BACKGROUND The validity of body mass index (BMI)-for-age for obesity diagnosis in Latin-American children may be limited due to observed cases of overweight without obesity (i.e. body fat excess), possibly due to certain physical characteristics. In the current study, we investigated whether the usefulness of BMI-for-age in the diagnosis of obesity among Mexican schoolchildren is modified by height, trunk length, muscle mass, body frame, or waist circumference. METHODS Our study cohort comprised 1,015 schoolchildren (aged 6-11 years) from Mexico City. Obesity diagnostics were derived from three classifications of BMI-for-age: percentiles of BMI according to the references of the Centers of Disease Control (CDC), the National Center for Health Statistics and the International Obesity Task Force. The area under the curve (AUC, through receiver-operating characteristic curves) and optimal cutoff points (by Youden index) of each classification were calculated. Body fat percentage, triceps skinfold thickness and blood pressure were used as standards. AUC and optimal cutoff point analysis were stratified according to height-for-age, sitting height, elbow breadth, arm muscle area (AMA) and waist circumference. RESULTS For the general population, the CDC reference had the highest values of AUC (0.94 for triceps skinfold thickness and 0.96 for body fat percentage), and the optimal cutoff point was the 85th percentile. Among schoolchildren with large body frames (measured through elbow breadth) or with high muscle mass (assessed by AMA), the optimal cutoff point was the 95th percentile of the CDC reference. CONCLUSIONS Our results suggest that the percentile cutoff to define obesity in children with high muscle mass or a large body frame should be the 95th percentile, while the 85th percentile can still be used for the other children.