Study design
Data for this two-arm parallel-group RCT was collected between October-December 2013. Treatment arms included a sit-stand workstation intervention group (each participant received a sit-stand workstation) and a control group (usual practice). Liverpool John Moores University ethically approved the trial. Participants and researchers were not blinded to group allocation.
Recruitment
Organisation level
Office workers from one organisation (Liverpool John Moores University, Liverpool, UK) were approached by the research team in August-September 2013. Consent was sought from 11 departmental managers for employee recruitment, installation of sit-stand workstations, study contact and laboratory visits during work time (Fig. 1). Departments were located across four buildings with varying office layout (open-plan, individual offices or a combination). Employees within the approached departments were predominantly administrative staff.
Individual level
Via an email from the research team, all employees in consenting departments received an overview of the study and participant information sheet, and were invited to a study information session (two sessions were organised per department). Employees were given 2 weeks to express interest. Interested employees were screened for the following eligibility criteria by the research team via telephone: a) full-time member of staff, b) access to a work telephone and desktop computer with internet, c) no cardiovascular or metabolic disease, d) not taking any medication, e) not pregnant and, f) no planned absence >1 week during the trial. If inclusion criteria were met, written informed consent was obtained and baseline assessments scheduled. There was no racial or gender bias in the selection of participants.
Group assignment and intervention
Following baseline assessments, participants were assigned by one member of the research team to a treatment arm using a randomised block design [23] and random number table. Departments served as blocks and participants within departments were randomly assigned at the individual-level to an arm [23]. Assignment of individual participants within each department alternated between arms (i.e. intervention, control, intervention, control…).
Intervention group
After baseline assessments, each participant had a sit-stand workstation installed on their existing workplace desk. A single (manufacturer’s suggested retail price £360) or dual (£375) monitor WorkFit-A with Worksurface + workstation was installed, dependent on the number of monitors the participant had. The computer monitor(s) and keyboard were housed on the workstation and the workstation could be quickly raised up and down by hand to enable seated or standing work. Participants were not prescribed an amount of time to use the station. Ergotron Ltd (www.ergotron.com) provided and installed the workstations in the standing position and gave participants basic face-to-face training and ergonomic information on correct workstation use. Participants received a web link to manufacturer ergonomic guidelines via an email from the research team (http://www.ergotron.com/tabid/305/language/en-AU/Default.aspx). No other behaviour change techniques were delivered, other than provision of the workstation. After end-intervention data collection, manufacturer staff uninstalled the workstations.
Control group
Participants were asked to maintain their normal work practices and received no intervention. Participants were offered the opportunity to have a sit-stand workstation installed for 8 weeks after all data collection.
Data collection
At baseline, 4 weeks (mid-intervention) and 8 weeks (end-intervention), participants’ office-based behaviours were assessed via ecological momentary assessment (EMA) diaries. At baseline and 8 weeks, participants attended University laboratories in the morning for individual assessments. Prior to laboratory visits, participants were required to fast for a minimum of 8 h, avoid the consumption of alcohol for 12 h, and avoid strenuous exercise for 24 h.
Outcome measures
Sitting, standing and walking time
The EMA diary assessed time spent sitting (primary outcome), standing, walking and in other activities during work hours over 5 days (Monday-Friday). At 15-minute intervals participants used a paper-based diary to record their main behaviour in response to the question: “What are you doing right now?” The behaviour options were sitting, standing, walking or other. If other was selected, participants were instructed to write the activity they were doing. EMA has been reported successfully in recent behavioural research [24, 25], including workplace research [26]. To promote compliance and minimise data loss, participants used the EMA diary to record the time they started and finished work each day, and consenting participants received one prompt to complete the diary via text and/or email at the start of each day from the research team.
Time spent in each behaviour per day (minutes/day: sitting, standing, walking, other) was estimated by multiplying the frequency of recordings by 15, based on the assumption that each behaviour episode occurred for the entire 15 min [25]. This method is assumed to provide valid estimates of time spent in behaviour categories when summed across a day, due to underestimation and overestimation errors cancelling each other out [27]. A diary day was considered valid if data entries were provided for ≥75 % of time spent at work (228/235 days at baseline; 218/235 days at 4 weeks; 193/230 days at 8 weeks). Time spent in each behaviour was calculated for each valid day and means were calculated from valid days. In accordance with previous trials [11, 12, 22] outcomes were standardised to an 8-hour work day to control for variations in work schedules [standardised minutes = outcome minutes * 480/observed workplace minutes]. To be retained for analyses, participants had to provide ≥2 valid days at each time point (met by 23 intervention and 21 control participants).
