Disease phenotype prediction in multiple sclerosis
https://www.diva-portal.org/smash/recor ... swid=-1635
Early identification of patients at risk for early conversion to progressive multiple sclerosis (PMS) would be of high value for making rational treatment decisions and patient counselling.
In two MS cohorts with a total of 179 individuals, a panel of 28 cerebrospinal fluid (CSF) metabolites that could differentiate PMS from relapsing-remitting MS (RRMS) on a group-level (AUC= 0.94), was identified using mass spectrometry and machine learning.
To assess prediction confidence on a patient-level, the classifier was complemented with conformal prediction that provided a valid measure of certainty for each patient. The conformal predictions showed that 88% of the patients could be correctly classified with a 90% confidence. One out of eight RRMS patients that developed manifest PMS within three years was predicted as PMS at study onset, whereas six had highly uncertain (<50%) predictions, thus indicating a transitional stage. As a proof-of-concept, this methodology was also applied to a longitudinal collection of CSF from 22 patients in a phase 1b clinical trial of rituximab for PMS.
Seventeen of these patients were classified as PMS at onset, of which twelve patients decreased their prediction certainty after treatment, indicating a treatment effect. Also, on a group-level, a significant (p-value<0.01) treatment effect could be seen after a twelve months follow-up. In conclusion, we demonstrate a methodology that can be used to differentiate PMS from RRMS patients, to monitor transition to PMS and to monitor treatment effects on a patient-level.
This conceptual work may also be useful in clinical practice for personalized treatment strategies and in future trials.
- Similar Topics
- Last post