Body fat was measured in 46 elderly subjects by dual-energy x-ray absorptiometry [BF(DEXA)], bioelectrical impedance analysis (BIA), and anthropometry. Equations for prediction of the body fat (by DEXA) of elderly people by BIA and/or anthropometry were developed. The prediction of body fat (by DEXA) by anthropometric variables alone gave an r2 of 0.94 and the corresponding SEEs were 1.61 kg for men and 2.43 kg for women. When BIA variables were added as predictors, r2 increased by 2-5% (P less than 0.05) and the corresponding SEE decreased by 25% (P less than 0.05). The prediction of body fat (by DEXA) in elderly subjects and by BIA or anthropometry with equations developed in populations of young healthy adults (adapted from the literature) was generally not good although the correlation coefficients were high (r greater than 0.9, P less than 0.001), which suggests that our equations may improve prediction of body fat in elderly subjects.