Modeling leg sections by bioelectrical impedance analysis, dual-energy X-ray absorptiometry, and anthropometry:...

This study aimed to assess the value of different DXA and BIA models for predicting muscle volume in mid-thigh segments obtained by MRI. Three DXA models were used: in model A, muscle was taken to be equivalent to fat-free soft tissue; in model B the thigh segment was divided into its constituent tissues using fixed assumptions about tissue composition; in model C the assumptions were similar to model B, but with variable distribution of fat and fat-free soft tissue, depending on body mass index. The two BIA models (both parallel tissue resistance models) involved impedance measurements at 50 kHz, and assumptions about either the specific resistivities of all the constituent tissues (model A), or resistivities of only adipose tissue and muscle (model B). Anthropometric estimates (thigh circumference and skinfold thickness) assumed that both limb and muscle circumference were circular. Compared to MRI estimates of muscle mass, those obtained by DXA model A (fat-free soft tissue) were not as good as those obtained using models B and C, although the standard deviations of the differences were similar with all three models. The BIA models were superior to the anthropometric estimates of muscle volume (relative to MRI) with respect to bias, but the standard deviations of the differences were large for both. The intraobserver repeatabilities for muscle volume were < 0.5% for MRI, < 1% for DXA, 1.8% for BIA, and 1.7% for anthropometry (interobserver value for BIA was 3.8% and for anthropometry 3.5%). The study suggests that DXA modeling provides a promising approach for assessing muscle mass in thigh segments, and suggests the potential value of parallel BIA models for groups of individuals but not for individual subjects, possibly because muscle resistivity is influenced not only by its composition but also by the direction of current flow in muscle.