MRI-based body composition predicts incident type 2 diabetes mellitus

Background Body composition is a key determinant of metabolic health, yet the combined impact of adipose distribution and muscle quality on type 2 diabetes mellitus (T2DM) risk remains unclear. Purpose To evaluate the predictive value of MRI-derived adipose and muscle parameters for incident T2DM and to identify distinct body composition phenotypes associated with differential risk. Methods We analyzed 8,208 UK Biobank participants without baseline T2DM who underwent MRI. Measurements included subcutaneous adipose tissue (SAT), visceral adipose tissue (VAT), muscle fat infiltration, lean tissue mass, and thigh fat-free muscle volume. K-means clustering identified body composition phenotypes, and multivariable Cox regression with Kaplan–Meier analyses evaluated incident T2DM. Results Four phenotypes were identified: Healthy (low adiposity, high muscle mass; n=1,964), Lean (very low VAT, low muscle mass; n=2,880), Visceral obesity (high VAT, high muscle mass; n=1,647), and Sarcopenic obesity (high adiposity, high muscle fat infiltration, low muscle mass; n=1,717). During follow-up, T2DM incidence differed significantly across groups (P＜0.0001). Compared with the Healthy phenotype, the Visceral obesity group showed a markedly increased T2DM risk (HR 3.43; 95% CI 2.14–5.51; p<0.001), while the Sarcopenic obesity group had the highest risk (HR 4.92; 95% CI 2.35–10.3; p<0.001). No significant association was observed in the Lean group (p=0.6). Conclusions MRI-derived body composition phenotypes, particularly those with adverse muscle quality and high visceral adiposity, strongly predict incident T2DM.