[Reproducibility and validity of radiographic measurements for the calculation of proximal femur morphometric indices predictive of hip fractures].

Cross-sectional-retrospective study with 120 anteroposterior pelvic radiographs. Eight measurements were taken by three raters in two times to calculate the morphological-cortical (MCI), cortico-metaphyseal (CMI), and cortico-diaphysaire (CDI) indices. Intra/interobserver reproducibility was evaluated with the Bland-Altman method, and hip fracture occurrence was predicted by comparing cut-off points between raters maximizing sensitivity/specificity. The proportion of measurements outside ± 2 SD was similar for diameters/corticals (~5%), but exceeded 1 mm in some corticals, representing 10-20% of distances < 1 cm. Cut-off points fluctuated between 2.45 and 2.53 for MCI (3% difference), between 0.37 and 0.41 for CMI (12% difference), and between 0.23 and 0.33 for CDI (44% difference).