Effects of Desk Pedaling Work Rate on Concurrent Work Performance among Physically Inactive Adults: A Randomized Experiment

Abstract

Purpose:

Under-desk pedaling devices could help reduce health risks associated with the global decline in work-related energy expenditure. However, the optimal pedaling work rate to facilitate concurrent work performance among physically inactive adults is unclear. We examined the effects of two light-intensity pedaling work rates on physically inactive adults’ work performance.

Methods:

We recruited equal numbers of older (45–65 years) vs. younger (20–44 years); male vs. female; and overweight/obese (BMI, 25–35 kg/m²) vs. normal weight (BMI, 18.5–24.9 kg/m²) participants. Using a Graeco-Latin Squares design, participants (n = 96) completed a laboratory experiment to evaluate the effects of using an under-desk pedaling device at two seated light-intensity work rates (17 and 25 W), relative to a seated non-pedaling condition on objectively-measured typing, reading, logical reasoning, and phone task performance. Ergonomic comfort under each pedaling work rate was also assessed. Equivalence tests were used to compare work performance under the pedaling vs. non-pedaling conditions.

Results:

Treatment fidelity to the 17 and 25 W pedaling work rates exceeded 95%. Mean work performance scores for each pedaling and non-pedaling condition were equivalent under alpha = 0.025. Age, sex, and BMI did not significantly moderate the effect of pedaling on work performance. Participants reported greater ergonomic comfort while completing work tasks at the 17 W relative to the 25 W work rate.

Conclusions:

Physically inactive adults obtained similar work performance scores under the 17 and 25 W pedaling and the non-pedaling conditions; suggesting that either pedaling work rate could help reduce health risks of sedentary worktime. The 17 W work rate yielded greater ergonomic comfort, and may be an appropriate starting point for introducing diverse inactive workers to under-desk pedaling.

INTRODUCTION

Most working-aged adults spend the majority of their waking hours in public or home-based workspaces (1), and these spaces significantly shape daily energy expenditure (2). Since 1960, increased workspace automation has led average work-related energy expenditure to drop by more than 100 calories per day; with approximately 80% of U.S. jobs now predominantly sedentary (2). The COVID-19 pandemic has further constrained physical movement to, from, and within workspaces; exacerbating the decline in work-related energy expenditure (3, 4). The health risks of prolonged sedentary worktime; including obesity, cardiovascular disease, type 2 diabetes, some cancers, and premature mortality; are especially pronounced among physically inactive adults—who may benefit from preventive interventions (5, 6).

Under-desk pedaling devices, space-efficient stationary cycles that fit under a desk or tabular surface, could contribute to preventive initiatives to reduce health risks of sedentary worktime (7–9). When pedaled while seated, under-desk pedaling devices can increase energy expenditure by about 70–90 kilocalories/hour over sedentary sitting, and facilitate weight gain prevention—without requiring employees to leave their desks (10, 11). Increasing light intensity activity, such as under-desk pedaling, by about 30–90 min/day may also improve cardiometabolic health (12, 13), and add about 3 years to physically inactive adults’ lifespans (14). As under-desk pedaling devices are available across rural and urban regions at a cost (~U.S. $50–$350) similar to an office chair, are quiet to operate, and can be safely used by deconditioned adults, they may contribute a scalable strategy to help address health risks of sedentary worktime (11).

To maximize the health benefits of using under-desk pedaling devices, while minimizing concurrent work-task interference, it is important to use under-desk pedaling devices at an optimal dose, or work rate. Appropriate pedaling work rates may help maximize long-term participation, satisfaction, and health-related outcomes across diverse workers, while preventing productivity- or health-related harms (8, 15). Implementing achievable and reinforcing under-desk pedaling work rates may be especially important for sustaining under-desk pedaling among physically inactive adults (16). However, dose-related intervention features such as pedaling work rate are often implemented based on educated guesses, and are rarely refined or rigorously tested prior to being deployed in larger-scale randomized trials (17).

