1. 2008;11;311-326Accuracy of MRI for Diagnosis of Discogenic Pain
   Observational Report
   Marina Kurgansky, PhD, Jay Kaiser, MD, Winnie Lau, BS, and Conor ONeill, MD.

BACKGROUND: Previous studies have compared MRI parameters to the results from discography. However, none have evaluated the overall diagnostic performance of MRI, taking into account that many MRI characteristics may be correlated.

OBJECTIVE: Determine the accuracy of MRI for diagnosis of discogenic pain, taking into consideration the interdependence of MRI parameters.

STUDY DESIGN: An observational report.

SETTING: Sample of 143 patients, 92 male and 51 female in a spinal pain speciality center. Discography classification and scorings for MRI parameters were collected as outcome measures.

METHODS: MRI and discography data were collected from patients with chronic low back pain. Five MRI characteristics were defined: high intensity zone, nuclear signal, disc height, disc contour, and bone marrow intensity change. On discography, each disc was classified as either positive or negative. The accuracy of MRI was evaluated using receiver operating characteristic curves.

RESULTS: MRI parameters are correlated with each other and with discography findings, and these correlations affect the accuracy of MRI. Overall, nuclear signal alone is as accurate as any of the other MRI parameters, or combination of parameters, in the diagnosis of discogenic pain. While there is no difference in overall accuracy between nuclear signal and the other MRI parameters, these parameters do influence test performance when there is a moderate loss of nuclear signal . Moderate loss of nuclear signal and disc bulge has the best combination of sensitivity (79.8%) and specificity (79.3%). Adding moderate loss of disc height improves specificity (82.0%) slightly, and decreases sensitivity (73.6%) slightly, while incorporating high intensity zone grade II further improves specificity (92.6%) and decreases sensitivity (54.7%). High intensity zone grade I and bone marrow intensity change have minimal influence, even when there is moderate loss of nuclear signal.

CONCLUSIONS: MRI parameters are correlated with each other and with discography findings, influencing the diagnostic performance of MRI. Combining MRI parameters improves the diagnostic performance of MRI, but only in the presence of moderate loss of nuclear signal. When there is either normal nuclear signal or severe loss of nuclear signal the other MRI parameters have no influence on test performance. The practical implication for physicians that use discography is that the most important single MRI parameter to consider is nuclear signal. If nuclear signal is normal the disc is very likely to be negative on discography, while if there is severe loss of nuclear signal it is very likely to be positive. Discography will be most useful in discs with moderate loss of nuclear signal, particularly if there are no other MRI abnormalities present.