Choosing between predictors of fractures

Area of Science:

• Osteoporosis research
• Biostatistics
• Gerontology

Background:

• Accurate identification of individuals at high risk for osteoporotic fractures is crucial for effective clinical intervention.
• Traditional statistical methods for comparing fracture predictors have limitations.
• Receiver-operating-characteristics (ROC) curves have been previously employed for predictor assessment.

Purpose of the Study:

• To present a formal statistical approach for comparing individual and sets of predictors of osteoporotic fractures.
• To contrast newer methods like bootstrapping with traditional ROC curve analysis.
• To evaluate the utility of different bone mass measurements in predicting fractures.

Main Methods:

• Utilized bootstrapping methods for statistical comparison of fracture predictors.
• Applied time-to-fracture data from a study of 521 subjects with up to 12.5 years of follow-up.
• Compared bone mineral density (BMD), bone mineral content (BMC), and bone mineral apparent density (BMAD) as predictors.

Main Results:

• Bootstrapping demonstrated advantages in comparing fracture prediction models.
• Bone mineral density (BMD) was a significantly better predictor of fractures than bone mineral content (BMC) in free-living subjects.
• BMAD did not improve fracture prediction compared to BMC and BMD in the studied populations.

Conclusions:

• The bootstrapping approach provides a robust method for comparing osteoporotic fracture predictors.
• The predictive value of bone mass measurements varies depending on the population and measurement type.
• Further research is needed to refine fracture risk prediction models using advanced statistical techniques.