The main objective of this study was to assess the effectiveness of an obesity prevention intervention which included nutrition education and physical activity applied to low income Chilean children in a longitudinal three-year follow-up study. Participants included all 4-7 year old children from 7 public schools located in a low socioeconomic district of Santiago. The main aspects of the intervention were: training of teachers so they could apply an educational program on healthy eating, increase in physical education classes from 3 to 4 per week, and improvement of their quality. Weight, height, waist circumference (WC) and the 6-minute walk distance (6MWD) were assessed yearly in 597 children. We calculated BMI, BMI Z, % normal (N), overweight (OW) and obesity (OB), WC > 90th percentile (NHANES III) and fitness (6MWD/height). Annual changes in BMI Z, WC, 6MWD and fitness were assessed, using repeated measures ANOVA and the test of proportions. Knowledge in healthy eating was assessed during 2 of the 3 years. Results showed that % OB remained unchanged for 2 years (17%) but increased to 19.3% at follow-up. BMI Z increased from 0.3 to 0.38 (p = 0.052) in the N, remained unchanged in the OW, while decreasing significantly in the OB (2.73 to 2.41 p < 0.0001). % WC > 90 th percentile decreased in the OW and OB; 6 MWD was higher in the OW, but the increase in distance overtime was greater among the N. Overall fitness improved, however it was only significant among the N (p = 0.0002). There was a significant increase in food knowledge. We conclude that in spite of a decrease in BMI Z of the OB, an improvement in fitness in the N and food knowledge in all the children, obesity increased at follow-up. This study shows that in school-based obesity programs it is necessary to apply more intense interventions with another type of methodology; otherwise this condition will continue to rise.