A previous review suggested that the inconsistent effects of school-based interventions on children’s PA may be caused by variation in the implementation of the interventions . Our results suggest that aligning the core mandates and curriculum-based obligations of a teacher with EOtC is a viable way to implement and thereby increase PA for boys in grades three through six. Furthermore, the intervention was implemented with low cost and effort, as the two-day seminar on EOtC practice was the only component of the intervention that required allocation of additional time and money. Otherwise, the intervention was designed to be time- and cost-neutral throughout the school year. We found no significant interaction between age and group, which indicates that EOtC had a positive impact on MVPA even for the older children, who are generally considered less active than younger children [31, 32]. However, it should be noted that PA still declined with age for both groups; a finding already well-establish in the literature [31, 32].
While our findings suggest that EOtC may be a cost-effective, supplementary strategy for creating a more active and varied school day, boys spent approximately double the amount of time in MVPA compared to girls in the EOtC group, suggesting that EOtC may contribute to the gap in levels of PA between sexes that is already well established in the literature [32,33,34]. This gap between sexes increases during the transition from childhood to adolescence as girls’ MVPA declines more, with onset at younger age and from a lower starting point at onset, compared to boys’ . In addition, findings from meta-analyses of the effectiveness of interventions targeting PA promotion in girls [36, 37] showed positive, but small, effects. This indicated that affecting behaviour change in the form of higher levels of PA is challenging in these groups. The same meta-analyses reported lower effectiveness of long-term implementation for girls of increasing age, indicating a need to develop interventions that provides positive and sustainable effects on girls’ PA behaviour starting from an early age [36, 37].
The greater effect of EOtC among boys is in line with research showing that the largest sex differences in children’s PA levels are seen in institutional contexts for self-organised PA compared to adult-led and structured activities, such as organised sport and PE . Similarly, a qualitative study of lived experiences set within an intervention aiming to increase PA during recess, concluded that the least active children increased recess PA through the inclusion of teacher-led play, but not free play . Generally, boys’ motivation for engaging in PA is intrinsic and girls’ to a larger extent driven by both intrinsic and extrinsic motivation . Having a teacher to create a social and supportive environment in which children are asked to engage in PA may therefore increase the extrinsic motivation for girls to be active. A large proportion of the least active children are girls in our study, and girls as group therefore could benefit more from adult-structured activities aiming to increase PA, such as PE or activities specifically aiming to integrate PA into school time .
During EOtC, the primary aim and the primary activities are not PA per se [15,16,17]; however, the outdoor environment and the structure of the teaching do provide children with more opportunities for being physically active of their own accord than the indoor classroom [20, 21]. A recent study in Australian children found that living in neighbourhoods with more green space was associated with higher odds of choosing physically active pastimes and lower odds of not enjoying PA for boys but not for girls . In our study, EOtC was often practised in green areas in the participants’ neighbourhood, which, in part, might explain the different effect on MVPA found between boys and girls. On this basis, we hypothesize that the opportunities provided for engaging in PA by the current EOtC practice in Denmark are mainly self-structured and therefore better suited to boys’ tastes, abilities, values, and motivation regarding physical activities. However, the previously mentioned meta-analyses targeting effectiveness of interventions on girls’ PA also showed that targeting multiple components, e.g. education and a change of environment, resulted in larger effect sizes [36, 37]. As such, a teacher instructing EOtC activities that specifically include sessions of activities with PA at a location providing better opportunities for PA than the classroom, might lead to positive effects of implementing EOtC practice on girls’ PA. Future analyses of PA segmented into domains, i.e. time spent in EOtC, recess, PE, and leisure time, and collecting qualitative data may provide a deeper understanding of how PA is accumulated for boys and girls engaging in EOtC practice.
The present study was the first quantitative, large-scale, controlled study investigating the effect of EOtC on PA. Strengths of the study included 1) parallel class design to obtain pairs for comparison that were alike regarding personal characteristics, 2) monitoring of EOtC practice to determine intervention fidelity and facilitate statistical analyses on both an ITT and a PP sample, and 3) valid and objective PA measurements using a strict inclusion criteria of seven days of 24 h measurements per participant. Limitations were that 1) we were unable to randomize class pairs to the EOtC and comparison groups, 2) the group of participants who were excluded had slightly different characteristics compared to those included in the analyses, 3) we had a higher non-compliance rate for children in the comparison classes compared to EOtC classes, 4) the effect of EOtC on PA was evaluated based on weekly PA, and thereby included all activities throughout the measured week, 5) seasonality was not included, and 6) we did not collect baseline data.
