Serum metabolic profile in early multiple sclerosis as predictor of long-term disease progression in BENEFIT
Marianna Cortese1, Xiaojing Peng1, Kjetil Bjornevik1, Clary B. Clish2, Gilles Edan3, Mark Freedman4, Hans-Peter Hartung5, Xavier Montalban6, Rupert Sandbrink7, Ernst-Wilhelm Radue8, Frederik Barkhof9, Eva-Maria Wicklein10, Ludwig Kappos11, Kassandra Munger12, Alberto Ascherio1
1Harvard T.H. Chan School of Public Health, 2Broad Institute of Massachusetts Institute of Technology and Harvard, 3CHU Hôpital Pontchaillou, 4University of Ottawa, 5Heinrich Heine University Medical Faculty Departme, 6Vall Hebron University Hospital-Multiple Sclerosis Centre of Catalonia, 7H. Lundbeck A/S, 8Medical Image Analysis Center, 9Image Analysis Center, 10Bayer AG, 11RC2NB, University Hospital Basel, 12Harvard School of Public Health
Objective:
To examine whether serum metabolite levels measured at clinical onset of multiple sclerosis (MS) are associated with long-term disease activity and progression.
Background:
Single metabolites have been suggested to be dysregulated in MS, however, it is unknown whether serum metabolites in early MS predict long-term outcomes.
Design/Methods:
We conducted a prospective study among 276 participants enrolled in the BENEFIT study at the time of their clinically isolated syndrome and followed them for 11 years. We measured 552 known metabolites in serum samples collected at baseline using liquid chromatography-mass spectrometry. Concentrations were log-transformed and standardized by sex. Using elastic net logistic regression with 5-fold cross-validation, we integrated data on all metabolites to identify a metabolic signature predictive of changes in clinical/radiological measures from month 6 to year 11 (above median changes), indicating worse MS progression. Models were adjusted for age, sex, treatment allocation, steroid use, multifocal symptom onset, number of T2 lesions, and body mass index at baseline, and further for mean 25-hydroxyvitamin D and smoking based on repeated serum cotinine levels within the first 2 years from CIS.
Results:
We identified metabolic signatures indicating a worse outcome by year 11 for total relapses (number of metabolites: 12), T2 lesion volume (24), brain volume (40), EDSS (7), and MSFC (14). The signatures were similar when accounting for vitamin D and smoking. When including the top 10 contributing metabolites according to variable importance in the projection score in multivariable-adjusted logistic regression models, we found them to be moderately predictive of worse disease activity and progression at year 11 (AUCs ranging from 0.55 to 0.7).
Conclusions:
These results suggest that metabolic signatures in early MS may moderately add to the predictive value of known predictors.