According to study results published in Brain, a new machine learning model integrating MRI data with serum neurofilament light chain (sNfL) levels offers a biologically-informed framework for subtyping multiple sclerosis (MS) into early and late-sNfL types. Ultimately, this approach might pave the way for more targeted therapeutic strategies and improved patient outcomes

The machine learning model was trained using MRI data and sNfL levels of individuals with relapsing-remitting MS (RRMS) (n=161) and secondary progressive MS (n=28) from a phase 2 clinical trial (NCT02975349) that evaluated the safety and effectiveness of evobrutinib (Merck KGaA, Darmstadt, Germany). Researchers selected MRI features based on their correlation with Expanded Disability Status Scale (EDSS) scores. The model was validated using data from the phase 3 REFLEX (NCT00404352) and REFLEXION (NCT00813709) clinical trials, which included individuals with newly diagnosed MS or clinically isolated syndrome (CIS) (n=445). Participants were grouped by treatment exposure, and longitudinal data were analyzed to compare outcomes across subtypes and assess the added value of incorporating sNfL.

Key results include the following:

  • The early-sNfL group showed elevated sNfL levels, corpus callosum injury, and rapid early lesion accrual, indicating a more inflammatory and neurodegenerative disease course.
  • The late-sNfL group demonstrated earlier cortical and deep grey matter volume loss with later sNfL elevation.
  • Adding sNfL to MRI-based models improved correlation with EDSS scores in both training and test cohorts.
  • The early-sNfL group had a 144% increased risk of new lesion formation (hazards ratio [HR] 2.44; 95% CI, 1.38 to 4.30; P<.005) and greater brain atrophy over time compared with the late-sNfL group.
  • The early-sNfL group demonstrated a more pronounced treatment-related reduction in gadolinium-enhancing lesions and a higher likelihood of new lesion formation.

Although these findings support the potential utility of integrating sNfL into routine imaging-based monitoring for individuals with MS, researchers note that a key limitation of the study involves the use of clinical trial populations that may not reflect the full diversity of the MS population, including those with comorbidities or primary progressive MS, and limited representation of late-stage disease. 

Source: Willard C, Puglisi L, Ravi D, et al. Combined magnetic resonance imaging and serum analysis reveals distinct multiple sclerosis types. Brain 148(12), 4578-4591. doi.org/10.1093/brain/awaf331