Abstract
Spinal cord involvement is a hallmark feature of multiple sclerosis, neuromyelitis optica with AQP4 antibodies and MOG-antibody disease. In this cross-sectional study we use quantitative spinal cord MRI to better understand these conditions, differentiate them and associate with relevant clinical outcomes. Eighty participants (20 in each disease group and 20 matched healthy volunteers) underwent spinal cord MRI (cervical cord: 3D T1, 3D T2, diffusion tensor imaging and magnetization transfer ratio; thoracic cord: 3D T2), together with disability, pain and fatigue scoring. All participants had documented spinal cord involvement and were at least 6 months post an acute event. MRI scans were analysed using publicly available software. Those with AQP4-antibody disease showed a significant reduction in cervical cord cross-sectional area (P = 0.038), thoracic cord cross-sectional area (P = 0.043), cervical cord grey matter (P = 0.011), magnetization transfer ratio (P ≤ 0.001), fractional anisotropy (P = 0.004) and increased mean diffusivity (P = 0.008). Those with multiple sclerosis showed significantly increased mean diffusivity (P = 0.001) and reduced fractional anisotropy (P = 0.013), grey matter volume (P = 0.002) and magnetization transfer ratio (P = 0.011). In AQP4-antibody disease the damage was localized to areas of the cord involved in the acute attack. In multiple sclerosis this relationship with lesions was absent. MOG-antibody disease did not show significant differences to healthy volunteers in any modality. However, when considering only areas involved at the time of the acute attack, a reduction in grey matter volume was found (P = 0.023). This suggests a predominant central grey matter component to MOG-antibody myelitis, which we hypothesize could be partially responsible for the significant residual sphincter dysfunction. Those with relapsing MOG-antibody disease showed a reduction in cord cross-sectional area compared to those with monophasic disease, even when relapses occurred elsewhere (P = 0.012). This suggests that relapsing MOG-antibody disease is a more severe phenotype. We then applied a principle component analysis, followed by an orthogonal partial least squares analysis. MOG-antibody disease was discriminated from both AQP4-antibody disease and multiple sclerosis with moderate predictive values. Finally, we assessed the clinical relevance of these metrics using a multiple regression model. Cervical cord cross-sectional area associated with disability scores (B = −0.07, P = 0.0440, R2 = 0.20) and cervical cord spinothalamic tract fractional anisotropy associated with pain scores (B = −19.57, P = 0.016, R2 = 0.55). No spinal cord metric captured fatigue. This work contributes to our understanding of myelitis in these conditions and highlights the clinical relevance of quantitative spinal cord MRI.