Comparisons of phantom-based and phantomless QCT bone mineral density for predicting subsequent fractures

This study aimed to compare the agreement between phantom-based (PB-QCT) and phantomless (PL-QCT) quantitative CT bone mineral density (BMD) measurements, and their predictive utility for subsequent fractures in postmenopausal women following percutaneous kyphoplasty (PKP) for osteoporotic vertebral compression fractures (OVCFs).

In this prospective cohort study, 185 OVCF patients underwent lumbar CT with synchronous phantom calibration. BMD was measured via PB-QCT, PL-QCT (using subcutaneous adipose tissue (SAT), psoas muscle, or dual-tissue calibration), and vertebral Hounsfield units (vHU). Agreement was assessed via Bland-Altman analysis, and fracture prediction via Cox regression.

Dual-tissue PL-QCT showed the highest agreement with PB-QCT (absolute bias: 0.04 to 2.18 mg/cm³). However, PB-QCT demonstrated superior predictive power: per 1-SD decrease, it predicted 20% to 22% increased risk of spinal fractures (adjusted HR = 1.21, p = 0.032) and any-site fractures (HR = 1.20, p = 0.043), maintaining significance after multivariable adjustment.

Despite high agreement - especially between PB-QCT and dual-tissue PL-QCT - PB-QCT demonstrated modest but consistent advantages in predicting subsequent fractures over PL-QCT and vHU. These results reinforce the essential role of standardized phantom calibration for accurate fracture risk stratification in post-PKP patients, while also supporting the potential use of PL-QCT and vHU as opportunistic screening tools in settings where phantom calibration is not feasible.