RESUMO
The aim of this study was to investigate the diagnostic performance of new dual-energy X-ray absorptiometry (DXA) technologies in the detection of vertebral fractures (VFs). Sixty-eight patients were submitted to DXA and conventional radiography (XR) on the same day. Lateral images of the spine were independently evaluated by three radiologists with different experience in skeletal imaging, in two sessions with 7 days between evaluations of the same anonymous images. The most expert physician repeated the analysis in a subsequent reading session after further 7 days. Results from expert XR evaluation were considered as gold standard. A semiquantitative approach was used to interpret images and morphometric analysis was performed when a VF was suspected. Seventy vertebrae (70/884, 7.9%) were excluded from the lesion-based analysis, as not evaluable: 11/70 (15.7%) missed by XR only, 56/70 (80.0%) missed by DXA only, 3/70 (4.3%) missed by both techniques (upper thoracic spine). Forty "true" fractures were detected (4.9% out of 814 vertebrae) in 26 patients (38.2% of the 68 studied patients). Twenty-five (62.5%) were mild fractures. DXA sensitivity and specificity were 70.0% and 98.3% on a lesion-based analysis, 73.1% and 90.5% on a patient-based analysis. Intra-observer agreement was excellent with no significant difference between the two techniques. Inter-observer agreement among the 3 observers was higher for XR (k=0.824 versus 0.720 in the detection of VFs, p=0.011). DXA accuracy was not influenced by radiologist experience; T4-L4 assessability as well as reproducibility and repeatability of the two techniques and accuracy of DXA were independent from sex, age, body mass index, grade of arthrosis. However DXA sensitivity was affected by mild VFs. Vertebral level did not affect the diagnostic performance with exception of vertebral body assessability. Latest improvements make VFs assessment by DXA competitive with traditional radiographic gold standard, providing consistent advantages and attractions. Few limitations still affect DXA performance and need to be overcome.
