The present study showed an intervention effect on overall physical activity at the 5% alpha level. The intervention effect appeared to be more profound among girls, and among participants in the low-activity group compared to boys and to participants in the high-activity group, respectively. Further, the intervention appeared to have a stronger effect among normal weight participants and participants with parents reporting 13–16 years of education compared to their counterparts.
With an intervention effect at alpha level 0.05 there is a degree of uncertainty to the results that needs to be considered. There is a 5% chance that the findings are not attributed to the intervention, which means the greatest value of uncertainty conventionally accepted before the findings are dismissed as non-significant. Keeping this in mind, the intervention effect on total physical activity is somewhat in contrast to results from the KISS intervention; a Swiss cluster randomized controlled school based physical activity programme. The KISS study, while comprising a bit younger participants, showed a favourable intervention effect on moderate to vigorous activity at school and all day, and also on total physical activity at school, but no effect on overall daily physical activity
]. No intervention effect on overall physical activity was shown in the Danish CoSCIS study either, with an intervention including a doubling of time for PE among 6–7 year olds
]. Compared to the KISS programme and the CoSCIS study, the HEIA intervention had less promotion of high intensity activities but focussed on increasing overall physical activity. While the HEIA study used a multi-facetted approach to increase physical activity including several small reminders and opportunities to increase all day physical activity level, the KISS study was oriented toward PE and using expert PE teachers and extracurricular mandatory PE. The CoSCIS study also used PE as their main intervention component, including a doubling of lessons per week, teacher training and an upgrade of PE and playing facilities. From the effect analyses it is not possible to disentangle specific intervention components to account for our findings. Some intervention components may have been more effective than others, or results may reflect synergistic effects of the intervention program as a whole. Thus, in concordance with suggestions in recent reviews
], the HEIA study aimed to affect physical activity in adolescents through multiple components and by combining personal, social and physical environmental factors. The increase in physical activity from baseline to post intervention in the control group was unexpected, as previous literature has shown decreasing physical activity with increasing age in youth
]. Since both groups increased, an increase as a result of the intervention may have been harder to detect, yet the intervention showed a positive effect. The intervention group was significantly less active than the control group at baseline, and it can be argued that the intervention group had a larger potential for change. However, these issues were taken care of by controlling for baseline-values in the effect analyses.
The relatively large increase in physical activity from baseline to post intervention in both groups can be attributed to seasonal variation. The baseline physical activity assessment was conducted during fall and post intervention assessment during spring. Kolle et al. (2009) observed seasonal variations in physical activity among 9 year old Norwegian children, but not among 15 year olds
]. The intervention effect should, however, not be affected by season, as both groups were measured simultaneously. The increase might also be a result of contamination effects of being the control group in a study aimed at increasing physical activity. When recruiting schools, most schools stated that they were hoping to become an intervention school to receive the intervention efforts. This could have stimulated the control schools to initiate their own “intervention”.
The overall increase in physical activity from baseline to post intervention was seen both on weekdays and weekend days, but with a larger increase on weekend days. The larger increase during weekend days may reflect the larger potential for change since the baseline values within that period of the week were considerably lower than during weekdays. The intervention components addressed both weekday and weekend day activity. The finding that the physical activity level was higher during weekdays than weekend days is consistent with earlier cross-sectional findings from Norwegian 9 and 15 year olds
The participants’ mean distribution of activity in our study differed between the two time points. Physical activity during school hours declined and physical activity after school hours and during weekend days increased for both groups and both genders. The decline in physical activity at school might be due to more demanding school curricula in 7th grade than 6th grade, and happened despite several intervention efforts aimed at increasing physical activity at school. A reason for the demonstrated decline in physical activity during school hours may also be a lack of facilities perceived as attractive by the adolescents as they grow older. Nettlefold et al. (2011) studied physical activity during the school day in Canadian 8–11 year olds and observed low physical activity during parts of the school day
]. The authors pointed out an urgent need to increase the intensity of activity during PE, and to provide more and/or facilitated opportunities for physical activity during school breaks. Haug et al. (2010) found that outdoor facilities in Norwegian secondary schools were associated with students’ daily physical activity participation during school breaks
]. Students in schools with many facilities had significantly higher odds of being physically active compared to students in schools with fewer facilities
]. The activity increase in both groups after school hours is hard to explain. A possible reason may be increased volume of exercise in leisure time sports activities with increasing age. The participants may also have been stimulated to increase leisure time physical activity in line with the HEIA study aims. There was, however, no intervention effect on these outcomes. Concerning time spent at different intensity levels, no intervention effect was seen for time spent in MVPA. Nevertheless, this was not a targeted aim of the study. However, reducing sedentary time was a clear aim of the study but no intervention effect was seen for the total sample. Exploring subgroups, boys appeared to have higher overall physical activity on all time points than girls, but the difference in increase from baseline to post intervention was significantly higher among girls in the intervention group compared to girls in the control group. The gender difference in intervention effect was also seen with time spent at different intensity levels as outcome. Girls in the intervention group increased significantly less in sedentary time from baseline to post intervention than girls in the control group. This is promising, as a recent comprehensive systematic review revealed a dose–response relationship between increased sedentary behaviour and unfavourable health outcomes in school-aged children
]. When the intervention strategies were planned and developed, the study group had a particular focus on making sure that it should appeal to inactive girls. By offering low threshold activities the aim was to make the physically less active participants want to take part rather than fear to take part. Intervention strategies aimed to target certain groups have earlier showed diverging results
]. These results suggest that having an inclusive approach but focusing on certain subgroups within the intervention can be successful. However, when interpreting the findings one should be aware of the lack of significant interaction between condition and gender. When an interaction term shows p
0.1 subgroup analysis is conventionally required for statistical reasons. We based our subgroup analyses on pre-specified hypotheses based on the nature of the study and previous findings
]. To evaluate the credibility of subgroup analyses Sun et al. (2010) have suggested eleven criteria
]. By meeting most, but not all these criteria, we find support for doing these secondary investigations, but we also acknowledge a degree of uncertainty of these exploratory findings.
