Differentiating epidural fibrosis from disc herniation on contrast-enhanced and unenhanced MRI in the postoperative lumbar spine.

Abstract

To determine diagnostic confidence and inter-observer/intra-observer agreement in differentiating epidural fibrosis from disc herniation and lumbar spinal stenosis parameters on magnetic resonance images (MRI) in postoperative lumbar spines with (Gad-MRI) and without (unenhanced MRI) intravenous gadolinium-based contrast agent. N = 124 lumbar spine MRI examinations of four groups were included: 1-6months, 7-18months, 19-36months, more than 37months between lumbar spine surgery and imaging. Two radiologists evaluated Gad-MRI and unenhanced MRI: diagnostic confidence was determined as confident or unconfident. Inter-observer and intra-observer agreement were assessed in differentiating epidural fibrosis from disc herniation and for lumbar spinal stenosis parameters on MRI. Fisher's exact test and Cohen's kappa served for statistics. Diagnostic confidence in differentiating epidural fibrosis from disc herniation was significantly higher on Gad-MR images compared with unenhanced MRI at 1-18months for observer 1 and at 1-6months postoperatively for observer 2 (p values: 0.01-0.025). Inter-observer agreement at 1-6months postoperatively for identification of epidural fibrosis was higher on Gad-MRI (kappa values: 0.53 versus 0.24). Inter-observer and intra-observer agreement for identification of disc herniation and for assessment of lumbar spinal stenosis parameters revealed inconsistent data, without a trend for higher inter-observer or intra-observer agreement on Gad-MRI compared with unenhanced MRI (kappa values: 0.17-0.75). Gad-MR images compared with unenhanced MRI improved diagnostic confidence and agreement in differentiating epidural fibrosis from disc herniation for both observers in the first 6months and for one observer in the first 18months after lumbar spine surgery. After 18months, Gad-MR images compared with unenhanced MRI did neither improve confidence nor agreement.