Inter- and intraobserver variability in the interpretation of computed tomography of the lumbar spine

In order to assess the variability in interpreting lumbar CT-scans, two radiologists and two rheumatologists examined the same set of 40 CT-scans and evaluated the presence of primary abnormalities using pre-established criteria. Inter- and intraobserver concordance was assessed using kappa statistics. Interpretation of herniated nucleus pulposus appeared reliable in this study (interobserver and intraobserver kappa statistics approximately 0.7 and 0.9, respectively). Conversely, significant variability of interpretation was seen in many findings often considered important in benign low-back pain or sciatica. Particularly low levels of agreement (interobserver kappa statistics lower to 0.20) were found for facet joint osteoarthritis and spinal stenosis. Since herniated nucleus pulposus appeared as the only reliable CT finding, lumbar CT ordering should therefore be currently restricted to confirmation and localization of herniated nucleus pulposus, especially when surgery or any other invasive intervention is planned to treat prolonged sciatica.