Several small pilot studies with predominantly healthy, younger, non-sedentary adults suggested that seated, under-desk pedaling work rates of approximately 17 to 30 W are feasible for supporting concurrent office work (10, 18, 19), and can result in work performance that is similar to that attained while seated without pedaling (18, 19). Pedaling work rates <17 W appeared to induce ergonomic discomfort (19); while pedaling work rates ≥40 W appeared to induce greater perspiration (20), and be less acceptable for office settings. There is a lack of adequately powered trials that have investigated if pedaling work rates previously shown to permit concurrent office work among non-sedentary adults would be similarly feasible among physically inactive adults—who comprise approximately 76% of U.S. adults (21). Clarifying optimal pedaling work rates to enable concurrent ergonomic work experiences among physically inactive adults could help better engage a large proportion of adults in under-desk pedaling.

We aimed to conduct a randomized equivalence trial to examine the effects of seated under-desk pedaling at approximately 17 and 25 W, relative to a seated, non-pedaling condition, on concurrent work performance among physically inactive adults varying on age, sex, and body mass index (BMI). Based on prior research (18, 19), we hypothesized that work performance during each seated pedaling condition would not differ meaningfully from work performance while seated without pedaling.

METHODS

Participants

To facilitate generalization of findings to diverse physically inactive adults, we recruited equal numbers of older (45–65 years) vs. younger (20–44 years); male vs. female; and overweight/obese (BMI, 25–35 kg/m²) vs. normal weight (BMI, 18.5–24.9 kg/m²) participants. Inclusion criteria for all participants were: physically inactive with <150 minutes/week of reported moderate-vigorous physical activity and ≥6 hours of sitting on a usual weekday based on the International Physical Activity Questionnaire, short form (22); native English speaker; comfortable typing without viewing the keyboard (touch typist) to ensure participants were measured on similar performance requirements; hearing described as “good” or “excellent” to enable accurate measurement on auditory tasks; not pregnant; and no risk factors on the Physical Activity Readiness Questionnaire (23) that could compromise pedaling safety.

To recruit participants, a list of households within a 20-mile radius of the State College, Pennsylvania study site with ≥1 person within the study’s age range was generated from a commercial marketing database (InfoUSA). A total of 37,840 recruitment flyers were mailed to the identified households; with up to two repeat mailings sent to households that InfoUSA identified as having ethnic/racial minority residents. The flyers directed recipients to a web-based screener (Qualtrics, LLC). Recruitment information was also distributed via e-mail and in community settings (e.g., churches, food banks). Participants were compensated with the option of either a U.S. $50 Amazon or Walmart gift card. The study was approved by the Pennsylvania State University Institutional Review Board, and no study activities occurred before written informed participant consent.

Study Design and Randomization

We used a Graeco-Latin Squares design (24), to assess if seated under-desk pedaling at 17 and 25 W, relative to a seated non-pedaling condition, results in equivalent performance on typing, reading, logical reasoning, and phone tasks during an ~2-hour laboratory session. To control for confounding effects related to the order in which participants completed the pedaling and non-pedaling conditions, and the concurrently-performed work tasks, participants were randomly assigned to one of six pedaling sequences and task-type sequences (Fig. 1) using stratified permuted block randomization (block size of 6 with equal allocation; determined by statistician). Study staff used a randomization list to assign each participant, upon enrollment, to the next available sequence. Within each sequence, participants completed all pedaling and non-pedaling conditions and all work tasks—the only treatment difference was the randomized order of the conditions and tasks. Participants were blinded regarding the random allocation of the pedaling and task-type sequences.

Figure 1.

Study design and protocol. Two participants in each of the eight strata (2 × 8 = 16) were randomly assigned to each pedaling sequence and task-type sequence. A, Pedaling sequence: Using a Graeco-Latin Squares design, each participant completed three treatment periods: no pedaling, 17 W pedaling, and 25 W pedaling; the sequence of the treatment periods was randomized. As participants performed typing, reading, and combined logic-phone tasks across all three treatment periods, three parallel-form versions of each task with equivalent difficulty were required, and are denoted as Task Set 1, 2, and 3. B, Task-type sequence: The sequence of the task types; typing, reading, and combined logic-phone, were also randomized.

Laboratory Setup

The study was based in Pennsylvania State University’s environment-controlled Human Factors Laboratory. We designed a simulated office (Fig. 2) consisting of: (1) the DeskCycle under-desk pedaling device (3D Innovations, Greeley, CO); (2) an electrically height-adjustable desk (Jarvis Desk, Ergo Depot, Portland, OR); (3) four non-wheeled, non-swivel chairs (Bevco 1200 Plywood Chair, Bevco Precision Manufacturing, Waukesha, WI) selected to reduce excess chair motion, and preadjusted to 15, 16, 17, and 18 inch seat heights, respectively, to accommodate diverse anthropometric characteristics (19); (4) a rubber mat and a wooden anchor to constrain extraneous motion from the under-desk pedaling device during cycling; (5) a standard computer monitor, keyboard, and mouse; and (6) an iPad for the phone task.

