Purpose: This study was done to evaluate the diagnostic accuracy of dual-energy X-ray absorptiometry (DXA) compared with conventional radiography for identifying vertebral fractures.
Materials and methods: A total of 930 postmenopausal women underwent conventional radiography and DXA imaging of the spine. The images were evaluated by two expert skeletal radiologists using the semiquantitative (SQ) method for conventional radiography and the morphometric vertebral fracture assessment (VFA) for DXA.
Results: The SQ method for radiography (SQ-Rx) analysed 99.1% of vertebrae, identifying 442 vertebral fractures; VFA analysed 97.5% vertebrae, detecting 420 vertebral fractures. Agreement between SQ-Rx and VFA reached 98.76%, and the κ-score was 0.96 [95% confidence interval (CI), 0.95-0.98]. Assessing the grading of vertebral fractures, agreement reached 97.5% and the κ-score was 0.841 (95% CI, 0.821-0.891). Considering SQ-Rx method as "gold standard", VFA had a sensitivity of 97.85 % and a specificity of 99.81%. The negative (NPV) and positive (PPV) predictive value for VFA were 99.83 % and 98.15%, respectively. Fractures were identified in 251 (27 %) and 242 (26 %) of patients on SQ-Rx and VFA, respectively. On a per-patient basis, the agreement between the two methods was 97% and the κ-score was 0.95 (95% CI, 0.920-0.968). The diagnostic parameters for VFA were 97.23% sensitivity, 98.86% specificity, 97.60% PPV and 98.84% NPV.
Conclusions: This study demonstrated that VFA with DXA may reach a high level of accuracy for diagnosing vertebral fractures, suggesting that VFA should be introduced in the screening of individuals with a risk of osteoporosis and in the follow-up of osteoporotic patients receiving treatment.