Comparison of three different methods for measurement of cervical cord atrophy in multiple sclerosis

BACKGROUND AND PURPOSE: Evidence is mounting that spinal cord atrophy significantly correlates with disability in patients with multiple sclerosis (MS). The purpose of this work was to validate 3 different measures for the measurement of cervical cord atrophy on high-resolution MR imaging in patients with MS and in normal control subjects (NCs). We also wanted to evaluate the relationship between cervical cord atrophy and clinical disability in the presence of other conventional and nonconventional brain MR imaging metrics by using a unique additive variance regression model. MATERIALS AND METHODS: We studied 66 MS patients (age, 41.2 +/- 12.4 years; disease duration, 11.8 +/- 10.7 years; Expanded Disability Status Scale, 3.1 +/- 2. 1) and 19 NCs (age, 30.4 +/- 12.0 years). Disease course was relapsing-remitting (34), secondary-progressive (14), primary-progressive (7), and clinically isolated syndrome (11). The cervical cord absolute volume (CCAV) in cubic millimeters and 2 normalized cervical cord measures were calculated as follows: cervical cord fraction (CCF) = CCAV/ thecal sac absolute volume, and cervical cord to intracranial volume (ICV) fraction (CCAV/ICV). Cervical and brain lesion volume measures, brain parenchyma fraction (BPF), and mean diffusivity were also calculated. RESULTS: CCAV (P <.0001) and CCF (P =.007) showed the largest differences between NCs and MS patients and between different disease subtypes. In regression analysis predicting disability, CCAV was retained first (R-2 = 0.498; P <.0001) followed by BPF (R-2 = 0.08; P =.08). Only 8% of the variance in disability was explained by brain MR imaging measures when coadjusted for the amount of cervical cord atrophy. CONCLUSIONS: 3D CCAV measurement showed the largest differences between NCs and MS patients and between different disease subtypes. Cervical cord atrophy measurement provides valuable additional information related to disability that is not obtainable from brain MR imaging metrics.