Fracture risk prediction, treatment thresholds, and the role of bone mineral density: a comparative analysis of FRAX and SHAFRE.

Abstract

Introduction: Accurate prediction of fracture risk is important for treatment decisions, yet evaluations of commonly used thresholds are limited, and the added value of volumetric bone mineral density (vBMD) remains uncertain. We conducted the first head-to-head comparison of the SHAFRE and FRAX algorithms using identical predictor information.

Materials and methods: We included 3525 community-dwelling Swedish men and women (mean age 71.4 years) who underwent a health examination with femoral neck areal BMD (aBMD) and radial vBMD. Incident fractures were retrieved from the National Patient Register. Model performance was evaluated using threshold-specific sensitivity and specificity.

Results: Over a mean follow-up of 8.7 years, 559 participants sustained a fracture. SHAFRE predicted a mean 10-year fracture risk of 21.9% in those who fractured versus 15.6% in the remaining cohort (ROC-area: 68%, 95% CI: 66-71%). Corresponding FRAX values were 16.7% and 12.6% (ROC-area: 66%, 95% CI: 63-68%), and FRAX slightly underestimated fracture risk in this cohort. Adding radial vBMD on top of aBMD did not materially improve discrimination for SHAFRE (70%, 95% CI: 67-72%) or FRAX (68%, 95% CI: 65-70%). Threshold-specific analysis identified an optimal predicted risk threshold of 13-17%, consistently below the commonly recommended 20%.

Conclusions: Predictive ability for major fractures remained modest for both algorithms and was not improved by adding vBMD. The estimated optimal threshold for treatment initiation was lower than the commonly recommended 20%. These findings reinforce that discrimination is substantially stronger for hip fracture than for major osteoporotic fractures.