The Prediction of Body Composition in African Americans From Spine and Hip Dual-Energy Absorptiometry

Abstract

Body composition, the makeup of the body's fat and lean tissue, is associated with important health outcomes and provides useful clinical information. Although body composition can be measured with total body dual-energy X-ray absorptiometry (DXA), this is rarely performed. As an alternative to total body DXA measurement, methods for estimation of body composition have been developed. These methods use soft tissue measures from spine and hip DXA to predict body composition and include prediction equations previously published by Leslie and proprietary equations within General Electric densitometry software. However, these estimates have not been tested in African Americans (AA), an ethnicity with a different distribution of fat than Caucasians (CA). Therefore, we examined the performance of the existing models in 99 CA and 162 AA subjects over the age of 40 who had total body, spine, and hip DXA measurements. We observed that existing models estimated body composition well in CA but underestimated fat mass and overestimated lean mass in AA. AA subjects were then randomly divided into 2 equal-sized subgroups—the first to develop new prediction equations and the second to independently validate them. We found that body composition can be more accurately estimated using either a new model that we derived in AA subjects using backward stepwise elimination or by adding a fixed offset for AA to the previously published model. Our results demonstrate that body composition estimates from spine and hip DXA require consideration of race/ethnicity.