Comparison of a dual-frequency bio-impedance analyzer with dual-energy X-ray absorptiometry for assessment of body composition in geriatric patients.

BACKGROUND AND OBJECTIVES: Measuring body composition is relevant in geriatric medical patients at high risk of sarcopenia (loss of muscle mass/strength) for diagnosis and monitoring efficacy of treatment interventions. Dual-energy X-ray absorptiometry (DXA) and bio-impedance-analysis (BIA) are widely used in research/clinical practices, but their agreement is unknown in this population. Therefore, it was investigated how a dual-frequency (DF)-BIA compares to DXA regarding replicability, direct comparisons, and monitoring of total/segmental body composition in geriatric medical patients. METHODS: Measurements were taken while admitted to the geriatric medical ward and repeated 12 weeks after discharge for monitoring. Total population and gender-specific analyses were made. RESULTS: Thirty-one participants were included (age: 82+/-6 years, BMI: 26.2+/-4.3), and ten lost to follow-up. Replicability was high for both methods (R(2) ; 0.956, 0.999, P<0.0001, n=31), however, with wide 5th-95th-percentile ranges for individual agreement. Bland Altman plots revealed significant fixed systematic and negative proportional biases for body composition variables - both for the direct comparisons (e.g. total LBM, 1.2 kg higher with DF-BIA, P<0.05, n=31) and for monitoring (total LBM, kg; men (n=15): DXA (0.96+/-2.61) vs. DF-BIA (-0.35+/-3.52) and women (n=16): DXA (-0.39+/-1.51) vs. DF-BIA (-1.05+/-2.24)). Generally, correlations for direct comparisons were higher than for monitoring (R(2) ; 0.870, 0.947 vs. 0.048, 0.717 (head-region not included), n=31). CONCLUSION: DF-BIA and DXA cannot be used interchangeably in geriatric medical patients. However, high replicability shows that both methods may be used for monitoring under standardized conditions. Results indicate gender-specific differences, and segmental analysis can only be recommended with DXA using clear anatomical references.