literature has successfully included a diversity of workplaces and countries and both healthy and higher‐risk populations. In terms of interventions to be investigated, we suggest that more studies are needed that have been designed to identify the core effective components of multi‐component interventions. Pedometers may not be the highest priority component to test for future research, especially given the increased availability of low‐cost accelerometers. Approaches to increase the sustainability of and engagement with the intervention over longer terms should be considered. If the aim of physical activity interventions is to prevent obesity and chronic disease, then the theoretical basis for intervention components and the role of other lifestyle components such as diet could also be considered.

Future research should consider measuring sustained outcomes after completion of the active intervention period and should include more consistently meaningful measures of physical activity (such as total metabolic equivalents (METs)), as well as important outcomes that indicate that increased physical activity is achieving desired outcomes (including anthropometry, blood pressure, quality of life, and reduced risk of adverse effects).

Future studies should ensure that they are designed to achieve low risk of bias. Particular challenges involving research in physical activity include the difficulty of blinding (although this is less important if the intention is to measure the effect of introducing a physical activity programme in a real workplace setting, where uptake would be variable) and high levels of attrition (which may be linked to issues with sustainability of the intervention), both of which could be considered in future intervention designs. Studies should be sufficiently powered to enable subgroup analyses (including intervention components or population categories) in this context of high attrition.

What's new

Date	Event	Description
14 June 2020	New citation required and conclusions have changed	Updated, with ten new studies included
14 June 2020	New search has been performed	Updated, with ten new studies included

History

Protocol first published: Issue 7, 2011
Review first published: Issue 4, 2013

Date	Event	Description
13 October 2013	Amended	Provided more detailed information on studies awaiting classification that were identified in the updated search in March 2013

Appendices

Appendix 1. CENTRAL search

We ran the following search strategy initially on 1 Feb 2012, and updated on 12 March 2013, 3 December 2014, and 13 December 2016 using appropriate date limits.

#1	(work* OR occupat* OR company* OR offic* OR busines*):ti,ab,kw
#2	MeSH descriptor: [Work], this term only
#3	MeSH descriptor: [Workplace], this term only
#4	MeSH descriptor: [Occupational Groups] explode all trees
#5	MeSH descriptor: [Walking] explode all trees
#6	step* near/5 count*
#7	step* near/5 daily
#8	(pedometer* OR manpometer OR "manpo meter" OR step next/1 meter* OR stepmeter*):ti,ab,kw
#9	(walking OR "10,000 step" OR "10,000 steps" OR "10000 steps" OR "10000 step" OR "10 000 step" OR "10 000 steps"):ti,ab,kw
#10	(#1 OR #2 OR #3 OR #4)
#11	(#5 OR #6 OR #7 OR #8 OR #9)
#12	(#10 AND #11)
#13	#12 in Trials

Appendix 2. CINAHL search

We ran the following search strategy initially on CINAHL through EBSCOhost on 6 Feb 2012, and updated on 14 March 2014, 10 December 2014, and 8 December 2016, using appropriate date limits.

S1	TX work or TX works* or TX work’* or TX worka* or TX worke* or TX workg* or TX worki* or TX workl* or TX workp* or TX occupat* or TX company* or TX (offic* OR busines* )
S2	Walking or TX Walking or MW Walking
S3	TX step or TX steps or TX “10,000 step*” or TX “10000 step*”
S4	TX count or TX counts or TX counting or TX counter or TX counters or TX counting or TX daily
S5	S3 and S4
S6	TX pedometer* or TX manpo‐meter or TX "manpometer" or TX "manpo meter"
S7	S2 or S5 or S6
S8	TX randomized controlled trial or TX controlled clinical trial or AB placebo or TX clinical trials or AB randomly or TI trial or TX intervent*
S9	S1 and S7 and S8
S10	TX animal not TX human
S11	S9 not S10

Appendix 3. MEDLINE search

We ran the following search strategy initially on MEDLINE through PubMed on 30 Janurary 2012, and updated on 11 March 2013, 1 December 2014, and 9 December 2016.

#1	Search work[tw] OR works*[tw] OR work'*[tw] OR worka*[tw] OR worke*[tw] OR workg*[tw] OR worki*[tw] OR workl*[tw] OR workp*[tw] OR occupat*[tw] OR company*[tw] OR offic*[tw] OR busines*[tw]
#2	Search "Walking"[Mesh] OR walking[tw] OR "10,000 step"[tw] OR "10,000 steps"[tw] OR "10000 steps"[tw] OR "10000 step"[tw] OR "10 000 step"[tw] OR "10 000 steps"[tw]
#3	Search (step[tw] OR steps[tw]) AND (count[tw] OR counts[tw] OR counting[tw] OR counter[tw] OR counters[tw] OR daily[tw])
#4	Search (pedometer* OR manpo‐meter[tw] OR "manpometer"[tw] OR "manpo meter"[tw] OR "step meter"[tw] OR "step meters"[tw] OR stepmeter*[tw])
#5	Search #2 OR #3 OR #4
#6	Search (randomized controlled trial[pt] OR controlled clinical trial[pt] OR randomized[tiab] OR placebo[tiab] OR clinical trials as topic[mesh:noexp] OR randomly[tiab] OR trial[ti] NOT (animals[mh] NOT humans[mh]))
#7	Search intervent*
#8	Search #6 OR #7
#9	Search #1 AND #5 AND #8

Appendix 4. Embase search

We ran the following search strategy initially on Embase through Embase.com on 31 Janurary 2012, and updated on 11 March 2013, 1 December 2014, and 13 December 2016.

#1	(work*or occupat*or company*or offic*or busines*).de,ab,ti.
#2	('10,000 step' or '10,000 steps' or '10000 steps' or '10000 step' or '10 000 step' or '10 000 steps').ab,ti.
#3	'walking'/exp OR walking {No Related Terms}
#4	(count* OR daily) NEAR/5 step* {No Related Terms}
#5	pedometer* OR 'manpometer' OR 'manpo meter' OR 'step meter' OR 'step meters' OR stepmeter* {No Related Terms}
#6	#2 OR #3 OR #4 OR #5
#7	'controlled clinical trial'/exp {No Related Terms}
#8	intervention* {No Related Terms}
#9	#1 AND #6
#10	#7 AND #9
#11	#8 AND #9
#12	#10 OR #11 {No Related Terms}
#13	'nonhuman'/exp {No Related Terms}
#14	#12 NOT #13

Appendix 5. OSH UPDATE

We searched the following reference databases.

• Health and Safety Information Centre of The International Labour Office in Geneva, Switzerland (CISDOC).

• UK Health and Safety Executive Information Services and Health and Safety Commission (HSELINE).

• INTERNATIONAL BIBLIOGRAPHIC DATABASE.

• Occupational Safety and Health Research Institute (IRSST).

• US National Institute for Occupational Safety and Health (NIOSHTIC, NIOSHTIC‐2).

• Ryerson International Labour Occupational Safety and Health (RILOSH).

We ran the following search strategy initially on OSH UPDATE on 31st Janurary 2012, 11 March 2013, and 1 December 2014. For the 2016 update, it was decided not to include OSH UPDATE, as the results of previous searches had not revealed additional included studies.

#1	GW{pedometer* OR manpo‐meter OR "manpometer" OR "manpo meter" OR "step meter" OR "step meters" OR stepmeter*}
#2	GW{walking OR 10,000 step OR 10,000 steps OR 10000 steps OR 10000 step OR 10 000 step OR 10 000 steps OR foot steps}
#3	GW{(step OR steps) AND (count OR counts OR counting OR counter OR counters OR daily)}
#4	#1 OR #2 OR #3
#5	GW{intervent* OR random*} OR TW{trial*}
#6	#4 AND #5
#7	DC{OUBIB OR OUCISD OR OUHSEL OR OUISST OR OUNIOC OR OUNIOS OR OURILO}
#8	#6 AND #7

Appendix 6. Web of Science search

We ran the following search strategy initially on Web of Science through apps.isiknowledge.com (including SCI‐Expanded, SSCI, and A&HCI databases on 31 Janurary 2012, and updated on 11 March 2013, 5 December 2014, and 9 December 2016.

#1	TS=(work* OR occupat* OR company* OR offic* OR busines*)
#2	TS=walking
#3	TS=((step OR steps) AND (count OR counts OR counting OR Counter OR counters OR daily))
#4	TS=("10,000 step" OR "10,000 steps" OR "10000 steps" OR "10000 step" OR "10 000 step" OR "10 000 steps")
#5	TS=(pedometer* OR "manpo‐meter" OR "manpometer" OR "manpo meter" OR "step meter" OR "step meters" OR stepmeter*)
#6	#2 OR #3 OR #4 OR #5
#7	#1 AND #6
#8	TS=("randomized controlled trial" OR "controlled clinical trial" OR placebo OR "clinical trials" OR randomly OR intervent*) OR TI=trial
#9	#7 AND #8
#10	TS=(animal* NOT human*)
#11	#9 NOT #10
#12	#11

Appendix 7. Clinicaltrials.gov search

We ran the following search strategy on clinicaltrials.gov on 26 April 2017.

(pedometer OR “10,000 steps” OR “10000 steps” OR “10 000 steps” OR "step count") AND (work OR workplace OR worksite OR employee OR employer OR company OR business OR office) AND (randomized OR randomised).

Appendix 8. WHO International Clinical Trials Registry Platform (ICTRP) search

We ran the following search strategy on the WHO International Clinical Trials Registry Platform (ICTRP), via http://apps.who.int/trialsearch/AdvSearch.aspx, on 26 April 2017.

• Title: (work OR workplace OR worksite OR employee OR employer OR company OR business OR office).

• Intervention: (pedometer OR 10,000 steps OR 10000 steps OR 10 000 steps OR step count).

• Recruitment status is: ALL.

Data and analyses

Comparison 1. Pedometer intervention vs minimal intervention.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Physical activity: combined (follow‐up after completion)	6		Ratio of means (IV, Random, 95% CI)	Totals not selected
1.1.1 Follow‐up after completion (medium, 3 to 10 months)	6		Ratio of means (IV, Random, 95% CI)	Totals not selected
1.2 Physical activity: combined (at completion of intervention)	10		Ratio of means (IV, Random, 95% CI)	Subtotals only
1.2.1 At completion of intervention (medium duration, 10 weeks to 6 months)	9	1672	Ratio of means (IV, Random, 95% CI)	1.37 [1.11, 1.71]
1.2.2 At completion of intervention (long duration, 2 years)	1	1299	Ratio of means (IV, Random, 95% CI)	0.91 [0.63, 1.31]
1.3 Sedentary behaviour (min/d)	3		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.3.1 At completion of intervention (medium duration, 3 to 6 months)	2	227	Mean Difference (IV, Random, 95% CI)	‐44.09 [‐99.51, 11.33]
1.3.2 At completion of intervention (long duration, 2 years)	1	1354	Mean Difference (IV, Random, 95% CI)	16.03 [‐61.11, 93.18]
1.3.3 Follow‐up after completion (medium, 6 months)	1	172	Mean Difference (IV, Random, 95% CI)	‐33.00 [‐84.28, 18.28]
1.4 CVD risk factor: body mass index (BMI; kg/m²)	8		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.4.1 At completion of intervention (medium duration, 10 weeks to 6 months)	7	751	Mean Difference (IV, Random, 95% CI)	‐0.67 [‐1.28, ‐0.06]
1.4.2 At completion of intervention (long duration, 2 years)	1	1405	Mean Difference (IV, Random, 95% CI)	0.67 [‐0.17, 1.51]
1.4.3 Follow‐up after completion (medium, 3 to 6 months)	3	486	Mean Difference (IV, Random, 95% CI)	‐0.64 [‐1.45, 0.17]
1.5 CVD risk factor: waist circumference (cm)	7		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.5.1 At completion of intervention (medium duration, 10 weeks to 6 months)	7	694	Mean Difference (IV, Random, 95% CI)	‐1.67 [‐3.85, 0.51]
1.5.2 Follow‐up after completion (3 to 6 months)	3	439	Mean Difference (IV, Random, 95% CI)	‐0.77 [‐1.91, 0.37]
1.6 CVD risk factor: systolic blood pressure (mmHg)	6		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.6.1 At completion of intervention (medium duration, 10 weeks to 6 months)	6	571	Mean Difference (IV, Random, 95% CI)	‐1.56 [‐3.66, 0.53]
1.6.2 Follow‐up after completion (medium, 3 to 6 months)	2	315	Mean Difference (IV, Random, 95% CI)	‐0.08 [‐3.26, 3.11]
1.7 CVD risk factor: diastolic blood pressure (mmHg)	6		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.7.1 At completion of interventions (medium duration, 10 weeks to 6 months)	6	571	Mean Difference (IV, Random, 95% CI)	‐0.71 [‐2.57, 1.14]
1.7.2 Follow‐up after completion (medium, 3 to 6 months)	2	315	Mean Difference (IV, Random, 95% CI)	0.70 [‐1.34, 2.73]
1.8 CVD risk factor: cholesterol: LDL	2		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.8.1 At completion of intervention (medium duration, 3 to 4 months)	2	127	Mean Difference (IV, Random, 95% CI)	‐3.58 [‐10.76, 3.59]
1.9 CVD risk factor: cholesterol: HDL	2		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.9.1 At completion of intervention (medium duration, 3 to 4 months)	2	127	Mean Difference (IV, Random, 95% CI)	‐0.94 [‐3.77, 1.89]
1.10 CVD risk factor: triglycerides	2		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.10.1 At completion of intervention (medium duration, 3 to 4 months)	2	127	Mean Difference (IV, Random, 95% CI)	‐10.83 [‐33.84, 12.18]
1.11 Quality of life: mental health component	2		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.11.1 At completion of intervention (medium duration, 3 months)	2	168	Mean Difference (IV, Random, 95% CI)	3.18 [‐1.32, 7.68]
1.11.2 Follow‐up after completion (medium, 3 months)	1	58	Mean Difference (IV, Random, 95% CI)	1.30 [‐1.80, 4.40]
1.12 Quality of life: physical health component	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only
1.12.1 At completion of intervention (medium duration, 3 to 6 months)	2	310	Std. Mean Difference (IV, Random, 95% CI)	0.21 [‐0.29, 0.70]
1.12.2 Follow‐up after completion (medium, 3 to 6 months)	2	310	Std. Mean Difference (IV, Random, 95% CI)	0.08 [‐0.14, 0.30]
1.13 Adverse effects	2		Odds Ratio (IV, Random, 95% CI)	Subtotals only
1.13.1 At completion of intervention (medium duration, 6 months)	1	180	Odds Ratio (IV, Random, 95% CI)	0.87 [0.39, 1.94]
1.13.2 Follow‐up after completion (6 to 9 months)	2	286	Odds Ratio (IV, Random, 95% CI)	0.50 [0.30, 0.84]

Comparison 2. Pedometer intervention vs alternative physical activity intervention.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Physical activity: combined	4		Ratio of means (IV, Random, 95% CI)	Subtotals only
2.1.1 At completion of intervention (short duration, 1 week)	1	60	Ratio of means (IV, Random, 95% CI)	1.26 [0.96, 1.66]
2.1.2 At completion of intervention (medium duration, 3 to 6 months)	3	273	Ratio of means (IV, Random, 95% CI)	1.04 [0.92, 1.17]
2.1.3 Follow‐up after completion (medium, 3 months)	1	143	Ratio of means (IV, Random, 95% CI)	1.07 [0.97, 1.18]
2.2 CVD risk factor: body mass index (BMI; kg/m²)	2		Mean Difference (IV, Random, 95% CI)	Subtotals only
2.2.1 At completion of intervention (medium duration, 3 to 6 months)	2	144	Mean Difference (IV, Random, 95% CI)	‐1.23 [‐2.85, 0.39]

Comparison 3. Subgroup analyses.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 Phyical activity: subgroup analysis by participant risk profile of Analysis 1.1 (follow‐up after completion)	5		Ratio of means (IV, Random, 95% CI)	Subtotals only
3.1.1 Only sedentary participants	3		Ratio of means (IV, Random, 95% CI)	1.55 [0.86, 2.79]
3.1.2 Open participation	2		Ratio of means (IV, Random, 95% CI)	1.03 [0.88, 1.21]
3.2 Physical activity: subgroup analysis by participant risk profile of Analysis 1.2.1	9		Ratio of means (IV, Random, 95% CI)	1.37 [1.11, 1.71]
3.2.1 Only high‐risk participants	3		Ratio of means (IV, Random, 95% CI)	1.20 [1.02, 1.41]
3.2.2 Only sedentary participants	3		Ratio of means (IV, Random, 95% CI)	1.49 [0.87, 2.56]
3.2.3 Open participation	3		Ratio of means (IV, Random, 95% CI)	1.32 [1.12, 1.55]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Aittasalo 2012.

Study characteristics
Methods	Randomised controlled trial Aim: to evaluate a 6‐month intervention to promote walking in office workers using pedometers and email messages
Participants	Population description: insufficiently physically active employees at 20 office‐based work sites (specifics not described) Intervention group: 123 participants Control group: 118 participants Location: Southern Finland Inclusion criteria: respondents to the baseline questionnaire were eligible if they volunteered for the study and were insufficiently physically active for cardiorespiratory health (less than 150 minutes of moderate‐intensity physical activity or less than 75 minutes of vigorous‐intensity physical activity per week accumulated from fewer than 3 days a week) and perceived no restrictions for physical activity Recruitment: 10 occupational healthcare units (OHCs) recruited 20 work sites for which they provide services. Baseline questionnaires were circulated to all 2230 employees. 646 responded, of whom 241 met the eligibility criteria Demographics: age: mean (SD) for control 45.3 (9.1), for intervention 44.1 (9.4) years Gender: control 44% male, intervention 29% male Highest level of education: control basic 9%, polytechnic or vocational school 64%, university degree 27%, intervention basic 6%, polytechnic or vocational school 64%, university degree 30%
Interventions	Duration: 6 months Intervention: the STEP programme consisted of 2 phases Pre‐intervention phase: a 1‐hour preliminary meeting at each work site held by a researcher and providing information on the intervention as well as on health benefits and recommendations for physical activity (PA) and walking. The use of stairs was emphasised from the aspect of health and easy applicability. Employees were supplied with walking leaflets, pedometers (Omron, Walking Style II), and printed logbooks. Employees were instructed to assess their average daily steps with the pedometer over 3 days including 2 working and 1 non‐working day. The average number of daily steps was then used as the baseline for further step goals, rather than using a general goal such as 10,000 steps per day Intervention phase: 1. Self‐reporting Self‐monitoring of physical activity with pedometer and logbook, which included prompts for further step goals 2. Emails Monthly emails using the Health Action Process Approach model were sent from the OHC, beginning within 2 weeks of the preliminary meeting. The first email, for example, included positive outcome expectations, pre‐action self‐efficacy and action planning, focusing on the benefits of integrating short bouts of physical activity into daily routine, and encouraging employees to add 2000 steps to their baseline on 2 days of the week. Employees were advised to choose the particular days of the week, and the remaining 5 days had a goal of the baseline step count. Full details of the messaging in each of the 6 emails is described in Aittasalo 2012. The ultimate goal was to add 4000 steps approximating 30 minutes of moderate‐intensity walking to the daily baseline on 5 self‐selected days of the week. The goal was approached progressively until the fifth email by setting smaller goals of increase in a sequence of 2000 steps Control: participated in data collection only until after the 12 months' follow‐up measurement, when a 1‐hour seminar was offered at each work site to provide feedback to all employees and to supply control participants with pedometers, logbooks, and walking leaflets. As a gesture of appreciation, participating OHCs were also provided point‐of‐choice stair posters and were offered a 2‐hour training session on physical activity and health for health personnel
Outcomes	Physical activity Adapted from the International Physical Activity Questionnaire (IPAQ Long) for a usual week (questions on walking from IPAQ, questions on activity at work and during transportation limited to walking, and additional question on stairs added). An employer‐representative named at each work site delivered the baseline questionnaires to employees, who completed and returned them by mail to the research institute. Final measures were walking at work, walking for transportation, walking stairs, walking for leisure, and total walking, reported as minutes/week and number of people who walked in each category. Vigorous‐ and moderate‐intensity leisure physical activity during leisure time was measured but not reported Sedentary behaviour Sitting during working and non‐working day adopted from the IPAQ Long. An employer‐representative named at each work site delivered the baseline questionnaires to employees, who completed and returned them by mail to the research institute. Reported as minutes/d Adverse effects Self‐reported responses to the question: "Has physical activity caused you any acute or chronic pains, injuries or health hazards during the past six months?" Response alternatives were "No" and "Yes, what kind? _____" Other outcomes not reported in this review RE_AIM factors (Reach, effectiveness, adoption, implementation, maintenance) and costs were measured using questionnaires, process evaluation, and interviews
Statistical analysis	Imputation of missing data: not reported by study authors Sample size calculation: "according to the power calculations (significance level of 0.05, power of 80%), 175 participants in each group totaling 350 participants were needed to detect the 30% between‐group difference in change in the weekly minutes of total walking" Statistical analysis: self‐reported data on different types of walking were analysed by using generalised linear mixed models (GLMMs). Due to the excess number of zeros (no walking) in the data, analysis was performed in 2 parts as suggested by Tooze et al [33]: 1. Logistic regression To analyse the between‐group difference in change in the probability (odds ratio ‐ OR) of walking from baseline to 2‐, 6‐, and 12‐month follow‐up. Data were coded dichotomously into zeros 2. Linear model To estimate between‐group changes in non‐zero responses. In this model, measurements indicated repetitions with the assumption that the correlations between repeated measurements were constant (compound symmetry). As most of the distributions were skewed, logarithm transformation was performed and geometric mean ratio (GMR) was used as an indicator of group differences. Work site was included in the model as a random effect (intercept), along with gender, taking care of children younger than 18 years of age (yes.no), age (continuous), and self‐reported body mass index at baseline (continuous). Random errors were assumed to be independent between the 2 model components
Notes	Funding sources include the Finnish Work Environment Fund and the Juho Vainio Foundation for financial support
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Eligible respondents at each work site were then allocated equally to either a pedometer (STEP) or a comparison group (COMP) using computer‐generated randomisation lists (stratified randomisations with random allocation sequences to ensure closely balanced groups within each work site)"
Allocation concealment (selection bias)	Unclear risk	Sites were recruited, employees were invited to participate, and baseline data collection was completed before randomisation was conducted. Eligibility was determined by defined criteria measured through the baseline survey (current levels of physical activity and any restrictions preventing physical activity); however no information was provided on whether study personnel or site staff could have known or altered the recruitment or allocation of individual participants before individual participants were enrolled or received the intervention
Blinding of participants and personnel (performance bias) All outcomes	High risk	Interventions included meetings, emails, and physical activity interventions, and participants were not blinded. Intervention participants were asked to not discuss the intervention with colleagues, as control participants were present in the same workplace. No blinding of other personnel was discussed. Personnel based in the workplace would likely have been aware of the identify of participants due to baseline meeting and role in sending motivational emails. Participants and personnel were aware of their allocation and could easily have changed behaviour in relation to physical activity
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	High risk	Self‐reported via questionnaire. Participants were aware of the expectation that intervention would increase exercise levels, and of specific targets identified. Both groups were likely to exaggerate levels of exercise, but intervention groups were more likely to be affected. Participants self‐reporting exercise data were likely to be influenced by knowledge of expectations of the intervention
Blinding of outcome assessment (detection bias) Disease risk factors	Low risk	BMI was objectively measured and was unlikely to be affected by lack of blinding
Blinding of outcome assessment (detection bias) Adverse events	High risk	Self‐reported via questionnaire. Participants' recall and attribution of adverse events to physical activity were likely to be influenced by knowledge of the intervention
Incomplete outcome data (attrition bias) Adverse events	High risk	Intervention: randomised: 123. Analysed at 6 months: 86, at 12 months: 89 Control: randomised: 118. Analysed at 6 months: 94, at 12 months: 87 Sensitivity analysis based on plausible assumptions about the data indicated that the effect estimate is very susceptible to change
Incomplete outcome data (attrition bias) All other outcomes	High risk	Intervention: analysed for walking time at 6 months: 86, at 12 months: 84. Analysed for sedentary time at 6 and 12 months: 87 Control: analysed for walking time at 6 months: 94, at 12 months: 80. Analysed for sedentary time at 6 months: 94, at 12 months 85. Sensitivity analysis based on plausible assumptions about the data indicated that effect estimates are very susceptible to change
Selective reporting (reporting bias)	Unclear risk	Study protocol could not be obtained. Although some planned outcome measures were not reported (study authors advised that quality of life was planned but was not assessed), multiple measures for some outcomes were reported and did not appear to have been selected for statistical significance
Other bias	Low risk	None was identified

Audrey 2015.

