Background: Studies in persons without HIV infection have compared adipose tissue measured by dual-energy X-ray absorptiometry (DXA) and magnetic resonance imaging (MRI), but no such study has been conducted in HIV-infected (HIV+) subjects, who have a high prevalence of regional fat loss.
Objective: We compared DXA- with MRI-measured trunk, leg, arm, and total fat in HIV+ and control subjects.
Design: A cross-sectional analysis was conducted in 877 HIV+ subjects and 260 control subjects in FRAM (Study of Fat Redistribution and Metabolic Change in HIV Infection), stratified by sex and HIV status.
Results: Univariate associations of DXA with MRI were strongest for total and trunk fat (r > or = 0.92) and slightly weaker for leg (r > or = 0.87) and arm (r > or = 0.71) fat. The average estimated limb fat was substantially greater for DXA than for MRI for HIV+ and control men and women (all P < 0.0001). Less of a difference was observed in trunk fat measured by DXA and MRI, but the difference was still statistically significant (P < 0.0001). Bland-Altman plots showed increasing differences and variability. Greater average limb fat in control and HIV+ subjects (both P < 0.0001) was associated with greater differences between DXA and MRI measurements. Because the control subjects had more limb fat than did the HIV+ subjects, greater amounts of fat were measured by DXA than by MRI when control subjects were compared with HIV+ subjects. More HIV+ subjects had leg fat in the bottom decile of the control subjects by DXA than by MRI (P < 0.0001).
Conclusions: Although DXA- and MRI-measured adipose tissue depots correlate strongly in HIV+ and control subjects, differences increase as average fat increases, particularly for limb fat. DXA may estimate a higher prevalence of peripheral lipoatrophy than does MRI in HIV+ subjects.