Methods
Study design
This study was a randomised crossover trial. This manuscript focuses on secondary outcomes related to sleep quantity and quality and patterns of physical activity and sedentary time. The primary outcome has been published previously, see Gale et al.21 For further details, see attached the study protocol in online supplemental file 1.
Participants
This study was conducted in Dunedin, New Zealand. Thirty participants aged 18–40 years were recruited by word of mouth. A sample size of 30 participants was estimated to provide 80% power (5% significance) to detect a difference of 0.4 SD in glucose total area under the curve (which was the primary outcome of this study). Eligible participants were: non-smokers, not taking medications or supplements known to impact glucose or triglyceride metabolism, able to speak and understand English, without intolerances or allergies to gluten or dairy (these components were present in the test meals) and those who self-reported habitual sedentary time of more than 5 hours (work) and 2 hours (evening) per day. Participants were asked to obtain medical clearance if their responses to the Physical Activity Readiness Questionnaire indicated that physical activity may not be appropriate (n=1). Participants from across the body mass index categories (minimum 18.5 kg/m2, no upper limit) were recruited to ensure representation from all groups given the relationship between obesity and glycaemic control. All participants provided written informed consent.
Preliminary measures
Participants attended an introductory session at the University of Otago to confirm eligibility for enrolment. Blood pressure was measured using an automated sphygmomanometer (OMRON HEM-907; Omron Healthcare; Kyoto, Japan) and a correctly sized cuff. Participants were excluded if their systolic or diastolic blood pressure readings were greater than 130 mm Hg and 90 mm Hg, respectively. Standard height and weight were measured in duplicate following standard procedures. Experimental protocols were discussed, and participants watched a video that demonstrated the exercises. Participants practiced the required exercises under supervision from the study research assistant (Registered Dietitian) who was instructed on how to observe and correct technique by a member of the research team who has a degree in Exercise Science (MCP). On completion of primary measurements, participants were fitted with an ActiGraph GT3X+ (ActiGraph, Pensacola, Florida, USA) accelerometer to be worn continuously (24 hours per day) on their non-dominant wrist for seven consecutive days to capture habitual physical activity and sleep patterns. Participants were provided with a wear time diary to record non-wear time, what times they retired to bed, attempted to sleep and woke up. Participants were also asked to record any physical activity performed while not wearing the accelerometer (eg, swimming or contact sport) or to record activities known to be inaccurately identified by the accelerometer (eg, stationary cycling, certain resistance-based exercises or yoga).
Randomisation
Participants were randomised to complete the two experimental conditions in one of two possible orders (figure 1), stratified by weight status. The randomisation sequence was generated by MCP prior to recruitment using Stata (V.16; StataCorp, College Station, Texas, USA) and concealed electronically. The randomisation sequence was revealed and assigned on the afternoon prior to each participant beginning their first experimental condition. Participants were informed of their allocated sequence on arrival.
Figure 1CONSORT study flow chart. BMI, body mass index; CONSORT, Consolidated Standards of Reporting Trials.
Pre-intervention standardisation protocols
To minimise diet-induced variability on experimental days, participants were provided with a standardised breakfast, morning tea, lunch and additional snacks to be consumed before 14:00 hours on each experimental day. A detailed summary of the standardised meal protocol is reported elsewhere.21 Participants were fitted with an ActiGraph GT3X+ accelerometer for continuous wear on their non-dominant wrist from the morning of the intervention day to 48 hours after the intervention. In the 24 hours prior to the first experimental condition, participants were asked to avoid all moderate-intensity to vigorous-intensity physical activity. Participants verbally self-reported compliance with all pre-intervention protocols before each experimental session.
