Comparison of prognostic value of different MRI classifications of signal intensity change in cervical spondylotic myelopathy

BACKGROUND CONTEXT: Signal intensity (SI) changes of the spinal cord on magnetic resonance imaging (MRI) in cervical spondylotic myelopathy (CSM) are thought to be a predictor of surgical outcome. However, the clinical significance of SI change remains controversial. Although several classifications exist for SI change, there are no previous studies comparing their prognostic significance. PURPOSE: To determine the MRI classification of SI changes in patients with CSM that is useful for prognostication of surgical outcome. STUDY DESIGN: Retrospective case study. PATIENT SAMPLE: Patients who underwent cervical laminectomy for CSM between the time period of January 2000 and December 2005. OUTCOME MEASURE: Clinical outcome was measured by the recovery rate (RR) and the postoperative Nurick grade. METHODS: We retrospectively studied 35 of the 77 CSM patients (mean age, 57.8 years; range, 30-69; preoperative symptom duration, 9.3 months) who underwent cervical laminectomy and who met the inclusion criteria. Postoperative MRIs were performed at a mean of 51.3 months postsurgery to assess for resolution of preoperative signal changes. The pattern of spinal cord SI was classified in three different ways: based on high SI on T2-weighted images (T2WI) (Grade 0 absent, Grade 1 obscure, and Grade 2 intense); based on the extent of SI change on T2WI into focal (confined to one disc level) and multisegmental (more than one disc level); and based on TI-weighted image (T1WI) and T2WI changes into Group A (MRI normal/normal), no intramedullary SI abnormality on T1WI or T2WI; Group B (MRI normal/high SI), no intramedullary SI abnormality on T1WI and high intramedullary SI on T2WI; Group C (MRI low/high SI changes), low-intensity intramedullary signal abnormality on T1WI and high-intensity intramedullary signal abnormality on T2WI. Preoperative clinical findings and MRI abnormalities were correlated with outcomes (Nurick scores, RR) after surgical intervention. RESULTS: Preoperative MRI studies demonstrated the following: Grade 0=1, Grade 1=13, Grade 2=13; focal = 18, multisegmental = 16; Group A = 1; Group B=29; and Group C=5. Resolution of signal changes in T2WI was seen in most patients; however, four patients developed low SI in T1WI in the postoperative MRI. There was no significant difference in the RRs of patients with different grades in the T2WI or with focal or multisegmental SI changes (p=.47 and .28, respectively). In contrast, patients with low SI changes in T1WI were associated with a poor surgical outcome (p<.001). The linear regression model performed using low-intensity signal changes as a dependent variable and the RR as an independent variable confirmed the significance (p<.001) of low SI changes on T1WI as a predictor for surgical outcome. CONCLUSIONS: A classification system of MRI signal changes that accommodates both T1WI and T2WI is more predictive of surgical outcome than those that include T2W SI changes alone. Postoperative MRI is useful to identify late onset of low T1W intensity changes in patients with poor neurological recovery. (C) 2010 Elsevier Inc. All rights reserved.