Machine learning shows promise for predicting multiple sclerosis progression

Machine learning models can reliably inform clinicians about disability progression in multiple sclerosis, according to a study published in an open access journal. PLOS Digital Health by Edward de Brouwer of the Catholic University of Leuven in Belgium and his colleagues.

Multiple sclerosis (MS) is a chronic progressive autoimmune disease that follows a complex pattern of progression, recovery, and relapse, causing severe disability over time. Over the past decade, the global prevalence of MS has increased by more than 30%. However, few tools are available to predict MS progression and help clinicians and patients make life-planning and treatment decisions.

For the new study, de Brouwer and his colleagues used data from 15,240 adults with at least three years of MS who were being treated at 146 MS centers in 40 countries.

Data on each patient's two-year disease progression was used to train a state-of-the-art machine learning model to predict the likelihood of disease progression over the following months and years. The model was trained and validated using strict clinical guidelines, enhancing the applicability of the model in clinical practice.

Although individual models performed differently across patient subgroups, the average area under the ROC curve (ROC-AUC) of the models was 0.71 ± 0.01.This study found that disability progression history was better at predicting future disability progression than treatment history or relapse history.

The authors conclude that the model developed in this study has the potential to significantly enhance planning for people with MS and could be evaluated in clinical impact studies.

De Brouwer added: “Using the clinical histories of over 15,000 patients with multiple sclerosis, we have trained a machine learning model that can reliably predict the likelihood of disability progression over the next two years. This model is broadly applicable because it only uses routinely collected clinical variables.”

“Our rigorous benchmarking and external validation support the great potential of machine learning models to help patients plan their lives and clinicians optimize treatment strategies.”