Study characteristics
Methods	Cluster‐randomised controlled trial Aim: to develop a pilot employer‐led scheme to increase walking to work, and to test the feasibility of implementing and evaluating the intervention in a full‐scale trial
Participants	Population description: employees who were professional, scientific, and technical; manufacturing; transportation; education; accommodation and food services; public administration; financial and insurance activities Randomisation: workplaces were paired with another as similar as possible with respect to total number of employees (up to 50, 51 to 250, 250+), location characteristics, and type of business. To allow for seasonality, data collections took place concurrently in each ‘matched’ pair of workplaces Intervention group: 7 workplaces (4 small; 3 medium; 2 large), 100 participants Control group: 10 workplaces (5 small; 3 medium; 2 large), 87 participants Location: Bristol, England Inclusion criteria: employees who were willing to incorporate some walking into their daily commute, with initial recruitment targeted at employees living within 2 miles of the workplace Participants were excluded if they cycled or used ‘other’ modes of travel to work Recruitment: workplaces were approached through the Bristol Chamber of Commerce via their mailing list of employers and a publicly available list of major employers. Initial contact included an information leaflet about the study and expectations, with contact details for the research team. Paper copies were also sent, with a reply‐paid envelope for initial expressions of interest, including willingness to allocate employee time for study activities. Those that did express an interest were asked to complete a short questionnaire about the size and type of business, and to identify how many of their employees lived within 2 miles of the workplace. This process was aided by the research team supplying the first 4 digits of postcodes likely to contain employees living within the required range and an instruction leaflet on how to calculate distance using Walkit.com. Workplaces that provided this information were recruited to the study and were ‘matched’ into pairs, with each pair containing workplaces as similar as possible with respect to total number of employees (up to 50, 51 to 250, 250+), location characteristics, and type of business Participants were contacted by e‐mail or by letter, as appropriate to the workplace Demographics: age: overall mean (SD) 37.8 (12) (range 17.3 to 67 years). Control mean (SD) 36.8 (12.4), intervention 38.7 (11.7) years Gender: control 52.9% male, intervention 43.0% male Highest level of education: control missing 10.3%, other 0%, no formal qualifications 0%, GCSE grades A to C or equivalent 9.2%, BTEC national or equivalent 5.8%, GCE A level NVQ level 3 8.0%, BTEC higher or equivalent 4.6%, degree NVQ level 4 39.1%, PhD masters degree NVQ level 5 23.0%; intervention missing 12.0%, other 1.0%, no formal qualifications 2.0%, GCSE grades A to C or equivalent 9.0%, BTEC national or equivalent 2.0%, GCE A level NVQ level 3 13.0%, BTEC higher or equivalent 3.0%, degree NVQ level 4 41.0%, PhD masters degree NVQ level 5 17.0% Ethnicity: control: white British 74.7%, white other 12.6%, mixed ethnic group 1.1%, Asian or Asian British 3.4%, Chinese 0%, missing 8.1%; intervention: white British 79.0%, white other 8.0%, mixed ethnic group 1.0%, Asian or Asian British 1.0%, Chinese 1.0%, missing 10.0% Income: control: < £10,000 6.9%, £ 10,000 to 20,000 8.1%, £20,000 to 30,000 8.1%, £30,000 to 40,000 16.1%, £40,000 to 50,000 12.6%; > £50,000 27.6%, don't know 10.3%, missing 10.3%; intervention: < £10,000 0%, £10,000 to 20,000 9.0%, £20,000 to 30,000 15.0%, £30,000 to 40,000 15.0%, £40,000 to 50,000 13.0%, > £50,000 26.0%, don't know 6.0%, missing 16.0% Occupation activity level: control: sedentary 65.5%, standing 23.0%, manual 3.5%, missing 8.0%; intervention: sedentary 83.0%, standing 4.0%, manual 3.0%, missing 10.0%
Interventions	The Walk to Work intervention was designed to encourage employers and employees to consider and address facilitators and barriers at each level of the socio‐ecological model, with 9 targeted behaviour change techniques based on taxonomy of multiple frameworks Duration: 10 weeks Intervention: 1. Organisational action Promoters: 10 Walk to Work promoters (‘champions’), volunteers, or nominated by participating employers were trained by the research team about the health, social, economic, and environmental benefits of walking to work and how to identify and promote safe walking routes for employees (half‐day training at the University of Bristol or individual sessions for those who could not attend). Their roles were to encourage participating employees in their workplace to walk to work; to help to identify walking routes; to encourage goal‐setting; and to be ‘role models’ and the recognised ‘points of contact’ for the Walk to Work intervention at their place of work. The aim was a maximum of 25 participants to each Walk to Work promoter Participating employees were contacted by the Walk to Work promoter and were given a Walk to Work pack including an information booklet, travel diary, and pedometer (pedometer brand and type not specified). They were also encouraged to complete diary sheets and to record whether or not they had walked and, for those using pedometers, how many steps had been registered 2. Goal‐setting Short‐, intermediate‐, and long‐term goals for incorporating walking into the journey to and from work set by participants with the support of the promoters 3. Encouragement/Extrinsic motivation Further encouragement was provided through 4 contacts from the Walk to Work promoter over 10 weeks (face‐to‐face, email, or telephone as appropriate to the workplace size, resources, and work routines). Walk to Work promoters were prompted and were encouraged in their role by 4 email/telephone contacts from the research team during this time. (Further details are available in the published paper) Control: completion of data collection rounds only
Outcomes	Physical activity Overall physical activity (measured in counts per minute (cpm)) MVPA Temporal pattern of physical activity (when activity has increased any compensatory decrease) Physical activity related to the commute to work Adverse events Self‐reported commute‐related adverse events (questionnaire), incidents, or accidents resulting from walking to work (e.g. aches and pains, blisters, road traffic incidents, street crime, anti‐social behaviour). Defined as any unexpected effect or untoward event affecting a participant in the study, including non‐serious adverse events and a separate category of serious adverse events (resulting in death of the participant, life‐threatening, requiring inpatient hospitalisation or possibly jeopardising the participant, resulting in persistent or significant disability/incapacity) Other outcomes not reported in this review Self‐reported health service use for specific adverse events (questionnaire) – reported total health service use only Self‐reported total health service use in the last 4 weeks (questionnaire) and cost incurred for using the health service (estimated by the research team) (GBP £)
Statistical analysis	Imputation of missing data: not reported by study authors Interview transcripts: transcripts were read and reread, and textual data were placed in charts focusing on key research questions. Charts were scrutinised, and each unit of data was coded Sample size calculation: this was a pilot study with the primary aim of testing feasibility, and it was not powered to detect differences in outcomes
Notes	The project was funded by the National Institute for Health Research (NIHR) Public Health Research programme (project number 10/3001/04) The work was undertaken with the support of The Centre for the Development and Evaluation of Complex Interventions for Public Health Improvement (DECIPHer), a UK Clinical Research Collaboration Public Health Research Centre of Excellence. Joint funding (MR/KO232331/1) from the British Heart Foundation, Cancer Research UK, Economic and Social Research Council, Medical Research Council, the Welsh Government, and the Wellcome Trust, under the auspices of the UK Clinical Research Collaboration, is gratefully acknowledged
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Randomised controlled trial". “Assignment of workplaces to the intervention group occurred within these matched pairs and employed computer‐generated allocation"
Allocation concealment (selection bias)	Low risk	Assignment was performed “by an independent member of the co‐applicant group”
Blinding of participants and personnel (performance bias) All outcomes	High risk	No information was provided and wait‐list for control was not mentioned. Unlikely to be blinded at the cluster level. Individuals within the clusters may not have been aware of differences in the intervention. Researchers identified the likelihood of a ‘Hawthorne effect’, increasing physical activity due to knowledge of the intervention, which they felt would be present in both groups, and was likely to reduce the observed difference between groups
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	Low risk	No information was provided; however physical activity was recorded objectively by an accelerometer
Blinding of outcome assessment (detection bias) Adverse events	Low risk	Self‐reported by participants. Intervention group may have been more likely to report adverse events, but no events were reported in either group
Incomplete outcome data (attrition bias) Adverse events	High risk	Intervention group: 9 workplaces were randomised, 2 withdrew before baseline data collection (1 due to time constraints, 1 due to "ample parking"). 100 individuals were recruited. 27 withdrew before 1‐year follow‐up (8 left workplace, 5 maternity leave, 1 retired, 10 declined, 3 sick) Participants remaining at 1‐year follow‐up = 73 Control group: 10 workplaces randomised. 1 eliminated after sole participant retired 87 individuals were recruited. 28 withdrew before 1‐year follow‐up (15 left workplace, 1 retired, 10 declined, 2 sick) Participants remaining at 1‐year follow‐up = 59 No events were observed, so effect estimate was not robust to small quantities of missing data
Incomplete outcome data (attrition bias) All other outcomes	High risk	Intervention group: as above Responses for physical activity = 71 Control group: as above Responses for physical activity = 57 Sensitivity analysis based on plausible assumptions about the data indicated that effect estimates were very susceptible to change
Selective reporting (reporting bias)	Low risk	Study protocol was published and all relevant outcomes and analyses were reported as planned
Other bias	High risk	As this is a cluster‐randomised trial, and recruitment of participants occurred after randomisation, it is possible that knowledge of the intervention affected the types of participants recruited. Given the substantive difference in numbers recruited, this could mean only that more self‐motivated participants joined the control arms

Carr 2013.

Study characteristics
Methods	Randomised controlled trial Aim: to test the efficacy of a multi‐component technology intervention for reducing daily sedentary time and improving cardiometabolic disease risk among sedentary, overweight university employees
Participants	Population description: middle‐aged, primarily female, physically inactive, overweight but without major medical problems, and employed in sedentary jobs at a large university Intervention group: 25 participants randomised Control group: 24 participants randomised Location: southeastern USA Inclusion criteria: healthy but physically inactive (self‐reporting less than 60 minutes of moderate‐ to vigorous‐intensity physical activity per week), overweight (body mass index (BMI) ≥ 25 kg/m²), working in full‐time (reporting a minimum of 35+ hours/week) at sedentary/desk‐dependent occupations (reporting a minimum of 75% of working time spent sitting). Participants were required to gain permission from their supervisor prior to enrolment. Participants were excluded if they had (1) limitations with or contraindications to ambulatory exercise; (2) acute illness or injury; (3) cognitive impairment, psychosis, or other diagnosed psychological illness (with the exception of depression and anxiety); (4) currently using psychotropic drugs; or (5) with diagnosis of a chronic condition such as heart failure or cancer Recruitment: via email advertisements placed on an electronic mailing list to 5392 employees Demographics: age: mean (SD) control 47.6 (9.9) years, intervention 42.6 (8.9) years Gender: control 5.9% male, intervention 13.1% male Education: college graduate: control 71.0%, intervention 86.0% Income > $40 000: control 62.5%, intervention 63.6% Job category: control professional/executive 35.0%, administrative 65.0%; intervention professional/executive 52.0%, administrative 48.0% Ethnicity: non‐Hispanic white: control 76.5%, intervention 63.6%
Interventions	Duration: 12 weeks Intervention: 1. Machine Access to a portable‐pedal machine (MagneTrainer, 3D Innovations, Greely, Colorado, USA) at the work site. The pedal machine is a portable (18" height, 20" duration) device that has been demonstrated as acceptable for use during sedentary office work, allowing employees to engage in light‐intensity activity (active sitting) for long periods throughout the day without causing them to perspire. The machine is accompanied by a PC interface and a software package that allows for objective monitoring of individual pedal activity. This software also provides the user with real‐time feedback through a display monitor on pedal time, distance, speed, and caloric expenditure. The research team delivered the pedal machine to each participant’s work site, downloaded the pedal tracking software to the participant’s work computer, and worked with the participant to identify the most feasible set‐up. Participants were asked to keep the pedal machine connected to their PC during all working hours. Participants were required to gain clearance to use the pedal machines and software at their work prior to participation. No additional interaction between research staff and participant’s supervisors occurred during the course of the study 2. Website Access to a motivational website (Walker Tracker, Portland, Oregon, USA) to receive tips and reminders focused on reducing sedentary behaviours throughout the day. The website was individually customised to the local culture of the work site of participants who were recruited (e.g. through posting local images and messages specific to the local institution), and participants were able to post profile photos, post status updates on a news feed, and send messages to members of their small groups further fostering social support, on a forum similar to Facebook 3. Pedometer use A pedometer to use in conjunction with the website (Omron HJ‐150) A virtual competition (aimed at building social support) in which small groups of intervention participants (4 to 5 per group) collectively travelled across the USA Specific goals were not set for intervention participants; rather, participants received advice on how to set goals and suggestions for daily pedaling time 4. Email Participants were emailed 3 motivational messages each week focused on reducing time spent sedentary. Messages were theory‐based, targeting constructs of the social‐cognitive theory, including self‐monitoring, social support, self‐efficacy, and perceived environment. Example messages included, "Let's try to pedal an extra 5 min during your lunch break today" and "Did you know standing up burns more calories than sitting? Maybe it's time for a break!" Most messages targeted time spent at work, although some messages broadly targeted sedentary time in general and could have impacted sedentary time outside work Participants were prompted through daily email messages to self‐monitor their daily pedal time and daily steps (through a pedometer) on the website Cost of the intervention was $180 (pedal machine and software, pedometer, access to website) per participant Control: Participants randomised to the wait‐list control group were asked to maintain their current behaviours for 12 weeks, at which time they were given the option to receive the intervention treatment materials Participants were not compensated for participation in the study
Outcomes	Physical activity Estimated aerobic fitness collected as a secondary outcome (sedentary behaviour was primary): a single‐stage submaximal treadmill walking test to measure VO₂ (mL/kg/min). Categorised as sedentary (0 steps/min), light (1 to 45 steps/min), moderate (46 to 75 steps/min), and vigorous (76+ steps/min) intensity. Note: this study was not powered to detect differences in measured cardiometabolic risk factors Sedentary behaviour Daily sedentary time, objectively measured by the StepWatch physical activity monitor (Orthocare‐Innocations, Mountlake‐Terrace, Washington, USA). All wakeful hours for 7 consecutive days and keeping track of wear time were recorded in an activity log. Days on which participants wore the monitors for less than 10 hours were excluded from the final analysis. Measured in minutes sedentary (min/d) and percentage of time sedentary (e.g. for light, moderate, and vigorous) Cardiovascuar disease and type 2 diabetes risk factors Body mass index (BMI). Measured to the nearest 0.1 kg and height to the nearest 0.1 cm using a professional grade digital medical scale and height rod (Seca 769, Hanover, Maryland, USA) Waist circumference. Measured in duplicate with a standard Gulick measuring tape according to standard procedures Blood pressure and heart rate. Measured with a stethoscope and a sphygmomanometer using standard techniques. Heart rate was monitored with a Polar heart rate monitor and chest strap Fasting blood lipoprotein. Low‐density lipoprotein (LDL) cholesterol, high‐density lipoprotein (HDL) cholesterol, and triglycerides were assessed using a fingerstick and point‐of‐care analysis (Cholestech LDX analyser) that has been shown to be an accurate and precise measure of HDL cholesterol (‐2.74% and 1.05%, respectively) and triglycerides (2.11% and 2.65%, respectively) Other outcomes not reported in this review Fasting blood lipoprotein. Total cholesterol Estimated aerobic fitness. A single‐stage submaximal treadmill walking test. Unit of measurement is VO₂ (mL/kg/min)
Statistical analysis	Imputation of missing data: not reported by study authors Sample size calculation: a sample size of 40 was necessary to detect, with 80% power, a 30 min/d difference in daily sedentary time. This study was not powered to detect differences in measured cardiometabolic risk factors. These measures were collected as secondary outcomes and were used to inform future trials Statistical analysis: a paired samples t‐test was used to determine any within‐group differences at baseline and post intervention. Analysis of covariance (ANCOVA) was used to test for differences between groups post intervention. Baseline values of interest were included as covariates in the model for all continuous variables, consistent with the recommended statistical procedures. The underlying assumption of no between‐group differences at baseline was confirmed for all measures by 1‐way ANCOVA
Notes	Funded by Oak Ridge Associated Universities grant #212112. Study authors declared no other conflicts of interest
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"A 1:1 random allocation sequence was generated by the principal investigator using an online random sequence generator"
Allocation concealment (selection bias)	Low risk	"Participants were assigned to one of two groups by a research staff member not involved in data collection based on the order in which they enrolled into the study"
Blinding of participants and personnel (performance bias) All outcomes	High risk	Intervention involved physical activity supported by a desk pedaling device and pedometer, compared to a wait‐list control who was asked to maintain usual behaviour. Although "Participants were located in 18 different buildings across the campus. No participants worked within visible proximity of each other", participants would have been aware of their own group allocation. Trial staff would have been aware of allocation through active communication by regular email intervention. Participants' supervisors were required to give permission to participate in the study and would have been aware. Physical activity behaviours could easily have been modified in response to participation in the trial. The control group would be more likely to modify behaviours in response to participation in the trial, despite being asked to maintain existing behaviours, whereas intervention group participants' responses would be considered a response to the intervention
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	Low risk	Physical activity was measured objectively by a computerised device, the Stepwatch physical activity monitor
Blinding of outcome assessment (detection bias) Disease risk factors	Low risk	"All measures were collected at baseline and postintervention (12 weeks) in a controlled laboratory setting by two staff members blinded to the participant’s group assignment. The two staff members were provided specific measurement duties to ensure that each measure was collected by the same staff member at baseline and postintervention." Measures were objective, although outcome assessors may have been unblinded by participants
Incomplete outcome data (attrition bias) All other outcomes	Low risk	9 participants were missing (2 of 25 from the intervention group, 7 of 24 from the control group), and the reason given for all was 'did not complete' All outcomes appeared robust to sensitivity analysis based on plausible assumptions about missing data
Selective reporting (reporting bias)	Low risk	Trial registration information was obtained from ClinicalTrials.gov. Although primary outcome appears to have changed from total physical activity to measures divided into intensity level, and heart rate was a planned outcome but was not reported, multiple measures are reported and none are statistically significant
Other bias	Low risk	None was identified

Dishman 2009.

