Modic changes on MR images as studied with provocative diskography: clinical relevance--a retrospective study of 2457 disks.

Abstract

To assess the value of vertebral body endplate signal intensity (Modic) changes on magnetic resonance (MR) images in predicting a painful disk, with provocative diskography as the reference standard. Lumbar spine diskograms and prediskogram MR images of 736 patients (2457 disks) were retrospectively reviewed as part of an institutional review board-approved HIPAA-compliant protocol. Each disk was assigned a Modic subtype: type 0, normal; type 1, nonfatty high signal intensity; type 2, fatty; and type 3, sclerosis. Statistical analysis involved use of a 2 x 2 contingency table of diskogram results for each of the subtypes. The subtype sensitivity, specificity, positive predictive value (PPV), negative predictive value, and confidence intervals (CIs) were calculated. Similar analyses of other MR variables (such as disk herniation, high signal intensity zone, and spondylolisthesis) and type 1 signal intensity changes also were performed. Type 1 changes (n = 155) had a high PPV (0.81; 95% CI: 0.74, 0.87) for a provocative diskogram. Type 2 changes (n = 126) had a lower PPV (0.64; 95% CI: 0.55, 0.72) for a positive diskogram. Type 3 changes (n = 21) had a PPV (0.57; 95% CI: 0.34, 0.78) that was not significant for a positive diskogram. The PPV of an endplate with a type 1 change (hereafter, type 1 endplate) for a tear in the annulus fibrosis of the disk was also insignificant (0.14; 95% CI: 0.09, 0.20). A similar analysis between a type 1 endplate and the presence of a disk herniation (PPV, 0.26; 95% CI: 0.19, 0.34) and between a type 1 endplate and vertebral body spondylolisthesis (PPV, 0.28; 95% CI: 0.20, 0.35) were significant. Type 1 signal intensity changes on MR images have a high positive predictive value in the identification of a pain generator.