Objective. To assess the validity of recommendations for use of the 85th and 95th percentiles of body mass index (BMI) of the population in the United States of America as a screening tool to assess overweight/obesity in adolescents. Methods. We investigated the relation between BMI and percent body fat in 1 540 adolescents (717 males and 823 females) aged 10 to 17.9 years old from a private high school in Niterói, a city in the state of Rio de Janeiro, Brazil. We used bioelectric impedance, with the appropriate equations for adolescents, to estimate percent body fat, which served as the gold standard (30% for girls and 25% for boys) to calculate the sensitivity and specificity of the 85th and 95th percentiles of the United States and Brazilian distribution curves of BMI. Results. Sensitivity and specificity were high (above 80%) for the Niterói boys, except for the 85th percentile of the Brazilian curve (specificity = 61.8%) and for the 95th percentile of the United States curve (sensitivity = 55.4%). For the Niterói girls, the 85th- and 95th-percentile BMI cutoff points, from both the United States and Brazilian curves, showed low sensitivity, and that sensitivity decreased with age. Specificity was high for the girls, and much higher than it was for the boys. Conclusions. These data suggest that using BMI to screen for overweight/obesity in adolescents can generate a high percentage of false-positives for Niterói boys and an even higher percentage of false-negatives for Niterói girls. A more universal approach to using anthropometric measures to screen for overweight/obesity should be developed, preferably linked to stages of maturation.