Study characteristics
Methods	Cluster‐randomised controlled trial Aim: to evaluate the efficacy of Move to Improve, a socio–ecological intervention delivered at the workplace to increase leisure time physical activity
Participants	Population description: employees without major medical problems from Home Depot, Inc., based at divisional offices, subsidiaries, call centres, and distribution centres, none of which dealt directly with customers Intervention group: 8 work sites, 885 participants (mean cluster size 111, range 49 to 387) Control group: 8 work sites, 557 participants (mean cluster size 70, range 42 to 126) Location: USA (Arizona, California, Colorado, Florida, Georgia, Illinois, Louisiana, Maryland, Texas) and Canada (Toronto) Inclusion criteria: employees without overt cardiovascular, pulmonary, or metabolic disease Recruitment: 20 sites in diverse regions were identified as eligible for the study because they could be paired on number of employees and nature of work. Sixteen work sites agreed to participate, were paired, and were randomly assigned. Recruitment of volunteers within each work site was performed via e‐messages, onsite flyers, interoffice mail, face‐to‐face meetings, and posters developed and delivered by site co‐ordinators, who recruited and supervised team captains. All employees who completed baseline testing received an incentive (e.g. t‐shirt, lunch cooler) Demographics: Age: range 19 to 64 years, mean (SD) 36.2 (9.8) years Gender: 31% male Ethnicity: white (60%), black (25%), Asian (3%), Pacific Islander or Native American (1%), or other (11%); 7% identified themselves as Hispanic or Latino Highest level of education: high school graduate (9%), some college or technical training (34%), associate degree (12%), bachelor degree (31%), postgraduate work or degree (14%). Job title: non–manager/supervisor (45%), supervisor (8%), manager/senior manager/director (12%), other (35%)
Interventions	A collaborative effort with the Building Better Health (BBH) programme, a pre‐existing health promotion intervention operating at approximately 1700 Home Depot locations Duration: 2‐month pre‐intervention phase, followed by a 12‐week intervention Intervention Pre‐intervention phase: project staff consulted with senior management at each work site to discuss project objectives and to review site selection criteria and expectations for participation. An employee was selected as a site co‐ordinator during the first month of installation Intervention phase: adapted from the Director’s 50th Anniversary Physical Activity Challenge implemented at the CDC in Atlanta 1. Organisational action Senior management endorsement received at the beginning of the project. Middle managers were encouraged to support employee participation Joint employee–management steering committees established at the sites, consisting of 8 to 10 employees, including a site co‐ordinator and a representative from each major participating work unit. The committee met regularly during both phases and was responsible for implementation of the intervention components Site co‐ordinators: received orientation, project requirement training, a handbook that served as the implementation manual, and weekly contact with project staff throughout both phases 2. Goal‐setting Individual and team goals were self‐set, specifically regarding performance and time, and were challenging but realistic and attainable and were easily assessed Participant handbook: detailed components, benefits, and incentives; participant responsibilities; and timing of the intervention. It contained 6 sequential, bi‐weekly tools to guide and assist the participant through the intervention: (1) goal‐setting, (2) overcoming obstacles, (3) sedentary temptations, (4) avoiding relapse, (5) staying motivated, and (6) keeping on moving. Each tool was a practical application of behaviour modification principles built around goal‐setting theory Personal goals: graduated increases in accumulation of 10‐minute blocks of moderate to vigorous physical activity (MVPA) and pedometer (style Yamax SW 200) steps each week, evaluated and adjusted weekly. Targeted toward meeting or exceeding established public health recommendations for physical activity: ≥ 150 minutes each week of MVPA and/or ≥ 10,000 pedometer steps each day Team goals: employees formed teams, usually based on organisational and work group structures. Team size ranged from 5 to 20 members with a mean of 9 members. Team captains were responsible for motivating participants to set goals and earn points for their team, while serving as liaisons between participants and site co‐ordinators. Team captains collected worksheets including outcome measures and revised goals every 2 weeks. Posters that recorded and compared team goal attainment were displayed in break rooms and were updated every 2 weeks by the site co‐ordinator Organisational goals: established by the steering committee at each work site. Participation objective at each work site was 50% of all employees. Goal attainment objective was that 75% of participating employees would accumulate 150 minutes of MVPA and 10,000 steps per day, or both, at least 9 of the 12 weeks during the intervention 3. Incentives Participants received a lunch bag if they completed bi‐weekly goal‐setting and assessments until the 6‐week mid‐point and an intervention t‐shirt if they did so through all 12 weeks Each member of every team that had 75% of its members reach the goal attainment target received an embroidered “winning logo” t‐shirt as an incentive Team captains received another incentive if their teams met this goal  Site co‐ordinators received incentives based on site participation, and 1 site received a recognition plaque and a free catered lunch for employees having the greatest participation 4. Environmental prompts Signage that encouraged physical activity and its health benefits, emphasised target goals for minutes and steps, and illustrated opportunities to be active, such as parking and walking, taking walk breaks, and climbing stairs Posted throughout the work site in places with high employee traffic (e.g. break rooms, bathrooms, points of decision such as elevators and stair wells) Changed bi‐weekly to vary messages within the same themes Control: usual treatment control condition, including completion of the CDC health‐risk appraisal and monthly newsletters describing the health benefits of physical activity. This provided a minimal treatment comparator for the intervention that has been shown to have modest or no effects on physical activity. Control sites had a programme director who dispensed monthly educational messages after baseline data collection
Outcomes	Physical activity International Physical Activity Questionnaire (IPAQ) short form. Hourly participation each week in activities rated according to multiples of metabolic equivalent of task (MET) units. MET is a measure of energy consumption, and 1 MET is equal to the energy produced at a standard resting metabolic rate attained during quiet sitting (Ainsworth 2000). Can assess frequency and duration of moderate (≥ 4 METs) and vigorous (≥ 8 METs) physical activity and walking. Reliability and criterion validity judged against accelerometry were comparable to other self‐report measures Number of people meeting US Healthy People 2010 recommendations for moderate or vigorous physical activity Other outcomes not reported in this review Perceived management support (Likert scale) Employee involvement (Likert scale) Physical activity diary and pedometer (style Yamax SW 200) steps (intervention group only) Satisfaction, confidence, commitment, and intention (1 to 4 scale, intervention group only)
Statistical analysis	Imputation of missing data: latent growth modelling imputation and latent transition analysis were undertaken using full‐information likelihood procedures for selected variables within those returning at follow‐up. Imputation was not undertaken for those lost to follow‐up. The imputed data were not used in this Cochrane Review Sample size calculation: study authors reported that the sample size provided adequate statistical power for latent growth model tests Adjustment for clustering: as there was no substantive difference in models after covariate adjustment using the Huber–White sandwich estimator procedure, unadjusted models were presented. Hence, raw data presented were not adjusted for clustering
Notes	Supported by Health Protection Research Initiative grant DP 000111 from the CDC. Study authors stated no other financial interests
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"16 work sites agreed to participate and were paired and randomly assigned". "Each of the paired sites was randomly assigned... to the intervention or a health education control condition using computer‐generated random numbers. Based on past work site intervention studies, the expected intervention enrolment and retention rates were approximately 35% and 50%, respectively, so recruitment goals at intervention and control sites were set at a 2:1 ratio"
Allocation concealment (selection bias)	Unclear risk	Recruitment of sites occurred before randomisation to enable pairing, although it is unclear if study personnel could have influenced allocation
Blinding of participants and personnel (performance bias) All outcomes	High risk	No information was provided and wait‐list for control group was not mentioned. Although work sites were geographically dispersed and contamination was unlikely, knowledge of participation in the study could have influenced behaviour in the control group or led to changes in the usual health‐related interventions offered at control work sites
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	High risk	All outcomes were self‐reported by participants. Participants receiving the intervention would have been aware of goals set and the intention of the intervention and may have been likely to overestimate outcomes
Incomplete outcome data (attrition bias) All other outcomes	High risk	Intervention: randomised: 885. Analysed for PA: 564 Control: randomised: 557. Analysed for PA: 265. "Percent loss to follow‐up ranged from 25% to 47% at intervention sites and 32% to 56% at control sites." Reasons for loss to follow‐up were not given. Study authors reported that in the control group, participants who were lost to follow‐up had slightly higher baseline physical activity scores than those who remained. This level of missing data places outcomes at high risk of bias
Selective reporting (reporting bias)	Unclear risk	All outcomes described in the published papers were reported. Study protocol was not available
Other bias	High risk	Following cluster‐randomisation, recruitment of individuals at each work site was voluntary. Volunteers were permitted to join the study after allocation and baseline measurement, at which point the nature of the intervention would have been clear. More participants joined the control group (138) than the intervention group (23) after baseline. Volunteers for the intervention may have been more motivated to change than volunteers for the less active control intervention

Linde 2012.

Study characteristics
Methods	Cluster‐randomised controlled trial Aim: to (a) conduct a simultaneous test of environmental approaches to healthy weight choices (including eating, physical activity, and weight‐monitoring behaviours), and (b) extend the test of these approaches to a more sustained time period than is typically observed
Participants	Population description: 6 work sites, including 2 community colleges; a regional insurance office, a beauty industry corporate headquarters with an attached manufacturing and distribution centre, a utility company office, and a national headquarters for a health‐related non‐profit organisation Intervention group: 3 work sites, 752 participants Control group: 3 work sites, 995 participants Location: metropolitan area, USA Inclusion criteria: work sites were eligible if they had between 250 and 1000 employees, within the study location; had food service on site, were located in a building with at least 2 floors (to ensure presence of stairs), with minimal seasonal fluctuation of employees, were expecting stability of location and workforce over the next several years, and were willing to provide employees’ work contact information. Employees were eligible if they were employed at least 50% of the time on‐site during a daytime shift Recruitment: work sites were identified using a business directory and were contacted by phone to ascertain eligibility. A letter was then sent to human resources (HR) personnel (identified in the initial screening call) to provide basic study information and to request contact to further discuss the study and eligibility requirements. The next telephone contact confirmed basic site eligibility criteria and explained the purpose and nature of the study. The principal investigator and the project director then visited sites for 1‐hour visits with site executives and HR staff, to present the study design and aims and field questions. For those sites then agreeing to participate in the trial, HR provided work site contact information (email and telephone) for all eligible employees. Study staff distributed a company‐wide email announcement, via company email address, to all eligible employees. The message described the partnership, gave a brief description of the study, notified employees that a study staff member would contact them within 2 weeks, and gave the option to call study staff in advance to opt out or enrol in the study. Staff followed a protocol of up to 9 contact attempts (by either telephone or email) for each employee during the scheduling process. Following baseline data collection for consenting participants, work sites were paired and randomised to the intervention or control group according to time of study entry Demographics: Age: mean 42.9 years, range 18 to 75 Gender: 37.4% male Marital status: 68.5% married or cohabiting Race: 88.6% white, 2.1% Hispanic
Interventions	The HealthWorks programme. Participants were compensated $10 for completion of each measurement visit and $10 for completion of each survey Duration: 2 years Intervention: intervention components were primarily targeted at making changes at the work site level: Food environment: baseline food data were used to determine targets for healthy food pricing and availability. Benchmarks for calorie smart food presence (i.e. 50% or greater) were determined from prior research and were communicated to work sites by intervention staff, primarily the lead interventionist, who had extensive food service experience, who worked directly with site food managers and vending delivery drivers to facilitate changes. Targets were: To increase the availability of 'calorie smart foods' to at least 50% of all cafeteria and vending machine offerings, as defined based on guidelines from the California School Nutrition Association standards (http://www.calsna.org/) for healthy portion sizes, according to the following categories: entrée (e.g. meat, egg, large soup, vegetarian meat substitute; ≤ 500 calories), side dish (e.g. small soup, bread, potato, rice, vegetable; ≤ 250 calories), snack (e.g. chips, granola bar, yogurt; ≤ 150 calories), beverage (e.g. coffee, dairy drinks, fruit drinks, soda; ≤ 150 calories), desserts (e.g. cookies, pie, cakes, ice cream, sweet rolls; ≤ 200 calories), condiments (e.g. ketchup, salad dressing; ≤ 40 calories), and combination foods (e.g. packaged meals consisting of entrée and side dish, with or without beverage, for a reduced meal price; ≤ 750 calories) To reduce the price of calorie smart foods by 15% while increasing the price of non‐calorie smart foods by 15% To offer smaller portion sizes as substitutes (e.g. 12 oz to replace 20 oz soda) To label and promote calorie smart items at the point of purchase (e.g. table tents in cafeteria, posters near vending machines) Physical activity environment: primary aims of the activity environment intervention were to promote walking at work (via organised group walks, competition between co‐workers, and activity monitoring) and to encourage stair use. Participants were provided with pedometers (pedometer brand and type not specified) and access to a free online step tracking site (http://www.americaonthemove.org) for use throughout the intervention. Up to seven 6‐ to 8‐week walking challenges were implemented, with input from study staff. Employees were grouped into competitive teams (self‐selected or based on work site units) and tracked step counts collected during challenge periods; other walks were staged around charitable giving events or fun activities at work (e.g. games played outdoors while walking). In addition, regular walking was encouraged as a means of meeting activity goals during the workday (e.g. by promotion of walking meetings, by taking time from lunch to walk, by walking before or after shifts). Motivational signs, decorative posters, and music were placed in select stairwells to enhance the stair environment and promote use Body weight tracking environment: weight scales, BMI charts, and weight tracking forms were placed at 4 accessible yet private locations (e.g. restroom, break room) to promote knowledge of healthy weight parameters and self monitoring. Up to 3 weight tracking competitions, framed around maintaining current weight (e.g. during the winter holidays), were held to encourage social support Health media environment: in addition to placement of signs and posters previously mentioned, a 2‐page digital monthly newsletter was distributed via email for 24 months. The first page of the newsletter addressed general information related to healthy eating, activity, or other relevant behaviours, with a section for site‐specific information regarding upcoming events. The second page reported recent site‐specific activities (e.g. competition results, co‐worker testimonials) Advisory panels: 8 to 11 work site employees at each site, including the work site liaison, were instituted as advisory panels and met every second month to provide guidance and ongoing feedback to study staff on planning, implementation, and acceptability of all intervention activities. Efforts were made to ensure that the advisory panel represented a cross‐section of employee classifications and organisational units Control: no contact was made except to engage in evaluation procedures. Following the last round of data collection, control sites were offered a DVD containing intervention materials (e.g. poster templates, newsletter content, descriptions of intervention activity procedures) and an opportunity to ask questions of intervention staff as needed. Although 2 of 3 control sites requested and received materials following study completion, this was not considered a wait‐list control group
Outcomes	Physical activity Physical activity (IPAQ total METs min/week, via online survey) Stair use (infrared beam sensor count) Sedentary behaviour Minutes sitting/d Cardiovascular disease and type 2 diabetes risk factors Body mass index (BMI) Body weight Percentage of participants overweight Adverse events Not defined Other outcomes not reported in this review Blood pressure and cholesterol self‐reported by participants as whether they had ever been told by a health professional that they had high blood pressure or cholesterol, and whether or not they were currently taking medication for the condition Dietary intake: online survey Weight awareness: frequency of self‐weighing Food inventory: available foods, price, portion size, and calories per portion Health media environment: presence of media (signs, posters, magazines, videos, etc.) related to eating, physical activity, or other health behaviours
Statistical analysis	Imputation of missing data: not reported or not undertaken by study authors Sample size calculation: the primary outcome examined here was BMI change between baseline and follow‐up; based on the known weight gain trajectory of free‐living adults without intervention, the study was powered to detect a 1.5‐kilogram difference in body weight between intervention and control sites at 2 years, accounting for approximately 15% attrition from work sites and adjusted for the work site component of variance (80% power, P < 0.05, 1‐tailed)
Notes	Funding: National Institutes of Health
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"...block randomisation at the work site level (block size = 2), using computer‐generated algorithms and performed by the study data manager, was used to assign work sites to intervention or no‐contact control conditions". Worksites were paired based on time of entry into the study and were randomised in those pairs
Allocation concealment (selection bias)	Low risk	Randomisation occurred "after completion of baseline site evaluations and participant data collection within each site pair. Sites were recruited by the principal investigator and project manager and randomised by the data manager…"
Blinding of participants and personnel (performance bias) All outcomes	High risk	"There was no blinding of participants or study staff to assignment of work sites to intervention versus control condition." Participants may have modified their usual behaviour in response to knowledge of the study, in particular, participants in the control group, although the impact on behaviour of the control group may not have been sustained over time. Involvement in the study may have led to changes in workplace policies or health‐related interventions at control sites
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	High risk	Physical activity and sedentary behaviour were self‐reported based on the IPAQ scale. Participants were aware of the expectation that intervention would increase activity levels. Both groups were likely to exaggerate levels of exercise, but intervention group was more likely to be affected. Participants self‐reporting exercise data were likely to be influenced by knowledge of expectations of the intervention
Blinding of outcome assessment (detection bias) Disease risk factors	Low risk	BMI and waist circumference were objectively measured by researchers
Blinding of outcome assessment (detection bias) Adverse events	Unclear risk	It is unclear how adverse effects were defined or measured during the study
Incomplete outcome data (attrition bias) Adverse events	High risk	As zero events were reported, this outcome was at high risk of bias, given any level of missing data
Incomplete outcome data (attrition bias) All other outcomes	High risk	Intervention: Randomised: 3 sites (752 individuals) Available at 24 months: 611 (115 left workforce or ineligible; 8 declined, 18 no contact) Analysed: physical activity: 573. Sedentary behaviour: 555. Disease risk factors: 611 Control: Randomised: 3 sites (995 individuals) Available at 24 months: 795 (180 left workforce or ineligible, 8 declined, 11 no contact) Analysed: physical activity: 726. Sedentary behaviour: 706. Disease risk factors: 706 Sensitivity analysis based on plausible assumptions about missing data indicated that these outcomes were very susceptible to change
Selective reporting (reporting bias)	Low risk	Multiple outcomes were measured but not reported in the main paper (including physical activity and sedentary behaviour), although data were obtained from the study authors. Blood pressure and high cholesterol were measured as self‐reported high levels only and could not be meta‐analysed with other results
Other bias	Low risk	Recruitment of participants within clusters was completed before study assignment

Mansi 2013.

Study characteristics
Methods	Randomised controlled trial Aim: to examine whether a pedometer‐driven walking intervention can improve health‐related quality of life and increase ambulatory activity in a population of meat processing workers when compared to a control group receiving educational material alone
Participants	Population description: physically inactive but physically able employees in a large, local meat processing plant, with 900 people working in 12 departments Intervention group: 29 participants Control group: 29 participants Location: meat processing plant located in the South Island of New Zealand Inclusion criteria: currently working, male or female aged 18 to 65 years, with a sedentary lifestyle and/or low levels of physical activity (fewer than 7499 steps per day); able to walk continuously for at least 10 minutes; able to read and sign an informed consent form and questionnaires; willing to participate in the full study. The physical ability of the participant to participate in walking programme will be screened using the physical activity readiness questionnaire (PAR‐Q) Recruitment: participants were recruited through advertisements (posters) at different work sites including health clinic, plant administration, cafeterias, and all department notice‐boards until a target sample (n = 60) was achieved Demographics: Age range 18 to 65 years: mean (SD) control 40 (12.2), intervention 43 (14.9) years. Gender (n males (%)): in control 14 (48.3%), in intervention 10 (34.5%) Height (cm) (mean (SD)): in control 165 (10.2), in intervention 164.4 (11.2) Body weight (kg) (mean (SD)): in control 76.9 (13.9), in intervention 80.2 (16.9)
Interventions	Duration: 12 weeks Pre‐intervention: all participants received pre‐baseline screening phase by wearing the pedometer (Yamax Digi‐walker SW‐200, Yamax, Tokyo, Japan) for 7 consecutive days. Participants were instructed on how to use the pedometer, to wear the pedometer on the waistband of their clothing for 7 days, and to reset the pedometer to zero at the beginning of each day, and to remove it at the end of each day, record on a step calendar the date and time the pedometer was attached and also removed, and the total number of steps displayed on the pedometer at the end of each day. After randomisation, all participants attended a 30‐minute education session on the health benefits of being physically active and received standardised educational material (physical activity justification booklet of “Walk into Health”, Toronto Public Health; www.toronto.ca/health/walkintohealth) that consists of written and graphical information describing the importance of walking as a physical activity for health benefits and for prevention of disease Intervention: based on the self‐regulation theory (SRT), goal‐setting, feedback, educational material, and use of a step calendar for self‐monitoring were used as measurements 1. Goal‐setting Participants received an email reminder at the beginning of each week about their step count goal for that week, based on their baseline walking activity level The goal was to gradually increase the level of activities by 5% from the previous goal‐setting target with an aim to reach at least 10,000 steps per day at the end of the 12‐week period. Targets were based on international guidelines for walking interventions Those who reached 10,000 steps per day at any time during the programme were encouraged to maintain and increase their physically active lifestyle 2. Step count and feedback Participants received step count feedback by looking at the digital display on the pedometer monitor Participants receive personalised weekly emails about daily average step count and additional health information, to encourage adherence with the programme 3. Step calendar Participants were given a diary to record their walks and to note each day whether they were adhering to the programme, time of day, duration of the walk, the week’s step count goal, and the number of steps taken at the end of each day Control: participants were encouraged to read the educational activity material and were asked to record any exercise performed over the 12 weeks. At completion of 12 weeks, and at follow‐up at 24 weeks, these participants again wore the pedometer for 1 week to establish a weekly step count for comparison to baseline scores
Outcomes	Physical activity Steps (pedometer) International Physical Activity Questionnaire short form (IPAQ‐SF). Developed as an instrument to measure health‐related physical activity in work age populations; a valid and reliable measure for monitoring population levels of physical activity. The questionnaire consists of 7 items that provide information within various intensity levels including aerobic activities of vigorous intensity, cycling activities of moderate intensity, walking activities, and sitting time in the last 7 days Cardiovascular disease and type 2 diabetes risk factors Body mass index Body fat percentage. Formulaically measured using skinfold thickness (Harpenden Skinfold Caliper W/Software) taken from 4 sites (triceps, biceps, subscapular, and suprailiac) according to recommended locations and technique. Three measurements to the nearest 0.1 mm were averaged. Body fat calculated by Linear Software, which is valid for people between 17 and 68 years old: body fat = (triceps + biceps + subscapular + suprailiac skinfolds) according to the participant’s body weight (kg) and age Waist circumference. Measured using plastic tape by placing it around the waist at the level of the umbilicus (iliac crest) Height and weight. Measured without shoes and light clothing using commercially available digital bathroom scales (Terraillon Lovely Classic Electronic Bath Scale, Terraillon, Paris, France) to assess body weight, which can weigh in 100G increments up to a maximum of 150 kg, and a standard laboratory stadiometer (Seca 213 Portable Stadiometer, Seca, Hamburg, Germany) to measure height Blood pressure. Measured with mOmron MX3 Plus Blood Pressure Monitor (HEM‐7200‐E) on 3 occasions with a rest period of 1 minute between measurements Quality of life Short Form 36 version 2 (SF‐36v2) questionnaire. Widely used to measure quality of life in general and specific populations. It has 8 domains of health‐related quality of life: physical functioning, role limitations resulting from physical health problems, bodily pain, social functioning, general mental health, role limitations resulting from emotional problems, vitality, and general health problems. The SF‐36 questionnaire has been validated and is a reliable measure of physical and mental health that can be completed in 5 to 10 minutes Other outcomes not reported in this review Functional exercise capacity (6‐minute walk test) Physical activity self‐efficacy scale (5‐point scale assessing participants’ beliefs or their confidence in their physical ability to successfully achieve their goals in different situations) Feasibility of pedometer use (satisfaction survey questionnaire)
Statistical analysis	Imputation of missing data: not reported by study authors Statistical analysis: independent t‐tests were used for nominal data to test for significant differences between experimental and control groups at baseline A repeated‐measure, mixed‐model ANOVA was used to examine pre‐ and post‐ between‐group differences in all outcomes at each follow‐up time point of the study. Bonferroni 95% confidence interval (CI) was used on estimated marginal means at each follow‐up time point to show the range of variation for between‐group interactions. Within‐group changes were assessed using pairwise comparisons for each variable. Effect sizes were calculated using standardised effect sizes for Cohen d values: 0.2 for small effect, 0.5 for medium effect, and 0.8 for large effect. Statistical analysis was performed using SPSS software 20.0
Notes	Funded by Mark Steptoe Memorial Trust, School of Physiotherapy, University of Otago, New Zealand
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	“participants will be randomised to one of the two groups”. Method of random sequence generation not described
Allocation concealment (selection bias)	Low risk	"After successfully completing the baseline assessment and signing the informed consent form, randomizations to one of the two groups will be performed using sealed envelopes. Participants will be invited to choose an envelope from a basket containing envelopes that allocate 50% of the sample for the intervention and the other 50% for the control groups: each will contain the group name for allocation, and the timetable of the study"
Blinding of participants and personnel (performance bias) All outcomes	High risk	“Researchers and participants will be not be blinded to group allocation.” Participants may have modified their usual behavior in response to knowledge of the study, in particular, participants in the control group
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	High risk	Physical activity and quality of life were self‐reported and were at risk of bias due to lack of blinding
Blinding of outcome assessment (detection bias) Disease risk factors	Low risk	"A registered nurse, blinded to group allocation, performed assessment at baseline, 12 weeks (at the conclusion of the intervention), and three months post‐intervention." Outcomes were objectively measured
Blinding of outcome assessment (detection bias) Adverse events	High risk	Adverse effects were self‐reported and were at risk of bias due to lack of blinding
Incomplete outcome data (attrition bias) All other outcomes	Low risk	Intervention: randomised: 29. Completed all assessments: 27 Control: randomised: 29. Completed all assessments: 26 All missing participants withdrew due to work commitments. "An intention to treat protocol was performed by replacing missing values with the group mean at 12 and 24‐week follow‐up time points”. All outcomes were robust to sensitivity analysis using plausible assumptions about missing data
Selective reporting (reporting bias)	Low risk	Study protocol was published. All outcomes appear to be reported as planned
Other bias	Low risk	None was identified

Maruyama 2010.

