Predicting fracture risk after non-recent high risk fracture: improving accuracy with simple modifiers to FRAX

Ye, Carrie, Morin, Suzanne N, Lix, Lisa M, McCloskey, Eugene V, Johansson, Helena, Harvey, Nicholas C, Kanis, John A and Leslie, William D (2026) Predicting fracture risk after non-recent high risk fracture: improving accuracy with simple modifiers to FRAX. Journal of Bone and Mineral Research, [zjag049]. (doi:10.1093/jbmr/zjag049).

Abstract

Prior fracture is among the strongest predictors of future osteoporotic fracture. However, the Fracture Risk Assessment Tool (FRAX®) incorporates it as a dichotomous variable, potentially obscuring important heterogeneity related to fracture site and multiplicity. Although recent (≤2-year) fractures are recognized as conferring imminent risk, it is unclear whether non-recent high-risk fractures (nrHRFs)-specifically hip, vertebral, or multiple fractures-continue to elevate long-term risk beyond that captured by standard FRAX inputs. 

In this population-based retrospective cohort study using the Manitoba Bone Mineral Density Registry linked to comprehensive provincial administrative health data, adults ≥40 years undergoing first DXA between 1996-2018 were included (N=88,653). Individuals who had a prior recent fracture (occurring less than 2 years prior to the index date) were excluded. Prior non-recent fractures (>2 years pre-index) were identified. Ten-year major osteoporotic fracture (MOF) and hip fracture probabilities were calculated using Canadian FRAX. Using Cox models adjusted for baseline FRAX probability, we evaluated residual risk associated with nrHRFs. A 2:1 train-test split was used to derive and internally validate recalibration multipliers based on observed-to-predicted fracture ratios incorporating competing risk of death. 

FRAX stratified fracture risk better in those without prior nrHRF compared to those with nrHRF. In the derivation cohort, despite including prior fracture as a FRAX input, observed-to-predicted ratios indicated underestimation of MOF risk for non-recent vertebral (1.32) and multiple fractures (1.34), and underestimation of hip fracture risk for non-recent multiple fractures (1.70). Derived multipliers (×1.3 for MOF; ×1.7 for hip fracture) demonstrated good calibration in the validation cohort. Applying adjustments reclassified 13.3% and 3.5% of individuals with prior nrHRFs into high-risk MOF and hip fracture categories, respectively. 

FRAX underestimates 10-year fracture risk in individuals with specific nrHRFs. Simple adjustment factors markedly improve calibration and enhance identification of patients who may benefit from fracture-preventive therapy.

This record has no associated files available for download.

Contributors

Author: Nicholas C Harvey

Download statistics