Reliability of readings of magnetic resonance imaging features of lumbar spinal stenosis

Study Design. A reliability assessment of standardized magnetic resonance imaging (MRI) interpretations and measurements. Objective. To determine the intra- and inter-reader reliability of MRI features of lumbar spinal stenosis (SPS), including severity of central, subarticular, and foraminal stenoses, grading of nerve root impingement, and measurements of cross-sectional area of the spinal canal and thecal sac. Summary of Background Data. MRI is commonly used to assess patients with spinal stenosis. Although a number of studies have evaluated the reliability of certain MRI characteristics, comprehensive evaluation of the reliability of MRI readings in spinal stenosis is lacking. Methods. Fifty-eight randomly selected MR images from patients with SPS enrolled in the Spine Patient Outcomes Research Trial were evaluated. Qualitative ratings of imaging features were performed according to defined criteria by 4 independent readers (3 radiologists and 1 orthopedic surgeon). A sample of 20 MRIs was reevaluated by each reader at least 1 month later. Weighted kappa statistics were used to characterize intra- and inter-reader reliability for qualitative rating data. Separate quantitative measurements were performed by 2 other radiologists. Intraclass correlation coefficients and summaries of measurement error were used to characterize reliability for quantitative measurements. Results. Intra-reader reliability was higher than inter-reader reliability for all features. Inter-reader reliability in assessing central stenosis was substantial, with an overall kappa of 0.73 (95% CI 0.69-0.77). Foraminal stenosis and nerve root impingement showed moderate to substantial agreement with overall kappa of 0.58 (95% CI 0.53-0.63) and 0.51 (95% CI 0.42-0.59), respectively. Subarticular zone stenosis yielded the poorest agreement (overall kappa 0.49; 95% CI 0.42-0.55) and showed marked variability in agreement between reader pairs. Quantitative measures showed inter-reader intraclass correlation coefficients ranging from 0.58 to 0.90. The mean absolute difference between readers in measured thecal sac area was 128 mm(2) (13%). Conclusion. The imaging characteristics of spinal stenosis assessed in this study showed moderate to substantial reliability; future studies should assess whether these findings have prognostic significance in SPS patients.