Study characteristics
Methods	Randomised, cross‐over study ‐ first phase only (second phase not used due to likely carryover effect) Aim: to investigate the effectiveness of a work site‐based intervention, Life Style Modification Program for Physical Activity and Nutrition (LiSM10!®), on metabolic parameters in middle‐aged male Japanese white collar workers requiring health guidance based on regular health check‐up results
Participants	Population description: office workers with risk factors for developing metabolic syndrome belonging to the health insurance association of the Nichirei Group Corporation, aged 30 to 59 years Intervention group: 52 participants Control group: 49 participants Location: Tokyo and surrounding area, Japan Inclusion criteria: male, office employees of the Nichirei Group Corporation, aged 30 to 59 years, with risk factors for developing metabolic syndrome, including 1 or more abnormalities involving serum lipoprotein, glucose levels, and blood pressure, with visceral obesity (umbilical circumference: ≥ 85 cm) and/or BMI ≥ 25. Abnormal levels were defined as triglyceride (TG) ≥ 150 mg/dL and/or HDL cholesterol (HDL‐C) ≥ 40 mg/dL, systolic blood pressure ≥ 130 mmHg and/or diastolic blood pressure ≥ 85 mmHg, fasting glucose ≥ 110 mg/dL and/or HbA1c ≥ 5.5%" Recruitment: individuals at risk were identified at regular medical check‐ups conducted by the Tokyo Health Service Association (Shinjuku‐ku, Tokyo, Japan), which was not involved in the study. 800 male employees were informed about the study, 319 showed interest in participating and agreed to use of their data. After receiving a detailed explanation of the intervention, 115 agreed to participate Demographics: Age: range 30 to 59 years, mean (SD) for control 35.5 (8.1), for intervention 43.1 (7.7) years Gender: 100% male
Interventions	Duration: 4 months Intervention: LiSM10!® programme was designed to promote healthy dietary habits and physical activity 1. Professional contact/counselling Monthly individual contact with a dietician and a physical trainer, both certified health councillors for this programme Baseline: 20‐minute session with dietician, including self‐assessment of consumption of beneficial foods (Group A: fish, soy products, green/yellow/white vegetables, mushrooms/seaweed/konnyaku) and foods recommended to be decreased (Group B: large servings of grains, confectionaries, sweet drinks, fatty meats, butter/margarine/dressing, eggs/liver, fried foods, pickles, soup, alcohol). The dietician gave advice on the impact of consumption of each food group, and participants were assisted to identify and record an action plan to monitor and change dietary habits or both, targeting specific foods based on stages of change theory. Ten‐minute session with physical trainer, including discussion of baseline physical activity measurements. Participants were assisted to identify and record an action plan to increase physical activity based on pedometer steps or other lifestyle changes. Participants wore a pedometer (Walking style HJ‐7101T Omron Health Care Co., Ltd., Kyoto, Japan) at the waist from the time they woke up until they went to bed Months 1 and 2: 10‐minute session with each counsellor, who assisted the participant to review the month's achievements against his or her action plan, to consider reasons for the results and effective strategies to improve, and if necessary, to revise the plan Month 3: counsellors provided advice via the website described below, including review of progress and revision of goals 2. Personal web page Enter current body weight; targeted food intake and physical activity; upload data from computer‐linkable pedometer for self‐monitoring throughout the study. Data obtained were automatically presented in figures Discuss awareness of their lifestyles for self‐monitoring throughout the intervention period Family members and counsellors could make comments and note their impressions of the data on the self‐monitoring page 3. Goal‐setting Dietary action plans: in food group A, 1, 2, and 3 items were selected to be increased by 20 (38.5%), 24 (46.2%), and 3 (5.8%) participants, respectively. Top items were white vegetables, green/deep‐yellow vegetables, and mushrooms/seaweed/konnyaku. In food group B, 1, 2, 3, and 4 items were selected to be decreased by 20 (38.5%), 19 (36.5%), 3 (5.8%), and 2 (3.8%) participants, respectively. Top items were confectionaries, alcoholic drinks, sweet drinks, large servings of grain and butter/margarine/dressing/mayonnaise Physical activity action plans: all participants decided to count steps. 32 (61.6%) decided to walk more than 10,000 steps daily Control: no intervention. Participants were given pedometers for 2 periods (at baseline and at follow‐up) of 1 week to allow outcome measurement. At the end of the week, pedometers were returned to study staff
Outcomes	Physical activity Pedometer steps (average steps per day over 1 week) Cardiovascular disease and type 2 diabetes risk factors Measured by Tokyo Health Service Association (Shinjuku‐ku, Tokyo, Japan), which was not involved in the study Body mass index Waist circumference (cm): umbilical circumference measured during the late exhalation phase in the standing position Blood pressure: mmHg, measured using an automatic blood pressure manometer with the participant in the seated position Cholesterol: high‐density lipoprotein (HDL: mg/dL, direct method); low‐density lipoprotein (LDL: mg/dL, Friedewald equation) Triglycerides (TGs): mg/dL, enzymatic method Fasting plasma glucose (PG): mg/dL, hexokinase‐UV method Other outcomes not reported in this review Additional biochemical measures: total cholesterol (TC: mg/dL, enzymatic method); HbA1c: %, enzymatic method; fasting insulin: μU/L, chemiluminescence immunoassay; insulin resistance, homeostasis model assessment (HOMA‐IR), calculated as PG (mg/dL) × insulin (IRI) (μU/L) ÷ 405); aspartate aminotransferase (AST), alanine aminotransferase (ALT), and gamma‐glutamyl transferase (γ‐GTP): IU/L, UV, and L‐γ‐glutamyl‐3‐carboxy‐4‐nitroanilide substrate methods; uric acid: mg/dL, uricase method Food intake: current targeted food intake entered at a website for self‐monitoring Lifestyle: a questionnaire on lifestyle, habitual food intake, stages and self‐efficacies of changes in habitual food intakes, and efforts to increase physical activity
Statistical analysis	Imputation of missing data: not reported by study authors Sample size calculation: calculated to detect a 10% change within the group and between groups, using 0.05 for the alpha and 0.20 for the beta error. The necessary sample size was 45 participants in each group Statistical analysis: between‐group comparisons at baseline were made using the Mann–Whitney U test for continuous data and the Chi² test for proportional data. The difference from baseline until the end of the study period for each group was examined using the paired t‐test. The difference in the programme's effectiveness was examined as an intergroup difference in intragroup change, using repeated‐measures analysis of variance
Notes	Funded by a grant from the International Life Sciences Institute Japan. Study authors acknowledged collaborative efforts of the Meiji Dairies Corporation, Suntory Holdings Limited, and Nichirei Foods, Inc. Study authors stated no financial interest in subject matter, materials, or equipment
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"A randomization code with equal numbers of alternative groups was generated from a list of all participants, using software SPSS (ver.15) at Waseda University"
Allocation concealment (selection bias)	Low risk	"The Nichirei Inc staff members managing the study and contacting participants were not involved in this randomisation process"
Blinding of participants and personnel (performance bias) All outcomes	High risk	"…as the participants received detailed explanations of the objectives and other aspects of this study, blinding to group assignments was not possible." No information was given about blinding of personnel, but it is unlikely that this was undertaken. Awareness of the purpose of the study may have led control group participants to behave differently during the study, which may have affected the outcomes
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	Low risk	Participants in the intervention group uploaded pedometer data electronically via a website. Control group participants returned the pedometer to study staff, and it is likely that results were electronically recorded. Due to electronic linkable pedometers, uploading of incorrect pedometer steps is unlikely to be influenced by lack of blinding
Blinding of outcome assessment (detection bias) Disease risk factors	Low risk	No additional information was given about blinding of outcome assessors. Anthropometric and blood test results were objectively measured and were unlikely to be affected
Incomplete outcome data (attrition bias) All other outcomes	High risk	Intervention: 52 participants randomised 48 completed the study (4 lost to follow‐up) 44 provided pedometer steps at end of intervention. Control: 49 participants randomised 39 completed the study (8 lost to follow‐up, 2 excluded after baseline due to abnormal blood tests indicating possible hyperlipidaemia) 22 provided pedometer steps at end of intervention Study authors did not conduct intention‐to‐treat analysis. Reasons for most missing data were not available. Sensitivity analysis indicated that these outcomes were at high risk based on plausible assumptions about missing data
Selective reporting (reporting bias)	Unclear risk	All outcomes described in the published papers were reported. Study protocol was not available
Other bias	Low risk	This cross‐over study offered the intervention to the control group as part of a second phase. Due to the likely carryover effect on behaviour of the first phase intervention, these second phase data were not used

Morgan 2011.

Study characteristics
Methods	Cluster‐randomised controlled trial Aim: to evaluate the feasibility and efficacy of a workplace‐based weight loss intervention that targeted overweight and obese male shift workers
Participants	Population description: overweight or obese male adult shift workers without a history of major medical problems at Tomago Aluminium, one of Australia's largest producers of aluminium, employing around 1200 staff Intervention group: 2 clusters of multiple crews (15 crews across both groups), 65 participants Control group: 2 clusters of multiple crews, 45 participants Location: the industrial suburb of Tomago, 13 km northwest of Newcastle, New South Wales, Australia Inclusion criteria: overweight or obese (BMI between 25 and 40 kg/m²) men aged 18 to 65 from Tomago Aluminium without a history of major medical problems such as heart disease in the last 5 years, diabetes, orthopaedic or joint problems that would be a barrier to physical activity, recent weight loss ≥ 4.5 kg, or taking medications that might affect body weight. Participants over 40 years of age were required to receive a doctor's clearance to participate Recruitment: individual recruitment via a staff email from the internal Health Service Department and through promotion at crew meetings by crew leaders. Participating crews were allocated to clusters prior to randomisation based on timing and rotation of shifts worked, to avoid contamination within the work site Demographics: Age: range 18 to 65 years, mean (SD) 44.4 years (8.6) years Gender: 100% male Socio‐Economic Indexes for Areas (SEIFA, based on residence postcode): 1 to 2 (lowest category) 7.9%, 3 to 4 18.0%, 5 to 6 52.8%, 7 to 8 18.0%, 9 to 10 (highest) 3.4%
Interventions	Duration: 3.5 months (14 weeks) Intervention: The Workplace POWER (Preventing Obesity Without Eating Like a Rabbit) intervention is based on social‐cognitive theory and behaviour change strategies. Adapted from a previous Internet‐based weight loss intervention, 'SHED‐IT' 1. Professional contact/counselling Information session by male researcher: 1 × 75‐minute face‐to‐face session 60 minutes covering education about energy balance, challenges of shift work related to diet and physical activity, weight loss tips, and behaviour change strategies including self‐monitoring, goal‐setting, and social support 15‐Minute technical orientation during information session to familiarise and teach participants how to use the website 2. Personal web page Study website: publicly accessible, allowing group support, free weight loss site (http://www.calorieking.com.au). Weekly, enter weight, submit online daily eating and physical activity diaries for the first 4 weeks, for 2 weeks in the second month, and for 1 week in the third month Website user guide Weight loss handbook Pedometer (style Yamax SW 200) 3. Feedback Website data given by the research team in 7 weekly individualised feedback documents via email over 3 months Each sheet gave weekly summary of results and suggested personalised strategies to address weight loss, reduce energy intake, and increase energy expenditure A research team email was available for questions, which were answered weekly by 2 research assistants with qualifications in health and physical education or nutrition and dietetics 4. Incentives Group‐based financial incentive Crews with the highest mean percentage weight loss after 1 month and at the conclusion of the intervention were given an AUD 50 gift voucher per person to be spent at a local sporting equipment store Control: received the intervention at 14 weeks (wait‐list control group)
Outcomes	Physical activity Leisure‐time physical activity. Self‐reported. Measured using a modified version of the Godin Leisure‐Time Exercise Questionnaire. "How many times per week do you engage in strenuous, moderate, and mild physical activity for a minimum of 10 minutes per session?" The total leisure activity score was calculated by (N·MET) moderate +(N·MET) strenuous +(N.MET) mild, where N = (number of bouts per week lasting ≥ 10 minutes multiplied by the time in minutes) for each category 'Workday' and 'usual' physical activity. Self‐reported. "(i) How much do you incorporate physical activity into your workday (during breaks, active commuting to and from work)?" scored on a 5‐point scale from 1 ‐ none to 5 ‐ a great deal; and "(ii) Is the amount of activity you did in the past month less, more, or about the same as your usual physical activity habits?" scored from 1 ‐ I am now much less active to 5‐ I am now much more active     Cardiovascular disease and type 2 diabetes risk factors Body weight (kg): measured with men wearing light clothing, without shoes on a digital scale to 0.1 kg (Model no. UC‐321PC, A&D Company Ltd., Tokyo, Japan) Body mass index (BMI): weight (kg)/height (m)², measured to 0.1 cm using a stadiometer (model KaWe 44440; Medizin Technik, Mentone Education Centre, Morrabin, Australia) Waist circumference (cm): measured level with the umbilicus with a non‐extensible steel tape (KDSF10‐02, KDS Corporation, Osaka, Japan) Blood pressure and resting heart rate: mmHg and beats per minute, measured using an NISSEI/DS‐105E digital electronic blood pressure monitor (Nihon Seimitsu Sokki Co. Ltd., Gunma, Japan) Quality of life Health‐related quality of life: 12‐Item Short Form Health Survey physical and mental scales Adverse effects Injuries at work: on‐site incident and injury recording system at Tomago Aluminium for the 12‐month period before and after intervention commencement. All work‐related injuries are reported by employees and recorded in an electronic database. Injuries can be the result of a single workplace exposure or event, or the result of cumulative exposure over time. Injuries excluded were those that occurred on the journey to and from work Other outcomes not reported in this review Selected dietary variables: specific foods (fruit, vegetable, bread) and beverages (milk, cola, soda, diet, and alcohol) Physical activity and dietary cognitions: self‐efficacy, pros and cons, behavioural intention, attitudes, stage of change Daytime sleepiness: Epworth Sleepiness Scale, which is a valid measure of general daytime sleepiness Workplace productivity or presenteeism: Work Limitations Questionnaire (WLQ) short form (the degree to which health problems interfere with performance of job tasks and estimate of related productivity loss. The WLQ generically assesses presenteeism, is validated, and is highly reliable Absenteeism: personal illness or non‐work‐related injury was recorded in an electronic database, presented as hours of leave. Carer's leave was excluded from the analysis. Absences for the 3‐month period before and after intervention commencement Feasibility: recruitment (achievement of target sample size), retention (retention rates at follow‐up), and attendance (at information sessions) Adherence to self‐monitoring: calculated from website usage data Research team emails: frequency and topic
Statistical analysis	Imputation of missing data: not performed by study authors Sample size calculation: based on 90% power to detect significant weight loss (primary outcome) difference between groups of 3 kg, assuming SD = 5 (P = 0.05, 2‐sided), and a correlation between pre‐ and post‐ scores r = 0.80; a sample size of 41 participants for each group was needed Statistical analysis: analyses were performed using PASW Statistics 17 (SPSS Inc. Chicago, Illinois, USA). Linear mixed models were used to assess all outcomes (except injuries at work) for the impact of group (Intervention and Control), time (treated as categorical with levels at baseline and 14 weeks), and group‐by‐time interaction; these 3 terms formed the base model. This approach was preferred to using baseline scores as covariates, as baseline scores for participants who dropped out at 14 weeks were retained, making this an intention‐to‐treat analysis. The intention‐to‐treat analyses reported are used in this Cochrane Review. Injuries at work were modelled using a generalised linear mixed model with a Poisson distribution, rather than a normal distribution, and to identity link function Adjustment for clustering: to examine potential clustering of effects at the crew level, crew was nested within both treatment and treatment‐by‐time terms as fixed effects, and these terms were used in the final models. These adjusted results are used in this Cochrane Review
Notes	Funded by Tomago Aluminium and the Hunter Medical Research Institute. Tomago had no involvement in study design, analysis and interpretation of data, or the decision to submit the manuscript for publication. Simon Mitchell from Tomago reviewed the drafted manuscript for accuracy and organised the data collection at Tomago. No other potential conflicts of interest were stated by study authors Australian New Zealand Clinical Trials Registry Number: ACTRN12609001003268
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"The random allocation sequence was generated by a computer‐based random number‐producing algorithm to ensure an equal chance of work crews being allocated to each group, without restriction." "As crews were randomly allocated based on crew shift clusters, we had an uneven number of men in intervention and control conditions"
Allocation concealment (selection bias)	Low risk	"To ensure concealment, the sequence was generated by a statistician. Randomization and participant study arm assignment was completed by a researcher who was not involved in the assessment of participants and the allocation sequence was concealed when enrolling participants by work crew. Participants were enrolled by Health Services staff at Tomago"
Blinding of participants and personnel (performance bias) All outcomes	Low risk	“Men were randomly allocated to one of two groups: the Workplace POWER (Preventing Obesity Without Eating like a Rabbit) intervention or a 14‐week wait‐list control group. Men worked in crews (n = 15) and were randomly allocated in four crew clusters based on the timing and rotation of shifts worked, to avoid contamination within the worksite.” Participants and assessors were blinded to group allocation at baseline assessment. The wait‐list control and steps to minimise contamination may have reduced the likelihood of blinding being broken
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	High risk	Physical activity and quality of life outcomes were self‐reported. It is possible that participants in the intervention group, knowing that the intervention had begun and the direction of expected change, may have more favourable outcomes, although obvious outliers were excluded from the analysis
Blinding of outcome assessment (detection bias) Disease risk factors	Low risk	"Participants and assessors were blind to group allocation at baseline assessment." It is unclear who performed these outcome measures, and whether they were blinded at follow‐up; however these outcomes are sufficiently objective to present low risk of bias
Blinding of outcome assessment (detection bias) Adverse events	Low risk	"Injury data were sourced from an on‐site incident and injury recording system at Tomago Aluminium for the 12‐month period before and after intervention commencement"
Incomplete outcome data (attrition bias) Adverse events	Low risk	Injury data appeared to be complete, as they were sourced from the company's on‐site incident and injury recording system
Incomplete outcome data (attrition bias) All other outcomes	High risk	Intervention: Randomised: 65. Followed up: 54 (10 unavailable for testing, 1 employee termination) Analysed for PA: 47. Missing data were imputed for all other outcomes Control: Randomised: 45. Followed up: 36 (9 unavailable for testing) Analysed for PA: 34. Missing data were imputed for all other outcomes Reasons for unavailability were not described. "Six men were identified as outliers in the total MET minutes variable and were omitted from the physical activity analyses as their reported physical activity levels were not plausible" At least 18.2% of the baseline sample were missing, but in some analyses, up to 31.8% were missing, This level of incomplete data is enough to pose high risk to the results under plausible assumptions about missing data For anthropometric analyses, study authors undertook a mixed‐model approach "as the baseline scores for subjects who dropped out at 14 weeks were retained, making this an intention‐to‐treat analysis." Imputation of follow‐up scores for missing participants was not described. Sensitivity analysis based on plausible assumptions about missing data indicates that these outcomes are highly susceptible to change
Selective reporting (reporting bias)	Unclear risk	All relevant outcomes described in the published papers and at trial registration were reported at 3 months, with the exception of pedometer steps, which were planned but not measured due to safety rules preventing pedometers from being worn in some areas of the workplace. Trial registry data stated that outcomes would be measured at 6‐month follow‐up, and that the wait‐list control group would be offered the intervention immediately following that point. However, the study reported that wait‐list control received the intervention at 14 weeks, and no results at 6 months were reported, so we assumed that 6‐month data collection was not undertaken. No information about the analysis plan was available
Other bias	High risk	As cluster‐randomisation was utilised for a small number of clusters, there is an increased chance of baseline imbalance between randomised groups in terms of clusters or individuals

Parry 2013.