Figure 2.

Under-desk pedaling device with pedaling speed monitor in simulated office setting.

The DeskCycle was kept at a resistance level of “2” throughout the study, and is factory-calibrated to ensure that torque values vary less than 2% between each cycling unit at fixed speeds of 30, 60, and 120 revolutions per minute (RPM) for each resistance level (3D Innovations, personal communication). We independently confirmed the DeskCycle’s calibration by comparing the power requirement (watts) of a DeskCycle used for approximately one year with a newly purchased DeskCycle. Each DeskCycle was set at a resistance level of “2” and the pedal crank was mechanically rotated at 40, 50, 60, 70, 80 and 90 RPM while computing the power requirements as a product of the electric voltage and current. In the validation setup, (1) a voltage transformer was used to control the electric voltage and current that rotated the shaft of the DeskCycle; (2) a laser tachometer was used to measure RPM; and (3) a torque transducer was used to measure torque. There was less than 1% difference (0.49%) between the mean power requirement of the two DeskCycles, indicating adequate calibration.

Experimental Conditions

In the standard non-pedaling condition, participants completed work tasks while seated at a desk without pedaling. In the 17 W pedaling condition, participants remained seated and completed work tasks while concurrently pedaling the DeskCycle at approximately 17.3 watts ±16% (equivalent to 67.2 RPM). In the 25 W pedaling condition, participants again remained seated and completed work tasks while concurrently pedaling the DeskCycle at approximately 25.0 watts ±16% (equivalent to 89.5 RPM). The 17 and 25 W work rates were selected based on our prior pilot work demonstrating that approximately 95% of participants could maintain each work rate within a ±16% error margin (19). To ensure sufficient separation between each pedaling work rate, we did not assess additional work rates in between 17 and 25 W.

To confirm the DeskCycle manufacturer’s specifications that pedaling at resistance level 2 at 67.2 and 89.5 RPM yielded a power requirement of 17.3 and 25 W, respectively, we measured the power requirements of the DeskCycle with a power output measurement system. This system (see Supplemental Figure, Supplemental Digital Content 1) recorded the DeskCycle’s torque and cadence using a: (A) variable transformer (POWERSTAT, Superior Electric, Bristol, CT, USA), (B) high torque drill (Bosch 1034VSR Drill, Bosch, Stuttgart, Germany), (C) rotary torque sensor (RSS-20, Transducer Techniques LLC, Temecula, CA, USA), (D) digital display (DPM-3, Transducer Techniques LLC, Temecula, CA, USA), (E) DeskCycle, (F) microcontroller (Arduino Uno R3, Ivrea, Italy) with hall effect sensor A3144 module, and (G) Arduino serial output for RPM measurements. We confirmed that the 17.3 and 25 W power requirements at resistance level 2 were consistent with the manufacturer’s specifications.

Treatment fidelity.

We developed a reliable sensor-based pedaling speed monitor to help participants maintain a consistent pedaling cadence at each work rate (25). A 3 mm disc neodymium magnet was attached to the flywheel of the DeskCycle and a Hall effect sensor A3144 module along with a microcontroller (Arduino Uno R3, Ivrea, Italy) was used to count every revolution as the magnet passed the sensor. The microcontroller computed the DeskCyle’s RPM, using the time difference between consecutive revolutions, and transmitted this information to the computer. We developed an app (Objective-C, Apple Inc., Cupertino, CA) to read the RPM data and presented a visual pop-up notification in the top right corner of the computer screen immediately after participants deviated from the targeted pedaling cadence ±16% (Fig. 2). The pop-up notification remained visible until the participant pedaled at the designated cadence for three consecutive revolutions. If participants did not adjust their cadence after ≥10 revolutions, the pop-up notification was followed by an auditory signal to further prompt fidelity to the planned pedaling cadence.

