Background: The prevalence of obesity among children is increasing, but treating obesity is difficult and has poor outcomes. We assessed the 4-year impact of a school-based educational intervention on physical activity, body mass index (BMI) and the prevalence of overweight and obesity. Methods: A 6-year cluster-randomised study was used to evaluate an intervention programme for children in primary school in 2006–08. The intervention promoted healthy eating habits and physical activity in the school setting through the investigation, vision, action and change educational methodology that involves children as active participants in healthy change. BMI and physical activity were measured in 2006, 2008, 2010 and 2012. Multilevel mixed-effects linear regression was used to assess change in BMI over time. Results: The effect of the intervention was maintained 4 years after completion. Average increase in BMI was greater in the control than in the intervention group (3.85 ± 2.82 vs. 2.79 ± 2.37); the mean difference was 1.06 (95% confidence interval: 0.14–1.97). In the intervention group, the prevalence of overweight and obesity decreased (1.4 and 3.7%, respectively). In the control group, the prevalence of overweight increased (9.4%) and the prevalence of obesity decreased (1.6%). The control group spent more time in sedentary activities. No differences were observed in physical activities. Conclusions: Control of weight gain was sustained 4-years after an educational intervention during the first 2 years of primary school. Future interventions to prevent obesity should build on principles viewing children as active agents of healthy change. Trial registration: ClinicalTrials.gov NCT01156805.