Reliability of MRI findings in candidates for lumbar disc prosthesis

Limited reliability data exist for localised magnetic resonance imaging (MRI) findings relevant to planning of treatment with lumbar disc prosthesis and later outcomes. We assessed the reliability of such findings in chronic low back pain patients who were accepted candidates for disc prosthesis. On pretreatment MRI of 170 patients (mean age 41 years; 88 women), three experienced radiologists independently rated Modic changes, disc findings and facet arthropathy at L3/L4, L4/L5 and L5/S1. Two radiologists rerated 126 examinations. For each MRI finding at each disc level, agreement was analysed using the kappa statistic and differences in prevalence across observers using a fixed effects model. All findings at L3/L4 and facet arthropathy at L5/S1 had a mean prevalence < 10% across observers and were not further analysed, ensuring interpretable kappa values. Overall interobserver agreement was generally moderate or good (kappa 0.40-0.77) at L4-S1 for Modic changes, nucleus pulposus signal, disc height (subjective and measured), posterior high-intensity zone (HIZ) and disc contour, and fair (kappa 0.24) at L4/L5 for facet arthropathy. Posterior HIZ at L5/S1 and severely reduced subjective disc height at L4/L5 differed up to threefold in prevalence between observers (p < 0.0001). Intraobserver agreement was mostly good or very good (kappa 0.60-1.00). In candidates for disc prosthesis, mostly moderate interobserver agreement is expected for localised MRI findings.