Study characteristics
Methods	Cluster‐randomised controlled trial Aim: to determine if participatory workplace interventions could reduce total sedentary time, sustained sedentary time (bouts of 30 minutes), increase the frequency of breaks in sedentary time, and promote light‐intensity activity and moderate/vigorous activity (MVPA) during work hours
Participants	Population description: sedentary clerical, call centre, and data processing workers (n = 62, aged 25 to 59 years) in 3 large government organisations Intervention group: intervention: 3 clusters, 30 individuals randomised Control group: 6 clusters, 103 individuals Location: Perth, Australia Inclusion criteria: workers participating in office‐bound duties for 6 or more hours per day and working 4 or more days per week were invited to participate in the study Recruitment: it is unclear how the 3 large government organisations were recruited. Potential participants were recruited at regular monthly staff meetings attend by 20 to 30 staff Demographics: Age (mean years (SD)): analysed participants 43.5 (6.4); non‐analysed participants 39.3 (11.8) Gender (n (%) female): analysed participants 50 (80.6); non‐analysed participants 59 (83.1) BMI (mean kg/m² (SD)): analysed participants 28.0 (6.4); non‐analysed participants 28.7 (6.4) Wear time workday (mean minutes (SD)): analysed participants 921.9 (83.8), non‐analysed participants 862.5 (87.3) Wear time work hours (mean minutes (SD)): analysed participants 501.8 (65.3), non‐analysed participants 495.7 (42.8) Gender (proportion male): total 10%, control 5.9%, intervention 13.1%
Interventions	Duration: 12 weeks All groups: participants from all 3 interventions were asked to attend 2 structured meetings with members of their own intervention group at their workplace to discuss and develop interventions. A participatory approach to intervention development was used so that workplace interventions could be tailored to the specific needs of the workplace, and employee participants had ownership of the intervention. Structured meetings were run by a facilitator (SP). During the first meeting, participants ‘brain stormed’ options to promote their specific intervention (active office, physical activity, or office ergonomics). Between meetings, participants were encouraged to think about specific strategies. At the second meeting, 2 to 3 weeks following the first meeting, participants shared their ideas and rated potential strategies in terms of feasibility and effectiveness. An action plan was developed, and the facilitator communicated with team leaders and management to help implementation. Within 4 to 6 weeks of the second meeting, strategies to be used were in place, and the intervention phase was considered to have commenced. Throughout the intervention period, to communicate with and motivate participants, tailored emails were sent to each participant by a facilitator every 2 to 3 weeks Intervention: 'traditional physical activity' 1. Pedometer challenge Participants were provided with a pedometer (Yamax Digi‐walker SW700, Tokyo, Japan) for use as a motivational tool 2. Group‐defined activities: Promote active transport ‐ walk instead of bus Walk and talk meetings Short frequent walks during breaks, lunchtime, to and from work Increase use of stairs Alternative physical activity intervention: for the purposes of this review, the following 2 groups were combined as a single active intervention comparison group Control 1: 'active workstation' 1. Active workstation Participants had access to either an electronic height‐adjustable desk with integrated treadmill (A7TR78928H, Steelcase, Sydney, Australia; Organisations 1 and 3) or a treadmill plus a stationary cycle ergometer (LF‐2850, Exertec Air Bike, Fairless Hills, Pennsylvania, USA; Organisation 2) It was recommended that the active workstation be used for short periods several times a day, starting at 10 minutes and building up to 30 minutes per session (goal‐setting aspects). The workstation was equipped with a computer terminal and a phone, so that normal office duties could be performed 2. Group‐defined activities Standing or exercise between calls/document processing Participating in walk‐and‐talk meeting Personally delivering information rather than sending an email Increasing incidental activity in and around workplace such as taking longer routes to the printer or scanner, etc. Control 2: 'office ergonomics' 1. Computer workstation setup 2. Active sitting Spending some time perching on edge of chair, encouraging movement during sitting 3. Group‐defined activities Taking breaks from sitting Standing meetings Using a "piano stool" – reinforcing active sitting Using an air cushion
Outcomes	Physical activity Duration of physical activity: combined (light, moderate, and vigorous) as a percentage of accelerometer wear time during workdays and non‐workdays. Measured by accelerometer ActiGraph (GT3X, Pensacola, Florida, USA) Sedentary behaviour Total sedentary time. Measured on workdays and during work time following the intervention period. Measured by accelerometer ActiGraph (GT3X) Sustained sedentary time. Bouts of 30 minutes. Measured on workdays and during work time following the intervention period. Measured by accelerometer ActiGraph (GT3X)
Statistical analysis	Imputation of missing data: not reported by study authors Statistical analysis • Independent t‐test evaluated differences in participant characteristics and baseline activity levels between participants who completed the study with sufficient data and those who did not • One‐way ANOVA or Chi² tests compared baseline differences between organisations and between intervention groups • Linear regression models (ANCOVA) were used to estimate the magnitude and corresponding 95% confidence intervals of intervention effects, with post‐intervention measures such as the dependent variable, the 3‐level categorical variables ‘organisation’ and ‘intervention’ as independent variables, and the corresponding baseline measure as a covariate
Notes	"These authors have no support or funding to report"
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Each organisation formed three groups of volunteers based on physical proximity." "Within each organisation the groups of physically proximal volunteers were randomly assigned to one of three interventions". "Simple randomisation with a 1:1:1 allocation ratio was used by drawing a sealed envelope containing the intervention allocation from a hat"
Allocation concealment (selection bias)	Low risk	As above, sealed envelopes were used. Participants were recruited and baseline measurements taken before randomisation
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding was not mentioned, but participants from all 3 intervention groups were asked to attend 2 structured meetings at their workplace to discuss and develop the interventions used. Although each group received an intervention, participants could be unhappy if allocated to an intervention they did not prefer. Participants may have modified their usual behavior in response to knowledge of the study
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	Low risk	"The researcher with primary responsibility for collection and analysis of accelerometer data (SP) had conducted the interventions and was not blinded to group allocation. The ActiGraph data were downloaded using the ActiLife 5 software...” Although the researcher was not blinded, physical activity and sedentary behaviour were objectively measured
Blinding of outcome assessment (detection bias) Disease risk factors	Low risk	BMI was measured objectively by research staff
Blinding of outcome assessment (detection bias) Adverse events	Unclear risk	Definition and method of obtaining information on adverse events used were unclear
Incomplete outcome data (attrition bias) Adverse events	High risk	Zero events were reported, making this outcome highly vulnerable to any quantity of missing data
Incomplete outcome data (attrition bias) All other outcomes	High risk	Pedometer intervention: Randomised: 30 Analysed: BMI: 18 (7 withdrew, 1 data set damaged, 4 did not want to complete follow up). Physical activity and sedentary behaviour: 14 (4 did not have sufficient workdays and non‐workdays recorded for analysis) Alternative exercise intervention (2 groups combined for this review): Control 1: Randomised: 49 Analysed: BMI: 29 (14 withdrew, 2 left workplace, 4 did no want to complete follow‐up). Physical activity and sedentary behaviour: 19 (10 did not have sufficient workdays and non‐workdays recorded for analysis) Control 2: Randomised: 54 Analysed: BMI: 39 (7 withdrew, 2 data sets damaged, 6 did not want to complete follow‐up). Physical activity and sedentary behaviour: 29 (10 did not have sufficient workdays and non‐workdays recorded for analysis) Sensitivity analysis based on plausible assumptions about missing data indicated that results were highly susceptible to change
Selective reporting (reporting bias)	High risk	In addition to data measured with an accelerometer, physical activity was measured with the IPAQ questionnaire and was not reported (although this measure would be less objective and would not have been used in this review). Waist circumference data were measured but were not reported. An additional measurement time point at 3 months after the end of the intervention was planned but was not conducted to minimise the burden on participants. Several variations on analysis were presented; no statistical analysis plan was included in the protocol
Other bias	Low risk	Cluster‐randomised trial. Individual as well as cluster recruitment took place before randomisation

Pillay 2014.

Study characteristics
Methods	Randomised controlled trial Aim: to examine the feasibility of a 10‐week pedometer‐based intervention complemented by regular motivational messages, to increase ambulatory PA; and to determine the minimum sample size required for a randomised controlled trial (RCT)
Participants	Populationdescription: employees without cancer and with physical ability from the Faculty of Health Sciences at a tertiary academic institution in KwaZulu‐Natal Province, South Africa Interventiongroup: 11 participants Controlgroup: 11 participants Location: KwaZulu‐Natal Province, South Africa Inclusioncriteria: all willing participants aged 21 to 49 years were eligible. Employees were excluded in the case of pregnancy, diagnosis or treatment of cancer, any other physical/clinical condition that made physical activity difficult, contract workers whose employment with the company would end before the 12‐week follow‐up measurement, or non‐compliance to a minimum of 3 days of blinded pedometer wear at baseline Recruitment: an advertisement was emailed to all staff members within the Faculty of Health Sciences, inviting participation in the study (convenience sampling) Demographics: Age: mean (SD) for control group 38.3 (7.7), and for intervention group 37.6 (8.6) years Gender: for control group 82% female, for intervention group 88% female
Interventions	The theoretical model for behavioural change supported the intervention, as it is recommended that the intervention is targeted towards participants at a specific stage of behavioural change Duration: 10 weeks Intervention Pre‐intervention phase: Participants were asked to wear a blinded pedometer (Omron HJ‐750 ITC) throughout the day and to follow their usual routine of daily activities, and to remove the pedometer only when bathing, showering, or swimming. Upon return of the pedometer at baseline, data regarding steps per day were electronically uploaded by the researcher according to the Omron Health Manager Software protocol. This information was provided to participants. The pedometer’s unique feature of providing an hourly representation of steps per day assisted in calculation of the total number of steps, as well as an understanding of times of high and low step count Intervention phase: 1. Pedometer feedback Participants were provided with an unblinded pedometer for 10 weeks. They were shown how to upload and interpret their pedometer data and were requested to provide the pedometer data to the researcher, via email, bi‐weekly 2. Email Participants were provided with individualised, emailed feedback and a generalised information sheet on ways to increase physical activity. Feedback included information on average daily steps per day accumulated, the highest number of steps per day accumulated by the individual over the past 2 weeks, the number of days (if way) that aerobic steps per day were accumulated, and the volume thereof in the form of a personalised email General supportive/motivational messages including a key message such as “be active everyday” or “walk tall” were coupled with a few strategies to increase physical activity such as “use the stairs instead of the lift/escalator” that were suggested 3. Goal‐setting Participants were encouraged to improve their physical activity levels steadily (e.g. by 10% per week) until the recommended 30 minutes of moderate physical activity at least 5 times a week was achieved Control: Control group participants were similarly provided with a general motivational message bi‐weekly but were not provided with a pedometer over the 10 weeks, and therefore received no pedometer feedback. Participants were encouraged to improve their physical activity levels steadily (e.g. by 10% per week) until the recommended 30 minutes of moderate physical activity of at least 5 times a week was achieved
Outcomes	Physical activity Total steps per day (steps/d): data regarding steps per day were electronically uploaded by the researcher according to the Omron Health Management Manager Software protocol, and the information was provided to participants Aerobic steps (steps/d): the pedometer output can illustrate steps accumulated as being ‘aerobic’ or ‘non‐aerobic’, according to a classification that integrates both intensity and duration. The number of steps classified as aerobic (> 60 steps/min; minimum duration 1 minute) and non‐aerobic (< 60 steps/min and/or duration < 1 minute) within the total steps per day record was therefore provided. Similarly, total time spent accumulating aerobic steps (in min/d; aerobic time) could be identified Aerobic time (minutes): as above Cardiovascular and type 2 diabetes risk factors Waist circumference: measured using a tape measure around the skin of the waist Body fat: the Omron Body Composition Monitor (BF500) used to measure percentage body fat based on the principles of bioelectrical impedance Body mass index (BMI) Body weight: measured using an electronic scale, allowing only a single layer of clothing. Values were rounded to the nearest 100 g Body height: measured using a height chart as the vertical distance from the floor to the vertex of the head. Participants stood barefoot with heels, buttocks, and head in contact with the wall and arms at their side Blood pressure: systolic and diastolic blood pressure (mmHg) recorded with a sphygmomanometer after the participant remained relaxed for 5 minutes. Two readings were taken approximately 5 minutes apart, and the average of the 2 readings was recorded. If the 2 were different from each other (> 5 mmHg), a third reading was taken. The average of the 2 nearest readings was then used
Statistical analysis	Imputation of missing data: none reported by study authors Sample size calculation: prospective: a power analysis for a 2‐group, independent sample t‐test was conducted using the GPower data analysis website (reference provided). Based on minimum improvement of 1000 steps/d, as established in this pilot study, a sample size of approximately 85 participants per arm of the study is required to ensure 80% statistical power and a P value < 0.05. Considering this possibility and the likelihood of performing subgroup analyses based on factors such as age and gender, as well as the possibility of loss to follow‐up, we estimate that a sample size of 150 participants in IG and CG, respectively, would be an appropriate target sample Author conclusions: based on improvement of 1000 steps/d (IG), a minimum of 85 participants in IG and CG, respectively, is required for a future RCT (80% power; P < 0.05). We recommend a minimum of 150 participants in each group to account for loss to follow‐up and to allow for subgroup analyses Statistical analysis: performed in STATISTICA version 8 (StatSoft Inc., Tulsa, Oklahoma, USA) with statistical significance set to P < 0.05. Repeated measures analysis was used to compare HR data in each of the walking trials, as a percentage of age‐predicted maximum HR. Regression analysis was used to predict a steps/min rate for MPA using the various trials conducted (i.e. the steps/min rate at which the participant would reach 60% of maximum age‐predicated HR). In the regression analysis, fitness emerged as a primary factor contributing to a difference in results. Data therefore were subsequently grouped by fitness (using the estimated VO₂max from the fitness test) and stratified by upper, middle, and lower tertiles. A multivariate analysis was subsequently used to determine factors that contributed to variance in the stepping rate corresponding to MPA. A post hoc Tukey HSD test was used to determine the trial/s that were similar to the self‐paced brisk walk (based on % maximum HR) The relationship between average number of steps/d and BP, %BF, BMI, WC, and VO₂max was assessed using Pearson‐Product‐Moment Correlation analysis. To differentiate between total steps/d and intensity of health and fitness outcomes, participants were grouped according to the number and intensity of steps: low (< 5000 steps/d, irrespective of intensity), high‐low (5000 steps/d with no aerobic activity), and high‐high (5000 steps/d with aerobic activity). The 5000 steps/d cut‐off is based on current PA classifications that categorise those accumulating fewer than 5000 steps/d as inactive. Analyses of covariance, adjusting for age, gender, and total steps/d, were used to compare groups, with Bonferonni post hoc analyses, to determine the between‐group effect of these categories for BP, %BF, BMI, waist circumference, and VO₂max Adjustment for clustering: none reported by study authors
Notes	Funded by Durban University of Technology (DUT) and the National Research Foundation (Thuthuka)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"participants were randomly allocated to an intervention group (IG) or a wait‐listed control group (CG)." "The allocation of participants into IG and CG was achieved by random selection of participants from a composite participant list (Microsfot Excel)." Method of random sequence generation not described
Allocation concealment (selection bias)	Unclear risk	No additional information provided
Blinding of participants and personnel (performance bias) All outcomes	High risk	Blinding of study personnel was not described. Blinding of participants unlikely as “two CG participants, upon being allocated to the CG, declined further participation in the study”. Although “the pedometer screen was covered to reduce the likelihood of participants observing their daily steps, which might have influenced habitual levels of physical activity and subsequently the accumulation of daily steps during the baseline measurement”, awareness of allocation may have affected behaviour during the study
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	Low risk	Steps per day were electronically recorded and uploaded by researchers from the pedometers. Unlikely to be affected by lack of blinding
Blinding of outcome assessment (detection bias) Disease risk factors	Low risk	Objectively measured by researchers and unlikely to be affected by any lack of blinding
Incomplete outcome data (attrition bias) All other outcomes	Low risk	Intervention: Randomised: 11. Analysed: 11 Control: Randomised: 11. Analysed: 8 (2 participants declined further participation after allocation to the control group, 1 participant did not complete the pedometer wear and follow‐up measures) All outcomes appeared robust to sensitivity analysis based on reasonable best and worst case assumptions about missing data
Selective reporting (reporting bias)	Low risk	No study protocol was available. Study appeared to report all outcome measures as planned. This was a pilot study not designed to identify statistically significant results, but to assist in calculation of sample sizes for a larger study, so selective reporting of analysis or outcomes is unlikely
Other bias	Low risk	None was identified

Ribeiro 2014.

Study characteristics
Methods	Randomised controlled trial Aim: a 4‐group randomised controlled trial evaluated the impact of distinct workplace interventions to increase physical activity (PA) and to reduce anthropometric parameters in middle‐age women
Participants	Population description: women aged 40 to 50 years who were physically inactive during their leisure time and employees from a large University Hospital that has more than 15,000 employees who were physically inactive at their leisure time Pedometer intervention: (interventions 2 and 3, PedIC and PedGC): 101 participants Alternative physical activity intervention: (intervention 4, AT): 47 participants Control group: (intervention 1, MTC): 47 participants Location: Brazil Inclusion criteria: participants were middle‐age females, physically inactive during their leisure time (< 30 minutes of physical activity in leisure time, evaluated by the question from the IPAQ long version). Women using pharmacotherapy for weight reduction and those with BMI > 40, with uncontrolled hypertension or diabetes mellitus, or who had functional limitations preventing them from walking, were excluded Recruitment: participants were recruited through pamphlets and posters distributed in the hospital between February 2010 and 2011. Women who were interested in participating in the study contacted the researcher by telephone or by email to schedule an evaluation to verify eligibility Demographics: Age: MTC 45 (3); PedIC 45 (3); PedGC 45 (3); AT 45 (3) Gender: 0% male (included only females) Marital status: married (%) MTC 23 (48); PedIC 26 (49); PedGC 32 (67); AT 25 (52) Education: high school education (%) MTC 43 (91); PedIC 43 (81); PedGC 42 (89); AT 34 (71) Employment: workers at day work shift (%) MTC 40 (85); PedIC 48 (90.5); PedGC 41 (85.5); AT 39 (83)
Interventions	Duration: 12 weeks Four health professionals with experience in physical activity counselling volunteered to participate in the MTC, PedIC, and PedGC protocols and received specific training before the study Pedometer intervention (interventions 2 and 3): Combined for this review from 2 study intervention groups Group 2 (PedIC): included 3 individual counselling sessions, 15 minutes each, once per month. During the sessions, participants were provided general advice about the benefits of physical activity and received a booklet with information on how to increase physical activity in their daily lives. Participants also received a pedometer (Yamax Digiwalker, SW‐200; Yamasa Tokei Keiki Co. Ltd., Meguro‐ku, Tokyo, Japan) to monitor their number of daily steps. The goal was set to an increase of a minimum of 2000 steps per day during the study. Women were also given a diary to record their total daily steps Group 3 (PedGC): participants received a pedometer, as well as 8 group counselling sessions, each lasting 60 minutes, with a maximum of 12 participants. The first 6 sessions were performed at 1‐week intervals, and the final 2 sessions at 2‐week intervals. Counselling aimed to achieve behavioural changes by identifying the benefits of improved physical activity, learning how to overcome daily barriers to increase physical activity, self‐monitoring physical activity using the pedometer, goal‐setting (an increase in the number of steps per day of a minimum of 2000 steps per day) and relapse prevention, and group walking (10 minutes of brisk walking and monitoring the number of steps). The target step number was the same as that used in the PedIC Alternative physical activity intervention (intervention 4, AT): 24 sessions, conducted twice per week, lasting 30 minutes (first month), 35 minutes (second month), and 40 minutes (third month). Aerobic exercise was performed on a treadmill, and after a 5‐minute warm‐up, exercise intensity was based as moderate to intense on perceived exertion (Borg scale from 11 to 13—somewhat hard) followed by a 5‐minute cool‐down Control (intervention 1, MTC): 3 individual counselling sessions as per Group 2 (PedIC), but without the pedometer component
Outcomes	Physical activity Steps. Changes in the total number of steps and steps performed at moderate intensity (frequency ≥ 110 steps per minute) were evaluated using a pedometer (Digiwalker, Power Walker Model, PW 610; Yamasa Tokei Keiki Co. Ltd., Tokyo, Japan). The minimum number of days measured to validate the data analysis was 3 weekdays and 1 weekend day, and the number of steps per day was averaged per week. The total number of steps was automatically recorded by the equipment for 7 days and then was recovered by researchers, and the number of moderate‐intensity steps was registered by the participant in a daily dairy. For analysis, the average of the number of total and moderate (≥ 110 steps per minute) steps was calculated Cardiovascular and type 2 diabetes risk factors Weight (kg) measured by a digital scale with 0.1 kg (Toledo, Brazil) with participants wearing minimal clothes and no shoes Waist circumference (cm) measured using a steel tape (Sanny, Brazil) passing midway between the last rib and the iliac crest
Statistical analysis	Imputation of missing outcomes: last observation carried forward method was used for those classified as lost to follow‐up Statistical analysis: one‐way ANOVA or Chi² tests compared baseline differences between organisations and between intervention groups. The Cohen effect size was calculated, and an intention‐to‐treat approach was performed and was classified as small (0.21 to 0.49), medium (0.50 to 0.79), or large (≥ 0.80)
Notes	This study was funded by the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) and by Conselho Nacional de Pesquisa (CNPq)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Using a Microsoft Excel program, a simple randomization sequence was computer‐generated by one of the investigators of the study who was not directly involved with either the assessments or treatment of the patients"
Allocation concealment (selection bias)	Unclear risk	"The allocation was performed using numbered, sealed, and opaque envelopes, and four sealed envelopes were prepared for every participant. Each envelope corresponded to one of the study groups, and an envelope was chosen by the participant after the baseline measurements were performed" Unclear how generation of a random sequence related to selection from among 4 envelopes by the participant, or whether personnel could have known which of the set of 4 corresponded to each group
Blinding of participants and personnel (performance bias) All outcomes	High risk	Trial registration information states “open label”. Participants may have modified behaviour in response to knowledge of allocation group, in particular, participants in the control group
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	High risk	Self‐reported by participants and may have been influenced in response to knowledge of allocation group
Blinding of outcome assessment (detection bias) Disease risk factors	Low risk	Measured objectively and unlikely to be affected by lack of blinding
Incomplete outcome data (attrition bias) All other outcomes	Low risk	Intervention groups (combined for this review): PedIC: randomised: 53. Analysed: 47 (1 did not receive intervention (lack of interest), 2 discontinued intervention (lack of interest), 3 were lost to follow‐up (refuse)) PedGC: randomised: 48. Analysed: 33 (2 did not receive intervention (lack of interest), 12 discontinued (8 lack of time, 3 health problem, 1 lack of interest), 1 lost to follow‐up (refuse)) Alternative exercise control: Randomised: 47. Completed: 26 (1 did not receive intervention (lack of time), 19 discontinued (13 lack of time, 3 health problems, 2 aesthetic surgery, 1 lack of interest), 1 lost to follow‐up (medical license)) Minimal intervention control: Randomised: 47. Completed: 45 (2 discontinued (1 lack of time, 1 health problem)) Follow‐up was completed with participants who did not receive or discontinued the intervention on an intention‐to‐treat basis. Only 4 individuals were lost to follow‐up. Missing data were imputed using the last observation carried forward. Although this assumption is likely to have been overly optimistic, the data were robust to a sensitivity analysis based on plausible assumptions about missing data
Selective reporting (reporting bias)	High risk	BMI, blood pressure and blood glucose, and quality of life were measured but follow‐up results were not reported. Study registration at ClinicalTrials.gov indicated that specified planned measures for primary outcome changed from planned steps/week to steps/d
Other bias	Low risk	None was identified

Swartz 2014.

