BACKGROUND & AIMS: The aim was to generate a predictive equation to assess body composition (BC) in children with obesity using bioimpedance (BIA), and avoid bias produced by different density levels of fat free mass (FFM) in this population. METHODS: This was a cross-sectional validation study using baseline data from a randomized intervention trial to treat childhood obesity. Participants were 8 to 14y (n = 315), underwent assessments on anthropometry and BC through Air Displacement Plethysmography (ADP), Dual X-Ray Absorptiometry and BIA. They were divided into a training (n = 249) and a testing subset (n = 66). In addition, the testing subset underwent a total body water assessment using deuterium dilution, and thus obtained results for the 4-compartment model (4C). A new equation to estimate FFM was created from the BIA outputs by comparison to a validated model of ADP adjusted by FFM density in the training subset. The equation was validated against 4C in the testing subset. As reference, the outputs from the BIA device were also compared to 4C. RESULTS: The predictive equation reduced the bias from the BIA outputs from 14.1% (95%CI: 12.7, 15.4) to 4.6% (95%CI: 3.8, 5.4) for FFM and from 18.4% (95%CI: 16.9, 19.9) to 6.4% (95% CI: 5.3, 7.4) for FM. Bland-Altman plots revealed that the new equation significantly improved the agreement with 4C; furthermore, the observed trend to increase the degree of bias with increasing FM and FFM also disappeared. CONCLUSION: The new predictive equation increases the precision of BC assessment using BIA in children with obesity.