Accuracy of densitometric vertebral fracture assessment when performed by DXA technicians--a cross-sectional, multiobserver study.

UNLABELLED: Six dual-energy X-ray absorptiometry (DXA) technicians reviewed lateral images of the spine for deformed vertebrae. The images were acquired with a DXA scanner in 235 patients referred for osteoporosis assessment. The outcome was compared to findings on spinal radiographs assessed by two radiologists. Three DXA technicians performed acceptable or better in identifying patients with fractured vertebrae.

INTRODUCTION: This is the first study to evaluate the accuracy of vertebral fracture assessment (VFA) when used by DXA technicians as a triage test to select patients with deformed vertebrae for spinal radiographs.

METHODS: Lateral single-energy scans and radiographs of the thoracolumbar spine (T4-L4) were acquired in 235 patients aged 65 years or more referred for osteoporosis assessment. Six DXA technicians evaluated lateral scans using dedicated software. The DXA technicians were trained to identify deformed vertebrae, but they did not assess the aetiology of deformity. Two radiologists evaluated the radiographs and their consensus evaluation served as the reference test for vertebral fracture. The main outcome was sensitivity and specificity of the DXA technicians' identification of patients with one or more grade II-III deformities according to Genant's classification.

RESULTS: The proportion of patients with one or more grade II-III vertebral fractures was 0.35. Sensitivity ranged from 0.61 to 0.83 and specificity ranged from 0.78 to 0.95 across the DXA technicians. In patients with grade II-III deformities on VFA, the mean probability of one or more grade II-III fractures was 0.74 (range 0.66-0.86). Conversely, in patients without such deformities, the mean probability of grade II-III fractures was 0.14 (range 0.10-0.18). Accuracy was lower for grade I-III deformities for all the DXA technicians.

CONCLUSION: Three of six DXA technicians achieved acceptable or better accuracy when using VFA to triage patients with grade II-III vertebral deformities for spinal radiographs. Heterogeneity between DXA technicians appears to be due to differences in subjective thresholds. VFA triage by DXA technicians to identify patients with grade II-III fractures is feasible.