Background: Interferon beta (IFNb) reduces relapse frequency and disability progression in patients with multiple sclerosis (MS).
Objectives: Early identification of prognostic biomarkers of IFNb-treated patients will allow more effective management of MS.
Methods: The IMPROVE study evaluated subcutaneous IFNb versus placebo in 180 patients with relapsing-remitting MS. Magnetic resonance imaging scans, clinical assessments, and blood samples were obtained at baseline and every 4 weeks from every participant. Thirty-nine biomarkers (32 transcripts; seven proteins) were studied in 155 patients from IMPROVE. Therapeutic response was defined by absence of new combined unique lesions, relapses, and sustained increase in Expanded Disability Status Scale over 1 year. A machine learning approach was used to examine the association between biomarker expression and treatment response.
Results: While baseline levels of individual genes were relatively poor predictors, combinations of three genes were able to identify subjects with sub-optimal therapeutic responses. The triplet CASP2/IRF4/IRF6, previously identified in an independent dataset, was tested among other combinations. This triplet showed acceptable predictive accuracy (0.68) and specificity (0.88), but had relatively low sensitivity (0.22) resulting in an area under the curve (AUC) of 0.63. Other combinations of biomarkers resulted in AUC of up to 0.80 (e.g. CASP2/IL10/IL12Rb1).
Conclusions: Baseline expression, or induction ratios, of specific gene combinations correlate with future therapeutic response to IFNb, and have the potential to be prognostically useful.
Keywords: Biomarker; RNA; bioinformatics; interferon beta; multiple sclerosis; prognostic.
© The Author(s), 2014.