Objective Physical Performance Predicts 10-Year Fracture Risk Without Densitometry in Older Men and Women:...

FRAX estimates 10-year fracture probability from clinical risk factors and, for best performance, DXA-derived bone mineral density (BMD), limiting its use where densitometry is unavailable. We tested whether a five-variable, laboratory- and imaging-free model (age, body-mass index, grip strength, gait speed and chair-stand rate; M5) predicts incident fracture and how it compares with FRAX. In community-dwelling older men (Osteoporotic Fractures in Men Study, n = 5994; 73.7 years) and women (Study of Osteoporotic Fractures, n = 9704; 71.6 years) with baseline physical-performance measures and adjudicated incident major osteoporotic fracture (848 and 2522 events), M5 was developed by five-fold cross-validation; we assessed time-dependent AUC at 5, 10 and 15 years, calibration and decision-curve net benefit. Comparators were FRAX in men and a conservative FRAX-surrogate model (in-cohort clinical risk factors) in women, since individually linked, licensed FRAX scores were not available for SOF. In men, M5 discriminated fracture better than FRAX without BMD (10-year ΔAUC + 0.052, 95% CI + 0.023 to + 0.080) and did not differ significantly from FRAX with BMD (+ 0.029, - 0.005 to + 0.063) in this cross-validated, in-cohort comparison; adding BMD to M5 gave the highest discrimination. In women, M5 did not differ significantly from the FRAX-surrogate comparator (- 0.008, - 0.018 to + 0.002), and BMD-based models performed best. M5 was well calibrated (calibration slope 0.97-0.98) and gave higher (men) or equivalent (women) net benefit. Death was a major competing event (72.7% of men, 61.4% of women) but changed 10-year incidence by ~ 1 percentage point. A five-variable, laboratory- and imaging-free model incorporating physical performance stratifies 10-year fracture risk at least as well as FRAX without BMD in men, without requiring imaging, laboratory tests or fracture history. Because M5 was developed within these cohorts whereas FRAX is an external, fixed algorithm, part of this margin reflects development asymmetry; the findings are promising but require external validation before clinical use, and do not replace measured BMD.