2.3.1 Data sources and search strategy
A systematic search of the literature was conducted using CINAHL (1981-2015), EMBASE (1974-2015), AMED (1985-2015), Scopus (1960-2015) and SPORTDiscus (1985-2015) databases. First, three keyword categorical searches were conducted: (i) ‘exercise’, or ‘physical activity’, or ‘aerobic training’, or ‘resistance training’, or ‘circuit training’; (ii) ‘workplace’, or ‘worksite’, or ‘employee’; (iii) ‘intervention’. Second, categories i to iii were combined using ‘and’, and studies were limited to ‘human’ and ‘English’. Third, duplicates were removed. Finally, each of the retrieved studies was manually searched for potential inclusion in the review. One reviewer (doctoral candidate) assessed study inclusion based on the title and abstract. Full texts were retrieved for those articles that met inclusion criteria from the title and abstract, and final decisions on inclusions were made from the full text. The reference lists of each of the studies and a number of review papers and position stands were also manually searched to extract further studies. Studies published up to January 2015 were retrieved (and
39 later updated to September 2016 from database alerts) and managed using Endnote X7.0.2 software. Data were extracted independently from each study for primary and secondary outcomes. This review was conducted in accordance with the preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) [176].
2.3.2 Study inclusion and exclusion criteria
Studies must have met the following criteria to be included in the review: i) cohorts were generally healthy adults above the age of 18 years or had a health condition for which exercise is recommended as a treatment modality (e.g. diabetes, depression). If other types of participants such as members of the public were also recruited, only data from employees were extracted; ii) exercise interventions were conducted in a workplace setting and designed to improve the physical activity participation, fitness and/or health of employees and measured at least one of the primary outcomes listed below. Population-based physical activity studies (i.e. large-scale pedometer studies) or studies that did not describe details of exercise prescription (e.g. volume, mode) were not included; iii) studies were published in English and were conducted as either a randomised controlled trial (RCT), non-RCT, or uncontrolled trial; iv) studies measured at least one of the following primary outcomes, however were not required to have measured any of the secondary outcomes.
Primary outcomes:
1. Physical activity-related outcomes: objective (e.g. accelerometer) and subjective (e.g.
physical activity recall questionnaires) measures of daily activity and/or exercise participation.
2. Fitness-related outcomes: cardiorespiratory fitness, muscular strength and muscular
endurance. Secondary outcomes:
40
3. Anthropometric outcomes: body mass, body mass index (BMI), waist circumference,
hips circumference, fat mass and lean mass.
4. Metabolic outcomes: blood pressure, lipid profile, glucose, insulin and inflammatory
markers.
2.3.3 Risk of bias of individual studies
Assessment of the possible risk of bias of each study was facilitated with information collected using Chapter 8 of the Cochrane Collaboration tool for assessing risk of bias [177] which covers: sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data (e.g. dropouts and withdrawals), selective outcome reporting and other identified sources of bias (not including a possible publication bias whereby editors may favour the publication of studies demonstrating statistically significant findings). A judgement as to the possible risk of bias on each of the seven domains was made from the extracted information and rated as ‘high risk’ or ‘low risk’. If there was insufficient detail reported in the study the risk of bias was judged as ‘unclear’. One reviewer (doctoral candidate) assessed the risk of bias rating for each item. A judgement as to the overall risk of bias for each study was then made based on the risk of bias ratings for the seven domains [177].
2.3.4 Assessment of the quality of evidence of included studies
Quality of evidence was assessed across the domains of risk of bias, consistency, directness and precision using the GRADE system as outlined in Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions [177]. Results were determined as consistent for a particular outcome if the estimates of effect for interventions either favoured the treatment or control group in > 65% of studies. A study was labelled as direct if the stated goal was to improve the physical activity participation, fitness and/or health of employees, otherwise it was
41 labelled as indirect. Evidence for a particular outcome was considered direct if > 90% of the involved studies were labelled as direct. The precision of each study was based on whether estimates of variability were reported for each outcome of interest. Evidence for a particular outcome was considered precise if > 90% of involved studies reported estimates of variability. Quality of evidence was then adjudicated as high if none of these elements (i.e. risk of bias, consistency, directness and precision) were downgraded (further research is very unlikely to change our confidence in the estimate of effect); moderate if one of these elements was downgraded (further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate); low if two of these elements were downgraded (further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate); or very low if three or more of these elements were downgraded (any estimate of effect is very uncertain) [177].
2.3.5 Data synthesis and analysis
A systematic synthesis is provided with information presented in the text and tables to summarise and explain the characteristics and findings of the included studies. Meta-analyses were conducted for RCTs on all outcomes (Review Manager version 5.3.5, Cochrane Collaboration, 2014), with the exception of C-Reactive Protein (CRP) as only two studies using different measures (i.e. high-sensitive vs. standard CRP) reported this outcome. Standardised mean difference were used for effects to account for differences in measurement techniques between studies for particular outcomes (e.g. maximal vs. submaximal CRF assessment), otherwise mean difference was used. Random effects models were used to account for heterogeneity between studies where required, defined as a Cochrane’s Q statistic (Chi square)
p value < 0.1 and Cochrane’s I2 value > 50% indicating the presence of moderate-to-high
42 the same units of measurement. Where possible, units of measurement were converted to a common unit to allow statistical comparison between studies.