GFAP and NfL as predictors of disease progression and relapse activity in fingolimod-treated multiple sclerosis

In multiple sclerosis (MS) patients under therapy, the increase of serum glial fibrillary acidic protein (sGFAP) concentrations is associated with the course of 'progression in absence of relapse' (PIRA). While serum neurofilament light chain (sNfL) reflects both response as well as insufficient or lack of efficiency of disease-modifying therapies (DMT), the longitudinal course of sGFAP levels as a drug response marker for future PIRA in relation to specific types of DMT is less clear. We aimed to compare the predictive capacity of sGFAP and sNfL for PIRA and relapse activity and the longitudinal course in people with MS (PwMS) treated with fingolimod, based on Z scores derived from normative values. Overall, 420 PwMS under fingolimod treatment with follow-up of 9.1 years (interquartile range: 7.0-11.0) from the Swiss MS Cohort, contributing 2935 longitudinal serum samples, were included. A reference data set for sGFAP established from 4297 healthy controls across three European and North American cohorts was used to calculate Z scores. The longitudinal course and the predictive capacity of biomarkers for time to PIRA and relapse were assessed by Cox proportional hazards and linear mixed-effects models. In controls, sGFAP concentrations were 13.6% higher in females than males and increased exponentially with age. Altogether, 31.0% of PwMS experienced ≥1 PIRA event. Elevated sGFAP Z scores (>0.75) were associated with increased risk of PIRA [hazard ratio (HR): 1.64; 95% confidence interval (CI): 1.16-2.32; P = 0.006], while this was not the case for sNfL. Conversely, elevated sNfL predicted relapses (HR: 1.58; 95% CI: 1.13-2.23; P = 0.008), while sGFAP did not. Both biomarkers decreased under treatment: sGFAP by 0.19 Z score units (ZSU)/10 years (95% CI: -0.27 to -0.11; P < 0.001) and sNfL by 0.16 ZSU/10 years (95% CI: -0.27 to -0.06; P = 0.002). Serum GFAP remained elevated in PwMS with future PIRA events (estimate: 0.29; 95% CI: 0.07-0.50; P = 0.009); no such association was found for sNfL. Serum GFAP and sNfL Z scores provide complementary predictive capacity for PIRA and relapse risk. The decrease of sGFAP under fingolimod is a feature not observed with other types of DMT and may hint to a specific anti-neurodegenerative effect of Sphingosine-1-phosphate-receptor modulators on astrocytes.