The WorkACTIVE-P study was a 3-month parallel RCT performed between June 2016 and May 2018 (ClinicalTrials.gov Identifier: NCT02561611) by Pennington Biomedical Research Center (Baton Rouge, LA). The center’s IRB approved the procedures and protocol of the WorkACTIVE-P study, and all outcome measurements were collected by Pennington Biomedical Research Center staff. All participants provided written informed consent prior to enrollment, and a data and safety monitoring board oversaw the study. All outcome assessors and investigators were blind to participant allocation, whereas participants, the project manager and interventionists prescribing and supervising the intervention were not blinded (single-blinded). Cessation of the trial occurred when the desired sample size had completed the study.

We approached corporations in Louisiana that we have worked with previously for similar office interventions (Tudor-Locke et al. 2014a). The workplaces comprised a mixture of open- and closed-based spaces that had adequate room for our pedal desks. After meetings with the facilities’ directors, human resources representatives, information technologists, department supervisors, and wellness coordinators, email and telephone communications were utilized within a site to recruit participants. No racial or gender biases occurred during the selection of corporations and participants.

We recruited office workers from five corporations whose job descriptions mainly required sitting-based administrative duties such as filing, scheduling, interacting via email and telephone, attending meetings, etc. To be eligible, participants had to be 18-64 years of age, overweight/obese, have a waist circumference > 102 cm (men) or > 88 cm (women), and satisfy at least one of the four defining criteria for metabolic syndrome, namely triglycerides ≥ 1.7 mmol/L, high-density lipoprotein (HDL) cholesterol < 1.0 mmol/L (men) or < 1.3 mmol/L (women), resting blood pressure ≥ 130 mmHg systolic and/or fasting glucose ≥ 5.6 mmol/L. Participants also had to be sedentary; that is, they averaged < 7500 steps/day at baseline (as measured by step counters, described below) and indicated in response to direct query that they “mostly sit during the day at work and do not walk about very much”. Exclusion criteria included high resting blood pressure (systolic > 179 mmHg and/or diastolic > 99mmHg), cardiometabolic disease—Type 1 or Type 2 diabetes, serious arrhythmias, cardiomyopathy, congestive heart failure, stroke, transient ischemic cerebral attacks, peripheral vascular disease with intermittent claudication, and uncontrolled angina—and other medical conditions that would significantly impede conduction of the intervention or collection of outcome measures.

Interested participants received orientation, run-in, and screening visits to determine eligibility and indicate whether they could cope with the rigors of the study (Figure 1). The orientation was a 1-hour visit that detailed the procedures of the trial. During the run-in visit, descriptive characteristics, blood pressure, and medical history were obtained. Participants deemed eligible at this stage completed a screening visit at which a fasting blood sample was taken. Participants were additionally provided an accelerometer (GT3X+, ActiGraph, LLC, Pensacola, Florida, USA) and were instructed to wear this for seven days. This served to examine the participants’ ability to comply with the wearing requirements of objective physical activity monitoring devices. If participants failed to wear the accelerometer for at least four days with at least ten hours of wear time recorded on each day, their compliance was deemed inadequate and they were excluded. Participants who complied with this requirement were deemed eligible and the data obtained from the 7 days of monitoring was utilized to determine if participants were sedentary (< 7500 steps/day).

Approximately 16 days after the screening visit, eligible participants completed a baseline visit at which all baseline measures (see below) were assessed. Participants were also randomized into a 3-month intervention or control arm. Randomization occurred via a computerized pseudo-random number generator by the study statistician, with equal numbers allocated to each group. Participants were notified of group allocation through sealed and numbered envelopes, which were provided by the project manager or interventionist.

Participants assigned to the control group were asked to maintain their typical work and lifestyle habits. By contrast, those assigned to the intervention arm received a combined behavior support program that aimed to promote use of a pedal desk and increase daily walking. The intervention focused on optimizing two components of the Social Cognitive Theory: self-efficacy and social support (Bandura 1986). Specifically, goals were set for both interventional components and these were scrutinized during weekly or biweekly phone contacts and scheduled group meetings with interventionists who had backgrounds in kinesiology or psychology and had experience delivering behavior change interventions. Individual meetings were also scheduled with interventionists in the participants’ workplace when necessary and convenient. During contact time with participants, interventionists utilized methods that developed self-monitoring and goal fulfillment, including the provision of objective data detailing use and performance on intervention tools, as well as instructions on how to monitor these as part of the intervention. To further enhance feedback during contact sessions, data from intervention tools were transferred via the internet to interventionists, enabling remote programmatic oversight and tracking. Participant goals were highly specified and accounted for variations in work and personal schedules. For example, a participant may have been set a goal to use their pedal desk for 1 accrued hour/day but over time lift their goal up to 4 or more hours/day, depending on work schedules that may have required attendance at events/meetings outside of their office on short notice.

