A new machine learning tool can now predict whether a person newly diagnosed with multiple sclerosis (MS) will experience worsening of their disability, even if they do not experience a relapse.
These artificial intelligence algorithms can identify the risk of “progression independent of recurrent activity” (PIRA) using simple clinical data, typically collected during routine doctor visits, according to a recent study.
By analyzing factors such as patients’ age of symptom onset, duration of illness, and initial disability score, researchers were able to predict which individuals were likely to experience PIRA within the first three years after diagnosis.
“Our results support the feasibility of the application. [machine learning] Techniques to determine PIRA predictions at an early stage [MS patients]”We used only data commonly available in clinical practice. Our results will allow for subsequent treatment adjustments and will certainly improve long-term prognosis,” the researchers wrote.
the study, “Applying machine learning analysis to predict early progression independent of relapse activity in multiple sclerosis patients” was published. European Journal of Neurology.
Understanding the progression of MS
Most people with MS are characterized by flare-ups or flare-ups, in which symptoms suddenly worsen, followed by periods of remission, in which symptoms ease.
To some extent, worsening disability in MS may be caused by symptoms that persist even after relapses have subsided. However, new research shows that most disability progression in MS occurs by gradual worsening, even when patients do not experience relapses. This is known as progression independent of recurrent activity.
Currently, there is no reliable way to predict the likelihood of PIRA in people newly diagnosed with MS. To address this, a team led by Italian scientists investigated whether machine learning could be used to predict short-term PIRA risk.
Machine learning is a type of artificial intelligence that works by feeding large data sets to a computer, along with mathematical rules (algorithms) that the computer uses to identify patterns in the data. Computers can then apply those patterns to understand other data.
For this analysis, the researchers used clinical and demographic data from 719 patients with MS who underwent regular evaluation during the first three years after the onset of MS. After 3 years, 13% of patients experienced PIRA.
To assess the accuracy of machine learning models, the researchers calculated a statistical measure called the area under the receiver operating characteristic curve (AUC). This measure assesses how well the test differentiates between two groups (i.e., PIRA and no PIRA). AUC scores range from 0.5 to 1, with higher values indicating better accuracy.
The researchers tested multiple machine learning models. The best performing model, the Random Forest model, achieved an AUC of 0.75 and exhibited relatively good accuracy. In this model, the most important factors for accurate prediction were patient age at symptom onset, disability score at 2 years of age, and delay between symptom onset and first assessment.
Notably, MRI data contributed relatively little to prediction accuracy.
Adjust predictions for specific patient groups
The scientists also showed that machine learning models can achieve higher accuracy by narrowing down a subset of patients. For example, when restricting to patients under 45 years of age, the AUC was slightly higher at 0.77. Additionally, for patients who showed no evidence of disease activity (meaning no recurrence, MRI activity, or worsening of disability) during the first 2 years, the model had an AUC of 0.8.
“Our study supports the feasibility of application. [machine learning] Technology to predict PIRA [people with] “MS is used in clinical routine with a good level of accuracy,” the team concluded.
The scientists stressed that further research is needed to validate and refine this approach, but said this type of machine learning analysis could one day be used to predict outcomes and guide treatment decisions for MS patients.
