Keyak et al. tested whether CT scan-derived finite element models could predict femoral fracture load in cadaveric specimens. Eighteen pairs of femora were loaded to failure under stance and fall conditions, and CT/FE predictions were compared against QCT densitometry. The central question: does this computationally intensive approach outperform the simpler densitometry methods already in clinical use?
In 1998, DXA and QCT densitometry were the standard tools for hip fracture risk assessment, but both measure bone density at a single region and cannot account for 3D geometry or patient-specific loading conditions.
This paper established the foundational methodology for CT-based finite element fracture prediction and set the performance benchmark that future refinements would need to beat: precision of −40% to +60% at 95% confidence, equivalent to what densitometry already offered.
For a trainee, the key lesson is understanding why densitometry alone is limited: it ignores cortical geometry, trabecular architecture, and how a specific patient loads their hip during a fall. CT/FE addresses all three — but in 1998, the added complexity did not yet translate to better individual-level prediction.
This work directly enabled subsequent generations of patient-specific bone strength modeling, including the QCT-FEA tools now used in osteoporosis drug trials and emerging clinical fracture risk calculators. The 1.1% additional variance from loading condition is a particularly board-relevant insight: bone structure, not just loading direction, dominates fracture risk.
Keyak et al. tested whether CT scan-derived finite element models could predict femoral fracture load in cadaveric specimens. Eighteen pairs of femora were loaded to failure under stance and fall conditions, and CT/FE predictions were compared against QCT densitometry. The central question: does this computationally intensive approach outperform the simpler densitometry methods already in clinical use?
In 1998, DXA and QCT densitometry were the standard tools for hip fracture risk assessment, but both measure bone density at a single region and cannot account for 3D geometry or patient-specific loading conditions.
This paper established the foundational methodology for CT-based finite element fracture prediction and set the performance benchmark that future refinements would need to beat: precision of −40% to +60% at 95% confidence, equivalent to what densitometry already offered.
For a trainee, the key lesson is understanding why densitometry alone is limited: it ignores cortical geometry, trabecular architecture, and how a specific patient loads their hip during a fall. CT/FE addresses all three — but in 1998, the added complexity did not yet translate to better individual-level prediction.
This work directly enabled subsequent generations of patient-specific bone strength modeling, including the QCT-FEA tools now used in osteoporosis drug trials and emerging clinical fracture risk calculators. The 1.1% additional variance from loading condition is a particularly board-relevant insight: bone structure, not just loading direction, dominates fracture risk.