BACKGROUND: Serum neurofilament light chain (sNfL) and glial fibrillary acidic protein (sGFAP) reflect axonal damage and astrocytic injury. Their clinical role in longitudinal real-world monitoring after immune reconstitution therapy (IRT) in multiple sclerosis (MS) remains insufficiently defined. We evaluated longitudinal sNfL and sGFAP dynamics in people with multiple sclerosis (pwMS) treated with alemtuzumab (ATZ) as a model of IRT to determine their prognostic and monitoring value in real-world care.

METHODS: PwMS initiating ATZ were prospectively followed up every three months for up to five years. sNfL and sGFAP levels were measured using single molecule array (Simoa) and converted to age- and BMI-adjusted Z scores based on healthy control datasets. Longitudinal trajectories were analysed with generalised linear mixed models adjusted for age, sex, and disease duration. Receiver operating characteristic (ROC) analysis with Youden’s index identified optimal cut-offs for disease activity. Logistic and Cox regression models assessed predictive values. Event-related analyses examined biomarker changes around relapses, MRI activity, progression independent of disease activity (PIRA), and retreatment.

RESULTS: Eighty-four pwMS (mean age 36.5 ± 9.0 years, 76% female) were included. Baseline sNfL Z scores were significantly higher in males and in those with recent MRI activity or treatment failure. sNfL rose transiently one month after the first ATZ course, declined by month 3, and remained stably reduced thereafter. Youden’s index-derived baseline sNfL Z scores ≥ 0.75 modestly predicted disease activity during year 1 (odds ratio: 5.10, 95% CI 1.79–14.49), and Z scores > 1.0 predicted relapses after the first ATZ course (hazard ratio: 2.96, 95% confidence interval 1.36–6.43, p  = 0.006). Compared to matched pwMS without disease activity, event-related analyses showed significant sNfL elevations around relapses ( p  = 0.004), MRI activity ( p  = 0.015), and retreatment ( p  = 0.002). A transient increase was observed in the 6 months before PIRA events, followed by normalization ( p  = 0.017). sGFAP levels remained overall stable over follow-up ( p  = 0.677) and showed no predictive value.

DISCUSSION: sNfL provided modest predictive and clear monitoring value in pwMS treated with ATZ and may complement individualized follow-up alongside clinical and MRI assessment. In contrast, sGFAP remained overall stable and did not associate with inflammatory events. However, a modest association with subclinical MRI activity was observed, suggesting further research is needed to fully understand the role of sGFAP in monitoring IRT-treated pwMS. These findings support the real-world clinical utility of high-frequency sNfL monitoring for early detection of breakthrough disease activity after IRT.