Despite the increasing use of tetrapolar whole-body bioelectric impedance (BI) analysis in the assessment of body composition, its usefulness in estimating fat-free mass (FFM) has not been evaluated in comparison with conventional skinfold anthropometry in children. We therefore compared 1) the intraobserver and interobserver reproducibility of BI and skinfold measurements and the derived FFM estimates, and 2) the predictability of FFM as calculated from measurements of total body potassium (TBK) using 40K spectrometry by equations based on either BI or skinfold measurements in 112 healthy children, adolescents, and young adults aged 3.9 to 19.3 y. A best-fitting equation to predict TBK-derived FFM from BI and other potential independent predictors was developed and cross validated in two randomly selected subgroups of the study population by stepwise multiple regression analysis. Although the technical error associated with BI measurements was much smaller than that of skinfold measurements, the reproducibility of BI-derived FFM estimates (intraobserver coefficient of variation [CV], 0.39%; interobserver CV, 1.23%) was only slightly better than that of FFM estimates obtained by use of weight and two skinfold measurements (0.62% and 1.39%, respectively). The cross validation procedure yielded the following best-fitting prediction equation: FFM = 0.65 x (height2/impedance) + 0.68 x age + 0.15 (R2 = 0.975, root mean square error = 1.98 kg, CV = 5.8%, 95% limits of agreement = -11.1% to +12.4%). Conventional anthropometry, using published equations to estimate FFM from skinfolds, slightly over-estimated TBK-derived FFM, but predicted FFM with precision similar to the best-fitting equation involving BI. Previously published FFM equations incorporating BI predicted TBK-drived FFM with variable predictive precision and accuracy.(ABSTRACT TRUNCATED AT 250 WORDS)