Multiple Sclerosis Research Repository

1. “Identifying multiple sclerosis subtypes using unsupervised machine learning and MRI data” by Arman Eshaghi et al

-we define MS subtypes as cortex-led, NAWM-led and lesion-led

-people w the lesion led subtype have the highest risk of confirmed disability progression (CDP) and the highest relapse rate

-people w the lesion-led MS subtype show positive treatment response in selected clinical trials

-our findings suggest that MRI-based subtypes predict MS disability progression and response to treatment and may be used to definite groups of patients in interventional trials

-grouping individuals on the basis of as similar appearance on a single time point MRI is not sufficient as patients belonging in the same subgroup would show different abnormalities as their disease evolves and would appear different

-the ability to distentable temporal and phenotypic heterogeneity

-Here, we aimed to redefine subtypes of MS based on a data driven assessment of the 

pathological changes visible on MRI scans, rather than the evolution of clinical symptoms, with a view to targeting therapies to subpopulations who share pathogenic mechanisms

-the lesion-led subtype had the highest EDSS, long disease duration, highest lesion load 

at baseline, highest lesion accrual over time, and the smallest cortical and deep grey 

matter volumes at baseline (all p less than 0.001); additionally, the lesion-led subtype 

sowed the highest SuStaIn stage at baseline and the highest annual increase in SuStaIn 

stage in both the training and validation datasets

-the most frequent subtype in both the training and validation datasets was the 

cortex-led subtype followed by the NAWM subtype in the training dataset and the 

lesion-led subtype in the validation dataset

-there were no differences in age and sex between the MS subtypes

-lesion-led had the most active disease and highest relapse rate

-no significant associations between the standard clinical phenotypes or baseline EDSS 

with the time to 24-week CDP suggesting that MRI-based subtypes were more strongly 

associated with the risk of disability progression than the standard clinical phenotypes

-no differences in the rate of EDSS worsening were observed between treated patients

and those on placebo/active comparator, who belonged to both the NAWM led and 

cortex-led subtype

-we found that a patient’s baseline subtype and stage was associated with the individual 

risk of disease progression

-our primary hypothesis was that a model based on MRI rather than solely on clinical 

data helps to improve a biological understanding of MS disease progression

-there were differences in the risk of disability progression, disease activity and treatment 

response across subtypes which suggested that they reflected different pathobiological mechanisms relevant to the manifestations of the disease 

-notably, the lesion-led subtype was the only subtype that showed a significant treatment response in both RRMS and progressive (PP and SP) MS trials

-when looking at the trajectory of MRI changes definition the cortex-led subtype, this group showed early cortical atrophy in the occipital and parietal regions, subsequent development of atrophy in the other grey matter regions and accumulation of T2 lesions and late NAWM abnormalities.  This suggests that the pathological underpinning of the cortex-led subtype is more insidious and may relate to neurodegeneration in the cortex and compartmentalised, chronic inflammation in the white matter, wish is not reflected by the visible lesions.  The concurrent development of cortical atrophy and accumulation fo lesions points to retrograde neurodegeneration of tracts transected in white matter lesions. 

-the majority of patients belonged to the cortex-led subtype in both the training and validation datasets; this subtype responded to a lesser extent to treatments

-this observations suggests that the neurodegenerative component of MS is clinically relevant and remains difficult to target with treatments

-it is also important to note that SuStaIn disentangles not only the sequence but also the severity of abnormality at each stage

-the parietal cortex showed the greatest early atrophy of all cortical regions, regardless of subtype

-we postulate that this reflects a shorter prodromal period in the lesion-led compared to other subtypes

-another possibility is that longer disease durations in the lesion-led subtypes reflect patients converting from another subtype

-the MRI-based subtypes and stages were more strongly associated with EDSS worsening than the baseline EDSS or clinical phenotypes

-the MRI-based subtypes can be used alongside clinical phenotypes to add value in prognosticating patient outcomes

-we found that the MRI-based subtypes predicted CDP whilst single MRI variables (lesion load and whole brain volume) did not, suggesting that a comprehensive model is necessary to achieve the difficult task to predict disability progression

-we were not able to use the most sensitive MRI measures; this was particularly the case for NAWM

-the spinal cord is affected from early stages and its atrophy is independently associated with disability