Study characteristics
Methods	Randomised controlled trial Aim: to assess change in sitting and physical activity behaviour in response to a workplace intervention to disrupt prolonged sitting time
Participants	Populationdescription: full‐time employees (employed ≥ 20 years) engaged in a sedentary occupation/clerical positions at a large, Midwestern American university through the university directory Interventiongroup 1: 40 participants Interventiongroup 2 (control): 38 participants Location: Milwaukee, USA Inclusioncriteria: participants were included if they sat for more than 60% of their workday, as measured through verbal self‐report Recruitment: not reported Demographics: Age (mean (SD)): all 44.3 (11.1), for stand 42.3 (11.6), for step (intervention) 46.1 (10.5) years Gender (proportion male (%)): all 32, for stand 40, for step 25 Ethnicity (proportion white (%)): all 89, stand 90, step 88 Completed college (%): all 92, stand 87, step 97
Interventions	Duration: 1 week (measurements during all working hours on the same 3 consecutive workdays during 2 successive weeks (baseline and intervention weeks)) Pre‐intervention phase: Participants attended a first meeting in a laboratory setting, where they provided consent and height and weight measures and a health history questionnaire. Participants were given instructions on how to wear the activPAL motion and postural assessment monitor (PAL Technologies, Glasgow, UK). The activPAL is a uniaxial piezoresistive accelerometer and inclinometer that is small and lightweight and measures both dynamic and static movements. It was attached to the midline of the front of the right thigh with a manufacturer‐supplied adhesive. The activPAL measures the posture of an activity according to the longitudinal axis of the thigh. Data were collected at a predetermined 10 Hz and at 15‐second intervals. Unit of measurement is steps Intervention group (step): The baseline week began at least 1 day but no more than 5 days after the laboratory visit. For 3 consecutive days during working hours, all participants wore the activPAL, a wrist watch (capable of prompting participants with beeps or vibrations), and a pedometer (Yamax SW‐200 Digi‐Walker Pedometr, Yamasa Tokei Keiki Co., Tokyo, Japan). Participants were asked to put on the activPAL immediately on arrival at the office, not to change their usual behaviour, and to remove the monitor just before going home. Participants were also asked to record in a paper log the exact times they put on and took off the activPAL. After the baseline monitoring period, research staff met each participant at his or her workstation and collected the monitor The intervention week began on the same day of the week during the week following the baseline period. Participants were asked to follow the intervention protocol, which included the same monitoring procedures as the baseline period. The intervention was designed to disrupt 60 continuous minutes of sedentary behaviour. Participants received a prompt from a wrist‐worn device and a desktop computer application to disrupt their sedentary time. A wrist‐worn device was set to prompt participants once per hour. Participants could select which device they preferred, either to beep (Armitron MD0346‐R(T)‐2, Armitron, Long Island City, New York, USA) or to vibrate (WobL Watch, PottyMD, Knoxville, Tennessee, USA). The computer prompt was a free download that provided a pop‐up message every hour: "Hello, please get up and walk 100 steps" Intervention group (stand): Baseline week identical to Step Group, except participants were not provided with a pedometer Intervention identical to the Step Group, except that the computer prompt stated, "Hello, please get out of your chair." Participants were not given further instructions on what to do while upright (how long to stand or do other activities)
Outcomes	Physical activity Steps/d and time spent stepping (minutes) recorded by the activPAL Time spent standing (minutes) recorded by the activPAL Sedentary behaviour Time spent sitting (minutes) recorded by the activPAL
Statistical analysis	Sample size calculation: a sample size of 30 per group was chosen a priori to achieve 80% to detect a minimum pre‐post change of 30 minutes Imputation of missing outcomes: not reported by study authors Statistical analysis: at baseline and post intervention, averages were estimated according to mixed‐effect models with random intercepts and random slopes for time, controlling for mean‐centred age, sex, and BMI. Analyses were performed with STATA 13. Significance was set at α level of P < 0.05
Notes	Funding sources include the Clinical and Translational Science Institute of South‐eastern Wisconsin
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Random number generation was used to assign participants to either the Stand group or Step group"
Allocation concealment (selection bias)	Low risk	"Assignments were written out and placed in sealed, numbered envelopes. The envelopes were opened sequentially by a researcher, participants were informed of group assignment, and a new monitor was issued to each participant." Unclear if envelopes were opaque, but probably done
Blinding of participants and personnel (performance bias) All outcomes	High risk	Unclear but unlikely given the nature of the intervention, and participants may have been aware of each other within the same workplace. Participants may have modified their usual behavior in response to knowledge of the study, in particular, participants in the control group
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	Low risk	“Recorded output from activPAL monitor was downloaded, processed, and classified into sitting, standing and walking by using manufacturer‐supplied activPAL software”
Incomplete outcome data (attrition bias) All other outcomes	High risk	Pedometer group: 40 randomised 38 participated in intervention (2 dropped out) 31 analysed (3 did not wear monitor for sufficient time, 3 equipment malfunction) Stand group: 38 randomised 37 participated in intervention (1 dropped out) 29 analysed (4 did not wear monitor for sufficient time, 3 equipment malfunction) Sensitivity analysis based on plausible assumptions about missing data indicated that results were highly susceptible to change
Selective reporting (reporting bias)	Unclear risk	Protocol not available, although all outcomes appear to have been reported as planned
Other bias	Low risk	None was identified

Talbot 2011.

Study characteristics
Methods	Randomised controlled trial Aim: to compare the effects of a pedometer‐based behavioral intervention (Fitness for Life, FFL) and the traditional army physical fitness (TRAD) intervention on physical activity, aerobic fitness, and chronic heart disease risk factors in healthy adult men and women in the Army National Guard (ARNG) who had failed the Army Physical Fitness Test (APFT)
Participants	Population description: volunteer, part‐time Army National Guard (ARNG) members, who had failed the 2‐mile run component of the Army Physical Fitness Test and had no history of major medical problems Intervention group: 84 participants Control group: 72 participants Location: Maryland and Washington, DC, USA Inclusion criteria: volunteer part‐time Army National Guard members who had failed the 2‐mile run component of the Army Physical Fitness Test (APFT), with 12 months or longer before re‐enlistment or retirement, and who had no history of CHD or stroke, were not currently taking hypertensive or cholesterol‐lowering medications, had not been pregnant within the previous 6 months, were not postmenopausal or currently taking hormone replacement therapy, and had no major musculoskeletal disorders. Postmenopausal women and women on hormone replacement therapy were excluded owing to their small numbers and potential confounding effects on serum lipoprotein Recruitment: from a pool of 261 ARNG soldiers who failed the run portion of the APFT, 156 met the criteria for inclusion in the study and were randomised into 2 groups Demographics: Age: mean (SD) for intervention 32.7 (10.1); for controls 32.8 (8.3) years Gender: 68.8% male for intervention, 80.4% for controls Race: intervention 50% white, 39% African American, 7% Asian/Pacific Islander. Controls 31% white, 50% African American, 7% Asian/Pacific Islander Education: intervention 27% high school graduate, 50% some college, 7% college graduate, 8% some postgraduate, 8% advanced degree. Controls 31% high school graduate, 50% some college, 2% college graduate, 15% some postgraduate, 2% advanced degree
Interventions	Duration: 24 weeks (12 weeks of conditioning and 12 weeks of maintenance) Pedometer intervention: the Fitness for Life (FFL) intervention was designed specifically for the reserve components of ARNG and Reserve to teach soldiers usually working a full‐time civilian job and a part‐time military job to incorporate moderate‐intensity physical activity (PA) into their daily lives 1. Professional contact/counselling Counselling sessions: discussed various activities to increase their daily step count Weeks 1 to 4: during brief weekly telephone counselling (< 5 minutes), pedometer logs were reviewed and feedback provided Weeks 5 to 12: weekly booster telephone calls, monthly support meetings Weeks 13 to 24: monthly maintenance meetings were continued; telephone calls were tapered to every 2 weeks, then monthly, to increase autonomy Monthly group meetings: held to provide support, emphasise relapse prevention, and encourage self‐monitoring of steps 2. Goal‐setting Pedometer (style Yamax SW 200) worn for self‐monitoring of daily steps. Central focus used to motivate and monitor steps through setting step goals, maintaining a daily step log, and promoting activities to increase steps Weeks 1 to 4: focused on accumulating daily steps through short bouts of walking combined with behavioural‐based physical activity counselling Weeks 5 to 12: focused on increasing the intensity of soldiers’ activities. Soldiers were taught to rate their perceived exertion while performing moderate‐ to high‐intensity daily activities in their target heart rate range, defined as 60% to 90% of predicted maximum heart rate, calculated as 220 − age. Using the Rating of Perceived Exertion scale, participants gauged the intensity of their activities by means of feedback from their target heart rate. The pedometer continued to be the central focus of behavioural strategies for setting step goals   Weeks 13 to 24: focused on sustaining gains in the number of steps and the intensity of physical activity. Participants were expected to continue using the pedometer to monitor their physical activity. Relapse prevention, self‐monitoring, and reinforcement were continually emphasised during the maintenance phase 3. Environmental prompts Motivational postcards: mailed weekly (Weeks 1 to 4) and then bi‐weekly to suggest various ways to increase steps 4. Feedback Through counselling sessions Taught how to gauge the intensity of their activities by means of feedback from their target heart rate Alternative physical activity intervention: the army physical fitness (TRAD) intervention follows Army Regulation 350‐41, with recommendations detailed in the army’s Field Manual 21‐20. The intervention consists of 12 weeks of high‐intensity conditioning, defined as 75% to 80% of maximum heart rate and 12 weeks of maintenance 1. Professional contact/counselling A Master Fitness Trainer, an ARNG soldier who had completed a 2‐week reserve component training course, oversaw the 12‐week training intervention 60‐minute briefing A brief reminder call was made before each monthly meeting Monthly group meetings 2. Goal‐setting Instructed to perform vigorous physical fitness training 3 to 6 days per week, including three 30‐minute sessions of aerobic training and three 30‐minute strength training sessions, performed unsupervised separately or combined, during the normal workday or during leisure time Supplementary booklet on the TRAD
Outcomes	Physical activity Seven‐day physical activity recall (PAR) interview. Amount of time spent asleep (1.0 metabolic equivalent of task (METs) units) and in moderate‐ (4 METs), hard‐ (6 METs), and very hard‐ (10 METs) intensity physical activity for the previous weekdays and weekend. The 7‐day PAR is moderately correlated with an accepted standard measure for cardiorespiratory fitness, the VO₂max, which is the maximum capacity of an individual's body to transport and use oxygen during incremental physical activity. It was assumed that they spent the remaining time in light activities (1.5 METs). To estimate energy expenditure per week, the average number of minutes at each activity level was multiplied by the respective MET value for an estimate of light, moderate, hard, and very hard physical activity in kcal/kg. Total physical activity was the sum of Moderate Intensity, Hard Intensity, and Very Hard Intensity physical activity because both interventions were designed to increase these 3 activity levels, but not Low Intensity Pedometer steps (assumed to be average steps per day over 1 week). A pedometer (Digiwalker, Yamax SW 200; New Lifestyles, Lees Summit, Missouri, USA) was used to count the number of steps in a day, which was recorded in a daily pedometer log. This pedometer’s accuracy is within 1% of the actual step count on a 4.88‐km sidewalk course. Participants wore the pedometer at the waist from the time they woke up until they went to bed. Control group participants were given the pedometers for only 2 periods (at baseline and at follow‐up, length of testing time unknown) to allow outcome measurement. At the end of the week, pedometers were returned to study staff. Intervention group participants retained the pedometer throughout the intervention Cardiovascular disease and type 2 diabetes risk factors Weight and height: measured using a digital scale, with participants in gym shorts and t‐shirt without shoes Body mass index (BMI): weight(kg)/height(m)² Waist circumference: measured during the late exhalation phase in the standing position Blood pressure: measured using an automatic digital monitor (Model 6009; American Diagnostic, Tokyo, Japan) on dominant arm at heart level while participants were seated. Three measurements were taken at 1‐ to 2‐minute intervals, and the mean of the 2 closest readings was reported. Cuff sizes reflected the circumference of the participant’s arm. Extreme values were checked by trained personnel, who repeated the digital recording and then recorded blood pressure manually Disease risk scores: 10‐year hard coronary heart disease (CHD) risk levels (recognised and unrecognised myocardial infarction, coronary insufficiency, CHD death) were calculated using the Framingham study risk equations Biochemical measures: fasting venipuncture from the anterior cubital fossa. Samples were allowed to clot at room temperature, then were centrifuged at 3000 rpm for 15 minutes, and the resulting serum was removed and stored at −80°C until analysis. The lipid panel was analysed using the Cholestech LDX System Analyser (Cholestech, Hayward, California, USA), with a sensitivity of 0.8%. All assays were conducted at Johns Hopkins Bayview Campus in the General Clinical Research Cente. Triglycerides; low‐density lipoprotein cholesterol (LDL‐C); high‐density lipoprotein cholesterol (HDL‐C); and TC:HDL‐C ratio were calculated Other outcomes not reported in this review Total cholesterol (TC) Army Physical Fitness Test (APFT): a standardised measure of cardiorespiratory fitness and muscular endurance according to standardised protocols detailed in Chapter 14 of the Army Field Manual 21‐20. This is a 3‐event physical performance test consisting of the number of standard Army push‐ups performed in 2 minutes; the number of standard Army sit‐ups performed in 2 minutes; and the time to complete a 2‐mile run. The APFT scoring is a normative‐based scale based on age and gender Cost‐effectiveness: total unit price, total cost, cost/consented soldier, and cost/completed soldier for each Trainer In‐person (per hour), trainer telephone (per hour), postcards (per card), paper (per sheet), and total
Statistical analysis	Sample size calculation: "based on the predicted effect of the intervention with projected 40% attrition, we estimated a total sample size of 156 ARNG soldiers to demonstrate a 10% improvement in APFT scores (effect size, d = 0.65) and physical activity (effect size d = 0.56) at an α of 0.05 and a power of 0.80" Imputation of missing data: "we used expectation‐maximisation for imputation estimates of missing data in the group of protocol completers, with SPSS Missing Value Analysis 16.0. The missing data for individual variables ranged between 1% and 19%. Missing data were determined to be missing at random, meeting expectation‐maximization assumptions." Hence, expectation‐maximisation was undertaken within those returning at follow‐up. Imputation was not undertaken for those lost to follow‐up Statistical analysis: statistical analyses were completed using PASW Statistics (SPSS) version 17.0. Descriptive statistics were calculated including frequencies, means, SDs, and percentages. Chi² test or Fisher’s exact test was used to examine relationships between group assignment and categorical variables. Independent sample t‐test statistics were used to determine differences between groups at baseline on the primary outcome variables (APFT and PA) and the secondary outcome variable (CHD risk). A repeated‐measures analysis of variance was used to test for main effects of the intervention (FFL vs TRAD) over time (baseline, 12‐week, and 24‐week follow‐up) and for the intervention‐by‐time interaction. The Geisser–Greenhouse correction 26 was applied if the statistical assumption of compound symmetry was not met. We used expectation‐maximisation for imputation estimates of missing data in the group of protocol completers, with SPSS Missing Value Analysis 16.0.27. Missing data for individual variables ranged between 1% and 19%. Missing data were determined to be missing at random, meeting expectation‐maximisation assumptions. A 2‐tailed P value ≤ 0.05 was set for statistical significance Based on the predicted effect of the intervention with projected 40% attrition, we estimated a total sample size of 156 ARNG soldiers to demonstrate 10% improvement in APFT scores (effect size, d = 0.65) and PA (effect size d = 0.56) at an α of 0.05 and a power of 0.80
Notes	Study authors acknowledged the Johns Hopkins Bayview General Clinical Research Center (which is funded by Department of Health and Human Services, National Institutes of Health (NIH), National Center for Research Resources, no. 5 M01 RR0279) for providing core laboratory and data management support and equipment, and the Intramural Research Program of the NIH, National Institute on Aging. Funding for the project was provided by Triservice Nursing Research Program, Johns Hopkins Bayview General Clinical Research Center, the Intramural Research Program of NIH, and the National Institute on Aging
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Study described as a "randomized controlled trial". No further information given on the method used to generate the random sequence
Allocation concealment (selection bias)	Unclear risk	No information provided
Blinding of participants and personnel (performance bias) All outcomes	High risk	No information provided. It is unlikely that personnel delivering the specific interventions would have been blinded. Participants may have been unaware of the differences between interventions on allocation, but being colleagues in the National Guard, it is possible that communication among participants would have occurred that identified the nature of each intervention. Although both interventions were active physical training interventions, it is possible that participants may have had a preference for the traditional training intervention (control) over the new (pedometer‐based) intervention, perhaps contributing to the higher dropout rate in the pedometer group
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	High risk	Physical activity outcomes were self‐reported. It is possible that participants in either study group, knowing that the intervention had begun and the direction of expected change, may have exaggerated activity levels. Although this applies to both groups, it may have applied more to 1 group than another, for example, if participants had stronger belief in the effectiveness of one intervention over another (e.g. traditional vs new intervention), or depending on participants' experience of the intervention. Participants in the pedometer arm may have felt greater pressure to report increased steps per day. Participants in the control arm may have felt greater pressure to report increased periods of higher‐intensity physical activity
Blinding of outcome assessment (detection bias) Disease risk factors	Low risk	Measured objectively by study personnel
Incomplete outcome data (attrition bias) All other outcomes	High risk	Intervention: Randomised: 84 Analysed: at 12 weeks: 51, at 24 weeks: 48 Control: Randomised: 72 Analysed: at 12 weeks: 53, at 24 weeks: 46 Reasons lost to follow‐up (not provided by group): military deployment (15), left National Guard (21), personal reasons (22), injury not related to the study (3), passed Army Physical Fitness Test (1) "ARNG soldiers who dropped out of the training intervention programme had lower baseline APFT scores, suggesting that less fit individuals require more motivation to complete such a program", indicating that data were not missing at random. "We used expectation‐maximization for imputation estimates of missing data in the group of protocol completers", but data were not imputed for participants who withdrew This level of incomplete data is enough to pose high risk for the results under plausible assumptions about missing data
Selective reporting (reporting bias)	Unclear risk	Study protocol was not available. Heart rate was measured but was not reported, but available outcomes (including additional data provided through personal communication) included a wide range of expected outcomes with a range of significant and non‐significant effects, favouring both interventions. It is unlikely that outcomes were selected based on results
Other bias	Low risk	None was identified

Viester 2012.

