Assessing the utility of magnetic resonance imaging-based "SuStaIn" disease subtyping for precision medicine in relapsing-remitting and secondary progressive multiple sclerosis.

BACKGROUND: Patient stratification and individualized treatment decisions based on multiple sclerosis (MS) clinical phenotypes are arbitrary. Subtype and Staging Inference (SuStaIn), a published machine learning algorithm, was developed to identify data-driven disease subtypes with distinct temporal progression patterns using brain magnetic resonance imaging; its clinical utility has not been assessed. The objective of this study was to explore the prognostic capability of SuStaIn subtyping and whether it is a useful personalized predictor of treatment effects of natalizumab and dimethyl fumarate. METHODS: Subtypes were available from the trained SuStaIn model for 3 phase 3 clinical trials in relapsing-remitting and secondary progressive MS. Regression models were used to determine whether baseline SuStaIn subtypes could predict on-study clinical and radiological disease activity and progression. Differences in treatment responses relative to placebo between subtypes were determined using interaction terms between treatment and subtype. RESULTS: Natalizumab and dimethyl fumarate reduced inflammatory disease activity in all SuStaIn subtypes (all p < 0.001). SuStaIn MS subtyping alone did not discriminate responder heterogeneity based on new lesion formation and disease progression (p > 0.05 across subtypes). CONCLUSION: SuStaIn subtypes correlated with disease severity and functional impairment at baseline but were not predictive of disability progression and could not discriminate treatment response heterogeneity.

Glial fibrillary acidic protein and multiple sclerosis progression independent of acute inflammation.

BACKGROUND: The clinical relevance of serum glial fibrillary acidic protein (sGFAP) concentration as a biomarker of MS disability progression independent of acute inflammation has yet to be quantified. OBJECTIVE: To test whether baseline values and longitudinal changes in sGFAP concentration are associated with disability progression without detectable relapse of magnetic resonance imaging (MRI) inflammatory activity in participants with secondary-progressive multiple sclerosis (SPMS). METHODS: We retrospectively analyzed longitudinal sGFAP concentration and clinical outcome data from the Phase 3 ASCEND trial of participants with SPMS, with no detectable relapse or MRI signs of inflammatory activity at baseline nor during the study (n = 264). Serum neurofilament (sNfL), sGFAP, T2 lesion volume, Expanded Disability Status Scale (EDSS), Timed 25-Foot Walk (T25FW), 9-Hole Peg Test (9HPT), and composite confirmed disability progression (CDP) were measured. Linear and logistic regressions and generalized estimating equations were used in the prognostic and dynamic analyses. RESULTS: We found a significant cross-sectional association between baseline sGFAP and sNfL concentrations and T2 lesion volume. No or weak correlations between sGFAP concentration and changes in EDSS, T25FW, and 9HPT, or CDP were observed. CONCLUSION: Without inflammatory activity, changes in sGFAP concentration in participants with SPMS were neither associated with current nor predictive of future disability progression.

Single-timepoint low-dimensional characterization and classification of acute versus chronic multiple sclerosis lesions using machine learning.

Multiple sclerosis (MS) is a chronic inflammatory and neurodegenerative disease characterized by the appearance of focal lesions across the central nervous system. The discrimination of acute from chronic MS lesions may yield novel biomarkers of inflammatory disease activity which may support patient management in the clinical setting and provide endpoints in clinical trials. On a single timepoint and in the absence of a prior reference scan, existing methods for acute lesion detection rely on the segmentation of hyperintense foci on post-gadolinium T1-weighted magnetic resonance imaging (MRI), which may underestimate recent acute lesion activity. In this paper, we aim to improve the sensitivity of acute MS lesion detection in the single-timepoint setting, by developing a novel machine learning approach for the automatic detection of acute MS lesions, using single-timepoint conventional non-contrast T1- and T2-weighted brain MRI. The MRI input data are supplemented via the use of a convolutional neural network generating "lesion-free" reconstructions from original "lesion-present" scans using image inpainting. A multi-objective statistical ranking module evaluates the relevance of textural radiomic features from the core and periphery of lesion sites, compared within "lesion-free" versus "lesion-present" image pairs. Then, an ensemble classifier is optimized through a recursive loop seeking consensus both in the feature space (via a greedy feature-pruning approach) and in the classifier space (via model selection repeated after each pruning operation). This leads to the identification of a compact textural signature characterizing lesion phenotype. On the patch-level task of acute versus chronic MS lesion classification, our method achieves a balanced accuracy in the range of 74.3-74.6% on fully external validation cohorts.