Gender aside, the intervention appeared to affect other subgroups differently as well. The intervention participants in the low-activity group demonstrated a significant increase in physical activity from baseline to post intervention. These results are encouraging, as increasing the activity level among the least active can cause larger health benefits than among participants already active
]. As a decline in physical activity with increasing age can be expected
], it is also noteworthy that we did not observe a significant decrease in the high-activity group. Among those overweight, the participants in the control group were more active at both time points and had a more positive development than participants in the intervention group. The issue of different responses on different groups are discussed by Brown and Summerbell (2009) in a comprehensive review on obesity-prevention in school-children
]. They suggest that particularly boys and girls and those differentiating in weight status in the age range of 10 to 14 seem to respond differently to different elements of the interventions
]. Participants from different parental education categories were also affected differently by the intervention. An intervention effect was observed only among participants with parents having a “mid-range” educational level. However, investigating other outcomes in the HEIA study, Bjelland et al. (2011) found no moderating effects of parental education for boys or girls with respect to intake of sugar-sweetened beverages, time used for watching TV/DVD and computer/game-use
]. The results of this intervention study are important to public health, as feasibility and sustainability were high priorities when designing the intervention. This has been recommended in previous studies and reviews
]. Although comprehensive, the intervention components were designed to be able to fit into current school curricula without substantial extra costs. With limited instructions and material provided by the study group, teachers were key deliverers of the intervention components. No extra personnel or costly material are needed to carry out such components in the current school system, and all components could easily be incorporated into existing curricula for this age group.
Strengths and limitations
The strengths of the present study include the study design and the large number of participants. The multicomponent intervention, lasting 20 months, was designed to be feasible to the school system and not financially demanding. Also, measures including objectively assessed anthropometric measures, pubertal maturation, self-reported parental education and whole sample measurement of physical activity by accelerometers are clear strengths of this study.
We acknowledge that our study has several limitations. Firstly, the use of two different generations of accelerometers (for practical reasons) represents an element of uncertainty compared to using only one kind. To address this issue we explored the potential difference between generations of monitors, and adjusted the values accordingly. Secondly, at baseline physical activity was assessed during fall and at post intervention physical activity was assessed during spring. However, the measurement month was adjusted for in the effect analyses, and this issue was also taken care of by the study design. Thirdly, according to the power-calculations of the study
], the number of participants providing valid accelerometer data at both time points was lower than opted for, and a higher number of participants with valid recordings may have made it easier to detect significant intervention effects on physical activity. However, the power-calculations on physical activity may also have been overestimated, since investigating change in such large groups objectively has rarely been done in previous studies. The large drop-out reduces the generalizability of the results. However, few differences were seen between those who provided accelerometer data at both time points and those who did not. Fourthly, the use of subgroup analysis is criticized by some and called for by others
]. We chose to include subgroup analyses based on the nature of the study where specific groups were targeted when planning the intervention efforts. Furthermore, the HEIA intervention components were primarily delivered through the teachers at the intervention schools. Unpublished process evaluation revealed that the degree of implementation differed between schools
], with a reduced dose of intervention received by the participants observed from mid-way to post intervention
]. Also, when investigating intervention effects of a multi-component intervention, it is not possible to sort out whether or how the components worked separately. Finally, the potential for generalization of our findings might be limited as the sample was recruited from a limited geographic area. However, comparing the HEIA study sample to nationally representative figures for 9 and 15-year-olds, the measures from the participants in the HEIA study lie adequately between the measures of the 9 and 15-year-olds when it comes to objectively measured height, weight and total physical activity