We were unable to randomize classes because of the way EOtC is organized in Denmark. Teachers, and thereby classes, were included in the EOtC sample because they were willing to practise EOtC regularly throughout the school year. We chose the parallel class design to minimize the risk of selection bias related to differing background characteristics between participants in the EOtC and comparison groups. Due to the random assignment into classes in grade 0 in Danish public schools , participants in each class pair should be comparable in terms of demographics, local area, overall school resources, and anthropometric characteristics. The data in Tables 3 and 4 comparing the anthropometric characteristics of participants included in the EOtC and comparison groups confirms the latter, as no statistical differences were found for age, sex and weight status between EOtC and comparison groups. However, 42% of participants in the EOtC group and 50% in the comparison group did not provide valid PA data. Also, participants excluded from analyses were more likely to be boys, older and overweight than the general study population. The retention rates observed may have caused the selection bias, as higher PA level, BMI percentile, and increases in age were negatively associated with lower back wear time of a tape-mounted accelerometer in our data collection . This is likely a consequence of the strict inclusion criteria and should be considered in the interpretation of the results , especially when it comes to older and overweight children who are generally considered less physically active than younger and normal weight children [31, 32].
We monitored the extent of implementation of the intervention in both EOtC and comparison classes during the PA measurements period and included both an ITT and PP analysis. Although not all EOtC classes reached the required minimum of 300 min of EOtC they were asked to do, and some comparison classes participated in activities categorized as EOtC, the EOtC classes averaged 175 more minutes of EOtC than comparison classes in the ITT analysis, and 288 more minutes in the PP analysis. The average duration of EOtC practiced during the week of PA measurements (239 min/week) was lower than the average reported across the school year for both EOtC classes (283 min/week) and control classes (64 compared to 98 min). This suggests that differences in weekly minutes of EOtC practiced during the week of PA measurements and an average week throughout the school year were similar and therefore a good representation of EOtC activities occurring throughout the intervention school year. Differences in the amount of EOtC performed in the EOtC and comparison group in combination with no differences in characteristics between EOtC and comparison participants, suggest that the differences in PA between groups can be attributed to EOtC. However, more detailed analysis of the impact of EOtC on PA could be conducted by comparing different days and contexts, such as days with EOtC, normal school days, days with PE, and weekend days. In addition, the amount of EOtC practiced was self-reported, and there might have been reporting errors. To reduce these possible errors we combined self-reported data on EOtC practice from two sources: class diaries and teacher reporting’s through the monitoring tool.
We successfully used tape-mounted monitors to obtain PA data with seven days of 24 h wear time and thereby removed or reduced the risk of shortcomings associated with current PA monitoring methodology in larger-scale studies . E.g. we did not have problems due to intentional non-wear, changes to the accelerometer’s position on the body and axis orientation, deciding and applying decision rules for detection of non-wear time, or the need to weight data due to differing number of days included. A study by Fairclough and colleagues  reported a need to include eight days for boys and 10 days for girls in order to achieve a 80% intra-participant reliability for whole day MVPA with a valid day set to include ≥10.1 h of wear time because of day-to-day variability. Data from our study showed a need to include 3.5 week days and 2.3 weekend days at 24 h wear time per day to achieve a 80% reliability for the PA construct “vector magnitude of three axes” . Based on these findings, our strict seven days of 24 h measurements inclusion criterion indicates a high reliability of measured PA.
Weather conditions impact PA accumulation of children with autumn and winter being associated with lower PA levels compared to spring and summer . The primarily outdoor nature of EOtC practice and our data collected only between November (late autumn) and April (early spring) may make our PA findings extra sensible to weather conditions. Also, we did not obtain baseline data and it is therefore unknown if children in the EOtC and comparison groups had different PA levels to begin with by chance, despite choosing a study design to achieve as similar groups as possible for comparison. Unfortunately, we were unable to apply a study design with baseline and follow-up measurement of PA for each class pair over the course of the entire school year.