The objective of this study was to validate an 8-electrode bioimpedance analysis (BIA(8)) device (BC-418; Tanita, Tokyo, Japan) for use in populations of European, Maori, Pacific Island, and Asian adolescents. Healthy adolescents (215 M, 216 F; 129 Pacific Island, 120 Asian, 91 Maori, and 91 European; age range 12-19 years) were recruited by purposive sampling of high schools in Auckland, New Zealand. Weight, height, sitting height, leg length, waist circumference, and whole-body impedance were measured. Fat mass (FM) and fat-free mass (FFM) derived from the BIA(8) manufacturer's equations were compared with measurements by dual-energy X-ray absorptiometry (DXA). DXA-measured FFM was used as the reference to develop prediction equations based on impedance. A double cross-validation technique was applied. BIA(8) underestimated FM by 2.06 kg (P < 0.0001) and percent body fat (%BF) by 2.84% (P < 0.0001), on average. However, BIA(8) tended to overestimate FM and %BF in lean and underestimate FM and %BF in fat individuals. Sex-specific equations developed showed acceptable accuracy on cross-validation. In the total sample, the best prediction equations were, for boys: FFM (kg) = 0.607 height (cm)(2)/impedance ( ohm) + 1.542 age (y) + 0.220 height (cm) + 0.096 weight (kg) + 1.836 ethnicity (0 = European or Asian, 1 = Maori or Pacific) - 47.547, R(2) = 0.93, standard error of estimate (SEE) = 3.09 kg; and, for girls: FFM (kg) = 0.531 height (cm)(2)/impedance ( ohm) + 0.182 height (cm) + 0.096 weight (kg) + 1.562 ethnicity (0 = non-Pacific, 1 = Pacific) - 15.782, R(2) = 0.91, SEE = 2.19 kg. In conclusion, equations for fatness estimation using BIA(8) developed for our sample perform better than reliance on the manufacturer's estimates. The relationship between BIA and body composition in adolescents is ethnicity dependent.