A comparative study of using non-hip bone density inputs with FRAX®

Summary: Use of lumbar spine T-score or minimum T-score as a bone mineral density ( BMD ) input to the FRAX® algorithm led to miscalibration compared with the recommended femoral neck input. Use of a weighted mean between the lumbar spine and femoral neck T-scores was found to provide an arithmetically equivalent result to a previously described offset adjustment. Introduction: FRAX assumes that the BMD input, when used in the calculation, is from the femoral neck. Use of other BMD inputs is not recommended, but there are no studies describing how this affects the performance of FRAX. Methods: Ten-year probabilities of a major osteoporotic fracture were calculated with different BMD inputs for 20,477 women and men aged 50 years and older from Manitoba, Canada. FRAX probability calculated with femoral neck BMD was designated the reference method. We also derived FRAX probabilities where the BMD input was based upon the lumbar spine T-score, minimum T-score ( lumbar spine or femoral neck ), weighted mean T-score ( lumbar spine or femoral neck ), or used an adjustment for the spine–hip T-score difference ( offset ). Fracture outcomes were assessed using a population-based administrative data repository. Results: All FRAX models showed good risk stratification with minimal differences. There was no consistent improvement in FRAX performance when lumbar spine or minimum T-score were used as inputs, but calibration was adversely affected due to higher mean fracture probabilities compared with the femoral neck. The weighted mean T-score was found to be equivalent to the spine–hip T-score offset adjustment, and both slightly improved risk classification without a change in calibration. Conclusions: The choice of BMD input to the FRAX model has a large effect on performance. The lumbar spine T-score or minimum T-score should not be used as inputs to the FRAX algorithm. Use of a weighted mean between the lumbar spine and femoral neck T-scores slightly improves risk classification.