Vascular outcomes
Endothelial dysfunction is an early and integral manifestation of atherosclerotic disease [28] that strongly and independently predicts cardiovascular events in asymptomatic adults [15], and can be measured in the peripheral conduit arteries non-invasively using flow-mediated dilation (FMD). To this end, B-mode images of the brachial artery were obtained in longitudinal section at a reproducible point in the distal third of the upper arm using high resolution ultrasonography (Terason, t3000, Teratech) and a 10–12-MHz probe. Ultrasonic parameters were set to optimize the luminal-arterial wall interface with the focal zone set to the near wall. Once set, these parameters remained constant throughout the session and the probe was held in a constant position. Endothelium-dependent vasodilation was assessed by measuring FMD in response to 5 min of forearm ischaemia [29]. Briefly, a 1-min baseline measurement was taken, then a pneumatic rapid cuff inflator (Hokanson, Bellevue, U.S.A.), placed around the forearm distal to the humeral epicondyle, was inflated to 220 mmHg for 5 min [30]. Recording of the image ceased on inflation of the cuff and recommenced 30 s prior to deflation. Recording continued for a period of 3 min post cuff deflation [30]. Data was analysed post hoc by custom designed automated edge-detection and wall-tracking software, the validity and reproducibility of which have been previously demonstrated [31].
Carotid artery intima media thickness (cIMT) is an early subclinical marker of structural atherosclerosis. Increased cIMT independently predicts cardiovascular events, is correlated with cardiovascular risk factors [32] and as such cIMT is commonly used as a surrogate marker of cardiovascular disease risk. cIMT was measured in the carotid artery using high resolution ultrasound from three different angles over six consecutive cardiac cycles. Subjects were instructed to turn the head laterally, by approximately 90° to the left, and the same investigator undertook all measurements to ensure head positioning and stability were maintained. The common carotid artery was measured 2 cm proximal to the bulbous. A two-dimensional image of the artery was obtained with the near and far wall of the artery displayed as two bright white lines separated by a hypoechogenic space. Post-test analysis was performed using custom designed automated edge-detection and wall-tracking software. A region of interest of at least 1 cm was taken on the first frame of every individual study, including both vessel walls. Detection of the near and far wall lumen edges and the far wall media-adventitia interface was performed on all subsequent frames. The distance from the leading edge of the first bright line of the far wall (lumen-intima interface) to the leading edge of the second bright line (media-adventitia interface) indicated the cIMT [33].
Blood sampling
Fasting blood samples were obtained from the antecubital vein of one arm via standard venepuncture technique (Vacutainers Systems, Becton-Dickinson). Samples were collected into vacutainers containing EDTA or lithium heparin and stored on ice until centrifugation for 15 min at 1500 g at 4 °C. Plasma aliquots were stored at −80 °C for subsequent analysis. Plasma glucose, triglycerides and total cholesterol concentrations were determined spectrophotometrically using commercially available kits (Randox Laboratories, Antrim, UK). Each sample was analysed in duplicate.
Musculoskeletal outcomes
Using a questionnaire adapted from a previous trial [26], participants rated their current level of discomfort or pain at three sites (lower back, upper back, neck and shoulders) on a Likert scale from 0 (no discomfort) to 10 (extremely uncomfortable).
Anthropometric, sociodemographic, work-related and office environment characteristics
Using standard techniques [34] stature was measured to the nearest 0.1 cm using a Leicester Height Measure and body mass to the nearest 0.1 kg using a calibrated mechanical flat scale (both Seca Ltd, Birmingham, UK). Participants wore light clothing and no shoes. Body mass index (BMI) was calculated as mass divided by stature (kg/m2). Participants self-reported sociodemographic, work-related and office environment characteristics at baseline (age, gender, ethnicity, marital status, education attainment, smoking history, employment history, job category, office layout, number of people in office).
