Prediction of hip fracture by high-resolution peripheral quantitative computed tomography in older Swedish women
Authors' conclusion
We conclude that the addition of HR-pQCT parameters improves the prediction of hip fractures in a cohort of older Swedish women.
Study context
- Population
- Community-dwelling women aged 75–80, Sweden; baseline DXA and HR-pQCT2,856 with complete HR-pQCT data
- Participants
- 3,028 women
- Exposure
- Cox model with HR-pQCT vs reference model vs FRAX risk score model
- Outcome
- Prediction of hip fracture (time-dependent AUC)
- Follow-up
- 5 yr
Methodological detailsprediction model development, Cox proportional hazards regression and machine learning algorithms
- Prediction model development
- Cox proportional hazards regression and machine learning algorithms
Main finding
As reported in the abstract
Hip fracture prediction, time-dependent AUC
HR-pQCT model
0.75
CI 0.64–0.85
Reference model
0.71
CI 0.58–0.81
FRAX model
0.70
CI 0.60–0.80
HR-pQCT model vs Reference model: P < 0.001 · HR-pQCT model vs FRAX model: P < 0.001
Observational data: associations, not effects.
Further findings
Incident fractures · median follow-up 8.0 yr
- Hip fractures
- 217
- Major osteoporotic fracture (MOF)
- 746
- Any fracture
- 1008
Strongest predictors of hip fracture (Cox, adjusted for age, BMI, clinical risk factors (CRFs), FN BMD)
- Tibia total volumetric BMD and cortical thickness were the strongest predictors
Cox vs neural network model accuracy
- Cox model had significantly higher accuracy than the neural network-based model (best machine learning algorithm) at clinically relevant sensitivity levels
Original publication
Prediction of hip fracture by high-resolution peripheral quantitative computed tomography in older Swedish women. J Bone Miner Res. Published 27 January 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →