Social pressure was associated with a change in intention suggesting that the intervention accomplished exactly what it was supposed to do: preparing children for secondary school.

One question is whether the transition to a different school instead of the intervention is responsible for the difference between the intervention and control students. Other findings indicated, among others, that students are more likely susceptible to smoking if they have two or more close friends who smoke, attend a school with a relatively high smoking rate among the older students or a school with less (endorsed) smoking restrictions (Leatherdale et al., learn more 2006 and Wakefield et al., 2000). If a larger part of the control students went to schools with a higher smoking rate, this change in school instead of the intervention might have caused the difference in smoking. Although we could not verify this school transition effect properly, we do not think that the effect of the transition Smad inhibitor to secondary school differs for intervention or control

students. First, in each participating region, we have randomized schools to the intervention or control group, meaning that an important part of the students in both conditions went to the same regional secondary schools. Secondly, there were no important differences in perceived non-smoking policies between the intervention and control group. The largest effect of the intervention is found in girls. Other studies already have shown that there are gender differences in smoking uptake in adolescence and that smoking is more prevalent either in girls than in boys (Rodham et al., 2005 and de Vries et al., 2003). Moreover, Mercken et al. (2010) found that particularly girls are influenced to smoke by their peers concluding that an intervention preparing girls to resist peer pressure might be more effective in girls than in boys. This might explain the larger effect of the present intervention among girls. The schools were randomly assigned to the intervention and control group

in order to reduce the chance of selection bias. In spite of the randomization procedure, differences between the groups at baseline were found. Chance confounding, due to randomization at school level, may explain these differences, so we adjusted for this in our analysis. Loss to follow-up was somewhat selective but seemed to have a limited effect on the results, while there were no significant differences in smoking behavior between the non-response of intervention and control condition. Moreover, intention-to-treat analyses by carrying the last observation of smoking behavior forward did not have different effects on smoking behavior. The response rate also did not differ between groups. Therefore, it is highly unlikely that selective response has affected the impact of the intervention. All measurements were self-reports, meaning that information bias could have occurred, especially in the intervention group.