To compare body composition assessment by bioelectrical impedance analysis (BIA) and dual-energy X-ray absorptiometry (DXA) in a multiethnic sample of adolescent girls.

Data were from a physical activity intervention study among 254 14-20-year-old sedentary American girls, including 69 whites, 74 blacks, 42 Hispanics, and 69 Asians. Height and weight were objectively measured. Body composition was assessed using a foot-to-foot BIA and a fan-beam DXA. We calculated ethnic-specific estimates of percentage body fat (BF%), fat mass (FM), fat mass index (FMI), fat-free mass (FFM), and fat-free mass index (FFMI) from BIA and DXA. We used Bland-Altman plots to examine ethnic-specific agreement between BIA and DXA, and used linear regression to test whether the BIA-DXA difference varied across the mean.

Compared to DXA, BIA estimates of fat measures (BF%, FM, and FMI) were lower and lean tissue measures (FFM and FFMI) higher. For example, the BIA-estimated BF% was lower than the DXA estimate by between 4.9% (95% CI: -5.9, -3.9) in blacks and 8.7% (-7.0, -5.0) in Asians, with large limits of agreement (-15.4 to -5.4 in blacks and -16.8 to -0.4 in Asians). Regression analysis showed that BIA-DXA differences were not constant across means for any body composition measure among Asians or for any measure except BF% among whites.

Compared to DXA, BIA yielded lower estimates of adiposity in a multiethnic adolescent sample. BIA-DXA differences varied by ethnicity, and across mean body composition values for some ethnicities.