Accuracy of a predictive bioelectrical impedance analysis equation for estimating appendicular skeletal muscle mass in a non-Caucasian sample of older people

Objective: To design and validate a predictive BIA equation to estimate ASM in a sample of non-Caucasian older people and to explore the accuracy of two previous single-frequency BIA published equations. Method: This is a cross-sectional study in 213 physically-independent and apparently healthy older men and women subjects from a central region of Mexico. Dual-energy X-ray absorptiometry, anthropometry, resistance, reactance, and other independent predictive variables were assessed. The total sample was randomized and split into two sub-samples: one group for the development of the predictive single-frequency BIA equation and the other for its validation. The equation was derived by step-wise regression, the goodness of fit was assessed by the highest determination coefficient (R-2) and precision by the lowest root mean square error. In the validation sample, the accuracy of the new and the published BIA equations was tested by pure error. In addition, Bland and Altman analysis tested the bias. Results: The selected model had a R-2 of 0.91 and a RMSE of 1.01 kg and fulfilled every regression assumption. The predictive variables included were: Ht(2)/R, sex, and body weight. The new BIA equation showed a PE close to RMSE of the estimate with no significant bias. The published BIA equations proved to be inaccurate and had significant bias. Conclusions: This new single-frequency BIA-derived equation was precise, accurate and free of bias. Published BIA equations to estimate ASM should not be used indistinctly in other ethnic groups without validation. (C) 2015 Elsevier Ireland Ltd. All rights reserved.