Probability of fractures predicted by FRAX® and observed incidence in the Spanish ECOSAP Study cohort

Abstract

To assess the ability of the Spanish version of the WHO fracture risk assessment tool (FRAX®) to predict the observed incident fractures in the ECOSAP Study cohort. 5201 women, aged 65 or older, were enrolled in a three-year, prospective study by a non-randomized sampling of consecutive cases in 58 primary care centers in Spain. Participants completed an osteoporosis and fracture risk questionnaire and attended follow-up visits every 6 months. All radiologically or surgically confirmed low-trauma, non-spinal fractures were collected. The individual 10-year absolute risks of hip and major osteoporotic fractures were calculated with the FRAX® algorithms for Spain without the inclusion of the bone mineral density (BMD) measurements. Calibration was evaluated by comparing the three-year estimated (E) fractures predicted with FRAX® with the number of observed (O) fractures, and their discriminative ability for the probability of new fractures with the area under the receiving operating characteristic (ROC) curves. Fifty (0.96%) women sustained an incident hip fracture, and 201 (3.81%) women presented with major osteoporotic fractures (hip, forearm or humerus). The E/O ratios for hip and major osteoporotic fractures were 1.10 and 0.66 respectively. Clinical vertebral fractures were not collected; therefore, the E/O ratio for major fractures should be expected to be lower. The difference between E and O cases reached statistical significance (χ(2), p<0.001). Areas under the ROC curves were 0.640 and 0.615 for hip and major osteoporotic fractures respectively. The Spanish FRAX® underestimates the risk for major osteoporotic fractures. The estimated risk for hip fractures was similar to the observed fractures; however the algorithm had only modest discriminative ability. These results should be interpreted in the context of the relatively low number of observed fractures, especially at the hip.