Dual energy X-ray absorptiometry produces larger measurement error in non-Hispanic Caucasians than Hispanics

Dual energy X-ray absorptiometry (DXA) is often used as a criterion measure in body composition research and in clinical settings for the estimate of body fat percent (%Fat). The accuracy of DXA for predicting %Fat has primarily been conducted in non-Hispanic populations.

The purpose of this study was to determine the agreement of DXA-derived %Fat in Hispanic and non-Hispanic Caucasian adults.

The sample consisted of Hispanic males (n = 96) and females (n = 102) and non-Hispanic Caucasian males (n = 145) and females (n = 161). The %Fat of a whole-body DXA scan was compared against a criterion 4-compartment (4C) model via constant error (CE = DXA - 4C model) and 95% limits of agreement. Also, a 2 × 2 factorial ANOVA, using CE as a dependent variable, was conducted to examine the main and interaction effects of sex and ethnicity.

When compared to the 4C model, DXA overestimated %Fat by 4.0% in Hispanics and 5.5% in non-Hispanic Caucasians (all p < 0.05). The 95% limits of agreement ranged from ±5.5% to ±5.9% for all group comparisons. The 2 × 2 factorial ANOVA indicated the CE was greater in non-Hispanic Caucasians than Hispanics (CE difference = 1.5%; p < 0.05).

Our findings revealed that DXA significantly overestimates %Fat in both populations (Hispanics and non-Hispanic Caucasians), when compared to a 4C model, regardless of sex (male or female). However, the error is more profound in non-Hispanic Caucasian adults. It is worth nothing that DXA may be useful for tracking changes in body composition that occur throughout a lifestyle intervention. Nonetheless, practitioners should be aware that the estimate of %Fat from DXA may be larger than the actual values obtained from a 4C model.