Intervention protocol
Details of the laboratory intervention sessions have been described previously.21 Each participant completed two 4-hour sessions, on the same day of the week, from 17:00–17:30 to 21:30–22:00 hours, separated by a minimum 6-day washout to eliminate carry-over effects (median 6 days, IQR 6–12 days). The intervention sessions were conducted on either Tuesday or Thursday evenings, to ensure the next day was a ‘typical’ weekday, rather than a weekend day. In the prolonged sitting condition, participants remained seated for the duration of the session. The regular activity breaks condition was identical, except participants interrupted sitting with 3 min of simple resistance exercises every 30 min. Each break involved three exercises (chair squats, calf raises and standing knee raises with straight leg hip extensions) for 20 s each over three rounds. Participants performed exercises in time with a video recording of a person performing the exercises in a time standardised manner, and included reminders about form and a timer. These simple, body weight resistance exercises were chosen as the mode of activity breaks for this study as they do not require equipment, can be performed on the spot and have been used previously.22 During the first session, participants were permitted to get up and use the bathroom as required and bathroom breaks were replicated during the subsequent session. While seated participants were able to watch television, read or work on a portable device during both conditions. Two standardised meals were provided during each condition at baseline and 2 hours. Sessions were supervised by two members of the research team. All participants completed every activity break, and no adverse events were reported during the breaks. Following the sessions, participants returned to their normal free-living environment with no further standardisation.
Physical activity and sleep data processing
For both periods of physical activity assessment (pre-trial habitual physical activity and the assessment of activity immediately prior to, during and after each intervention) time-stamped activity data were downloaded using ActiLife software (ActiLife V.6.13.4), saved in 15 s epoch and imported into Stata. Self-reported sleep and wake times were entered manually into ActiLife to constrain the Cole-Kripke algorithm23 that determined sleep period time (time between self-reported time attempted sleep and the wake time), wake after sleep onset (WASO (minutes spent awake between sleep onset determined by algorithm and end of sleep)), total sleep time (amount of time spent sleeping during sleep period time for example, sleep period time minus WASO), number of awakenings and sleep efficiency (how consolidated the sleep was). The intensity and duration of activity performed during self-reported non-wear time (eg, contact sport) were identified and manually overwritten in Stata. Sedentary time was classified as <2860 counts/min, with total physical activity represented by over this cut point (ie, ≥2860), which therefore combines light, moderate and vigorous activity.24 Valid wear time was classified as wear time ≥10 hours during waking hours.
Physical activity and sleep data were separated into two distinct time periods: intervention and post-intervention (online supplemental figure 1). The post-intervention period was defined as the 48-hour period following the end time of the experimental condition although each nocturnal period (defined based on self-reported attempted sleep and wake times) during the post-intervention period was analysed separately.
Statistical analysis
Thirty participants completed the study, however, two participants with missing data were excluded (n=1: accelerometer malfunction, n=1: removed accelerometer overnight). Twenty-eight participants were included in the analyses. To investigate differences between conditions, mixed-effects regression models were used with sleep and activity variables as outcomes, intervention condition as the independent variable and participant as a random effect. Mean differences, 95% CIs and p values were calculated. Residuals of models were plotted and visually assessed for homoskedasticity and normality. A p value of <0.05 was considered statistically significant. All analyses were carried out in Stata V.17.0 (StataCorp LLC, College Station, Texas, USA).
Time spent in physical activity and sedentary behaviour were reported in (1) absolute minutes and (2) proportions of the waking day. Both are reported because if sleep period time is different between conditions, then absolute minutes in activity and sedentary time would necessarily be different due to the 24-hour constraint of the day. In this situation, the difference in activity or sedentary time may not represent the effect of the intervention directly, but rather represent displacement of other time because of a change in sleep period time. Proportions, however, describe differences in time-use composition of the waking day, independent of sleep period time. Both are informative.
The first 24 hours was analysed as the primary time period to assess the acute effects of regular activity breaks in the evening. The full post-intervention period (48 hours) was analysed as the secondary time period to determine if any acute effects were apparent over 2 days.
As an increase in these sleep and activity variables can be either health promoting (sleep period time, total sleep time, sleep efficiency and physical activity) or not health promoting (WASO, number of night awakenings and sedentary time), a forest plot was created so that direction and strength of effects could be visually assessed more easily. For this, all mean differences and 95% CIs were standardised to be in units of SD.
Equity, diversity and inclusion statement
Our research and author team consist of women, junior, mid-career and senior researchers from different disciplines (Human Nutrition & Dietetics, Biostatistics Sleep and Exercise Sciences); however, all members are based at one University. We acknowledge that our study population is mostly well-educated, white women. We did not purposefully recruit marginalised communities, nor did we investigate the effects of merorganisation on the observed responses.