Objective To evaluate the effectiveness of anthropometric measures and physical activity level in the prediction of metabolic syndrome (MetS) in children. Design Cross-sectional study with children from public and private schools. Children underwent an anthropometric assessment, blood pressure measurement and biochemical evaluation of serum for determination of TAG, HDL-cholesterol and glucose. Physical activity level was calculated and number of steps per day obtained using a pedometer for seven consecutive days. Setting Vi篳a, south-eastern Brazil. Subjects Boys and girls (n 187), mean age 9繰 (sd 0緩 years. Results Conicity index, sum of four skinfolds, physical activity level and number of steps per day were accurate in predicting MetS in boys. Anthropometric indicators were accurate in predicting MetS for girls, specifically BMI, waist circumference measured at the narrowest point and at the level of the umbilicus, four skinfold thickness measures evaluated separately, the sum of subscapular and triceps skinfold thickness, the sum of four skinfolds and body fat percentage. Conclusions The sum of four skinfolds was the most accurate method in predicting MetS in both genders.