Observer variation in MRI evaluation of patients suspected of lumbar disk herniation

Our objective was to assess observer variation in MRI evaluation in patients suspected of lumbar disk herniation.

Two experienced neuroradiologists independently evaluated 59 consecutive patients with lumbosacral radicular pain. Per patient, three levels (L3-L4 through L5-S1) and the accompanying roots were evaluated on both sides. For each segment, the presence of a bulging disk or a herniation and compression of the root was reported. Images were interpreted twice: once before and once after disclosure of clinical information. Interobserver agreement was expressed as unweighted kappa values.

Without clinical information, interobserver agreement for the presence of herniation or bulging disk was moderate (full agreement, 84%; kappa = 0.63; 95% confidence interval [CI], 0.53-0.72). Of a total of 352 segments evaluated, there was disagreement on 58 segments (17%): bulging disk versus no defect in 26 (7.4%), bulging disk versus herniation in five (1.4%), and hernia versus no defect in 27 (7.7%). With clinical information, twice as many bulging disks were reported but no new herniations were detected. Agreement slightly decreased, but not significantly (full agreement, 77%; kappa = 0.59; 95% CI, 0.49-0.69; p = 0.12).

On average, more than 50% of interobserver variation in MRI evaluation of patients with lumbosacral radicular pain is caused by disagreement on bulging disks. Knowledge of clinical information does not influence the detection of herniations but lowers the threshold for reporting bulging disks.