Uncovering racial and genetic disparities in FRAX performance for fracture risk assessment in postmenopausal women

The Fracture Risk Assessment Tool (FRAX) is widely used in osteoporosis management, yet its performance across diverse racial, ethnic, and genetic populations remains uncertain. We evaluated the accuracy of FRAX in predicting fracture risk, with and without bone mineral density (BMD) measurements, among 27,512 postmenopausal women aged 50 to 79 years participating in the Women's Health Initiative (WHI) with 10 years of follow-up. FRAX-predicted fracture risks were compared to observed outcomes, stratified by race/ethnicity and genetic risk categories derived from a Genome-Wide Polygenic Score (GPS), calculated using 103,155 genetic variants via the LDpred algorithm and UK Biobank GWAS data. FRAX with BMD information substantially overestimated fracture risk for African American women by 82% (hazard ratio [HR], 0.18; 95% CI, 0.10-0.36) and Hispanic women by 48% (HR, 0.52; 95% CI, 0.31-0.87), compared with non-Hispanic White women (both P < .001). Conversely, FRAX with BMD information significantly underestimated fracture risk in women with the highest genetic risk (top 5%; HR, 2.15; 95% CI, 1.22-3.68; P < .001), compared with the low GPS reference group. Calibration analyses revealed systematic inaccuracies, highlighting racial disparities in overestimating fracture risk and genetic disparities in underestimating fracture risk predictions. Sensitivity analyses confirmed these findings persisted regardless of estrogen use. Our results underscore critical limitations of FRAX, suggesting that integrating genetic risk profiling and implementing race-specific recalibrations can substantially improve fracture risk prediction accuracy, promote equity in osteoporosis care, and support personalized clinical decision-making.