Childhood obesity has increased worldwide over the past four decades. This quasi-experimental study aimed to assess the effectiveness of a multicomponent and multilevel school-based intervention (POIBA) at 3 years of follow-up. The nutrition intervention focused on food groups, food pyramid, nutrients, portions, and balanced menus. In total, 3624 children participated in the study. Anthropometric measurements and information on food frequency and behavior, physical activity, and daily screen use were collected in the intervention (IG) and comparison group (CG). Positive unadjusted changes toward adherence to recommendations were found for water, meat, sweets, and fried potato consumption, proper breakfast, not having dinner in front of the TV, out-of-school physical activity, and daily screen use. Three scores were used to calculate the proportion of children making more than one change to improve healthy habits regarding physical activity (global Activity score), nutrition (global Nutrition score), and both (global score). Students exposed to the intervention had a significantly better global Activity score (16.2% IG vs. 11.9% CG; p = 0.012) and Global score (63.9% IG vs. 58.5% CG; p = 0.025). Intervention effects on obesity incidence at 3-year follow-up lost significance but maintained the positive trend. In conclusion, school-based interventions including a family component could be useful to address the childhood obesity problem.