Study characteristics
Methods	Randomised controlled trial Aim: to evaluate the effectiveness of a tailored intervention for improving lifestyle behaviour and for prevention and reduction of overweight and musculoskeletal disorders (MSDs) among construction workers, and to describe the evaluation study regarding its (cost)‐effectiveness
Participants	Population description: blue collar construction site workers and factory workers at a large construction company Intervention group: 162 participants Control group: 152 participants Location: The Netherlands Inclusion criteria: blue‐collar workers of a Dutch construction company who attended the voluntary periodic health screening (PHS) at the occupational health service between February 2010 and October 2011. Workers who had been on sick leave for > 4 weeks at baseline were excluded. Sickness absence data were collected over a 2‐year period, starting 12 months prior to baseline Recruitment: conducted through the usual communication channels of the company at a non‐compulsory periodic health screening (PHS). Together with the invitation for the PHS, all workers received a brochure about the study, an informed consent form, and an additional questionnaire to measure baseline variables not included in the PHS Demographics: Age: total: 46.6 (9.7); intervention: 46.3 (9.9); control: 47.0 (9.5) Gender: 100% male
Interventions	The 'VIP in Construction' intervention. An extensive development phase was conducted based on a preliminary survey and focus groups with a sample of workers, as well as theoretical models of behaviour change and a review of evidence on effective intervention models. An adoption and implementation plan was also developed to accomplish intervention adoption and implementation by influencing behaviour of individuals who will make decisions about adopting and using the intervention and of individuals who deliver the intervention. This includes engagement of company human resources and management staff and efforts to ensure compliance of participants, including communication via company channels and linking the intervention to the company's existing PHS intervention Duration: 6 months Intervention: the intervention consisted of materials and tailored information on physical activity and diet, personal health coaching, and training instruction. Materials and coaching protocols were developed to enable tailoring of the intervention to individual risk factors and wishes of participants based on baseline measurements and questionnaires Intervention materials were made attractive and recognisable for the target group by using a standard lay‐out and logo. The "VIP in Construction Toolbox” consisted of tailored brochures, a calorie guide, a pedometer (brand and type not specified), a BMI card, a waist circumference measuring tape, recipes, a knowledge test, an overview of the company health‐promoting facilities, PEP forms, and exercise card describing abdominal and dorsal strengthening and stabilisation exercises. It was intended that the exercises be performed 3 times a week and be easily fitted into daily life routines, performed at home without additional expense Personal health coaching (PHC) consisted of (1) feedback on participants' health screening and current lifestyle behaviour; (2) support for goal‐setting by helping participants to formulate a written personal motivation and action plan containing physical activities or healthy food choices or a combination, targeting behaviour that is not at the desired level, as well as providing information on the company’s health‐promoting activities and distributing the intervention materials; (3) feedback and evaluation of formulated goals during follow‐up contacts, and discussion of possible barriers or new goals or both; (4) instructions for self‐monitoring using PEP forms and materials; and (5) instructions on how to use the exercise card. Participants were coached face‐to‐face at the first visit (60 minutes). Follow‐up contacts (feedback and motivating) were conducted by telephone with a minimum of 2 and a maximum of 4 contacts (10 to 30 minutes), varying according to participants' stage of change. To ensure that a standardised protocol would be used by the PHCs, all coaches received a training session and a manual describing the protocol and goals in detail Control: care as usual, including the PHS intervention
Outcomes	Physical activity Physical activity: energy output quantity and quality measured as frequency of vigorous activities (PHS questionnaire) or as moderate physical activity (number of days per week activities performed, e.g. walking and cycling for at least 30 minutes) Short Questionnaire to Assess Health‐enhancing physical activity (SQUASH): validation, duration, frequency, and intensity of different domains of physical activity (active work transportation, occupational physical activity, household activities, and leisure time activities) (MET minutes/week) and accelerometry (random sample of participants (25 intervention, 25 control) over 7 days) Sedentary behaviour Sedentary behaviour: accelerometry (random sample of participants (25 intervention, 25 control) over 7 days). Insufficient data collected for analysis and only baseline data reported Cardiovascular disease and type 2 diabetes risk factors Weight: measured using a digital weight scale with participants standing without shoes and heavy outer garments Body mass index (BMI): weight(kg)/height(m)² Waist circumference: measured midway between the lower rib margin and the iliac crest with participants in standing position at the end of expiration with a Seca 201 waist circumference measure (Seca, Hamburg, Germany) and measurement protocol Systolic and diastolic blood pressure (mmHg): measured twice with a fully automated blood pressure monitor (type: OMRON M6, Omron Healthcare, Tokyo, Japan). The mean value of the 2 measurements was computed Cardiovascular disease risk indicators: European Systematic Coronary Risk Evaluation (SCORE, based on smoking, systolic blood pressure, and blood cholesterol levels (either total cholesterol or ratio of total/HDL cholesterol)) Other outcomes not to be reported in this review Blood cholesterol: total cholesterol (mmol/L) was measured with non‐fasting fingerstick samples analysed on a Cholestech LDX desktop analyser (Cholestech, Hayward, California, USA) Musculoskeletal disorders: prevalence of regular pain or stiffness in upper and lower extremity regions (dichotomous scale (yes/no)); prevalence of MSD during the past 3 months for different body regions (Dutch Musculoskeletal Questionnaire, validated), intensity of pain (Von Korff scales, average pain and worst pain experienced on an 11‐point numerical scale (0 to 10)) Health behaviour: smoking (yes/no) Dietary intake: energy intake quantity and quality measured as alcohol consumption (PHS questionnaire, average glasses per week), portion size at dinner, number of beverages and slices of bread, consumption of energy‐dense snacks, average weekly intake and daily portions of several food groups during a usual week during the past month (questions also used in the Health Under Construction study), fruit and vegetable consumption (Short Fruit and Vegetable questionnaire, validated, number of days per week and number of daily servings of fruit, vegetables, and fruit juice using 5 items on citrus fruit, other fruits, cooked vegetables, raw vegetables, and fruit juice) Determinants of energy balance‐related behaviour: knowledge, attitudes, self‐efficacy, and stage of change for physical activity and dietary behaviours Physical functioning: physical functioning, social functioning, role limitations (physical problem), role limitations (emotional problem), mental health, pain, general health perception, and health change (RAND‐36, validated, Dutch version) Fitness: non‐exercise test estimation model including age, BMI, resting heart rate, and self‐reported physical activity Workplace productivity loss: sickness absence (company records), presenteeism (reduced productivity while at work, WHO Work Performance Questionnaire (WHO‐HPQ), and PROductivity and DISease Questionnaire (PRODISQ) Work ability: Work Ability Index Work engagement, work satisfaction, and vitality: vitality (Utrecht Engagement Scale (UWES), high levels of energy and resilience, willingness to invest effort, not being easily fatigued, and persistence in the face of difficulties), organisational commitment, and work satisfaction Use of company facilities: use of company health‐promoting facilities (e.g. company‐sponsored fitness) Cost: intervention costs; other workplace health promotion costs (use of company facilities); healthcare costs (care by general practitioner, allied health care, medical specialist, complementary and alternative medicine, hospitalisation, and medications based on 3‐monthly retrospective questionnaires). Dutch standard costs will be used to value services or prices according to professional organisations if not available. Medication use will be valued using unit prices provided by the Dutch Society of Pharmacy; productivity‐related costs (i.e. work absenteeism and presenteeism, salaries of participants sued for the employer’s perspective, average salaries per gender, and 5‐year age group used for the societal perspective) and participant costs (self‐reported costs related to sports activities (membership fees and sports equipment costs)) Process outcomes: context, recruitment, reach, dose delivered, dose received, satisfaction about the intervention, and fidelity (RE‐AIM framework; Steckler and Linnan framework)
Statistical analysis	Imputation of missing data: not reported/undertaken by study authors Sample size calculation: sample size was based on detecting a difference in change in body weight between intervention and control groups. In each group (intervention and control), 130 participants will be needed, based on a power of 80% and an alpha of 5%, and an expected weight loss of 1.5 kg (SD 4.3 kg) as a result of the intervention. The standard deviation used was subtracted from previous work by the same research group, studying construction workers. Taking into account loss to follow‐up of 20%, 324 workers should be included in this study
Notes	This project is part of the research programme, "Vitality in Practice", which is financed by Fonds Nuts Ohra (Nuts Ohra Foundation)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	“...participants will be randomly assigned to either the intervention or the control group by a computer generated list using SPSS”
Allocation concealment (selection bias)	Low risk	“Following baseline measurement….the randomization was prepared and performed by an independent researcher”
Blinding of participants and personnel (performance bias) All outcomes	High risk	“Although participants were not informed of their allocation, participants and intervention providers could not be blinded.” Participants may have modified their usual behaviour in response to knowledge of the study, in particular, participants in the control group
Blinding of outcome assessment (detection bias) PA, sedentary and QoL	High risk	Physical activity and quality of life were self‐reported by participants who were not blinded
Blinding of outcome assessment (detection bias) Disease risk factors	Low risk	“...data collectors and data analysts were blinded for allocation." “To ensure standardization of measurements, [occupational physicians] and assistants were provided with measurement protocols." BMI, waist circumference, blood pressure, and cholesterol were objectively measured
Incomplete outcome data (attrition bias) All other outcomes	Low risk	Intervention group: 162 randomised 6 months: 139 (9 lost to follow‐up, 10 termination of employment, 1 health problems, 1 deceased, 1 unknown) 12 months: 130 (9 lost to follow‐up, 1 termination of employment, 7 no time/interest, 1 other) Analysed: 128 (complete case analyses – all follow‐up measurements and questionnaires) Exceptions: weight, BMI n = 127; waist circumference n = 119, cholesterol n = 116 Control group: 152 randomised 6 months: 138 (14 lost to follow‐up, 65 termination of employment, 9 no time/interest) 12 months: 130 (8 lost to follow up, 8 no time/interest) Analysed: 129 (complete case analyses – all follow‐up measurements and questionnaires) Exceptions: waist circumference n = 114; cholesterol n = 115 However, these outcomes were found to be at low risk based on plausible assumptions about missing data
Selective reporting (reporting bias)	High risk	Protocol was published. Outcome measures for physical activity (PA) changed between protocol (no. days per week with at least 30 minutes moderate physical activity, SQUASH questionnaire) and publication (no days per week with at least 20 minutes of vigorous physical activity, SQUASH physical activity for leisure time only as the full questionnaire was not applied). A statistically significant effect was found for this definition of vigorous activity. Data on moderate and overall physical activity were requested from study authors but were not provided Overall cardiovascular disease risk score was planned in the protocol but was not reported
Other bias	Low risk	None was identified

ALT: alanine aminotransferase.

ANCOVA: analysis of covariance.

ANOVA: analysis of variance.

AST: aspartate aminotransferase.

AT: aerobic training

BMI: body mass index.

BTEC: Business and Technology Education Council.

CDC: Centers for Disease Control and Prevention.

CHD: coronary heart disease.

CI: confidence interval.

cpm: counts per minute.

GCE: general certificate of education.

GCSE: general certificate of secondary education.

GMR: geometric mean ratio.

GTP: gamma‐glutamyltransferase.

HbA1c: glycosylated haemoglobin.

HDL: high‐density lipoprotein.

HDL‐C: high‐density lipoprotein cholesterol.

HOMA‐IR: insulin resistance, homeostasis model assessment.

HR: human resources.

IPAQ: International Physical Activity Questionnaire.

IPAQ‐SF: International Physical Activity Questionnaire‐Short Form.

IRI: immunoreactive insulin.

LDL: low‐density lipoprotein.

MET: metabolic equivalent of task.

MSD: musculoskeletal disorder.

MTC: minimal treatment comparator

MVPA: moderate to vigorous physical activity.

NVQ: national vocational qualification.

OHC: occupational healthcare unit.

OR: odds ratio.

PA: physical activity.

PAR‐Q: physical activity readiness questionnaire.

PEP: performance evaluation and planning.

PG: plasma glucose.

PHS: periodic health screening.

SD: standard deviation.

SF‐36: Short Form‐36.

SRT: self‐regulation theory.

TC: total cholesterol.

TG: triglyceride.

VO2: maximum rate of oxygen consumption.

WLQ: Work Limitations Questionnaire.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Aittasalo 2004	Pedometers were not used throughout the intervention period
Baghianimoghaddam 2016	Not a randomised controlled trial
Barrington 2012	Pedometers were not used
Bassey 1983	Pedometers were not used throughout the intervention period. Controls also received a pedometer
Binoy Mathew 2012	Pedometers were not used
Bort Roig 2012	Controls also received a pedometer (linked to Puig‐Ribera 2008)
Brehm 2011	Pedometers were not used throughout the intervention period. No physical activity was measured as an outcome
Brooke‐Wavell 1996	Pedometers were not used
Buman 2011	Controls also received a pedometer
Butler 2004	Not a workplace setting
Compernolle 2015	Accelerometers were used
Della 2010	Controls also received a pedometer
Erfurt 1991	Pedometers were not used. No physical activity was measured as an outcome
Finkelstein 2015	Accelerometers were used
Furukawa 2003	Accelerometers were used. Controls also received an accelerometer. Participants could not view step count
Gell 2015	Pedometers were not used
Gilson 2007	Controls also received a pedometer
Gilson 2008	Pedometers were not used throughout the intervention period
Hultquist 2005	Not in a workplace setting. Participants could not view step count
Hunter 2012	Not in a workplace setting. Pedometers were not used
Hunter 2012a	Pedometers were not used
Härmä 1988a	Pedometers were not used
Härmä 1988b	Pedometers were not used
Iwane 2000	Controls also received a pedometer
Johannesson 2010	Controls also received a pedometer. No physical activity was measured as an outcome
Kazi 2012	Controls also received a pedometer
Kennedy 2007	Pedometers were not used
Kim 2013	Not in a workplace setting. Pedometers were not used throughout the intervention period
Lara 2016	Not in a workplace setting (4% retired)
Lee 1997	Pedometers were not used.
Leibiger 2012	Not a randomised controlled trial. Excluded by German native speaker (Dario Sambunjak)
Mackey 2011	Controls also received a pedometer
Martin 2013	No physical activity was measured as an outcome
Miller 2015	Accelerometers were used
Molde 2003	Not in a workplace setting. Pedometers were not used
Moreau 2001	Not in a workplace setting
Motl 2005	Not in a workplace setting. Pedometers were not used
Murphy 2006	Pedometers were not used throughout the intervention period. Controls also received a pedometer
Mutrie 2002	Pedometers were not used
Naito 2008	Pedometers were not used throughout the intervention period
Oja 1991	Not in a workplace setting. Pedometers were not used
Petersen 2012	Not in a workplace setting
Polzien 2007	No physical activity measured as an outcome
Prestwich 2012	Pedometers were not used
Puhkala 2011	Not conducted in a workplace setting
Puig‐Ribera 2008	Controls also received a pedometer
Serwe 2011	Controls also received a pedometer
Slootmaker 2009	Accelerometers were used
Speck 2001	Controls also received a pedometer
Sternfeld 2009	Not in a workplace setting
Terry 2010	Pedometers were not used
Thorndike 2012	Accelerometers were used
Thøgersen‐Ntoumani 2010	Controls also received a pedometer
Torstensen 1998	Not in a workplace setting. Pedometers were not used
Tucker 2011	Participants could not view step count
Tudor‐Locke 2004a	Not in a workplace setting
Van Berkel 2011	Pedometers were not used. Accelerometers were used
Vincent 2009	Not in a workplace setting
Wing 1996	Pedometers were not used

Characteristics of studies awaiting classification [ordered by study ID]

Audrey 2019.

Methods	Multi‐centre cluster‐randomised controlled trial in 84 workplaces in southwest England and south Wales Aim: the overall aim of the research is to examine the effectiveness and cost‐effectiveness of an employer‐led scheme to increase walking during the commute
Participants	Population description: employed healthy volunteer adults; further description not reported Intervention group: 44 workplaces (31 participants) Control group: 43 workplaces (323 participants) Location: workplaces in southwest England and south Wales Inclusion criteria: employees of any age and gender; in small, medium, and large workplaces; in Bath, Swansea, and South Gloucestershire Exclusion criteria: employees who already always walk or cycle to work; disabled in relation to walking to work; employees whose job required driving to work or regular driving throughout the day (e.g. delivery drivers, sales representatives who set off from home in the work vehicle); workplaces with little direct communication between senior management and employees are not suited to this intervention (unless there is a local supervisor/manager with the authority to agree with the study activities) because of the need for employer support for recruitment of participants and Walk to Work promoters; workplaces with a large proportion of staff on short‐term or zero‐hours contracts are not suited because of the need for 1‐year follow‐up data collection; employees who are retiring before the 1‐year follow‐up data collection; workplaces with firm plans to significantly downsize or relocate during the study period; very small workplaces (< 5 employees) Recruitment: workplaces were identified through business directories and local authority lists of employers; were sent information about the study and an invitation to participate Demographics:  Intervention group: 43% male, mean (SD) age 41.2 (11.4) years, 91% white British, 62% tertiary degree or equivalent, 38% income > £50,000 Control group: 43% male, mean (SD) age 42 (11.3) years, 90% white British, 59% tertiary degree or equivalent, 38% income > £50,000
Interventions	Duration: 10 weeks Intervention: There are 3 main stages of the intervention 1. Workplace ‘Walk to Work promoters’ will be identified (volunteers, or those nominated by participating employers, with an interest in walking and the capacity, within their usual role in the workplace, to promote walking amongst their colleagues). Walk to Work promoters will be trained (at a group external event or onsite, as appropriate to the needs of the workplace) by expert members of the research team about the health, social, economic, and environmental benefits of walking during the daily commute, and how to promote increased walking by walking the entire route (mainly those within 2 miles of the workplace) or by mixing walking with public transport or ‘park and walk’. They will be given resource packs and will be trained to access relevant websites and toolkits 2. Participating employees will be contacted by the Walk to Work promoter and will be given a Walk to Work pack including a booklet and a pedometer. Benefits of increased walking will be discussed, barriers and solutions discussed, and safe, feasible routes identified. Goals for incorporating walking into the journey to and from work will be set 3. Further encouragement will be provided through 4 contacts from the Walk to Work promoter over the following 10 weeks (face‐to‐face, email, or telephone as appropriate). During this time, Walk to Work promoters will also be prompted and encouraged in their role by 4 email/telephone contacts
Outcomes	Physical activity Primary outcome: daily minutes of moderate to vigorous physical activity (MVPA), measured using accelerometers at baseline and at 1‐year follow‐up Secondary outcomes Overall levels of physical activity (cpm), measured using accelerometers Daily minutes of sedentary time Modal shift (number of days, over the previous 5 working days, when walking was the major mode of travel to/from work) Also, the study intended to measure other process‐related (e.g. facilitators and barriers to participation in the intervention) and economic evaluation outcomes (e.g. costs to employers, employees, and the public sector of implementing the intervention)
Notes	Statistical analysis: "the treatment effect will be estimated with its 95% confidence interval using regression methods, which will allow any between‐workplace variation in outcome to be incorporated as a random effect (an equivalent approach to multi‐level modelling), and will allow the minimisation variables to be included as covariates. In addition, baseline MVPA and imbalanced baseline measures will be included as covariates, according to a process pre‐specified in the study statistical analysis plan. Linear regression will be employed in the analysis of MVPA, and ordered logistic regression in the analysis of the number of days, over the previous 5 working days, when walking was the major mode of travel to and from work. If there is concern that a skewed MVPA distribution is causing problems for the linear regression, this will be investigated in a sensitivity analysis of log‐transformed MVPA" Sample size calculation: "using findings from the feasibility study, the sample size for the full‐scale trial was based on an average cluster size of 8, an ICC of 0.15, and participant attrition of 25%. With a total sample size of 678, we have 80% power with a 5% significance level to detect a 15% increase in mean MVPA" Trial registration: ISRCTN15009100

Bang 2016.

Methods	Randomised controlled trial Aim: to determine the physical and psychological effects of an urban forest‐walking programme for office workers
Participants	Population description: 54 office workers Location: Korea
Interventions	Duration: 5 weeks, from October to November 2014 Intervention: 5 weeks of walking exercise based on Information‐Motivation‐Behavioural Skills Model Control: not described
Outcomes	Physical activity Cardiovascular disease and type 2 diabetes risk factors Anthropometric measurements (waist size, BMI) Blood pressure Quality of life Other outcomes: Health promotion behaviour Depression Bone density
Notes

Mattila 2013.

Methods	Three‐arm randomised controlled trial Aim: overall aim not stated. Based on methods provided, the likely aim is to assess the benefit of an intervention that targets several behavioural health risk factors delivered face‐to‐face vs the intervention with additional technology support (including a pedometer) vs usual care
Participants	Population description: employees of the city of Espoo, Finland. This encompassed a range of workplaces, including teachers, engineers, and healthcare personnel Intervention group: 118 participants Intervention + technology (including a pedometer) group: 118 participants Control group: 118 participants Location: employees of the city of Espoo, Finland Inclusion criteria: aged 30 to 55 years, willing to participate in the intervention, willing to make lifestyle improvements in 1 of the targeted behaviours (i.e. weight management, eating habits, physical activity, sleep habits, smoking, or alcohol consumption) within the following 6 months, rated their work ability as 7, 8, or 9 on a scale of 0 to 10 (10 being their lifetime best work ability), and either increased risk of diabetes (score of 12 to 20 in the diabetes risk test) or at least 1 of (1) overweight (body mass index; BMI = 27 to 34 kg/m²), (2) low physical activity level (not meeting physical activity recommendations), (3) unhealthy eating habits (not eating vegetables daily and/or not eating during the working day), (4) sleeping difficulties (at least 2 hours of self assessed sleep deprivation), (5) risky alcohol consumption (score of 5 or more for men, 4 or more for women in the Alcohol Use Disorders Identification Test), and (6) daily or occasional smoking. Pregnant women were excluded Recruitment: a web‐based health questionnaire was sent to all employees, of whom 37.93% (n = 4134) responded. Of 783 eligible employees, 352 were randomly assigned to the 3 arms Demographics: aged 30 to 55 years. Details available only for the intervention + technology (including a pedometer) group (n = 114): 28.9% male, aged 44.6 + 7.1 SD years, 58.8% college/university or higher education
Interventions	Duration: 3 months, between February and June 2008 Intervention: Designed to motivate and empower individuals by teaching them generic strategies for improving their lifestyles, irrespective of their personal goals and health risks. The intervention was based mainly on the transtheoretical model and acceptance and commitment therapy 5 biweekly face‐to‐face meetings in groups of 7 to 12 participants. A leader was trained to perform the intervention from a manuscript and with guidance of the intervention developers. The following topics and strategies were covered during the course of the 5 meetings: personal analysis of values and good life and health and wellness (meetings 1 and 2), mindfulness skills (meetings 1 and 2), self‐monitoring (meetings 1 to 3), problem‐solving (meetings 3 and 4), healthy lifestyles and work ability (meeting 3), relaxation (meeting 4), and the transtheoretical model and preparation and planning for the future (meeting 5). Participants also received homework assignments Intervention + technology (including a pedometer): The intervention as above. Total duration of each meeting for the technology group was 2 hours, including a 90‐minute intervention, followed by a 30‐minute technology introduction The technology toolbox was designed to address the strategies, to provide additional support for behaviour change, and to help maintain intervention effects between meetings and after the active intervention period. Participants were encouraged to choose the technologies they considered most appropriate in supporting their personal goals, which could change The technology toolbox consisted of monitoring devices, mobile applications, and Web services Monitoring devices included weight scales (Seca Sensa 804, Hamburg, Germany), a pedometer (Omron Walking style II, Kyoto, Japan), and 3‐day loan of a heart rate belt (Suunto MemoryBelt; Suunto, Vantaa, Finland) Three mobile applications: The Wellness Diary (Nokia, Helsinki, Finland) enables manual self‐monitoring of 16 health‐related variables with automatic graphical feedback based on the entries. Mobile Coach (Firstbeat Technologies, Jyväskylä, Finland) creates an adaptive weekly exercise programme based on the user’s activity level and performed exercises and provides graphical and numerical feedback with comparison to user’s set targets. SelfRelax (Relaxline, Mantes La Jolie, France) is an audio‐guided relaxation application that can be personalised (e.g. duration, purpose, body position, background sounds) A password secure Web portal (the Portal; Nokia, Helsinki, Finland) was developed specifically for the study, providing messaging between intervention leaders and participants, access to information on healthy lifestyles, and 3 integrated wellness services: Wellness Diary Connected (Nokia, Helsinki, Finland) is a Web‐based version of the Wellness Diary. Hyperfit is a detailed food and exercise diary for weight management, which provides in‐depth information on eating and exercise habits and the quality of nutrition. Nutritioncode (Tuulia International, Helsinki, Finland) is a commercial service linked to a loyalty card of a Finnish grocery store chain for easy monitoring of the nutritional quality of groceries Control: Standard occupational health care
Outcomes	Physical activity Fitness test: submaximal bicycle ergometer test for evaluating maximal aerobic capacity. The test was performed in a laboratory on a stationary bicycle ergometer with an initial load of 40/30 W (male/female) that was increased every 2 minutes by 20/15 W with a target of reaching 85% of estimated maximum heart rate Self‐reported exercise habits: from the health questionnaire Cardiovascular disease and type 2 diabetes risk factors Anthropometric measurements were made by a research nurse and included data on height, weight, waist circumference, body fat percentage by bioimpedance, and blood pressure Blood lipids (e.g. total cholesterol, triglycerides) Risk score Diabetes risk test: the initial screening questionnaire provided the relevant components. It included participants' health (e.g. self‐estimated health on a scale from 1 = good to 5 = poor) and health behaviours (e.g. eating and exercise habits, smoking, sleep, stress)
Notes	Statistical analysis: analysis and findings not reported. Outcomes collected at baseline and at 12 months (even though the intervention duration was only 3 months) Sample size calculation: not reported

BMI: body mass index.I

cpm: counts per minutes.

MVPA: moderate to vigorous physical activity.

SD: standard deviation.

Characteristics of ongoing studies [ordered by study ID]

ACTRN12617001395325.

