Measurements of body composition are being made increasingly widely in pediatrics. Tetrapolar whole body impedance (BI) is particularly suitable as a method of estimating body composition in children and is therefore the subject of great interest at present. However, the ability of BI to accurately estimate fat-free mass (FFM) in children is unclear, and users of BI are faced with a growing choice of prediction equations for estimation of FFM. Studies in adults have suggested that choice of prediction equation can have a profound effect on the estimate obtained. The aim of the present study was to measure the ability of four published pediatric BI equations to predict FFM in 98 Caucasian prepubertal children (mean age 9.0 y). For three of the published equations, limits of agreement between predicted and reference FFM were wide and distinct biases were apparent. With mean FFM of 25 kg, the equation of L. Cordain et al. overestimated reference FFM (95% CI +2.1 to +3.1 kg), whereas those of P. Deurenberg et al. (95% CI -1.9 to -2.9 kg) and F. Schaefer et al. (95% CI -1.4 to -2.5 kg) systematically underestimated reference FFM. The equation of Houtkooper et al. (95% CI -0.2 to +0.8 kg) predicted FFM with negligible bias and had narrower limits of agreement relative to the reference method than the other three equations tested. We conclude that the ability of BI to predict body composition in children depends on the equation chosen and that the general applicability of BI equations cannot be safely assumed. Cross-validation of BI equations is recommended before they are used routinely for estimation of body composition in children.