ABSTRACT Introduction A comparison of body composition assessments using military circumferences to bioelectrical impedance analysis (BIA) and the reference standard dual-energy X-ray absorptiometry (DEXA) can gauge effectiveness of assessments. High-frequency (500 KHz) direct segmental multifrequency bioelectrical impedance analysis (DSM-BIA) accurately calculates total water mass and body fat% (BF%), but it is unknown whether higher frequencies (1,000 KHz) increase measurement accuracy. The purpose was to compare DSM-BIA 500, DSM-BIA 1000, the DoD Circumference Method (CM), and the reference-standard DEXA. Materials and Methods Design: Cross sectional, observational study. Participants/Setting: A total of 62 participants from the military healthcare system (n = 25 males, 38.8 ± 11.4 years, n = 37 females 43.7 ± 15.95 years) were measured in an outpatient clinic setting. Statistical Analysis: BF% was estimated via DEXA, DSM-BIA 500, DSM-BIA 1000, and CM to identify the relationship between methods using Pearson correlation, intraclass correlation coefficients (ICCs), and Bland-Altman plots. The study was approved by the IRB from Walter Reed National Military Medical Center at Bethesda and Concordia University Chicago. Results Circumference Method BF% was moderately correlated with DSM-BIA 500 (males r = 0.63, ICC = 0.76; females r = 0.77, ICC = 0.85), DSM-BIA 1000 (males r = 0.59, ICC = 0.74; females r = 0.77, ICC = 0.85), and DEXA (males r = 0.62, ICC = 0.62; females r = 0.73, ICC = 0.82). DSM-BIA 500 BF% was strongly correlated with DSM-BIA 1000 (males r = 0.99, ICC = 0.99; females r = 0.99, ICC = 0.99) and DEXA (males r = 0.93, ICC = 0.94; females r = 0.89, ICC = 0.89). Lastly, DSM-BIA 1000 BF% was also strongly correlated with DEXA (males r = 0.93, ICC = 0.94; females r = 0.84, ICC = 0.90) (P for each reported r < 0.01). Bland-Altman analysis confirmed an overall mean bias of −1.72% CM vs. DEXA in females, indicating the tendency of CM to underestimate BF% compared to DEXA limits of agreement from −14.24 to 10.8. There was an upward slope of the linear relationship between the bias and mean of the measures (Beta = 0.34, P = 0.01). In the full cohort, there was an overall mean bias of 1.14% of CM vs. DSM BIA 1000, with CM tending to overestimate BF% compared to DSM BIA 1000 with limits of agreement −11.13 to 13.41%. There is an upward slope line of the linear relationship between the bias and the mean of the measures (Beta = 0.17, P = .03). Conclusion This study found that CM BF% was moderately correlated with DSM-BIA 500 kHz, DSM-BIA 1,000 kHz BIA, and DEXA. Both DSM-BIA 500 and DSM-BIA 1,000 kHz strongly correlated well with DEXA implying that there was no further increase in correlation with increased frequency. Additionally, there was proportional bias in BF% in the female group between CM and DEXA and in the total group between CM and DSM BIA 1000.
Read full abstract