Work Task Selection

The typing, reading, logical reasoning, and phone tasks were selected to match common performance requirements for sedentary jobs (26). Because work performance was being repeatedly measured over the 17 W and 25 W pedaling, and non-pedaling conditions, we also selected work tasks meeting the following criteria: (1) multiple equivalent-form task versions available; (2) capacity for each task to be administered in ≤5 minutes to minimize the confounding effects of subject fatigue; and (3) work performance unlikely to markedly improve after completing two practice trials, based on prior research (25, 27, 28). The randomized study design provided further control for possible task practice effects.

Typing Task

Participants were given a 5-minute typing task using TypingMaster Pro Lite software (TypingMaster Inc., Helsinki, Finland). Participants viewed the typing passage in an upper window on the computer screen, while typing the passage into a lower window. Three equivalent-form typing passages; each at a moderately difficult syllabic intensity of approximately 1.3 (18), were selected from the TypingMaster software, and were randomized across the pedaling and non-pedaling conditions. As done previously, scores were computed as adjusted words per minute (AWPM) to account for the total words typed (speed) minus letter/punctuation errors (accuracy) (19, 29).

Reading Task

Participants were given up to 5 minutes to read a ~290-word passage, and answer five multiple-choice reading comprehension questions by computer. Three equivalent-form reading passages; each written at the average U.S. eighth grade reading level (30), were selected from Reading for Comprehension Level H (Continental Press Inc.; approved by New York State Textbook Law), and were randomized across the pedaling and non-pedaling conditions. Scoring was based on the number of seconds to read the passage and answer the questions (speed), and the percentage of questions answered correctly (accuracy/comprehension).

Combined Logic-Phone Task

Logic task.

The Baddeley Reasoning Test was used to assess the ability to apply logic and problem solve in unfamiliar situations (28). Participants were asked to respond “true” or “false” to a series of statements such as: “A follows B—BA,” “B precedes A—AB.” We prepared 128 items for computer administration (the original 32 items represented four times) to ensure a sufficient item quantity for the 5-minute experimental task. As done previously, we randomized the item sequence across the pedaling and non-pedaling conditions using a constrained randomization approach (28). Scoring was based on the number of items completed (speed) and the percentage of items answered correctly (accuracy). The Baddeley Reasoning Test has good test-retest reliability (r = .80–.82), construct validity, and sensitivity to environmental changes (28, 31).

Phone task.

We developed a tablet-based task that simulated phone call transfer requests in offices (iPad, Apple Inc., Cupertino, CA). To simulate office multi-tasking, the phone calls were programmed to occur once per minute during the 5-minute logic task. Each call arrived between the first 0–20 seconds (uniformly distributed) of each minute. Upon answering the call, a 5-second recorded message requested a randomly selected employee from a simulated phone directory with 9 departments and 37 names per department (e.g., “Hi, I’d like to speak to Arnold Perry in Administration please.”). Participants were given 20 seconds to locate and enter the requested employee’s department and 3-digit extension number, using the phone directory. After entering the requested information on the iPad, participants resumed the logic task. Scoring was based on the number of seconds required to locate and enter the requested information (speed), and the percentage of information entered correctly (accuracy).

Experimental Procedures

Participants were assessed individually, with two staff members present. We first measured participants’ height and weight, without shoes, to the nearest 0.1 cm and 0.1 kg, respectively, using a calibrated physician’s Detecto 439 scale (Detecto, Webb City, MO). Next, to optimize participants’ ergonomic set-up, participants: (1) sat in four chairs that varied only in their 15–18 inch seat height and selected their preferred chair (19); those selecting the lowest or highest chair had the option to further adjust its height; (2) placed the selected chair a comfortable distance from the DeskCycle; (3) adjusted the DeskCycle pedal straps (with staff assistance) to their comfort; (4) adjusted the desk height during slower- and faster-paced pedaling, to prevent their knees from bumping the desk; (5) adjusted the computer screen and keyboard position and made any other needed adjustments; and (6) were given the option to further adjust the desk height during subsequent non-pedaling work conditions.

