Observer variability in the assessment of disc degeneration on magnetic resonance images of the lumbar and thoracic spine.

Abstract

Intraobserver and interobserver reproducibility study. This study investigates the variability in the interpretation of degenerative disc findings using magnetic resonance imaging. Magnetic resonance imaging has been used for years in clinical diagnostics, primarily to investigate disc herniation and spinal stenosis. Less attention has been paid to other disc findings and their assessment reliability. Three independent readers evaluated magnetic resonance images of the lumbar and the lower and middle thoracic spines of 122 subjects by grading 12 aspects of the intervertebral discs and adjacent endplates using written definitions and example images. Images of 20 subjects were reevaluated for the assessment of intraobserver agreement. Agreement was highest in the lower lumbar and poorest in the middle thoracic spine. Intraobserver agreement was generally fair to excellent for almost all variables in the lumbar and lower thoracic spine (most intraclass correlation and kappa coefficients for these regions were above 0.70). Interobserver agreement was notably lower than intraobserver agreement, except for osteophytes and endplate defects in some regions. Intraobserver agreement in the evaluation of disc degeneration was at an acceptable level, in general, in the lumbar and lower thoracic spine. However, assessments were substantially more variable between readers, which limits comparisons of evaluations between different readers.