Trabecular bone score may improve FRAX® prediction accuracy for major osteoporotic fractures in elderly Japanese men: the Fujiwara-kyo Osteoporosis Risk in Men (FORMEN) Cohort Study.

ABSTRACT

UNLABELLED FRAX® is widely used to evaluate fracture risk of individuals in clinical settings. However, FRAX® prediction accuracy is not sufficient, and improvement is desired. Trabecular bone score, a bone microarchitecture index, may improve FRAX® prediction accuracy for major osteoporotic fractures in community-dwelling elderly Japanese men. INTRODUCTION: To improve fracture risk assessment in clinical settings, we evaluated whether the combination of FRAX® and Trabecular Bone Score (TBS) improves the prediction accuracy of major osteoporotic fractures (MOFs) in elderly Japanese men compared to FRAX® alone. METHODS: Two thousand and twelve community-dwelling men aged ≥65 years completed the Fujiwara-kyo Osteoporosis Risk in Men (FORMEN) Baseline Study comprising lumbar spine (LS) and femoral neck areal bone mineral density (aBMD) measurements, and interviews regarding clinical risk factors required to estimate 10-year risk of MOF (hip, spine, distal forearm, and proximal humerus) using the Japanese version of FRAX® (v.3.8). TBS was calculated for the same vertebrae used for LS-aBMD with TBS iNsight software (v.2.1). MOFs that occurred during the follow-up period were identified by interviews or mail and telephone surveys. Prediction accuracy of a logistic model combining FRAX® score and TBS compared to FRAX® alone was evaluated by area under receiver-operating characteristic curves (AUCs), as well as category-free integrated discrimination improvement (IDI) and net reclassification improvement (NRI). RESULTS: We identified 22 men with MOFs during 8140 person-years (PY) of follow-up among 1872 men; 67 men who suffered from fractures other than MOFs were excluded. Participants with MOFs had significantly lower TBS (p = 0.0015) and higher FRAX® scores (p = 0.0089) than those without. IDI and NRI showed significant improvements in reclassification accuracy using FRAX® plus TBS compared to FRAX® alone (IDI 0.006 (p = 0.0362), NRI 0.452 (p = 0.0351)), although no difference was observed in AUCs between the two. CONCLUSIONS: TBS may improve MOF prediction accuracy of FRAX® for community-dwelling elderly Japanese men.