Following the ergonomic adjustments, we administered two practice sessions. During the initial practice session, we first demonstrated how the sensor-based pedaling speed monitor operates to help maintain consistent pedaling speeds. Participants were then introduced to the typing, reading, and combined logic-phone tasks and asked to practice all tasks under the 17 W pedaling condition; while working as fast and as accurately as possible. To promote a steady pedaling pace while completing the work tasks, we asked participants to pedal for approximately 1 minute before beginning the full set of typing, reading, and combined logic-phone tasks and for approximately 30 seconds (while stretching their arms/back) in between each task. The typing and logic-phone tasks which could be programmed for specific durations were practiced for 1 minute, while the reading task (with a fixed passage length) was practiced for up to 5 minutes. Participants were given the option to practice each task as much as needed; 35.4% practiced at least one task at least one additional time while pedaling at 17 W. Next, participants took a 3-minute walking/standing stretch break and were given complimentary bottled water. Participants then practiced the typing, reading, and combined logic-phone tasks again under the 25 W pedaling condition, following similar procedures to those used for the 17 W condition. During both practice sessions we used work tasks that were separate from, but equivalent in difficulty to, those used in the main experiment. After completing the second practice session and before proceeding with the main experiment, participants took a 5-minute walking/standing stretch break; with more complimentary water provided.

During the main experiment, participants followed similar procedures to the practice sessions, but were given 5 minutes to complete each work task under the 17 W pedaling, 25 W pedaling, and seated conditions. Immediately after completing the 17 and 25 W conditions only, participants rated their perceived exertion (32) and ergonomic comfort while working. The ergonomic comfort questions were: (a) Are you comfortable typing while pedaling at this speed?; (b) Are you comfortable reading while pedaling at this speed?; and (c) Are you comfortable completing the logic/phone task while pedaling at this speed? Response options to each question were Yes/No. Following the main experiment, participants provided more information about their demographic characteristics via an online survey. Fig. 1 outlines the experimental protocol.

Statistical Design Considerations

Equivalence margins for pedaling vs. non-pedaling conditions.

As described previously (33), we defined equivalence, or feasibility for under-desk pedaling, a priori based on two standards: (1) the International Organization of Standardization ergonomic standard for computer keyboards indicates that average typing speeds with a new keyboard must not exceed 0.75 standard deviations of average speeds for standard keyboards (in the direction of poorer performance) to be acceptable (34); (2) in clinical research, a change of 0.50 standard deviations in health status sometimes is used as a basis for treatment modifications (35, 36). Using the approximate midpoint of these two standards, we defined equivalence, or feasibility, for the 17 and 25 W pedaling conditions as average work performance scores that do not exceed 0.60 standard deviations (in the direction of poorer performance) of average work performance scores during the non-pedaling condition (37).

Power analysis.

Sample size estimates were based on the primary outcome of AWPM typed. We assumed that an equivalence margin standardized by the standard deviation is 0.6 for the primary outcome (34–36). Using two one-sided t-tests for equivalence testing (38) with the significance level adjusted for multiple comparisons via the Bonferroni correction factor (alpha = 0.05/2 = 0.025), a sample size of 96 subjects yields >95% power to declare equivalence when the true difference between the means is zero.

Data Analysis

For equivalence tests, we set the significance level at alpha = 0.025 for each work performance measure to adjust for multiple comparisons between the 17 and 25 W conditions and the non-pedaling condition. All other statistical tests were considered statistically significant at alpha = 0.05. Analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC), and SPSS Statistics for Windows version 26 (IBM Corp., Armonk, NY).

Treatment fidelity.

We used 2-sided paired-sample t-tests to assess if, as planned, participants had more RPM and greater perceived exertion during the 25 W condition than during the 17 W condition, for each work task separately. We also used 2-sided paired-sample t-tests to assess differences between the 17 and 25 W conditions, for each work task separately, in the proportion of participants’ RPM within the target cadence range (±16%), and their comfort performing each task.

Effects of experimental conditions on work performance.

To test whether pedaling at 17 and 25 W resulted in equivalent work-task performance as compared with the non-pedaling condition, we applied the equivalence test of means based on two one-sided t-tests (38) for each continuous work performance measure. Linear mixed effects models (39) were used to estimate mean work performance outcomes for the 17 and 25 W pedaling conditions and the non-pedaling condition while adjusting for pedaling sequence, task-type sequence, task set, and period effects (Fig. 1). An unstructured covariance matrix was specified in the mixed effects models to account for the correlation between the repeated measurements within the same subject under each condition. We constructed the confidence interval for equivalence using the approach of Berger and Hsu (40). Specifically, we obtained the 97.5% confidence interval for equivalence by taking the 95% confidence interval for the difference from mixed effects models and setting the lower limit to zero if positive and the upper limit to zero if negative. We concluded equivalence at a significance level of 0.025 if the 97.5% confidence interval was contained within the equivalence margin (40).

