Prognostic value of changes in spinal cord signal intensity on magnetic resonance imaging in patients with cervical compressive myelopathy

Background context Signal intensity on preoperative cervical magnetic resonance imaging (MRI) of the spinal cord has been shown to be a potential predictor of outcome of surgery for cervical compressive myelopathy. However, the prognostic value of such signal remains controversial. One reason for the controversy is the lack of proper quantitative methods to assess MRI signal intensity. Purpose To quantify signal intensity and to correlate intramedullary signal changes on MRI T1- and T2-weighted images (WIs) with clinical outcome and prognosis. Study design Retrospective case study. Patient sample Patients (n=148; cervical spondylotic myelopathy, n=102 and ossified posterior longitudinal ligament, n=46) who underwent surgery for cervical compressive myelopathy and had high signal intensity change on sagittal T2-WI MRI before surgery between 2006 and 2010. Outcome measure Neurologic assessment was conducted with the Japanese Orthopedic Association (JOA) scoring system for cervical myelopathy. The rate of neurologic improvement was calculated with the use of preoperative and postoperative JOA scores. Methods Quantitative analysis of MRI signal on both T1- and T2-WIs via use of the signal intensity ratio (SIR; signal intensity of lesion relative to that at C7-T1 disc level) was performed. Correlations between SIR on T1- and T2-WIs and preoperative JOA score, JOA improvement rate, disease duration, and MRI morphologic classification (cystic or diffuse type) were analyzed. Multivariate regression analysis for JOA improvement rate was also analyzed. In a substudy, 25 patients underwent follow-up MRI starting from 6 months after surgery to analyze the relationship between changes in SIR on follow-up MRI and clinical outcome. Results SIR on T1-WIs, but not SIR on T2-WIs, correlated with postoperative neurologic improvement. The disease duration correlated negatively with SIR on T1-WIs and JOA improvement rate but not with SIR on T2-WIs. SIR on T2-WIs of "cystic type" was significantly greater than of "diffuse type," but SIR on T1-WI and JOA improvement rate were not different in the two types. Stepwise multivariate regression analysis indicated that SIR on T1-WIs and long disease duration were significant predictors of postoperative neurologic outcome. SIR on follow-up T1-WI and changes in SIR on T1-WI after surgery correlated positively with postoperative improvement rate. SIR on follow-up T2-WI and changes on T2-WI correlated negatively with postoperative neurologic improvement. Conclusions Our results suggest that low intensity signal on preoperative T1-WIs but not T2-WIs correlated with poor postoperative neurologic outcome. Furthermore, decreased signal intensity on postoperative T1-WIs and increased signal intensity on postoperative T2-WIs are predictors of poor neurologic outcome.