Disease phenotype prediction in multiple sclerosis
Herman S, Arvidsson McShane S, Zhukovsky C, Khoonsari PE, Svenningsson A, Burman J, Spjuth O, Kultima K..
Disease phenotype prediction in multiple sclerosis.
iScience. 26, 6 (2023). DOI: 10.1016/j.isci.2023.106906
Progressive multiple sclerosis (PMS) is currently diagnosed retrospectively. Here, we investigated an alternative route to the diagnosis. A selection of cerebrospinal fluid metabolites (n=15) were shown to be able to distinguish PMS from relapsing-remitting MS (AUC=0.93) in a validation cohort. Conformal prediction models showed that, at a 94% confidence level, 88% of patients could be correctly predicted, and four out of eight patients that developed PMS within three years were predicted as PMS. The methodology was further applied to PMS patients in a clinical trial where traditional endpoints were largely inconclusive. The methodology showed that 68% of the patients decreased their similarity to the PMS phenotype one year after treatment. Also on a group-level, a significant (P<0.01) biochemical effect could be seen. This conceptual work demonstrates a methodology that may be useful in monitoring disease progression and as an endpoint in future trials. Finally, 4-acetamidobutanoate showed significant associations to several clinical measures, further strengthening the evidence of its relation to MS.