Subgroup differences.

As individual factors including age, sex, and BMI may moderate the association between experimental conditions (17 and 25 W pedaling, non-pedaling) and work performance (41), we fitted linear mixed effects models by adding these individual factors as covariates, along with their interactions with experimental conditions, one at a time. Significant interaction terms indicated that these factors moderated the relationship between the experimental conditions and work performance.

RESULTS

Sample Characteristics

Ninety-six participants provided valid data and were included in analyses (Fig. 3). Data for 9 participants were excluded due to: (a) technical malfunction that led speed monitor to stall repeatedly during experiment (n = 6); and (b) participant factors, including random responding and not following study directions (n = 2); or unable to pedal at targeted intensity levels (n = 1). Most participants were Caucasian, had moderate-high education levels, and had high levels of sedentary behavior. Participants’ occupations spanned diverse occupational categories (42) (Table 1). No adverse events were reported for any participant.

Figure 3.

CONSORT diagram: participant flow through trial.

TABLE 1.

Participant characteristics.

Characteristic	Participants (n = 96)a
Age, (mean ± SD), yr
20–44	31.7 ± 5.7
45–65	55.8 ± 6.2
Male, n (%)	48 (50.0)
BMI (mean ± SD), kg/m2 b
18.5–24.9	23.2 ± 1.4
25.0–35.0	30.0 ± 2.8
Ethnic/racial minority, n (%)	11 (11.5)
Education, n (%)
Completed high school/GED or some college	18 (18.8)
Completed 4-year college, but not graduate degree	39 (40.6)
Completed graduate degree	39 (40.6)
Occupational classification, n (%)c
Managers (e.g., administrative, production, hospitality, retail)	18 (18.8)
Professionals (e.g., engineering, health, teaching, business)	35 (36.5)
Technicians and associate professionals (e.g., health, business, cultural)	27 (28.1)
Clerical support workers (e.g., general, customer service, data recording)	12 (12.5)
Sales and service workers (e.g., sales workers, personal service or care)	2 (2.1)
Unemployed	2 (2.1)
Moderate-to-vigorous physical activity, (mean ± SD), min/wk d	50.1 ± 35.0
Time spent sitting in usual day, (mean ± SD), h/day d	10.3 ± 3.0
Time spent working on computer in usual day, (mean ± SD), h/day	7.3 ± 2.4
Self-rated health, (mean ± SD)e	3.5 ± 0.7

^aBy design, the sample was half younger (20–44) vs. older (45–65), half male vs. female, and half normal weight (18.5–24.9) vs. overweight/obese (25.0–35.0).

^bObjectively-measured body mass index.

^cBased on International Standard Classification of Occupations (42).

^dBased on International Physical Activity Questionnaire, short form (22). Moderate-to-vigorous physical activity computed as the sum of all moderate (including walking) and vigorous activity minutes.

^eSelf-rated health: 1 = poor, 5 = excellent.

Treatment Fidelity

Adherence to the planned 17 and 25 W pedaling work rates exceeded 95% across all work tasks (Table 2); between 80%–100% adherence to a targeted standard is generally considered high treatment fidelity (43). In addition to receiving each planned pedaling work rate, participants differentially experienced each work rate (Table 2); with greater comfort reported at the 17 W relative to the 25 W work rate.

TABLE 2.

Treatment fidelity: pedaling work rate received and perceived (n = 96).

Treatment fidelityindices	17 W			25 W
	Type	Read	Logic-Phone	Type	Read	Logic-Phone
Work rate received
RPM (mean ± SD)a	64.2 ± 4.6*	65.5 ± 5.4†	66.9 ± 4.4‡	79.4 ± 5.1*	81.0 ± 5.8†	81.6 ± 5.8‡
RPM, % at target cadence (mean ± SD)	98.0 ± 3.7*	98.6 ± 1.6	97.0 ± 5.4	95.3 ± 6.0*	98.4 ± 2.3	97.0 ± 5.8
Work rate perceived
Comfortable doing task at this pedaling speed, % indicating “yes”	93.8*	96.9	94.8§	74.0*	93.8	85.4§
	17 W—across all task types			25 W—across all task types
Rating of perceived exertion, (mean ± SD)b	9.5 ± 1.7‖			11.4 ± 1.9‖

RPM, revolutions per minute.

