Think Globally, Act Locally: The Importance of Population-Specific Bioelectrical Impedance Analysis Prediction Equations for Muscle Mass Assessment.

Background Bioelectrical impedance analysis (BIA) is a convenient muscle assessment method, but its accuracy highly depends on population-specific aspects of the adopted equation. We aimed to develop appendicular lean mass (ALM) prediction models for older South Americans and to compare their performances to those of reference equations in the same sample. Methods Cross-sectional evaluation of 192 community-dwelling Brazilian subjects >= 60 years old from the COMO VAI? study. Using measurements from single-frequency and multifrequency devices (BIASF and BIAMF, respectively), new ALM prediction equations were developed (reference method: dual-energy x-ray absorptiometry [DXA]). Validity was assessed by bootstrapping. Four previously established equations were also tested, and the performances were compared using Bland-Altman analysis. Results Stepwise variable selection produced the following equations: ALM(SF-BIA)= (2.08 x sex) + (0.04 x weight) + (0.24 x RI50) + (0.07 x Xc(50)) - 0.16; ALM(MF-BIA)= (1.85 x sex) + (0.03 x weight) + (0.31 x RI50) + (0.04 x Xc(50)) + (0.01 x Z(5)) - 8.16, where ALM is estimated in kg; female sex = 0 and male sex = 1; weight is measured in kg; RI(50)is the resistance index at 50 kHz measured in cm(2)/omega); Xc(50)is the reactance at 50 kHz measured in omega; and Z(5)is impedance at 5 kHz measured in omega. The equations explained, respectively, 89% and 90% of the variability of ALM(DXA)in our sample, and their estimates were not significantly different from DXA measurements. Bland-Altman analysis revealed accurate and unbiased performances for both models, with similar limits of agreement (BIA(SF): +/- 2.58 kg; BIA(MF): +/- 2.48 kg), and their validity was considered adequate by the bootstrap method. The reference equations, however, systematically overestimated ALM in our sample. Conclusion The proposed equations might represent practical options to estimate ALM in older noninstitutionalized South Americans. Further external validation, though, is required to verify the reproducibility of our findings.