The objective of this study was to validate the estimation of body fat (%BF) by DXA (Dual-Energy X-Ray AbsorciomDPX-MD) against the four compartment model (4C) of body composition in 32 Mexican pubertal girls and boys (aged 9-14 y; F=16). The mean of the difference between DXA and 4C model was -3.5 %BF (p=0.171). The limits of agreement (95% = 2 SD) were +5% to -12%BF. The precision of estimated limits of y the confidence intervals were -1.9% to -5.1%BF (P = 0.050). The concordance correlation coefficient was p = 0.85. The test of accuracy for coincidence of slop intercepts between DXA and the 4C model showed no coincidence (p < 0.05). The precision by R2 explained 83% of the variance (SEE, 4.1%). The individual accuracy assess by the total error was 5.6%. The group mean accuracy by two way analysis of variance of body fat did not show interaction between method (DXA-4C model) and separate analysis of gender and overweight. However, there was an effect of method (p = 0.043) in the presence of overweight (p < 0.001). In conclusion, the estimation of percent of body fat by DXA was not precise and accurate in a group of Mexican children. However, results do not limit the utility of DXA for the measurements of body composition and its relation with health outcomes, especially in follow up studies.