Submitted to: The American Journal of Clinical Nutrition
Publication Type: Other
Publication Acceptance Date: 7/20/2008
Publication Date: 10/1/2008
Citation: Lukaski, H.C. 2008. In Search of a Practical Tool to Assess Regional Body Composition. The American Journal of Clinical Nutrition. 88:875-876.
Technical Abstract: Findings that regional accumulation or depletion of adipose tissue and fat signal increased risk of chronic disease (1,2) or are consequences of therapy (3) have prompted an advance in the use of methods to assess regional body composition in humans. This growing demand for accurate and sensitive methods of measurement poses unique challenges to the assumptions and physical bases of the available methods and to the need to develop diagnostic criteria that evaluate the impact of specific therapeutic regimens on the composition of these sites. In this issue of the Journal, Scherzer et al (4) describe the results of an observational study that examines the validity of dual x-ray absorptiometry (DXA) compared to magnetic resonance imaging (MRI) to estimate total body and regional fat mass in human immuno-virus positive (HIV+) and healthy adults. This study is important because it addresses the practical issue of the use of a generally available and relatively low-cost method, DXA, as a surrogate for MRI, the established research tool to measure regional adipose tissue volume and, by calculation, fat mass in clinical studies of human body composition. The principal finding is that DXA, regardless of manufacturer or type (pencil or fan x-ray beam), significantly overestimates total body and regional fat mass, as compared to MRI, in both HIV+ and control adults. There is a paucity of studies comparing measurements of regional fat depots determined with DXA and MRI. The few studies were limited by small sample sizes that included healthy adults; only one study reported a comparison of these methods in a limited number of HIV+ adults. Thus, the findings of the present study are a more robust comparison because they were derived from a multi-center trial with an appreciably greater number of HIV-negative and -positive adults than previously reported. Many of the potential problems associated with the use of multiple test locations have been controlled by using standardized measurement protocols and a single image reading center. Inclusion of a whole-body soft tissue phantom at the start of the study allowed a standardization of the individual DXA instruments. However, evidence of regular assessment of the accuracy and reliability of operation of each DXA scanner for the duration of the study is lacking. Although DXA-derived fat mass was highly correlated with MRI determinations, the magnitude of the bias or error of DXA estimates increased directly in proportion to fat mass. Importantly, the over-estimation of fat mass by DXA, when expressed as a percentage of the MRI-derived value, was greatest in the arms (up to 120%) compared to the legs (up to 70%) compared to the trunk (7%). This observation that DXA disproportionately over-estimates limb fat depots in regions of the body in which fat amounts are relatively low suggests that the basic assumption of DXA algorithms in sites such as the arms and legs may be inaccurate. Some explanations for the overestimation of fat by DXA include the interpolations for soft tissues located over bone (5) and the treatment of pixels containing a small portion of bone (6). Thus, use of proprietary algorithms to calculate soft tissue composition hampers an understanding of the potential errors in DXA determinations of regional fat. A novel approach to validation of DXA was the determination of the equivalence of DXA, relative to MRI, to assess the prevalence of a clinical outcome measure, peripheral lipoatrophy, in HIV-infected and control adults. As expected, the HIV+ subjects had lower distribution of leg fat than controls measured by using MRI and DXA. The prevalence of lipoatrophy by leg fat in the HIV-infected adults, operationally defined as the lowest decile of leg fat determined by DXA and MRI in controls, was significantly greater with DXA than MRI in men (69 vs 50%) and women (47 vs 33%). Th