Acceptability and feasibility
At 8 weeks, participants in the intervention arm completed a 19-item five-point Likert scale (1 strongly disagree, 2 disagree, 3 neutral, 4 agree, 5 strongly disagree) adapted from a previous trial [35] to assess the acceptability and feasibility of the sit-stand workstation. Purposive sampling was employed to invite participants of the intervention arm (n = 23) to attend a semi-structured interview to discuss their experiences and perspectives of using the workstation. Recruitment emails were sent to all participants in the intervention arm. Seven female participants responded and took part, with interviews facilitated by the second author. The seven participants interviewed did not significantly differ to other intervention arm participants for any baseline characteristic (p > 0.05). Focus groups were conducted using a semi-structured interview guide to ensure consistency in interview approach. The semi-structured interview guides were designed to allow freedom in response whilst also ensuring a degree of commonality across the transcripts [36]. Interview questions were developed based upon a review of established literature [37] and the identified aims of the study. Sample questions included; “Please provide a brief overview of your experience of using the sit-stand workstation?” (prompts included the use of the workstation during the working day in terms of choice of tasks to stand and complete, patterns of use, time of day selected to stand) and “Can you reflect on the influence of the workstation on your working practices?”. Interviews took place in a familiar work setting, during work hours and within a space where participants could be overlooked but not overheard. Interviews lasted on average 13 min (range 8–17 min), were audio recorded and later transcribed verbatim. Verbatim transcripts were read and re-read to allow familiarisation of the data. Content analysis techniques were used to identify core and common themes in the data [38]. The process involved reading and re-reading text and assigning broad thematic codes. The lead (LG) and second author (RM) discussed and debated emerging themes in the data with reference to acceptability and feasibility of workstation use. Therefore a combination of inductive and deductive techniques was used to generate codes. Key emergent themes and participant quotes have been used to ensure authenticity in the represented data.
Sample size
Allowing for small drop out, the study aimed to recruit 25 participants per arm, and retain 23 participants per arm. A sample size of 23 per arm was chosen a priori to achieve 90 % power (alpha 5 %; two-tailed) to detect a minimum difference of 60 min/8-h workday between arms for workplace sitting time (primary outcome: expected SD of 60 min/day). 60 min was selected based on a recent protocol paper [39] and differences observed in similar trials [22, 26]. Data collection for vascular and metabolic outcomes would provide effect size estimates for power calculations in subsequent trials.
Statistical analyses
Data was analysed using SPSS version 22 (IBM, New York, USA) with the alpha level set at p ≤0.05. Intervention effects were compared at 4 weeks (sitting, standing and walking) and 8 weeks (all outcomes) from baseline using analysis of covariance (ANCOVA). The variable change score (4 or 8 weeks minus baseline) was the dependent variable, with intervention arm (control vs intervention) the independent variable. In all analyses, covariates were the baseline value for the variable to control for any imbalances at baseline [40]. Anthropometric, sociodemographic, work-related and office-environment characteristics were tested as potential confounders for all outcomes. Confounders were entered as covariates if significant associations (p ≤0.05) were observed with changes in an outcome and the effect on the mean difference between groups exceeded 20 % [41]. For changes in sitting, standing and walking time, baseline values of the other two behaviours were tested as potential confounders, though no effects on the mean difference between groups exceeded 20 % [41]. Adjusted change scores and 95 % confidence intervals (CIs) for the difference in change between groups are presented unless stated otherwise. Acceptability and feasibility data are reported as medians and quartiles.
Missing data and intention-to-treat analysis
Due to participant withdrawal, lost EMA diaries or the inability to conduct assessments, data were missing for all outcomes (Fig. 1). Accordingly, a per-protocol analysis was conducted and participants were excluded from analyses for outcomes they were missing data for. For workplace sitting, standing and walking, the per-protocol analysis was compared with an intention-to-treat analysis, as a sensitivity analysis. To treat missing data, the fully conditional imputation technique and ten imputation sets were used due to a low rate of missing data [42]. Imputation was based on all 47 randomized participants.
Minimum important differences analysis
Inferential statistics were ran using minimum clinically important difference principles, described elsewhere [43, 44]. Briefly, this approach makes inferences based on meaningful magnitudes and is recommended alongside hypothesis testing [43, 44]. A spreadsheet (see http://newstats.org/generalize.html) computed the quantitative and qualitative probability that the true effects were beneficial, trivial or harmful, after the outcome statistic, its p value, and the smallest/minimal important difference was entered. Minimum important differences for sitting and standing were 60 min/day, and for walking 10 min/day, as guided by a recent protocol paper [39] and differences in similar trials [22, 26]. Minimum important differences for other outcomes were determined through a distribution-based method as a Cohen’s d (standardized difference between change scores between groups) of 0.2 between-subjects standard deviations (SDs) [45]. The SD of pooled baseline data was used to negate the possibility of individual differences from the intervention influencing the SD at 8 weeks. For each effect at 8 weeks, quantitative probabilities for benefit, trivial and harm, and qualitative descriptors are reported. Effects were interpreted as unclear if probabilities for benefit and harm were >5 % [46].