Cross-sectional and longitudinal agreement between two multifrequency bioimpedance devices for resistance, reactance, and phase angle values.

BACKGROUND: The use of raw bioelectrical variables, such as resistance (R), reactance (Xc), and phase angle (φ), has been advocated for evaluating physiological changes.
METHODS: Before and after 8 weeks of resistance training, adult females were assessed via multifrequency bioelectrical impedance analysis (MFBIA; Seca® mBCA 515/514) and bioimpedance spectroscopy (BIS; ImpediMed® SFB7). Data were analyzed to determine whether cross-sectional estimates and changes (i.e., Δ scores) of R, Xc, and φ differed between devices at 16 shared measurement frequencies ranging from 3 to 1000 kHz.
RESULTS: Cross-sectionally, strong correlations (r ≥ 0.96) were observed for R across all frequencies, although MFBIA produced values 9-14% greater than BIS. Strong correlations (r ≥ 0.92) for Xc and φ were observed up to frequencies of ~150 kHz. BIS produced greater Xc and φ values at lower frequencies, while MFBIA produced greater values at higher frequencies. In general, proportional bias was not observed, with the exception of Xc at high frequencies and φ at low frequencies. ΔR did not differ between devices at any frequency and was correlated at all frequencies. ΔXc and Δφ did not differ at any frequency and were correlated between devices for frequencies up to ~300 kHz. Proportional bias was generally not observed longitudinally. While individual-level errors were potentially acceptable cross-sectionally, they were concerningly high longitudinally.
CONCLUSION: Despite notable differences in the characteristics of the bioimpedance devices and cross-sectional disagreement, strong group-level agreement for detecting changes in R, Xc, and φ was generally observed. However, large errors were observed at the individual level.