The Pennington Pedal Desk™, which has been described before (Schuna et al. 2016), has been rated positively during computer-based tasks in office-based workers (Proença et al. 2018). Briefly, the pedal desk provides a maneuverable, fully automated tracking and intervention system that encompasses the user, the desk, and specialized pedal desk tracker software (Schuna et al. 2016). Participants randomized to the intervention group had a pedal desk placed in their usual workspace. To promote prolonged pedaling, a non-adjustable magnetic braking system provided a flywheel resistance of ≈0.30 kilopounds, which aimed to yield power outputs of 12-36 W when pedal rates were 30–90 revolutions per minute (Schuna et al. 2016). Furthermore, in accord with the intervention’s goal to allow participants to monitor their usage, the pedal desk was designed to provide a real-time display of revolutions per minute and the time of pedal desk use via a pop-up box that appeared on the computer monitor (Schuna et al. 2016). Such tracking measures are important, as other measures of physical activity do not effectively monitor seated activity (Nelson et al. 2016). To facilitate attainment of pedal-based goals and address low usage, participants were trained to independently use aspects of the pedal desk tracking software, including how to interpret graphical displays in the pop-up box that illustrated time of use and distance, which was estimated from a product of wheel revolutions and wheel diameter (Schuna et al. 2016). These illustrations could present daily, weekly, and monthly accrued data, and were readily available to pedal desk users. We intended to transfer and retain this data via the internet within the system so interventionists could direct participants to achieve pedaling goals and we could accurately characterize pedal desk use. However, our software experienced technical faults when capturing the revolutions per minute data, and 15 participants in the intervention group (75%) reported problems with the tracking software. We consequently were unable to obtain estimates of power and energy expenditure as we had intended. Instead, we have used time of use as our determinant of pedal desk use as this feature worked more consistently.

The walking component of the intervention was based on the success of previous studies (Chan et al. 2004) and aimed to increase physical activity by ≥ 3000 steps/day. Participants in the intervention group were provided with Fitbit devices (Fitbit Zip, Fitbit Inc., San Francisco, CA) and instructed to wear these for the whole intervention except for water-based activities, such as showering, and during sleep. Fitbits were connected to internet monitoring software, which transferred data to an on-board visual display to track average steps/day. Participants were taught to monitor these data during the trial to facilitate the attainment of individualized stepping goals that were provided by interventionists. Additionally, interventionists obtained steps/day data each week from the Fitbit through internet connectivity and used these data to track compliance and assist participants in meeting specific goals during contact periods.

The study had 80% power to detect a 21% difference in VAT between groups at the nominal 0.05 whilst accounting for a 10% attrition rate (De Souza et al. 2012). Linear mixed models examined the influence of the intervention on change in outcome measures, with group used as a fixed effect and between-group differences the primary hypotheses of interest. Age, sex, and race were included in preliminary change from baseline models as covariates. Without clear-cut data for clinically meaningful differences in the primary endpoint, absolute Cohen effect sizes (ES) were performed on change scores to supplement findings. Calculations were made by dividing the difference between the mean values by the pooled standard deviation (SD) from the model (Cohen 1988). Cohen ESs were considered small, moderate, and large if they spanned from 0.20-0.49, 0.50-0.79, and 0.80 or greater, respectively (Cohen 1988). Analyses were performed using SAS software version 9.3 of the SAS System for Windows (SAS Institute, Cary, North Carolina). Significance level for all analyses was set at the nominal 0.05, and unless otherwise stated, data are presented as least square (LS) means (95% confidence interval [95% CI]).

In total, after exclusions, 20 participants were assigned to each group and all participants completed the intervention and were analyzed for the primary outcome (Figure 1). Trial cessation occurred when the targeted sample size completed the outcomes of the trial. The majority of participants recruited were female (31 [77.5%]) and white (28 [70.0%]), with a mean age of 46.4 (SD 10.5) years and a mean BMI of 31.4 (SD 4.4) kg/m². The control and intervention groups were not significantly different with respect to baseline demographic, anthropometric, and adipose tissue measures (P ≥ 0.07; Table 1). Sedentary time, light physical activity, MVPA, and steps/day, as measured by the ActiGraph, were also similar between groups at baseline (P ≥ 0.39), as were energy and macronutrient intake (P ≥ 0.27) and cardiometabolic disease risk markers (P ≥ 0.19; Table 1).

Participants in the intervention group used the pedal desk for a mean time of 127 (SD 61) min/day (Figure 2A). However, mean steps/day, as assessed with the Fitbit throughout the intervention period, were 6313 (SD 2371), indicating that the intervention did not successfully increase daily walking by the targeted 3000 steps/day (Figure 2B). Similarly, ActiGraph data showed participants in the intervention group did not increase steps/day (−83 steps/day; 95% CI = −946, 780; P = 0.85), and no between-group differences were seen (P = 0.51; ES = 0.19; Table 2; Supplementary Figure S1). The ActiGraph also revealed no differences in changes in sedentary time, light physical activity, and MVPA between the control and intervention groups (P ≥ 0.27; ES ≤ 0.31; Table 2; Supplementary Figure S1).

Similar changes in SAT and TAT were seen between groups (P ≥ 0.58; ES ≤ 0.15), yet there was a tendency for a within-group effect for VAT in the intervention group (−0.10 kg; 95% CI −0.22, 0.01; P = 0.08) and the between-group comparison showed VAT was reduced in the intervention group versus control (−0.15 kg; 95% CI = −0.29, −0.01; P = 0.04; ES = 0.59; Table 3; Supplementary Figure S2). Change in weight, BMI, and waist circumference from baseline to month 3 did not, however, differ between the intervention and control groups (P ≥ 0.30; ES ≤ 0.29; Table 3; Supplementary Figure S2).

The intervention engendered a 0.29 mmol/L (95% CI = −0.51, −0.06) reduction in fasting blood glucose relative to control (P = 0.01; ES = 0.70), but changes in fasting insulin and HOMA-IR were comparable between groups (P ≥ 0.29; ES ≤ 0.30; Table 3; Supplementary Figure S3). Change in HDL, LDL, total cholesterol, triglycerides, HbA1c, systolic blood pressure, and diastolic blood pressure were likewise not different between the control and intervention groups (P ≥ 0.11; ES ≤ 0.45; Table 3; Supplementary Figure S3).

The changes in total energy, carbohydrate, fat, and protein intake during the trial were similar between the intervention and control groups (P ≥ 0.11; ES ≤ 0.47; Table 4; Supplementary Figure S4).