MR evaluation of epidural fibrosis: proposed grading system with intra- and inter-observer variability

The purpose of this study is to present a grading system for the Magnetic Resonance (MR) appearance of epidural fibrosis, and to present the inter- and intra-observer variability of the system. The study population was from a randomized, double blind, controlled multicenter clinical trial evaluating the safety and effectiveness of ADCON-L anti-adhesion barrier gel, in preventing epidural fibrosis and dural adhesions following single-level unilateral laminectomy/discectomy for lumbar disc herniations. MRI of the lumbar spine was obtained, without and with gadolinium enhancement, pre-operatively and at six months post-operatively. Patients having extensive epidural scar have been shown to be 3.2 times more likely to experience recurrent radicular pain than those patients with less extensive epidural scarring. New agents have appeared in the marketplace and in clinical trials that may be used intra-operatively to modulate the presence or absence of epidural scar tissue. Given these factors, the need to accurately describe the location and amount of epidural scar tissue has assumed more pressing importance. Two readers, blinded to clinical findings and to the other reader, independently evaluated the MR examinations in 50 post-operative lumbar spine surgery patients using the previously defined categorization of epidural scar for five levels in each patient, four quadrants per level. In addition, 114 examinations were separately evaluated by the same reader for evaluation of intra-observer variability. For the purposes of epidural fibrosis identification, only the axial T1-weighted images with and without contrast were utilized. The amount of epidural fibrosis was graded on a scale of 0-4 for each quadrant at each imaging slice encompassing the operative level: 0 = no/trace scar; 1 = > 0% and < or = 25% of quadrant filled with scar; 2 = > 25% and < or = 50% of quadrant filled with scar; 3 = > 50% and < or = 75% of quadrant filled with scar; 4 = > 75% and < or = 100% of quadrant filled with scar. Each reader evaluated a total of 1000 epidural quadrants for the inter-observer assessment. The estimated kw index is 0.68 with 95% confidence interval (CI) of [0.64, 0.71]; this is substantial agreement. There were a total of 2,280 quadrants evaluated overall for the intra-observer assessment. The estimated kw index is 0.94 with 95% CI of [0.93, 0.95]; this is almost perfect agreement. Substantial intra-observer and near perfect inter-observer agreement was achieved for evaluation of epidural scar using a relatively simple, semi-quantitative approach to the T1-weighted axial MR images. A standard grading system for epidural scar is proposed.