Comparison of DXA, BIA, and anthropometry for assessing subcutaneous, visceral, liver, and pancreas fat measured by MRI.

Correlations of adiposity indices from DXA, BIA, and anthropometry with MRI measures were strongest for subcutaneous fat, followed by visceral fat, and weakest for liver and pancreas fat.

• Study included 18,622 White participants from the UK Biobank imaging visit.
• A total of 27 indices from DXA, BIA, and anthropometry were selected to characterise overall and regional adiposity.
• All indices were standardised and Pearson correlation coefficients were used to assess linear association with MRI-determined fat depots.
• This pattern held generally across DXA (total, trunk, android, gynoid, arm, and leg fat), BIA (total, trunk, arm, and leg fat), and anthropometry (BMI, waist and hip circumference).

Correlation coefficients with MRI-based visceral fat were larger for indices reflecting abdominal adiposity compared to other regional measures.

• Indices with stronger correlations to MRI visceral fat included DXA-based visceral, android, and trunk fat, BIA-based trunk fat, and waist circumference.
• Measures of total adiposity (e.g., DXA- and BIA-derived total fat) also showed relatively larger correlations with MRI-based visceral fat.
• Analyses were conducted in subgroups stratified by age and sex.

Absolute agreement with MRI-based visceral fat was highest for DXA-based visceral, trunk, and android fat, followed by waist circumference.

• Intraclass correlation coefficients (ICC) for DXA-based visceral, trunk, and android fat with MRI-based visceral fat ranged from ICC = 0.73 to 0.94.
• Waist circumference showed ICC = 0.73–0.77 with MRI-based visceral fat.
• The remaining adiposity indices showed lower absolute agreement with MRI-based visceral fat than these abdominal measures.
• ICC was used to evaluate absolute agreement between corresponding fat measures.

Waist circumference was identified as a potentially appropriate surrogate for assessing visceral fat when considering cost and accessibility.

• DXA-derived visceral, trunk, and android fat and waist circumference appeared highly reflective of MRI-determined visceral fat.
• Despite DXA measures showing comparable or higher agreement with MRI visceral fat, waist circumference was highlighted for its cost-effectiveness and accessibility.
• The recommendation was made in the context of comparing DXA, BIA, and anthropometry as alternatives to MRI.

MRI-derived liver and pancreas fat showed the weakest correlations with all adiposity indices from DXA, BIA, and anthropometry.

• This pattern of weakest correlation for liver and pancreas fat was consistent across DXA, BIA, and anthropometric indices.
• The finding suggests that none of the commonly used non-MRI adiposity measures adequately capture ectopic fat in the liver or pancreas.
• Subcutaneous fat showed the strongest correlations with non-MRI adiposity indices across all modalities tested.