A new predictive index for vertebral fractures: The sum of the anterior vertebral body heights

Evaluation of osteoporotic vertebral fracture risk is currently based on measurement of bone mineral density (BMD), but bone strength depends also on bone quality parameters. Aim of this study was to evaluate the validity of a new vertebral morphometric index, the Anterior Vertebral Heights sum (AHs) in discriminating women at high risk of vertebral fracture, comparing its diagnostic accuracy with that of BMD measured at lumbar spine (LS-BMD) and femoral neck (FN-BMD). Materials and methods: A total of 163 Caucasian post-menopausal women (age range 46-74 years, mean age +/- SD = 63.8 +/- 7.1 years), who did not present prevalent fractures at baseline evaluation, were observed at longitudinal follow-up. X-ray of the thoracic and lumbar spine, LS-BMD and FN-BMD measurements were obtained in all patients at baseline and repeated at the second follow-up visit 18-24 months later (mean 21 +/- 1.7 months). Radiographs of spine were analysed in order to identify vertebral fractures using a visual semiquantitative method (SQ) and vertebral morphometry as well as by calculating the AHs morphometric index. Results: During follow-up, 21/163 patients (12.9%) sustained a new vertebral fracture; 95.2% (20/21) of fractured patients but only 4.9% (7/142) of non-fractured women had reduced AHs values. As regarding BMD, 66.6% (14/21) and 61.9% (13/21) of women with incident fracture were osteoporotic at lumbar spine and femoral neck baseline evaluation, whereas among non-fractured women, 38% (54/142) at LS-BMD and 33.1% (47/142) at FN-BMD were osteoporotic. Analyses of Receiver Operating Characteristic (ROC) curves showed that AHs discriminated vertebral fractures almost perfectly (AUC 0.97: 95% CI 0.950.99). On the other hand, the AUC for LS-BMD) was only 0.73 (95% CI 0.64-0.81) and for FN-BMD was 0.72 (95% CI 0.63-0.80), showing that the diagnostic accuracy of AHs was significantly higher compared to that of LS-BMD (p<0.001) or FN-BMD (p<0.001). A modified Poisson regression model for binary data was used to assess the independent role of AHs in predicting vertebral fracture. The effect of AHs remained statistically significant (p