AI Rewrites the Rules of Multiple Sclerosis: Study Recasts MS as a Disease Continuum

A landmark study published in Nature Medicine has upended decades of conventional wisdom about multiple sclerosis (MS), proposing that the condition should no longer be viewed as three distinct subtypes but rather as a single disease continuum. Using artificial intelligence to analyze data from more than 8,000 patients and over 35,000 MRI scans, researchers have created a data-driven model that redefines how MS evolves and how it should be treated.

For nearly 30 years, clinicians have classified MS into relapsing-remitting (RRMS), secondary progressive (SPMS), and primary progressive (PPMS) forms. These categories have guided both treatment and research, yet they rely on outward clinical features rather than underlying biology. As a result, they have offered limited predictive value for disease course or treatment response.

To address this, an international team led by researchers from the University of Oxford, the University Hospital of Münster, and Novartis used a probabilistic machine learning framework called the factor analysis hidden Markov model (FAHMM). The AI system synthesized multimodal data—clinical measures, MRI findings, and relapse records—to identify patterns of disease progression over time.

From this vast dataset, the model identified four key dimensions that define MS: physical disability, brain damage, relapse activity, and asymptomatic (subclinical) disease activity. These elements interact dynamically, producing a fluid spectrum that extends from early/mild/evolving MS (EME) to advanced MS.

The model revealed that most patients progress from early to advanced stages through intermediate “active” states—periods of inflammatory activity, either symptomatic (relapses) or silent (radiological lesions without symptoms). Once patients reach the advanced stage, marked by significant disability and brain atrophy, they rarely revert to earlier states. This trajectory was confirmed in two independent datasets: a Roche ocrelizumab trial cohort and the real-world MS PATHS registry, encompassing more than 4,000 additional patients.

The findings have sweeping implications for how MS is diagnosed, monitored, and treated. Notably, the study found no biological justification for distinguishing between SPMS and PPMS—a distinction that has long complicated both regulatory approvals and access to therapies. Instead, both forms appear to share the same underlying mechanisms once patients reach the advanced stage.

The analysis also highlighted the clinical importance of subclinical MRI activity, which predicted worsening even in the absence of relapses. Approximately 11% of patients progressed directly to advanced disease via these silent inflammatory episodes. The authors argue that monitoring and preventing such hidden activity should become a central treatment goal, particularly as modern therapies have reduced overt relapses.

Crucially, treatment with disease-modifying therapies (DMTs) significantly reduced the risk of transitioning from early to active disease states and prolonged stability within the EME phase. This reinforces the benefits of early intervention and continuous therapy to preserve brain integrity and delay progression.

By reconceptualizing MS as a biologically continuous disorder, the researchers hope to accelerate therapeutic innovation and streamline clinical trial design.