Ten-year re-validation of the fracture risk assessment tool (FRAX®) in Taiwan

The Taiwan-specific fracture risk assessment tool (FRAX) model is widely used to estimate 10-year fracture risk. However, FRAX with bone mineral density (BMD) may underestimate fracture risk, particularly in older adults.

To evaluate the predictive performance and calibration of the Taiwan-specific FRAX calculator using complete 10-year real-world outcomes and to establish data-driven, population-specific screening thresholds.

A total of 1,976 participants aged ≥ 40 years were recruited from two population-based cohorts in Taiwan (2009-2010) and followed for 10 years via linkage with the National Health Insurance Research Database. Ten-year probabilities of major osteoporotic fracture (MOF) and hip fracture (HF) were estimated using Taiwan-specific FRAX calculators. Model calibration was evaluated using observed-to-expected (O/E) ratios. Statistically optimal thresholds were determined via receiver operating characteristic analysis maximizing the Youden index.

During follow-up, 213 (10.8%) participants experienced MOF, and 93 (4.7%) experienced HF. FRAX without BMD demonstrated strong calibration for both MOF (O/E = 1.08; 95% CI: 0.94-1.24) and HF (O/E = 1.10; 95% CI: 0.89-1.34). Conversely, FRAX with BMD systematically underestimated risk, particularly among individuals aged ≥ 65 years (MOF O/E = 1.23; 95% CI: 1.06-1.42). Age, female sex, prior fracture, and low BMD were significant independent risk factors. Crucially, population-specific data-driven thresholds were identified via ROC analysis at 9.0% for MOF and 4.5% for HF using FRAX without BMD, and 8.5% for MOF and 4.0% for HF using FRAX with BMD, respectively.

FRAX® remains a useful and widely applied tool for fracture risk prediction in Taiwan. However, fracture risk may be underestimated, particularly in adults aged ≥ 65 years. Taiwan-specific, data-driven thresholds may improve risk stratification and support more targeted fracture prevention strategies.