^aSignificant difference in the mean RPM between the 17 and 25 W conditions supports treatment fidelity to the planned pedaling work rates.

^bSignificant difference in the mean rating of perceived exertion between the 17 and 25 W conditions supports treatment fidelity to the planned pedaling work rates; 9 = very light exertion, 11 = light exertion.

^{*, †, ‡, ‖}Across each row, items sharing the same symbol differ significantly at P < 0.0001.

^§Across each row, items sharing the same symbol differ significantly at P = 0.006.

Effects of Experimental Conditions on Work Performance

Adjusted mean work performance scores for the typing, reading, logical reasoning, and phone task outcomes were estimated and compared between the 17 and 25 W pedaling conditions and the non-pedaling condition (Fig. 4). All 97.5% confidence intervals for mean work performance outcomes were within the equivalence margin; indicating equivalence for each pedaling and non-pedaling condition under alpha = 0.025.

Figure 4.

Effects of pedaling vs. non-pedaling conditions on work performance (n = 96). A, Typing: adjusted words per minute; B, Reading: total reading time and percent of comprehension questions answered correctly; C, Logical reasoning: total questions done and percent answered correctly; D, Phone: total time to enter/transfer extension information, and percent entered correctly. Graphs A-D present mean scores adjusted for sequence and period effects; error bars indicate standard error of the mean. E, Equivalency evaluation: the mean difference in work performance scores between the pedaling vs. non-pedaling conditions, and the corresponding 97.5% confidence intervals, were all within the equivalence margin; indicating equivalence for each pedaling and non-pedaling condition under alpha = 0.025.

Subgroup Differences

The effects of the 17 and 25 W pedaling conditions vs. the non-pedaling condition on all work performance scores was not significantly moderated by demographic factors, including age (F-value range for moderation effects: 0.02–3.00; P-value range: 0.055–0.976), sex (F-value range for moderation effects: 0.18–2.35; P-value range: 0.101–0.840), and BMI (F-value range for moderation effects: 0.20–2.45; P-value range: 0.092–0.818).

DISCUSSION

To our knowledge, this is the first adequately-powered randomized experiment to investigate the effects of different under-desk pedaling work rates on concurrent work performance in physically inactive adults. As hypothesized, we found that pedaling at 17 and 25 W, relative to a non-pedaling condition, yielded equivalent work performance outcomes in a controlled experiment. Physically inactive adults obtained similar work performance scores on typing, reading, logical reasoning, and phone tasks, regardless of whether they were concurrently engaged in seated pedaling at 17 or 25 W, or remained seated without pedaling. As participants reported greater comfort at the 17 W work rate relative to the 25 W work rate for most work tasks, future desk-pedaling programs may benefit from recommending an initial work rate of ~17 W for physically inactive workers. Variations in age, sex, and BMI did not significantly moderate the effect of pedaling on concurrent work performance; suggesting that under-desk pedaling could help multiple subgroups of working-aged adults to reduce health risks associated with sedentary worktime.

Our finding that work performance was equivalent across the 17 and 25 W light-intensity pedaling, and the non-pedaling conditions, is consistent with results from prior pilot studies—which found small or null differences in work performance when comparing light-intensity pedaling to non-pedaling conditions (18, 19). Our study extends these prior studies by including only physically inactive adults—a key priority group to target in national and global workplace physical activity interventions (5–9). We also uniquely documented treatment fidelity to the planned pedaling work rates—which is important for clarifying the precise work rates tested, and for enabling these work rates to be replicated in both laboratory and real-world workspaces. The consistent pattern of equivalent work performance across the light-intensity pedaling and the non-pedaling conditions, and across multiple work-task types, increases confidence in the reliability of study findings. These findings contribute to the evidence-base needed to help identify a feasible range of desk-pedaling work rates for future interventions with physically inactive adults.

