Agreement in the interpretation of magnetic resonance images of the lumbar spine.

[eng] Background: Correlation between clinical features and magnetic resonance imaging (MRI) findings is essential in low-back-pain patients. Most previous studies have analyzed concordance in the interpretation of lumbar MRI among a few radiologists who worked together. This may have overestimated concordance. Purpose: To evaluate intra- and interobserver agreement in the interpretation of lumbar MRI performed in an open 0.2T system. Material and Methods: Seven radiologists from two different geographic settings in Spain interpreted the lumbar MRIs of 50 subjects representative of the general Danish population aged 40 years. The radiologists interpreted the images in routine clinical practice, having no knowledge of the clinical and demographic characteristics of the subjects and blinded to their colleagues' assessments. Six of the radiologists evaluated the same MRIs 14 days later, having no knowledge of the previous results. Data on the existence of disc degeneration, high-intensity zones, disc contour, Schmorl nodes, Modic changes, osteophytes, spondylolisthesis, and spinal stenosis were collected in the Nordic Modic Consensus Group Classification form. Intra- and interobserver agreement was analyzed for variables with a prevalence ≥10% and ≤90% by means of the kappa statistic. Results: Intra- and interobserver agreement was excellent for variables related to Modic changes, and fair to good for disc contour, high-intensity zones, and Schmorl nodes. The evaluations for disc degeneration and osteophytes were found to have fair to good intraobserver agreement and poor interobserver agreement. The agreement for the evaluations of spondylolisthesis and spinal stenosis was not analyzed because they were observed in