Gluteus maximus morphology shows sex-specific associations with type 2 diabetes risk in UK Biobank imaging

Background: The gluteus maximus (GM) is a major hip extensor essential for mobility and metabolic health. Most MRI studies rely on global measures, such as muscle volume or fat fraction, which can overlook spatially localised remodelling. Here, we integrate conventional volumetric and fat fraction metrics with 3D mesh-based shape phenotypes to provide a spatially resolved characterisation of GM morphology in relation to anthropometric, lifestyle, and cardiometabolic factors, with a focus on type 2 diabetes (T2D) and sex-specific effects. Methods: We analysed T1 Dixon MRI from UK Biobank participants to quantify GM muscle volume, fat fraction, and regional surface morphology using 3D meshes. Statistical parametric mapping was used to assess regional associations with anthropometric, lifestyle, and clinical variables Bi-directional causal mediation analyses were performed using GM volumetric and principal components (PCs) of shape variation. PCs were also tested for associations with prevalent and incident disease. Longitudinal changes in GM composition were evaluated in participants with repeated imaging evaluations. Results: GM muscle volume and fat fraction were strongly associated with age, adiposity, and physical activity. Shape analysis revealed spatially localised remodelling patterns not captured by global measures, with region-specific surface shrinkage linked to age, BMI, alcohol intake, grip strength, physical activity, frailty, osteoporosis, and cardiometabolic disease. T2D showed marked sex-differences, with regional shrinkage in men and relative expansion in women. PCA reduced high-dimensional shape variation into interpretable components. Mediation analyses indicated that T2D-related differences in GM morphology partly mediated increases in fat fraction, suggesting that disease effects manifest through spatially patterned shape changes rather than overall muscle size. PCs capturing variations in the central-upper posterior and anterior GM, differentiated between T2D cases from controls, and were associated with incident T2D risk (Men: PC6 HR per SD: 0.81 [0.70-0.95], false discovery rate (FDR)-adjusted p = 0.038, in left GM; 0.76 [0.65-0.88], p = 0.002, in right GM; women; PC5 HR = 1.32, [1.08-1.61], p = 0.032, in right GM). Conclusions: Integrated 3D quantification of GM composition and morphology provides spatially resolved biomarkers that go beyond muscle volume and fat fraction. By capturing region-specific GM remodelling, linked to anthropometric, lifestyle and cardiometabolic factors, this approach offers a more nuanced characterisation of muscle-fat phenotypes and enhances mechanistic insight and risk stratification in population-based imaging studies.