Machine Learning Model Enhances Real-World Studies for Predicting Disability Progression in Multiple Sclerosis


A machine learning model applied to real-world data in a multiple sclerosis study increased patient inclusion for future real-world studies on assessing patient outcomes and disability progression.

Carl Marci, MD, chief clinical officer and chief psychiatrist and managing director of Mental Health and Neuroscience at OM1

Carl Marci, MD

Credit: LinkedIn

Presented at the 2024 Americas Committee for Treatment and Research in Multiple Sclerosis (ACTRIMS) Forum, held February 29 to March 2, in West Palm Beach, Florida, a machine learning model applied to predict disability in a real-world data (RWD) source increased the number of patients available for real-world studies to assess disease progression and outcomes. These findings suggest that the model improves the utility of RWD sources for multiple sclerosis (MS) research and informs the understanding of disability during the transition to different disease stages.1

Among a cohort of 4366 patients, 3568 had relapsing remitting MS (RRMS), 556 had secondary progressive MS (SPMS) and 242 transitioned from RRMS to SPMS. Although participants in the RRMS subgroup were younger compared with those in the other groups, sex and race were similar across all of the subgroups. In the study, 3404 clinician-administered Expanded Disability Status Score (EDSS) scores were documented between 2013 and 2021.

Top Clinical Takeaways

  • The AI model applied to real-world data increased the availability of patient data, facilitating a more comprehensive understanding of disability progression in multiple sclerosis.
  • Transitioning from RRMS to SPMS showed a consistent disability progression, highlighting the potential for AI to improve insights into patient journeys with fidelity.
  • The study emphasizes the future potential of AI and big data for decision support tools in treating MS, aiming to provide the right treatment at the right dose and time.

“The main finding was that we were able to demonstrate the utility of an AI/machine learning model that we presented last year at ACTRIMS in a real-world study. We were able to amplify 3400 EDSS scores to 46000 estimated EDSS (eEDSS) scores,” Carl Marci, MD, chief clinical officer and chief psychiatrist and managing director of Mental Health and Neuroscience at OM1, told NeurologyLive®. “The surprise finding is that when you look at over 200 patients in our premium dataset who have documented transition from RRMS to SPMS, we see a surprisingly smooth but steady progression in their disability as measured by the combined EDSS/eEDSS scores.”

READ MORE: Safety Profile of Fenebrutinib in MS Similar to Other Autoimmune Conditions

Prior to this study, a machine learning model was created to predict EDSS scores at discrete time points using routinely-recorded unstructured clinical notes from neurologists. In a previous study, after applying RWD in patients with MS, it demonstrated a high performance.2

In the current study, Marci and colleagues investigated the feasibility of a new machine model to reduce the number of missing scores for EDSS to support the characterization and understanding of disability progression in patients with RRMS who transition to SPMS. This model was applied to the OM1 MS PremiOM Dataset, which is a RWD source that contains deidentified, deterministically associated clinical and administrative data between 2013 and 2021 on over 17,000 patients with MS managed by neurologists in the United States. Marci noted that in this analytic cohort, he included patients with RRMS and SPMS who had a clinician-administered or predicted EDSS score.

Applying the model resulted in an additional 46,644 predicted EDSS scores available for analysis, which Marci noted that it allowed for a more complete description of disability by age in patients with RRMS and SPMS. The findings demonstrated that patients who transitioned from RRMS to SPMS had an increased disability that led to the date of documented transition that continued over the course of the disease.

“The clear implications are that we can use AI and machine learning on unstructured clinical notes to fill a major gap in MS research - the ability to understand patient journeys with fidelity means we can evaluate key questions about MS disease improvement and progression. This will ultimately lead to AI and big data based decision support tools to help patients with MS get the right treatment at the right dose and the right time,” Marci told. “Next steps are to publish manuscripts using the AI/machine learning eEDSS scores that answer relevant clinical questions for our pharmaceutical partners.”

Click here for more coverage of ACTRIMS 2024.

1. Marci C. Characterizing Disease Progression in Multiple Sclerosis Subtypes Using RWD: Feasibility of Applying a Machine Learning Model to Address Missing Data. Presented at ACTRIMS Forum 2024; February 29 to March 2; West Palm Beach, Florida. P012.
2. Alves P, Green E, Leavy M, et al. Validation of a machine learning approach to estimate expanded disability status scale scores for multiple sclerosis. Mult Scler J Exp Transl Clin. 2022;8(2):20552173221108635. Published 2022 Jun 22. doi:10.1177/20552173221108635
Related Videos
Charbel Moussa, MBBS, PhD
4 KOLs are featured in this series.
4 KOLs are featured in this series.
4 KOLs are featured in this series.
5 KOLs are featured in this series.
5 KOLs are featured in this series.
Debra Miller
Charbel Moussa, MBBS, PhD
© 2024 MJH Life Sciences

All rights reserved.