Performance of FRAX in clinical practice according to sex and osteoporosis definitions: the Manitoba BMD registry.

Abstract

Some studies have reported that FRAX performance differs according to sex and/or osteoporosis definitions. We evaluated whether the performance of FRAX to predict incident MOF and HF in women and men was affected by the presence or absence of osteoporosis defined by World Health Organization (WHO) or National Osteoporosis Foundation (NOF) criteria. We studied women and men age ≥ 40years with baseline hip and spine DXA scans (1996-2013). Individuals were classified into four non-overlapping subgroups: osteoporosis by WHO criteria, osteoporosis exclusively by NOF criteria, high fracture risk by FRAX (MOF ≥ 20% or HF ≥ 3%, without osteoporosis), and low fracture risk (MOF < 20% and HF < 3% without osteoporosis). In each subgroup, we evaluated stratification (hazard ratios [HR]) and calibration (observed vs predicted 10-year fracture probability) for incident fracture. The population included 62,275 women (5345 MOF and 1471 HF) and 6455 men (405 MOF and 108 HF). FRAX scores were strongly predictive of MOF (HR per SD: women 2.12, 95% CI 2.06-2.18; men 1.89, 95% CI 1.73-2.08; sex interaction p value = 0.97) and HF (women 4.78, 95% CI 4.44-5.14; men 4.20, 95% CI 3.22-5.49; sex interaction p value = 0.71). FRAX scores gave similar HRs for MOF among the four subgroups (subgroup interaction p value 0.34 for women, 0.22 for men). Observed versus predicted 10-year MOF and HF probability for the defined subgroups demonstrated a high level of concordance for women and men (all r2 ≥ 0.9). FRAX was a strong and consistent predictor of MOF and HF in both women and men and performed well in those with osteoporosis whether defined by WHO or NOF criteria.