Predicting disease progression in multiple sclerosis with clinically accessible information and technology.

Area of Science:

• Neurology
• Artificial Intelligence
• Personalized Medicine

Background:

• Individualized prediction of disease progression is crucial for personalized medicine.
• Machine-learning tools were previously developed to estimate 5-year multiple sclerosis (MS) progression risk.
• Models must incorporate disease-modifying therapy (DMT) and objective outcome definitions.

Purpose of the Study:

• Develop and evaluate two complementary machine-learning tools for individualized 5-year risk estimation in people with relapsing-remitting multiple sclerosis (RRMS).
• Address bias and complex treatment histories in MS progression prediction.
• Define progression using both clinical and objective outcomes.

Main Methods:

• Retrospective multicenter case-control study of adults with RRMS at baseline.
• Developed two machine-learning tools: DAAE-M (transparency, bias mitigation) and ELIE (dynamic modeling, complex treatments).
• Defined disease progression using clinical (RRMS-to-progressive MS) and objective (late-stage confirmed progression independent of relapse activity) criteria.

Main Results:

• Among 34,510 people with RRMS, 9.8% and 21% experienced clinical and objective progression, respectively, over five years.
• Both DAAE-M and ELIE demonstrated good calibration (Brier scores 0.06-0.16).
• High-efficacy DMT was associated with approximately half the progression risk compared to low-efficacy DMT (risk-ratios: 0.42-0.59).

Conclusions:

• Two well-calibrated machine-learning tools were developed for predicting 5-year MS progression.
• These tools offer individualized risk stratification for clinical and objective outcomes.
• Findings support the use of these tools in personalized risk assessment and secondary prevention strategies for MS.