Article Impact Level: HIGH Data Quality: STRONG Summary of Nature Medicine. https://doi.org/10.1038/s41591-025-03901-6 Dr. Habib Ganjgahi et al.
Points
- An AI-driven analysis of over 8,000 patients reframes multiple sclerosis as a continuous disease process, challenging the long-held dogma of distinct relapsing and progressive subtypes.
- The new model identifies four key dimensions that define a patient’s state, including physical disability, brain damage, clinical relapses, and silent subclinical inflammatory activity.
- Disease progression to advanced stages is driven almost exclusively by the accumulation of brain damage from inflammatory states, whether they are clinically apparent or completely silent.
- This dynamic, state-based classification enables individualized risk assessment and underscores the critical need for early treatment, particularly in patients who exhibit only silent inflammatory activity.
- The findings were validated in external cohorts of more than 4,000 patients and provide a new framework for improving patient management and future drug discovery.
Summary
A study published in Nature Medicine presents a data-driven reclassification of multiple sclerosis (MS) that challenges traditional disease subtypes. Utilizing probabilistic machine learning, the investigators analyzed an extensive clinical trial database (the NO.MS cohort) encompassing approximately 8,000 patients, 118,000 patient visits, and over 35,000 magnetic resonance imaging scans. The objective was to create a model that better reflects the pathobiology of MS, which affects approximately 2.9 million people globally, moving beyond the limited prognostic value of classifications such as “relapsing” or “progressive.”
The resulting model defines MS disease evolution along four key dimensions: physical disability, brain damage, clinical relapses, and subclinical inflammatory activity. It conceptualizes MS as a continuous spectrum between two poles: Early/Mild/Evolving (EME) MS, with minimal impairment and brain damage, and advanced MS, characterized by high disability and radiological burden. The model demonstrates that transitions to advanced MS occur via brain damage accumulation driven by inflammatory states, with or without overt symptoms. Progression to severe disease without preceding inflammatory activity was found to be virtually nonexistent, establishing silent inflammation and relapses as central drivers of deterioration.
The model’s validity was confirmed using an independent clinical trial database and a real-world cohort, which together totaled more than 4,000 patients with MS. The authors advocate for a streamlined disease classification that views MS as a continuum, enabling individualized and dynamic risk assessment. This state-based approach emphasizes the need for early therapeutic intervention in patients with subclinical disease activity to prevent irreversible progression. The proposed framework has significant implications for improving patient management, enhancing the precision of drug discovery, and reforming regulatory logic for future therapies.
Link to the article: https://www.nature.com/articles/s41591-025-03901-6
References Ganjgahi, H., Häring, D. A., Aarden, P., Graham, G., Sun, Y., Gardiner, S., Su, W., Berge, C., Bischof, A., Fisher, E., Gaetano, L., Thoma, S. P., Kieseier, B. C., Nichols, T. E., Thompson, A. J., Montalban, X., Lublin, F. D., Kappos, L., Arnold, D. L., … Holmes, C. C. (2025). AI-driven reclassification of multiple sclerosis progression. Nature Medicine. https://doi.org/10.1038/s41591-025-03901-6
