Interobserver Reliability of Magnetic Resonance Imaging Predictors of Outcome in Cervical Spine Degenerative Conditions

BACKGROUND: Although recent work has focused on characterizing quantitative magnetic resonance imaging (MRI) markers that may predict outcome among patients with cervical degenerative conditions, little is known about their reliability. Measurement and reporting of these markers is time-consuming and nonstandardized, preventing routine use in clinical care. METHODS: We retrospectively analyzed cervical MRI among subjects prospectively enrolled in a health outcomes study of elective surgery for degenerative cervical spine conditions. Two radiologists independently reviewed MRI for presence or absence and length of cord signal change, level of worst cord compression, axial anteroposterior (AP) and lateral spinal cord diameter, midsagittal AP diameter, and kyphosis. Interobserver reliability was compared using kappa and intraclass correlation coefficient (ICC). RESULTS: Inclusion criteria were met by 209 patients who had MRI available for review. Most patients were female (58%) and middle-aged (mean age 51 years), and 54% had a diagnosis of myelopathy. Reliability was fair for cord signal change on T1 (k = 0.33) and good on T2 (k = 0.74) images. Among patients with T2 change (n = 22), reliability for signal change length was good (ICC = 0.67). For level of worst compression, reliability was good (k = 0.79). For AP cord diameter, reliability was very good (ICC = 0.82; T2/midsagittal) and good (ICC = 0.66; T2/axial). Reliability was moderate for lateral cord diameter (ICC = 0.55; T2/axial) and good for kyphosis (k = 0.76). CONCLUSIONS: Good and very good reliability observed in measuring T2-weighted spinal cord signal change, level of worst compression, AP cord diameter, and kyphosis support use of these markers in standardized reporting, which could be incorporated into routine clinical use.