New AI model challenges how multiple sclerosis is classified
Viewing MS as single disease, not by type, might improve management
A new model using artificial intelligence (AI) suggests that multiple sclerosis (MS) progression is better viewed along a single disease spectrum, rather than as distinct disease types — such as those now used in MS diagnosis and treatment — according to a study led by scientists in Europe.
The global research team says moving to this view could ultimately lead to better disease management for people with MS.
“Our data clearly show that MS cannot be characterized by subtypes such as relapsing or progressive MS, but is instead a continuous disease process with definable state transitions,” Heinz Wiendl, MD, a professor at the University of Freiburg and one of the study’s senior authors, said in a university news story outlining the study’s findings.
“Instead of categorizing [MS] patients, we should quantify their state and track it dynamically,” said Wiendl, a neurologist.
The study, “AI-driven reclassification of multiple sclerosis progression,” was published in the journal Nature Medicine.
In MS, the immune system mistakenly attacks healthy parts of the brain and spinal cord. The neurodegenerative disease is categorized into three main types based on its clinical presentation, meaning the symptoms seen among patients.
Relapsing-remitting MS (RRMS) is characterized by relapses, or periods in which symptoms worsen, interspersed with periods of remission, where symptoms ease or go away entirely. It is thought to be largely driven by inflammatory activity.
Progressive MS is marked by symptoms that steadily worsen over time, with neurodegeneration that happens even without relapses or signs of active inflammation. In primary progressive MS (PPMS), this progression starts right from disease onset, whereas secondary progressive MS (SPMS) follows RRMS.
Disease type now determines treatment strategies for patients
These classifications are used to determine treatment strategies and predict prognosis. However, evidence suggests that the distinctions may not be so clear-cut — and that they may have limited value for predicting the disease course and treatment responses.
“This raises the important question of whether the current MS categorization into distinct subtypes is justified or whether MS would be better described as a disease continuum from a focal inflammatory to a progressive disease course,” the researchers wrote.
With this in mind, the scientists used an AI-based approach to develop a new MS classification system that would be independent of the diagnosed disease type.
They built the AI model using data from more than 8,000 people with RRMS, SPMS, or PPMS who were included in a large Novartis-sponsored clinical trial database. The data came from nine Phase 2/3 clinical trials, and covered up to 15 years of follow-up, with approximately 120,000 neurological assessments and more than 35,000 MRI scans.
[Emerging evidence] raises the important question of whether the current MS categorization into distinct subtypes is justified or whether MS would be better described as a disease continuum from a focal inflammatory to a progressive disease course.
The researchers identified four aspects of MS that best capture its progression: physical disability, brain damage, clinical relapses, and silent inflammatory activity seen on MRI scans.
These variables were used to build a model that gives a probabilistic assessment of how likely people with MS are to transition between eight distinct disease states over time.
This included states 1-3, which are early phases of clinical stability with minimal disability, and stage 4, or asymptomatic MRI disease activity. State 5 corresponds to a clinical relapse, and states 6-8 are phases of advanced MS in which there’s less inflammation but greater neurodegeneration and disability. States 4 and 5 were both considered intermediate inflammatory stages.
According to this framework, patients don’t have to move linearly through these states. Instead, they can transition in any order or direction.
Still, the data indicated that it’s nearly impossible for a person to progress directly to advanced stages without first experiencing some type of intermediate inflammatory activity. Further, once patients are in advanced stages, they won’t return to an early clinically stable phase, the data indicated.
As most people diagnosed with either SPMS or PPMS were considered to be in advanced stages, the data don’t support distinguishing between the two entities, according to the researchers.
AI-based model now must be tested in clinical practice
The AI approach was then validated using data from a separate clinical trial and real-world databases.
Overall, the model was accurate in predicting a person’s risk of transitioning to advanced MS, according to the researchers. Data also showed that treatment with an MS disease-modifying therapy increases the odds that a person will stay in early, mild disease stages for longer.
The findings support a new way of describing MS, the team noted.
“Our … model results are more compatible with the view of MS as a disease continuum than with the traditional view of distinct [types],” the researchers wrote.
Ultimately, scientists hope this new framework will help patients access treatment more easily. The currently used classification system limits which patients can access DMTs, as most are indicated for relapsing disease types.
If the new system were to be adopted, it would allow for treatment to be initiated based on individual risk assessment, regardless of the diagnosed disease type.
“Patients with active but clinically silent inflammatory activity, in particular, require early treatment decisions, as the model strikingly demonstrates,” the news story noted.
Now, the model needs to be tested in clinical practice to see how useful it is for making MS treatment decisions in real-world settings, according to the team.
AI model could have implications beyond MS
Lutz Hein, MD, who is not one of the study’s authors but serves as dean of the faculty of medicine at the University of Freiburg, noted that the idea behind this AI model — to move past rigid disease categorizations — also has implications beyond MS.
“The principle is fundamental and pioneering—and it can also be applied to many other diseases, both within neurology and beyond,” Hein said.
About the Author
