BMI-stratified performance of native CT hounsfield units in the proximal femur for osteoporosis detection and fracture association

Opportunistic computed tomography (CT) bone density assessment using Hounsfield units (HU) is increasingly used. Body mass index (BMI) may influence dual-energy X-ray absorptiometry (DXA), computed tomography X-ray absorptiometry of the hip (CTXA), and CT attenuation values, potentially affecting HU thresholds and fracture discrimination. We quantified associations between BMI, femoral HU, hip T-scores, and prevalent fractures. We retrospectively analyzed 600 consecutive patients (65.3 ± 14.1 years; 81% female; BMI 26.7 ± 6.5 kg/m²) who underwent CTXA hip assessment. Trabecular HU were measured on coronal CT images using a circular ROI in the femoral head and an irregular ROI in the proximal femur. BMI subgroups followed WHO categories. Spearman correlations (ρ) and linear regression analyses (R²) were performed. Multivariable linear and logistic regression models were adjusted for age and sex. BMI correlated positively with femoral HU and hip T-score (ρ = 0.46; p < 0.001). In multivariable models, each 1 kg/m² increase in BMI was associated with higher HU values (+ 3.32 HU; 95% CI 2.61–4.02) and higher hip T-score (+ 0.081; 95% CI 0.064–0.098; p < 0.001). Fracture prevalence decreased across BMI categories. HU values discriminated patients with and without hip fracture (AUC up to 0.99). Given the cross-sectional design, these values reflect discrimination within the study cohort rather than prospective predictive validity. No proximal femur fractures occurred above the osteoporosis threshold of 94.77 HU. BMI was independently associated with CT attenuation and CTXA-derived bone density parameters and inversely associated with prevalent fractures in this cohort. BMI should be considered when interpreting opportunistic HU measurements and applying fixed thresholds.