Data from the 8 month mid-way assessment indicated that girls in the intervention group spent significantly less time on watching TV/DVD and using computer/games compared to the girls in the control group, and the intake of SSB during weekend days was significantly lower among the girls in the intervention group. Girls' WS did not moderate these findings. No significant differences between the intervention and control group were found for outcome variables among the boys with normal weight or the overweight/obese boys, but moderation effects were found for WS (TV/DVD and computer/games during week days). There were no moderating effects of parental education for neither boys nor girls with respect to any of the three behaviours. The process evaluation showed that parental awareness was significantly higher among the parents of girls, while the parents of boys were more satisfied with the fact sheets. No other differences in the parental process evaluation were found.
The effects found were both in a desired direction (girls) and an undesired direction (overweight/obese boys). However, it may be questioned whether the effects were large enough to have any public health impact. One review suggests that in children an imbalance over time of about 2% (125 KJ or 15 minutes of play replaced by TV-viewing) may lead to obesity . Based on these estimates two groups did benefit from the HEIA study. The decrease in intake of SSB among the overweight girls was 0.4 dl for week days and 0.5 dl for weekend days. This represents a decrease in calorie intake equal to 68-85 KJ per day (0.4 or 0.5 dl and 170 KJ/dl). By reducing the time used for TV/DVD during week days by 0.3 hours (about 18 minutes) and time used for computer/games during week days by 0.2 hours (about 12 minutes) among normal weight girls, the total sedentary screen time was reduced by 30 minutes, indicating a decrease in sedentary behaviour with a possible public health impact.
Further, we found that the overweight/obese boys in the intervention group had a non-significant tendency towards an undesired effect with regards to more time used for computer/games during week days compared to the overweight/obese boys in the control group (p = 0.06). By the use of magnitude-based inference as an alternative approach for this variable, we explored to what extent this change was of relevance. A confidence interval or p-value does not address the question of the clinical or practical importance of an outcome; a magnitude-based inference does . It was possible to estimate the chances or probabilities that the true effect was harmful, trivial or beneficial, and the chances were estimated using the same assumptions about the outcome statistic as when estimating p-values or confidence intervals. The result indicated that the intervention likely produced an increase in time used for computer/game-use in week days in overweight/obese boys. This was an effect of clinical/practical importance, however, it was an unintended and undesired consequence of the intervention.
We can only speculate in why the overweight/obese boys did not respond to the intervention in a desired direction and why the overweight/obese girls did so. The same goes for why we found an overall effect in girls and not in boys. With respect to the former, it could well be that boys being overweight/obese show evidence of reactance by responding with less functional strategies (becoming more sedentary) when confronted with messages concerning healthy eating and enhanced physical activity . As to the latter, one possible explanation is that both the development and implementation of the intervention were dominated by a female approach. The intervention was to a large degree developed by women (mainly female researchers and pedagogues involved), it was mainly women who implemented the intervention at school (mostly female teachers) and process evaluation findings indicated that mothers were more involved than fathers at home (more than 80% of the parent answering the process evaluation questionnaire were women). Furthermore, Haug et al.  found that boys across Europe and USA were more likely to be overweight than girls, indicating that preventive initiatives may be inadequate and/or less effective for boys. A third explanation may be difference in parental involvement. The process evaluation indicated that parents of girls were more aware of the project compared to parents of boys, which could result in more parental support for the girls. Finally, analyses of the pre-test data from the HEIA study indicate that the girls may have better role models in their mothers compared to boys with regard to weight . Parents, and in particular fathers, should be made aware of their potential to improve as role models [36, 37].
When comparing the results from our study with other intervention studies aimed at reducing the consumption of SSB among children/adolescents, only two of the four identified studies reported effects by gender [38–41]. Haerens et al.  found no effect, while Singh et al.  reported a significant lower intake in the intervention group both for girls and boys. In total, eight [41–48] out of nine  identified studies that aimed at reducing the time used for screen activities among both boys and girls aged 9-15 assessed the moderating effect of gender or reported effect in boys and girls separately. Five of the studies [42–44, 47, 48] reported effects both for boys and girls, while one reported effect for boys only . Harrison et al.  found no effect, while Salmon et al.  found an effect in the undesired direction in one of the intervention groups. Three of the studies checked the moderating effect of WS for sedentary behaviour [43, 45, 48]. Harrison et al.  found no interaction for screen time. In Planet Health , a reduction in TV-viewing predicted obesity change and mediated the intervention effect among the girls. Finally, obese children reported higher screen time at the post-test than overweight and normal weight adolescents in the study by Gentile et al. . These findings are inconsistent and no clear pattern in the behavioural measures emerges, as reported in recent reviews as well [3, 9]. The results from our study support that interventions work better for girls than for boys [3, 15, 16].
Because of the weak evidence of effective school-based obesity prevention interventions, Lytle  suggests that it may be time to re-evaluate where the research needs to move. Lytle points out that an investigation of how study participants receive the intervention rarely is examined . The process evaluation in the HEIA study indicates that girls to a larger extent "bring the project home" compared to boys. This result is supported by previous process evaluations [50, 51], and studies on gender differences in parent-child communication reporting that girls' self-disclosure at home about every day life is higher than for boys . Parents of boys are more dependent on getting information from others than their sons . This might explain why parents of boys appreciated the fact sheets more than parents of girls. Qualitative studies have found that parents are in need of effective communication strategies about ways to improve positive health behaviours, and that fact sheets may be a useful tool [54, 55].
Strengths and limitations
Our research has some limitations. The SSB consumption variables have not been validated, but our results are in line with data from a national representative study . The measures of sedentary behaviour consisted of single items, resulting in crude estimates only . Still, the mean behavioural outcomes are in line with the trends described by Marshall et al. . The gender differences in time spent on watching TV were small, but boys spent more time on computer/games compared to girls. Furthermore, the test-retest correlation coefficients for the outcome measures were moderate to high (r = 0.46-0.78) . The potential for generalization of our findings is limited because a local sample was recruited from a limited geographic area, mainly in small towns and their close surroundings. The recruitment of schools and participants may have caused a sampling bias, restricting the number of overweight/obese participants and resulting in reduced precision (larger confidence intervals). Finally, some degree of social desirability may be present in the data [59, 60]. Still, the effects found should be taken into consideration because of the design of the HEIA study. One of the strengths of the present study is the large sample with objective measures of weight and height. Another strength is that parental education was reported by the parents themselves, and that we were able to collect these data from nearly all the parents giving their adolescent consent to participate in the study, and not only from those parents answering a questionnaire.