Comparing DXA and MRI body composition measurements in cross-sectional and longitudinal cohorts

Detailed body composition assessment, through measurement of adipose tissue and muscle distribution, is essential for understanding population health trends, guiding targeted interventions, evaluating lifestyle effects, and monitoring age-related changes such as sarcopenia. Dual X-ray absorptiometry (DXA) provides estimates of fat and lean mass from low-dose X-ray images, while magnetic resonance imaging (MRI) offers three-dimensional measurements of adipose tissue and muscle distribution. Both methods have their strengths and weaknesses, but comparison between them at scale is missing.

We assessed the accuracy of DXA compared to MRI for evaluating body composition. Using data from 32,961 participants in the UK Biobank, including 1,928 re-scanned participants after about two and a half years, we examined cross-sectional and longitudinal agreements in DXA and MRI measurements within android and gynoid regions. MRI scans were acquired using a Siemens 1.5 T MRI scanner, and DXA scans on a GE iDXA.

Our results show that DXA is able to capture fat measurements, including visceral adipose tissue and fat mass, but overestimates lean mass compared to MRI, particularly in android regions for men (4.10 kg measured by DXA vs. 1.74 kg by MRI) and women (2.92 vs. 1.10 kg). Longitudinal MRI data reveal a 4-5% muscle and lean mass decrease, undetected by DXA, which shows lean mass increases in women at the follow-up visit.

Although DXA is practical for population-level fat assessments, it may not be as precise for lean mass assessments, especially in longitudinal studies. If economically and practically feasible, MRI remains the preferred method for detailed and precise longitudinal body composition analysis.