Machine Learning–Based Tools Predict MS Progression

A new study published in the Journal of Neurology¹ detailed the development of 2 machine learning–based tools that were able to predict multiple sclerosis (MS) progression for different individuals. These developments can help to provide risk stratification and prevention efforts for those with increased risk of MS as measured by the tests.

MS is an autoimmune condition that can cause disability over time, including numbness and weakness throughout the body, and can affect patients in cycles of remission and relapse.² Being able to predict the progression of the disease can help clinicians target the right care for each patient. In the past, these researchers had created the DAAE score, made up of disease duration, age, age at disease onset, and Expanded Disability Status Scale (EDSS), to assess risk of disease progression in the next 5 years. The researchers aimed to build on this score by developing 2 complementary tools using machine learning to assess 5-year risk in individuals: the DAAE–disease modifying therapy (DAAE-M) score and the Empirical Landmark-based Individualized Estimation (ELIE) of MS progression risk. The researchers aimed to validate both of these risk predictors against clinically and objectively defined progression.

The researchers used participant data collected from 1964 to 2023 that was collected in the MSBase registry. Participants were included if they had no other neurological disorder besides relapsing-remitting MS at baseline, had 2 or more clinical evaluations, were aged 18 years or older, had more than 3 years of clinical follow-up, had data not previously used in the development of the DAAE score, and had baseline neurologic testing of disability.

The DAAE score is measured on a scale of 0 to 12 points, and patients with MS were separated into groups based on their risk score. Clinical disease progression was defined as when a clinician judged that a patient had transitioned to secondary progressive MS. The MSBase-Lorscheider criteria were used to determine objective disease progression, which included an EDSS of 4.0 or higher, an EDSS increase of 1.0 or higher in patients with an EDSS of 5.5 or lower, or an EDSS increase of 0.5 or higher in patients with an EDSS of 6 or higher. Patients with objective disease progression also needed no relapses and had disability worsening that was confirmed after 3 months in EDSS.

There were 34,510 patients with MS who met the inclusion criteria and were included in the study, with 72.6% being women. The mean (SD) age of the participants was 37.1 (10.8) years, and disease duration was 5.8 (7.3) years. Clinical progression criteria were met by 9.8% of the participants, whereas 21% met objective progression criteria, both over 5 years.

Patient-level risk estimates were provided through the DAAE-M using monotonic risk escalation across risk groups. In patients who had unspecified therapy, the clinical disease progression was 3.1% in those with a very low DAAE-M score, 11.2% in those with a low score, 22.6% in those with a medium score, and 33.4% in those with a high score. Patients receiving high-efficacy disease-modifying therapy (DMT) had a lower risk when compared with patients receiving low-efficacy DMT across the different risk scores. Risk was also significantly higher in the very-low efficacy treatment group compared with the low-efficacy treatment group.

Objective disease progression was also calculated and increased stepwise across the DAAE-M risk groups, with patients with very low DAAE-M scores having 8.6% risk of progression, the low group having 14.5% risk, the medium group having 23.3% risk, and the high group having 38.8% risk. Risk of progression was significantly reduced in patients in the high-efficacy treatment group compared with those in the low-efficacy treatment group. Patients who went without treatment had a significantly higher risk of progression compared with the low-efficacy treatment group.

The ELIE model found similar results, as risk of clinical disease progression and objective disease progression increased by stratified risk deciles.

There were some limitations to this study. All participants were adults and the pediatric population with MS was not accounted for. MRI data was not included in either machine-learning tool. Only 7.2% of the patients were using high-efficacy therapy during this study. Subtle forms of disability could not be considered for either the DAAE-M or the ELIE.

“Both systems performed well in primary calibration analyses with predicted risk closely matching actual incidence, albeit with distinct trade-offs per system depending on what was optimized for in development,” the authors concluded. “These results support the use of such risk stratification systems for identifying target [patients with] MS for relevant secondary prevention and restorative treatment strategies.”

References

1. Fuchs TAN, Schoonheim MM, Strijbis EMM, et al. Predicting disease progression in multiple sclerosis with clinically accessible information and technology. J Neurol. 2026;273(5):281. doi:10.1007/s00415-026-13802-4
2. Multiple sclerosis (MS). National Institute of Neurological Disorders and Stroke. Updated December 17, 2025. Accessed April 20, 2026. https://www.ninds.nih.gov/health-information/disorders/multiple-sclerosis-ms