RESUMO
PURPOSE: The aim of the study was the presentation of osteoporotic fracture prediction in men. METHODS: Eight-hundred and one men at the mean age of 70.8 ± 9.31 years were examined. The 10-year fracture prediction was established, using the FRAX calculator and Garvan nomogram. RESULTS: The mean value for any fracture and hip fracture probabilities for FRAX were 7.26 ± 5.4% and 3.68 ± 4.25%, respectively. For Garvan fracture, risk values were 26.44 ± 23.83% and 12.02 ± 18.1%. The mean conformity for any fracture and hip fracture prediction for threshold of 20% (any fracture) and 3% (hip fracture) between Garvan and FRAX values was 55.8% (κ 0.041) and 79.65% (κ 0.599), respectively. ROC analyses showed the following areas under the ROC curves (AUC) for any fractures: FRAX 0.808 and Garvan nomogram 0.843 (p = 0.059). The AUC values for hip fractures were 0.748 for Garvan nomogram and for 0.749 FRAX, and did not differ. On the base of ROC data, the cut-off values with best accuracy to predict fractures for both methods were established. The conformity between methods for thresholds indicated by ROC analysis was 72.5% (κ 0.435) for any and 77.7% (κ 0.543) for hip fractures. CONCLUSION: The conformities between FRAX and Garvan in regard to hip fracture prediction were acceptable for a threshold of 3% and thresholds derived by ROC analysis, while for any fracture we recommend to use thresholds established by ROC analysis. This may suggest that the use of "universal" cut-off points is probably misleading.