There has been limited research on optimal pedaling work rates to facilitate ergonomic work performance during under-desk pedaling, despite the availability of more than 50 models of under-desk pedaling devices with adjustable intensity levels. Clarifying ergonomic pedaling work rates may help sustain desk pedaling among physically inactive adults, whose pedaling volume in desk pedaling interventions has typically declined over time (44–46). By demonstrating that physically inactive adults prefer a ~17 W pedaling work rate, and that work rates between ~17–25 W permit productive concurrent work-task performance, this study may provide a starting point to help inform future pedaling work rates that can be recommended, and further tested for their long-term effects on worker engagement. Future studies may benefit from exploring the effects of different combinations of pedaling resistance and cadence within a ~17–25 W range to help optimize pedaling ergonomics for different demographic subgroups. Within the current study, participants’ greater ergonomic comfort at the 17 W (67.2 RPM) relative to the 25 W (89.5 RPM) work rate likely reflected the lower pedaling cadence at 17 W. As pedaling resistance was kept at a constant level, the upper/lower pedaling resistance limits that permit concurrent ergonomic work performance remain unclear—and warrant further investigation. Although rarely examined, there is growing recognition that ergonomic factors, such as precisely fine-tuning pedaling metrics, may contribute to long-term intervention engagement (15, 17).

Efforts to engage diverse subgroups in under-desk pedaling may also benefit from understanding if demographic factors differentially influence the ability to pedal and perform concurrent work tasks (41). To our knowledge, this is the first study to investigate if age moderates of the effect of pedaling on concurrent work performance; with no significant moderation effect observed. Adults’ sex and BMI also did not moderate the effect of pedaling on concurrent work performance, which is consistent with prior desk pedaling experiments reporting no effect of sex (18) and BMI (47) on work performance; but inconsistent with experiments reporting enhanced typing speed for women, relative to men, when engaged in pedaling (10, 48). As only a handful of desk-pedaling trials have investigated effects of demographic moderators, future trials should further explore if there are unique effects for demographic subgroups that warrant tailored intervention approaches. Current findings support the feasibility of using under-desk pedaling devices to boost work-related energy expenditure among multiple demographic subgroups.

A key strength of this study was its emphasis on balancing external with internal validity. To maximize external validity and generalizability to real-world workspaces (49), we recruited a sample with diversity on age, sex, and BMI; measured diverse work-related subskills; and used a popular commercially-available under-desk pedaling device. Simultaneously, to maximize internal validity while using a consumer-grade pedaling device, we uniquely incorporated a sensor-based pedaling speed monitor to increase the precision with which different pedaling work rates were tested. Receipt of the planned 17 and 25 W pedaling work rates was confirmed based on objectively-measured treatment fidelity exceeding 95% for each work rate, and by participants’ differential perceived exertion and comfort at each work rate. Structured experimental protocols, with tracking for treatment fidelity were also used to ensure consistent treatment of each experimental participant. The study’s Graeco-Latin Squares design (24) further strengthened internal validity by enabling control for person-specific factors while randomizing the order of all independent (pedaling vs. non-pedaling conditions) and dependent (work tasks, and equivalent-form versions of tasks) variables. Finally, we used equipment to reduce extraneous motion from the chair and under-desk pedaling device during cycling—treatment confounders that received little attention in other desk-pedaling studies (19, 47).

Despite efforts to balance external and internal validity, our study has some limitations. First, participants were relatively highly educated with limited racial and ethnic diversity, and results require further confirmation with more diverse groups. Second, participants were identified as physically inactive based on their self-reported physical activity (22), and objective activity monitoring may have provided more complete information about participants’ activity levels. Third, the study was based in a controlled laboratory setting, and the degree to which findings may generalize to public and private workspaces warrants further exploration. Fourth, the study was not designed to investigate the interactive effects of different pedaling cadence and resistance settings, and future studies may benefit from investigating if different combinations of these variables could further optimize ergonomic comfort. Finally, because work tasks were done for only 5 minutes at a time, findings may not generalize to tasks performed over longer durations. We selected a 5-minute task duration because short and repeated 5-minute pedaling activity bouts can contribute to meeting physical activity guidelines (50); while remaining feasible for sedentary, deconditioned adults. However, viable time ranges for performing pedaling and concurrent work tasks should be further explored.

CONCLUSIONS

In conclusion, across two light-intensity desk-pedaling conditions and a non-pedaling condition, physically inactive adults obtained similar work performance scores on typing, reading, logical reasoning, and phone tasks. These findings suggest that light-intensity desk-pedaling activity could provide a time-efficient and feasible option to mitigate health risks of sedentary worktime among physically inactive adults. By demonstrating that ~17–25 W pedaling work rates enabled productive concurrent work-task performance, while ~17 W work rates conferred greater ergonomic comfort, this study can inform the selection of pedaling work rates in future efforts to disseminate and evaluate under-desk pedaling devices more broadly.