Assessing the Relative Importance of Imaging and Serum Biomarkers in Capturing Disability, Cognitive Impairment, and Clinical Progression in Multiple Sclerosis

Using machine‐learning analyses in two independent multiple sclerosis cohorts, spinal cord atrophy and cortical degeneration emerged as key predictors of disability and progression independent of relapses. Deep gray matter damage further improved prediction, while serum biomarkers of brain damage provided complementary information, highlighting the value of a multimodal approach to stratify disease severity and progression risk. ABSTRACT The heterogeneity of multiple sclerosis (MS) pathology calls for robust biomarkers to predict disability and progression, particularly progression independent of relapse activity (PIRA). Here, we aimed to identify the most informative MRI and serum biomarkers for predicting clinical outcomes in people with MS (pwMS), including disability severity, cognitive impairment, disease phenotype, and risk of PIRA. We applied a machine learning–based feature selection approach to cross‐sectional and longitudinal data from two independent pwMS cohorts. Cohort 1 (n = 120) included 57 MRI biomarkers, incorporating advanced quantitative MRI (qMRI). Cohort 2 (n = 279) included 35 MRI biomarkers derived from conventional MRI. Both cohorts obtained serum neurofilament light chain (sNfL) and glial fibrillary acidic protein (sGFAP) measurements. Spinal cord atrophy consistently emerged as the strongest predictor of disability severity and predicted PIRA, along with cortical thinning and subcortical atrophy – particularly in deep gray matter. sNfL, sGFAP, and qMRI metrics independently contributed to the prediction of PIRA and progressive disease phenotype. In conclusion, our findings show that spinal cord atrophy and cortical degeneration are the most robust and consistent predictors of MS severity and progression. Serum biomarkers of neuroaxonal and astrocytic damage, together with qMRI‐derived tissue metrics, provide independent and complementary value for outcome prediction. Using machine-learning analyses in two independent multiple sclerosis cohorts, spinal cord atrophy and cortical degeneration emerged as key predictors of disability and progression independent of relapses. Deep gray matter damage further improved prediction, while serum biomarkers of brain damage provided complementary information, highlighting the value of a multimodal approach to stratify disease severity and progression risk. ABSTRACT The heterogeneity of multiple sclerosis (MS) pathology calls for robust biomarkers to predict disability and progression, particularly progression independent of relapse activity (PIRA). Here, we aimed to identify the most informative MRI and serum biomarkers for predicting clinical outcomes in people with MS (pwMS), including disability severity, cognitive impairment, disease phenotype, and risk of PIRA. We applied a machine learning–based feature selection approach to cross-sectional and longitudinal data from two independent pwMS cohorts. Cohort 1 (n = 120) included 57 MRI biomarkers, incorporating advanced quantitative MRI (qMRI). Cohort 2 (n = 279) included 35 MRI biomarkers derived from conventional MRI. Both cohorts obtained serum neurofilament light chain (sNfL) and glial fibrillary acidic protein (sGFAP) measurements. Spinal cord atrophy consistently emerged as the strongest predictor of disability severity and predicted PIRA, along with cortical thinning and subcortical atrophy – particularly in deep gray matter. sNfL, sGFAP, and qMRI metrics independently contributed to the prediction of PIRA and progressive disease phenotype. In conclusion, our findings show that spinal cord atrophy and cortical degeneration are the most robust and consistent predictors of MS severity and progression. Serum biomarkers of neuroaxonal and astrocytic damage, together with qMRI-derived tissue metrics, provide independent and complementary value for outcome prediction. Advanced Science, EarlyView.