Background According to the operational definition provided by the World Health Organization, osteoporosis is diagnosed when bone mineral density (BMD) measured at lumbar spine or proximal femur is at least 2.5 standard deviations lower than the young adult mean. Currently, dual X-ray absorptiometry (DXA) is the most widely adopted method for osteoporosis diagnosis, since it is considered the gold standard for BMD measurements. Therefore, DXA outcomes are often used as the ground truth reference to assess the diagnostic performance of novel alternative approaches that are undergoing clinical validation. However, very recent literature  documented that high percentages of clinical routine DXA investigations are typically affected by specific errors altering the final measurement result. This may also have significant effects on the comparative evaluation of the effectiveness of the innovative diagnostic methods under validation.
Objectives To assess the actual diagnostic performance of the Osteoporosis Score (OS), a recently introduced ultrasound (US) parameter for osteoporosis diagnosis on the femoral neck, through a retrospective analysis of the DXA reports of the enrolled patients.
Methods In a previous study , we directly assumed the outcome of DXA investigations as the gold standard reference to assess the OS performance for osteoporosis diagnosis on the femoral neck in a population of 170 patients aged in 55–80 years, obtaining an overall accuracy of 84.7% in patient classification (osteoporotic, osteopenic, or healthy) coupled with a good correlation between DXA-measured BMD and corresponding OS-derived values (r =0.71, p<0.001). In the present work, we performed a retrospective check of the 170 DXA reports, excluding all those cases presenting a typical DXA error as identified by recent literature (e.g., incorrect patient positioning, presence of artefacts, improper image segmentation, etc.)  and re-assessing the actual diagnostic accuracy of our US approach by considering only those patients having a reliable DXA report.
Results 57 patients out of the initial 170 (33.5%) were excluded because of clear DXA inaccuracies. The overall agreement between DXA and OS-based diagnoses on the remaining 113 patients was 93.8% (r =0.79, p<0.001), ranging from a minimum of 85.7% (r =0.79, p<0.001) to a maximum of 95.6% (r =0.82, p<0.001) when single 5-year age intervals were considered. Therefore, undetected DXA errors had resulted in a significant underestimation of OS accuracy in our previous study.
Conclusions The actual potential of the OS-based approach for osteoporosis diagnosis on the femoral neck was previously underestimated. The adopted method for retrospective analysis of DXA reports could be also used to re-assess the performance of different US methods that assumed routine DXA investigations as the gold standard reference.
Messina et al, Eur Radiol 2015;25:1504–11.
Muratore et al, Osteoporos Int 2015;26(Suppl 1):P530.
Acknowledgement This work was partially funded by FESR PO Apulia Region 2007–2013 – Action 1.2.4 (grant n. 3Q5AX31: ECHOLIGHT Project).
Disclosure of Interest F. Conversano Shareholder of: Echolight srl, a National Research Council spin-off that may or may not benefit from the results of this study, M. Muratore: None declared, A. Greco: None declared, P. Pisani: None declared, M. Peccarisi: None declared, M. Aventaggiato: None declared, M. Di Paola: None declared, D. Costanza: None declared, A. Grimaldi: None declared, S. Casciaro Shareholder of: Echolight srl, a National Research Council spin-off that may or may not benefit from the results of this study