Study name	Activity for Wellbeing
Methods	Randomised controlled trial with mixed‐methods design Aim: to study the efficacy of using a need‐supportive, person‐centred physical activity programme for improving and maintaining physical activity participation and psychological wellbeing in front‐line aged care workers
Participants	Population description: target sample size: 30 Inclusion criteria: current employee of ACH Group, able to speak fluent English, at least 18 years of age, able to engage in moderate physical activity (as defined by the American College of Sports Medicine (ACSM)) and complete all outcome measures (including the Six‐Minute Walk), not currently meeting ASCM recommendations for physical activity (≥ 30 minutes of moderate cardiorespiratory exercise per day, 5 days per week) Exclusion criteria: volunteers with previously diagnosed conditions or signs or symptoms of disease as indicated on the Adult Pre‐exercise Screening System who need to provide medical clearance prior to participation in the study
Interventions	Duration: 12 weeks Intervention: active need‐support (support of basic needs of competence, autonomy, and relatedness) delivered by an accredited exercise physiologist, using a self‐determination theory‐based approach and applied through the following strategies One initial face‐to‐face motivational interviewing‐based session (approximately 1 hour), which will be used to investigate individual barriers and preferences for physical activity, and for collaborative development of a person‐centred activity plan (aimed at assisting the participant to meet ACSM recommendations for physical activity). During this session, participants will also be taught how to use affect and ratings of perceived exertion to regulate exercise intensity, and how to use a pedometer and interactive website for monitoring step counts Weekly contacts via the dashboard page of the programme website for providing feedback on goal‐setting, motivational statements, and active encouragement and support (12 contacts over the course of 12 weeks), or email contacts Control: wait‐list control
Outcomes	Reported at baseline and at 12 weeks and 6 months post intervention (9 months post baseline) Physical activity 2‐day activity recall (Multimedia Activity Recall for Children and Adolescents (MARCA) software) Accelerometry Cardiovascular disease and type 2 diabetes risk factors: Anthropometric measures: weight Blood pressure Quality of life AQOL‐8D Other outcomes Behavioural regulations (BREQ‐3) Exercise causality orientations (ECOs) Negative psychological states (psychological distress) as measured by the Kessler 10‐point Psychological Distress Scale Psychological need satisfaction in exercise (PNSE) Six‐Minute Walk Distance Perceived effectiveness and limitations of the programme for this population (survey of programme implementers) Reach (survey of all invited employees) Perceived autonomy support given by the exercise physiologist during implementation of the programme (Healthcare Climate Questionnaire (HCCQ))
Starting date	9 November 2017
Contact information	Ms Merilyn Lock Alliance for Research in Exercise, Nutrition and Activity School of Health Sciences Playford Bldg, City East Campus University of South Australia Adelaide, SA 5001 Australia Telephone: +61 8 8302 1752 merilyn.lock@mymail.unisa.edu.au
Notes	Sources of funding: ACH Group; University of South Australia

ACTRN12618001567213.

Study name	Not stated
Methods	Randomised controlled trial
Participants	Population description: education workforce Location: Hunter New England region, New South Wales, Australia Inclusion criteria: government primary schools that currently use the Skoolbag Communication app Exclusion criteria: schools located outside the Hunter New England Local Health District, with secondary students (central schools), catering exclusively for children requiring specialist care and/or already involved in an obesity prevention intervention trial. Participants younger than 18 years
Interventions	Duration: 6 months Intervention group: the intervention is guided by social‐influence and cognitive‐behavioural theories and utilises the Get Healthy@Work framework to support the school’s implementation of the following existing 2 programs Premier’s Sporting Challenge (PSC): to support school staff to increase their physical activity, schools will be supported to implement the NSW Department of Education’s Premier's Sporting Challenge. This 10‐week programme encourages staff at both an individual and school level to meet physical activity guidelines by monitoring and recording their own activity through the “PSC Tracker App”. The programme provides school staff with pedometers and access to online resources to help them achieve their physical activity goals. The PSC encourages participants to increase activity levels and tracks their progress through the app and an e‐wall chart. Award levels for the 10‐week challenge reflect a daily activity time commitment ‐ either as part of a team or as an individual. Once the challenge has begun, data entry should be maintained on a week‐to‐week basis to effectively monitor physical activity progress. All types of moderate to vigorous physical activities can count toward an award. Awards are based on the amount of time spent being physically active. For example, Bronze ‐ 30 minutes per day, Silver ‐ 45 minutes per day, Gold ‐ 60 minutes per day, and Diamond ‐ 80 minutes per day. The school communication platform app will be used to send messages and reminders to school staff regarding physical activity Healthy eating programme: to support school staff to increase their consumption of vegetables, fruit, and water and to decrease their consumption of discretionary foods and sugary drinks, schools will be supported to implement a healthy eating programme that will provide school staff with advice and resources on how to make simple healthier swaps for their lunches, snacks, and catering for staff gatherings, etc., from ‘less healthy’ foods to ‘healthier’ everyday items Control group: not described
Outcomes	Measured at baseline and at 6 months Physical activity Counts per minute (accelerometer) Moderate to vigorous physical activity (MVPA, accelerometer) Sedentary behaviour Accelerometer Other outcomes Dietary behaviours: items from the NSW Population Health Survey will be used to assess school staff consumption of (i) servings of fruits and vegetables, (ii) number of discretionary snacks, (iii) glasses of water, and (iv) number of sugar sweetened beverages per day Pain (5‐item tool to ask participants about muscle, joint, or bone pain over the last 12 months) Sleep patterns and satisfaction (7‐item composite insomnia severity index) Well‐being (Cohen’s Perceived Stress Scale, which is a widely used psychological instrument used to measure the perception of stress). The 10‐item tool asks participants about their feelings and thoughts during the last month
Starting date	15 October 2018
Contact information	Dr Nicole Nathan HNE Population Health Longworth Ave Wallsend NSW 2287 Australia +61 2 49 246 257 nicole.nathan@hnehealth.nsw.gov.au
Notes	Sources of funding: Teachers Health Fund

ACTRN12619000173190.

Study name	Virgin PulsE Global ChAllenge Study (VEGAS)
Methods	Randomised controlled trial
Participants	Inclusion criteria: provision of signed and dated informed consent form, stated willingness to comply with all study procedures and availability for the duration of the study, aged 18 and older, in good general health as evidenced by self‐reported medical history, employed by Australian organisations, agreeing to participate in the Virgin Pulse Global Challenge (VPGC) programme, physically able to participate in the VPGC programme (e.g. able to perform step count), agreement to adhere to Lifestyle Consideration (stated below) throughout study duration and to not discuss the VPGC programme with the team assigned to the control arm for the duration of the study and vice versa, fasted before any blood tests collected by participants’ respective physicians or pathologists, abstaining from caffeine, alcohol, or strenuous exercise 4 to 6 hours before each study visit assessment Exclusion criteria: previously enrolled in the VPGC programme, pregnant women or women who plan to conceive during the VPGC programme, weight ≥ 150 kg at baseline screening, unable to follow the VPGC programme due to physical limitations, unable to follow the VPGC programme due to language barriers, unable to commit to the study procedures and visits throughout the study duration
Interventions	Duration: 100 days Intervention group: Virgin Pulse Global Challenge (VPGC) programme, which is a well‐being solution that equips participants with the knowledge, tools, and support required to improve lifestyle behaviour. The VPGC is a holistic health and well‐being platform for participants that is accessible 24/7, 365 days a year. All participants are encouraged to achieve 10,000 steps daily, but this is not mandatory. Participants will be randomised into teams of 7, joining either the intervention group or the control group. As soon as the teams are finalised, all participants will receive the same activity step count tracker. They are then required to set up a login account on the Virgin Pulse website (https://www.virginpulse.com) or via the Virgin Pulse application (app) available at Google Store. Next, participants are asked to sync their tracker with the Virgin Pulse app to enable daily steps to be recorded automatically throughout the 100 days. All activities involving step counts are automatically recorded on the tracker, and options are available to find equivalent steps for activities such as swimming and cycling. The 100‐day VPGC programme is developed by Virgin Pulse; the entire programme is delivered digitally on devices such as personal computer (through the Virgin Pulse website) and the Virgin Pulse app. Also, Virgin Pulse has a built‐in daily email reminder system to remind participants to sync their daily steps. All data captured at the Virgin Pulse website or on the Virgin Pulse app are stored on the Virgin Pulse server in Australia Control group: not described
Outcomes	Measured at 20 days and at 3 months post intervention (4 and 6 months post baseline) Cardiovascular disease and type 2 diabetes risk factors: Blood pressure Quality of life Personal Well‐being Index ‐ Adult version (PWI‐A) Other outcomes Cognitive abilities: a selective computerised cognitive assessment using CogState will be performed by participants. It is designed and validated for clinical trials. Each of the CogState tests (see below) is designed to measure a specific area of cognition: Identification Test (attention and reaction time test); Set‐Shifting Test (executive function); Two‐Back Test (working memory) Depression anxiety stress scale (DASS‐21) score (emotional symptoms) Epworth Sleepiness Scale (ESS) score (daytime sleepiness) Pittsburgh Sleep Quality Index (PSQI) (measure of overall sleep dysfunction, global score)
Starting date	22 May 2019
Contact information	Dr Won Sun Chen Swinburne University of Technology Department of Statistics, Data Science and Epidemiology, Level 9 Internal Mail H24, PO Box 218 Hawthorn, VIC, 3122 Australia +61 3 9214 8437 wchen@swin.edu.au
Notes

Aittasalo 2017.

Study name	Not stated
Methods	Cluster‐randomised controlled trial Aim: to promote active commuting to work through environmental, social, and behavioural strategies in 2 urban areas
Participants	Population description: 1144 employees Intervention group: 8 workplaces Control group: 8 workplaces Location: Tampere, Finland Inclusion criteria: workplaces with a minimum of 10 employees were included to ensure that there would still be participants left in each workplace for follow‐up after dropout Exclusion criteria: none stated Recruitment: workplaces were identified using the database of Statistics Finland. After the information was gained, it was checked via Internet searches, visits to the area, and telephone calls to confirm that the workplaces were still operating, located in the area within the reach of the main walking and cycling trail, and employed more than 10 people. At this stage, the many car sales companies in the area were excluded because based on a few initial contacts, their interest towards ACW promotion seemed mild. Management was contacted by email or by telephone to inquire about their preliminary interest to participate in the study. Management determined the extent of each workplace’s participation. In small workplaces, all employees were included, but in larger workplaces, participation was limited to 1 or 2 units or departments to minimise the workload of the contact person in each workplace. Contact persons shared the names and email addresses of included employees with the researcher, who then was able to deliver personal study invitations to the employees. In some workplaces, personal invitations were delivered via the contact person due to a data privacy policy. A kick‐off meeting was organised in all participating workplaces to introduce the study to the employees Demographics: 52.7% of women in Area 1 (11 workplaces) and 54.3% in Area 2 (5 workplaces). Mean (SD) age in Area 1 was 44.5 (32.3) years, and in Area 2 was 44.1 (11.4) years. In Area 1, 42.9% had a university degree, and 65.2 in Area 2
Interventions	Duration: 6 months, from Fall 2016 to Spring 2017 Intervention: the intervention was based on the SE approach, which integrates multiple life contexts relevant to behaviour change. SE models offer different pathways and levels to reach population groups, which may be underrepresented or unconnected with single‐level interventions. Each workplace in EXP nominated a team to plan and carry out social and behavioural strategies. Each workplace nominated a team to plan and carry out social and behavioural strategies to promote active commuting to work (ACW). The team selected the strategies most suitable for their workplace using a workbook. Selection of strategies (n = 43) was categorised into organisational (n = 24), working unit (n = 9), and individual (n = 10) levels. Teams were obliged to select at least 1 strategy from each level and to make an implementation plan for each strategy. Teams met with researchers 3 times to plan. In the first meeting, teams were given the workbook, examples, and supportive material. In the second meeting, they were assisted to finalise selection of strategies. In the third meeting, teams were helped to agree on practical arrangements such as timing and responsibilities related to implementation Control: workplaces participated only in data collection but were offered the chance to receive the intervention after the study
Outcomes	Measured at 6 months post baseline Physical activity Distance and minutes spent walking/cycling to work (travel diaries, accelerometer) Sedentary behaviour Accelerometer Adverse effects Injuries due to active commuting to work (self‐reported) Other outcomes Self‐rated health (single question) Subjective well‐being at work (four single‐item questions from the Work Ability Index (WAI)) Motivation for active commuting to work (three questions) Use of the main walking and cycling trail (fixed‐point monitoring stations) Process evaluation: compliance with the study, delivery of intervention strategies, employee's perceptions and awareness of strategies Cost‐effectiveness, cost savings (Health Economic Assessment Tool for Cycling and Walking (HEAT))
Starting date	September 2014
Contact information	Minna Aittasalo  UKK Institute for Health Promotion Research PO Box 30, 33501 Tampere, Finland +358‐3‐282‐9267 minna.aittasalo@uta.fi
Notes	Statistical analysis: generalised linear mixed models will be used to analyse effects of the intervention Sample size calculation: a sample size of 8 individual workplaces per group with 64 participants per workplace achieves 80% power to detect a group‐wise difference of 1 in the change in number of daily walking or cycling sessions to work per week when the standard deviation is 3.0 and the intracluster correlation is 0.05

Pillay 2012.

Study name	Steps That Count!
Methods	Randomised controlled trial Based on previous feasibility study (Pillay 2014) Aim: to investigate the effectiveness of a 10‐week pedometer‐based work site health promotion intervention (Steps that Count!) and individualised email‐based feedback to effect physical activity behavioural change (protocol)
Participants	Population description: employed adults (further description not reported) Intervention group: not reported Control group: not reported Location: selected work site settings based in the province of KwaZulu‐Natal, RSA, South Africa Inclusion criteria: employees attending the wellness event and willing to participate were eligible if aged between 21 and 49 years; identified as being in the contemplation stage of the Transtheoretical Model towards improved physical activity; who have a contract with their employer until the end of the 12‐week measurement period. In addition, participants must not be pregnant; must not be under diagnosis or treatment of cancer; must not have any other condition that makes physical activity difficult or impossible; must be non‐compliant for 3 or more days during the pre‐intervention blinded wearing of a pedometer Recruitment: to be undertaken through a Health Risk Appraisal available to all employees attending a corporate wellness event Demographics: Age: mean 32 (SD 8) Gender: 50% male
Interventions	Duration: 3 months Intervention: Participants will be encouraged to steadily increase their steps by approximately 10% per week until the target of at least 30 minutes of aerobic steps is achieved and maintained until the end of the intervention Unblinded pedometer (Omron HJ 750 ITC); data to be uploaded bi‐weekly Bi‐weekly email: individualised feedback will be given via personalised email and will include information on average daily steps accumulated; number of days (if any) that aerobic steps were accumulated, and the volume thereof; highest number of steps per day accumulated by the individual over the past 2 weeks; the category within which average steps per day fall; general supportive and motivational messages; and a few strategies to achieve step goals Control: no intervention received
Outcomes	Physical activity Pedometer steps: participants (intervention and control groups) to wear a blinded pedometer (Omron HJ 750 ITC), attached to the left or right hip during Weeks 1 and 12. Data will be downloaded electronically, and pedometer output will be expressed as steps per day. Step counts will be classified as aerobic (at least 60 steps per minute, minimum duration of 1 minute) and non‐aerobic (fewer than 60 steps per minute, less than 1 minute in duration, or both). Total time spent accumulating aerobic steps in minutes per day (aerobic time) and number (in hours) of sedentary time will be calculated Cardiovascular disease and type II diabetes risk factors (specifics not reported) Body mass index (BMI) Percentage body fat (%BF) Waist circumference (WC) Blood pressure: systolic and diastolic
Starting date	Not reported. Still ongoing as of March 2016
Contact information	Julian David Pillay UCT/MRC Research Unit for Exercise Science and Sports Medicine Faculty of Health Sciences University of Cape Town Cape Town, South Africa Email: pillayjd@dut.ac.za
Notes	Statistical analysis: "linear regression analyses will be performed with the follow‐up value of the outcome measure as the dependent variable and adjustment for the baseline value. Assumptions of linear regression analysis will be verified with residual analysis. To assess whether the differences in the primary outcome between the groups are affected by random differences between them, an analysis of covariance (ANCOVA) will also be undertaken" Imputation of missing data: not reported Adjustment for clustering: not reported Sample size calculation: "a sample size of 30 participants per arm of the study is required to ensure 80% statistical power and with a P value set at < 0.05. However, if a modest improvement of 1500 steps/d is considered, a sample size of approximately 85 participants per arm is required. In order to achieve this, 1200 employees attending wellness events will be targeted. Of these, a minimum of 480 employees (40%) will be identified to be in the contemplation stage of the Transtheoretical Model"

ACSM: American College of Sports Medicine.

ACW: XXX.

ANCOVA: analysis of covariance.

AQOL‐8D: Assessment of Quality of Life, Eight‐Dimensional.

BF: body fat.

BMI: body mass index.

BREQ‐3: Behavioral Regulation in Exercise Questionnaire.

DASS‐21: Depression Anxiety Stress Scale.

ECOs: Exercise Causality Orientations.

ESS: Epworth Sleepiness Scale.

HEAT: Health Economic Assessment Tool for Cycling and Walking.

MARCA: Multimedia Activity Recall for Children and Adolescents software.

MVPA: moderate to vigorous physical activity.

PNSE: Psychological Need Satisfaction in Exercise.

PSC: Premier's Sporting Challenge.

PSQI: Pittsburgh Sleep Quality Index.

PWI‐A: Personal Well‐being Index ‐ Adult version.

SD: standard deviation.

VPGC: Virgin Pulse Global Challenge.

WAI: Work Ability Index.

WC: waist circumference.

Differences between protocol and review

Intervention duration categories in the Subgroup analysis and investigation of heterogeneity section have been amended to coincide with follow‐up duration categories listed as Primary outcomes. Two categories of time points have been added: follow‐up after completion of the intervention (our primary points of interest, as they indicate sustained effect) and measurements taken immediately at completion of the intervention period. In each category, time points are grouped into short term (< 1 month), medium term (1 month to < 1 year), and long term (1 year or longer). We clarified that reporting of secondary outcomes was not restricted to studies in which a positive effect was observed for physical activity, and we noted that outcomes would not be reported if they were measured only in a selected subgroup of participants.

Several changes were made to the planned Secondary outcomes. As other body composition measures were available, body weight was not assessed as a secondary outcome in Secondary outcomes. Given the opposite direction of benefit for HDL and LDL cholesterol, we decided not to include the additional combined measure of total cholesterol. We also changed our decision to report the physical health component of quality of life measures in the previous version of the review in favour of mental health components, as we felt this was a more important complement to the other measures of physical health reported.

Some outcome descriptors have been changed for clarity and accuracy. Body composition outcomes are now referred to as anthropometric. Hypertension outcomes are now referred to as blood pressure. Biomedical outcomes are now referred to as biochemical.

ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform (ICTRP) trials registers were added as additional electronic sources searched. We had initially planned to search the Cochrane Public Health Group and the Cochrane Heart Group trials registers for relevant studies, but we determined that our search was sufficiently broad (and already included the Cochrane Central Register of Controlled Trials), and that this additional searching was not required. OSH UPDATE was searched up until 1 December 2014, but was not included in subsequent updates due to lack of relevant studies identified. Websites of relevant organisations were included in the first search for the review, and authors of included studies were contacted to identify any additional studies, but these procedures were not repeated in subsequent searches.

As noted in the Excluded studies section, an additional study was excluded because, although random allocation was used, only one workplace cluster was allocated to each of the intervention and control arms, which was not considered adequate to reduce the risk of imbalance of confounders between these two arms (Racette 2009). This was an additional criterion not originally planned at the protocol stage.

In the 2016 update of the search, we used the Cochrane Crowd crowd‐sourcing platform to perform one of the initial screens of titles and abstracts, specifically to identify reports of randomised trials. The resulting reduced set of records was then screened against the other eligibility criteria as normal.

As outlined in the Quality of the evidence section, the quality of evidence was assessed using the GRADE system. We initially planned to conduct GRADE assessments of outcomes for which three or more studies contributed data, but in the final review we conducted GRADE assessments for all pre‐specified main outcomes at the longest available follow‐up after completion of the intervention (as presented in 'Summary of findings' tables).

In the original protocol and in early versions of this review (Freak‐Poli 2011; Freak‐Poli 2013), we specified a threshold for considering studies at high risk of bias ‐ "three or more, excluding blinding". Given advances in our understanding of risk of bias over time, we considered this threshold to be lenient. Although blinding was unlikely to have been met by most of the studies included in the review, lack of blinding can have a serious effect on study outcomes. Hence, the risk of bias criteria have been re‐defined as studies have failed to achieve low risk of bias in three or more domains including blinding of participants and personnel and blinding of outcome assessment.

Contributions of authors

RFP designed and co‐ordinated the review, provided subject matter expertise, conducted additional searching, conducted study selection, contacted study authors, undertook data extraction and risk of bias assessment, conducted analyses, and drafted the review text. MC assisted with co‐ordination; conducted additional searching; participated in study selection; undertook data extraction, risk of bias assessment, and GRADE assessment; and conducted analysis. LAB conducted study selection; undertook data extraction, risk of bias assessment, and GRADE assessment; contacted study authors; and assisted with manuscript amendments. SC conducted study selection and provided subject matter expertise. AP conducted study selection and provided subject matter expertise. All review authors reviewed, amended, and approved the final text.

Sources of support

Internal sources

• School of Public Health and Preventive Medicine, Monash University, Australia

  Salary support for RFP and MC

• School of Sport, Exercise and Health Sciences, Loughborough University, UK

  Salary support for SC

• Institute for Evidence‐Based Healthcare, Bond University, Gold Coast, Australia

  Salary support for LAB

• Institute for Health Transformation, Deakin University, Geelong, Australia

  Salary support for AP

External sources

• Heart Foundation, Australia

  RFP receives support through a Post‐Doctoral Fellowship (101927, from 2018)

• National Health and Medical Research Council, Australia

  AP receives support through a Centre for Research Excellent Grant (from 2018). RFP received support through an Early Career Fellowship (to 2018). Cochrane Australia receives infrastructure funding, which covered in part salary costs for MC (to 2018)

Declarations of interest

None of the researchers have a commercial interest in the outcomes of this review. Two review authors (RFP, AP) would like to disclose that they undertook an independent research study, titled the Global Corporate Challenge® (GCC®) Evaluation Study, which evaluated the impact of a workplace pedometer intervention, which was not included in this review, as it was not a randomised controlled trial. The GCC® study was partially funded by the Australian Research Council (ARC) and the Foundation for Chronic Disease Prevention™ in the Workplace, which is associated with the Global Corporate Challenge®.      

New search for studies and content updated (conclusions changed)