Secondary outcomes (tentative outcomes in the definitive trial)

Fig 2 gives an overview of the outcome measurement protocol that we planned to include in the definitive trial. We included multiple outcome measurements in the protocol at both baseline and follow-up. We included accelerometry for a one-week period (seven x 24 hours). The baseline and follow-up measurement of accelerometry period spanned eight days (the same days of the week twice during two separate weeks), including two weekend days, e.g. from Tuesday five pm to the following Tuesday five pm. Details regarding equipment and protocols for measurements other than accelerometry can be found elsewhere [20]. Briefly, participants collected saliva samples (three times in the morning and one time immediately before bedtime) on three consecutive days for later cortisol and cortisone assessment. The participants’ heart rate variability was measured for three consecutive days using the Firstbeat Bodyguard 2.0 device. Data from saliva sampling and heart rate variability measurements were included as indicators of physiological stress. Finally, we measured sleep using the Zmachine electroencephalography-based sleep monitoring system, for three consecutive nights. At baseline, we included an extra night of sleep measurement to get acquainted with the sleep protocol. Baseline and follow-up measurements were initiated on the same day of the week, whenever possible, to ensure comparability.

The planned primary outcome in the definitive full-scale SCREENS randomized controlled trial was children’s leisure non-sedentary time. A secondary outcome of the present study was therefore to examine compliance to physical activity measurements (accelerometry) in the context of the entire measurement protocol. Compliance to sleep assessment and physiological stress are reported elsewhere [22]. We employed accelerometry to objectively assess non-sedentary time and other physical activity measures using two Axivity AX3 (Axivity Ltd., Newcastle upon Tyne, United Kingdom) accelerometers, which measure acceleration in three planes. The accelerometers were worn in elastic belts around the hip and thigh for seven consecutive days at baseline and follow-up. The devices sampled at 50 Hz and the sensitivity of measurement was set to +/- eight g. The subjects were instructed only to remove the devices during water activities.

Based on algorithms developed by Skotte et. al. 2014 using a thigh-worn accelerometer in one-second epochs [23], we developed child and adolescent specific thresholds for the categorization of accelerometry into distinct daily body positions and physical activities (lying down, sitting, moving, standing, biking, running, and walking). The categorization via the algorithm into these activities based on these algorithms indicate high sensitivity and specificity (≥85.8% in all cases) in children similar in age to our study population [24]. We defined non-sedentary time as any activity not in a lying or sitting position (also including standing). We decided post hoc (a priori of knowledge of data in the ongoing full-scale SCREENS randomized controlled trial) that a valid day of measurement should not include >10 percent non-wear. Furthermore, we decided that a complete measurement period at baseline and follow-up should include at least four weekdays and at least one weekend day. To be compliant the data had to be complete at both baseline and follow-up. Non-sedentary time was summarized for all valid days then divided by number of valid days in those who met our compliance criteria, to get daily amounts, at both time points.

The identification of non-wear has been described elsewhere [20]. Briefly, using data on acceleration, temperature (individually estimated non-moving temperature) and predefined child awake time (06:00 AM to 10:00 PM), periods where the belts were not worn, were identified. During baseline and follow-up measurement, participants filled out a daily checklist, where schedule information (time of awakening, when arriving at and leaving work or school, as well as bedtime) was reported. Based on checklist data we were able to time annotate the accelerometry data (as well as SDU DT and TV data) and restrict to only leisure and awake time. A description of the handling of missing schedule data can be found elsewhere (under “Study documents” at registration {"type":"clinical-trial","attrs":{"text":"NCT04098913","term_id":"NCT04098913"}}NCT04098913 at clinicaltrial.gov). We then computed hours of non-wear at baseline and at follow-up, as well as the number of non-wear bouts (number of sessions where non-wear time was accumulated). Then the proportion of non-wear during each day was computed. The software OmGUI version 1.0.0.37 was used to set-up, extract, re-sample and convert the data. Raw accelerometry was processed using Matlab (Mathworks Inc., Natick, Massachusetts, US) release R2019a version 9.6.0.1099231.

In the survey used to recruit participants into the trial, highest educational attainment was obtained. Based on this information we categorized the individuals according to the International Standard Classification of Education.