Investigating comparability of quantitative computed tomography with dual energy x-ray absorptiometry in assessing bone mineral density of patients with chronic spinal cord injury

Study design Psychometric study using retrospectively collected data. Objectives We investigated the comparability of quantitative computed tomography (qCT) in assessing bone mineral density (BMD) with dual energy X-ray absorptiometry (DXA). We evaluated how well previously suggested normal values for spinal Hounsfield units (HU) correlated with routine DXA results in patients with chronic spinal cord injury (SCI). Furthermore, we investigated inter/intra-observer reliability of measuring HU in the spine. Setting Academic medical center in Tehran, Iran. Methods Spinal CT scans of 44 male participants with chronic SCI who had undergone DXA studies on the same day were selected. The main outcome measures were sensitivity, specificity, and area under curve (AUC) of HU at each spinal region against DXA results of areal BMD. The secondary outcome was inter/intra-observer reliability of measuring HU in the spinal column. Results We found no significant difference between qCT and DXA results (p-value = 0.237, R = 0.188). However, the two methods showed overall unfavorable comparability, with a sensitivity of 0%, 0%, and 80%, specificity of 50%, 90%, and 85%, and area under curve (AUC) of 0.27, 0.53, and 0.83 for cervical, thoracic, and lumbar spine, respectively. The best comparability was achieved at the lumbar region although not statistically significant (p-value = 0.072). Measuring HU was reliable (inter/intra-observer reliability >98%). Conclusions This study demonstrates that currently proposed normal values result in unfavorable comparability in the cervical and thoracic regions; however, as the agreement improved at the lumbar spine, it is possible that qCT could become an indicator of bone strength with further research.