Premium
Evaluation of a Novel Bioimpedance Analysis Equation Development Approach
Author(s) -
Zheng Jolene,
Bourgeois Brianna,
Heymsfield Steven B
Publication year - 2017
Publication title -
the faseb journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.709
H-Index - 277
eISSN - 1530-6860
pISSN - 0892-6638
DOI - 10.1096/fasebj.31.1_supplement.643.32
Subject(s) - foot (prosody) , population , linear regression , medicine , regression analysis , mathematics , bioelectrical impedance analysis , dual energy , limits of agreement , stepwise regression , body mass index , statistics , zoology , nuclear medicine , biology , bone mineral , osteoporosis , philosophy , linguistics , environmental health
Traditional bioimpedance analysis (BIA) prediction equations developed at 50 kHz for %fat often include population‐specific predictor variables such as age, sex, and race. Recent studies suggest that good generalizable %fat prediction equations can be developed without population‐specific variables if segmental impedance (Z) measurements at variable frequencies are used in model development. The current study tested this hypothesis using the 8‐electrode Selves Healthcare X Scan PLUS 970 (Seoul, South Korea) multifrequency segmental BIA system in a two‐phase study in 161 ethnically mixed adults (9 African American, 109 white, 43 other). Subjects ranged widely in age (15–72 years) and BMI (16.5–40.4 kg/m 2 ). Using dual‐energy X‐ray absorptiometry (DXA, GE Lunar Prodigy) as the reference for fat‐free mass (FFM), 86 subjects had segmental impedances measured at 5, 50, and 250 kHz in phase I. Stepwise multiple linear regression analysis was then used to develop the FFM prediction model. The developed %fat (= 100 × [weight‐FFM/weight]) model was then validated in 75 phase II subjects. The developed model has 6 segmental Z predictor variables (hand‐hand, foot‐foot, left hand‐left foot, left hand‐right foot, right hand‐right foot, right hand‐left foot) indexed to either height or weight; age, sex, and race did not enter the final model. Predicted BIA %fat and DXA %fat in phase II subjects were highly correlated (R 2 =0.91, p<0.0001) and the Bland‐Altman plot slope was non‐significant. Comparable prediction models were developed for other similar multifrequency BIA systems (e.g., Selves Healthcare 357s). Highly predictive BIA %fat models can thus be developed from multiple frequency segmental impedance measurements that capture variance related to age, sex, and race. These models do not require the addition of potentially population‐specific covariates to %fat prediction formulas.