ORIGINAL ARTICLES  BODY COMPOSITION 

The Journal of Sports Medicine and Physical Fitness 2015 January-February;55(1-2):68-75

Copyright © 2015 EDIZIONI MINERVA MEDICA

lingua: Inglese

Comparisons of accuracy of estimating percent body fat by four bioelectrical impedance devices with different frequency and induction system of electrical current

Demura S. ¹, Sato S. ² ✉

¹ Natural Science and Technology, Kanazawa University, Kakuma, Kanazawa, Ishikawa, Japan; ² Life‑long sports core, Kanazawa Institute of Technology, Ishikawa, Japan

AIM: This study aimed to compare the accuracy of estimating present body fat (%BF), using dual-energy X-ray absorptiometry (DXA) as a reference, among four BIA devices; foot-to-foot bioelectric impendence (FF-BIA), hand-to-hand BIA (HH-BIA), foot-to-hand single-frequency BIA (FH-BIASF), and foot-to-hand multiple-frequency BIA (FH-BIAMF).
METHODS: Forty-four healthy college students (21 males, 172.9±5.5 cm and 65.8±9.1 kg and 23 females, 160.7±6.6 cm, 52.6±6.17 kg) volunteered for this study. The relationship with the reference was determined by Pearson’s correlation coefficient. Standard error of estimation (SEE) was calculated by regression analysis when estimating the reference measures from the predictor (BIAs). The technique of Bland-Altman was used to determine estimation bias and the limit of agreement for %BF between the reference and the predictor.
RESULTS: When referenced to DXA, FF-BIA and FH-BIASF significantly overestimated %BF. The correlations with DXA were 0.90 or over, except for HH-BIA (r=0.88, P<0.05). When estimating %BFDXA by the %BF of the other methods, the lowest and highest SEE values were found in FH-BIASF (2.14%) and in HH-BIA (2.65%), respectively. The value of the limits of agreement was 10% or under in FH-BIASF and FH-BIAMF, but was 10% or over in FF-BIA and HH-BIA.
CONCLUSION: Among four BIA devices, the method with the least estimation error from DXA is FH-BIASF. In comparing single-frequency BIA devices, margin of error tended to be small in FH-BIASF, and estimation bias tended to be small in FH-BIASF and HH-BIA.