Agreement and systematic difference between bioelectrical impedance analysis and dual-energy X-ray absorptiometry for appendicular lean mass in 1617 community-dwelling older adults: The Bunkyo Health Study.
Nakagata T, Yamada Y, et al. • Clinical nutrition ESPEN • 2026
InBody 770 and Hologic DXA provide highly correlated ALM estimates but show a systematic method/device difference (~10% on average), which materially impacts low muscle mass screening prevalence, though recalculation from impedance variables, a simple ~1.10 scaling, or InBody-specific cut-offs can improve agreement with DXA-referenced estimates.
Key Findings
Results
InBody 770 ALM showed a strong correlation with DXA ALM but systematically underestimated it by approximately 9.2% on average.
Pearson correlation coefficient between InBody and DXA ALM was r = 0.95
InBody ALM was ~9.2% lower than DXA on average across the full cohort
Sex-specific underestimation was ~7.0% in men and ~10.7% in women
Study population included 1617 community-dwelling older adults aged 65-84 years (680 men, 937 women)
Results
Bland-Altman analysis revealed a mean systematic difference and proportional bias between InBody 770 and Hologic DXA for ALM.
Mean difference (InBody - DXA) was -1.64 kg
95% limits of agreement ranged from -4.14 to 0.87 kg
A proportional difference was detected with a slope of -0.03 (p < 0.001), indicating the bias was not constant across the range of measurements
Measurements were taken after overnight fasting, with MF-BIA preceding DXA in the morning
Results
Low muscle mass prevalence was substantially higher when assessed by InBody 770 original software output compared to DXA using AWGS 2019 criteria.
LMM prevalence was 38.1% by InBody (original software) vs 9.1% by DXA
LMM was defined using Asian Working Group for Sarcopenia (AWGS) 2019 criteria
The nearly fourfold difference in prevalence demonstrates the material clinical impact of the systematic device difference
Results
Three correction approaches—recalibration by scaling, Yamada's equation, and InBody-specific cut-offs—each brought LMM prevalence estimates close to DXA-based estimates.
A simple recalibration of InBody values multiplied by 1.10 yielded an LMM prevalence of 9.7%, close to the DXA-based 9.1%
Recalculation using Yamada's equation from impedance variables (Z5, Z50, Z250; Ht2/Z50, Z250/Z5, and 1/Z50) yielded a similar prevalence to DXA
Applying InBody-specific cut-offs (<6.6 kg/m2 for men and <5.0 kg/m2 for women) also yielded a prevalence of 9.5%, similar to DXA
The authors note that findings are device-/method-specific and should be externally validated
Background
Previous studies examining the validity of BIA-estimated ALM against DXA have produced inconsistent results, and no prior study had directly compared LMM prevalence assessed by DXA and BIA in the same cohort.
The inconsistency of prior validation studies motivated the current investigation
This study is described as, to the authors' knowledge, the first to directly compare LMM prevalence by DXA and BIA in the same cohort
The study used a segmental multi-frequency BIA device (InBody 770) and a DXA device (Hologic Discovery)
Nakagata T, Yamada Y, Tabata H, Someya Y, Kaga H, Kawamori R, et al.. (2026). Agreement and systematic difference between bioelectrical impedance analysis and dual-energy X-ray absorptiometry for appendicular lean mass in 1617 community-dwelling older adults: The Bunkyo Health Study.. Clinical nutrition ESPEN. https://doi.org/10.1016/j.clnesp.2026.102961