Refining Treatment Monitoring in Patients With Relapsing Multiple Sclerosis Based on Changes on Neurofilament Light Chain Z-Scores

BACKGROUND AND OBJECTIVES: Serum neurofilament light chain (sNfL) reflects inflammatory axonal damage in relapsing multiple sclerosis (RMS). Longitudinal sNfL dynamics may aid treatment response assessment, although data on expected changes under disease-modifying treatments remain limited, particularly in real-world practice. This study evaluated whether on-treatment changes in sNfL were associated with subsequent evidence of disease activity (EDA) and whether they provided complementary information to established treatment response scoring systems (TRSS). METHODS: This prospective, real-world cohort study was conducted at the Multiple Sclerosis Center of Catalonia (Cemcat). Adult RMS patients initiating disease-modifying treatments (DMTs) with serum samples at treatment initiation (baseline) and after 1 year (Y1) were included. The primary exposure was change in sNfL z-scores (ΔzNfL = zNfLY1 - zNfL-baseline) analyzed using data-derived thresholds (ΔzNfL ≥0 and <0.5-point reduction). At Y1, Expanded Disability Status Scale progression, clinical/radiologic activity, and TRSS (Rio [RS], modified Rio [mRS], and MAGNIMS) were assessed. The primary outcome was EDA between years 1 and 2 (EDA-Y2). Multivariable logistic regression models evaluated associations between ΔzNfL and EDA-Y2, as well as the added value of ΔzNfL across TRSS strata. RESULTS: A total of 329 patients (66.3% female; median age 39 years) initiated 352 DMTs; 342 remained on treatment for ≥1 year and 182 had complete EDA-Y2 data. Mean (SD) ΔzNfL at Y1 was -0.62 (1.2), with greater reductions in monoclonal antibody-treated and treatment-naïve patients. Lack of zNfL reduction over the first year (ΔzNfL ≥0) was associated with higher odds of EDA-Y2 in the overall cohort (OR [95% CI], p: 3.07 [1.47-6.39], 0.003), in treatment-naïve cases (14.04 [2.47-79.93], 0.003), and in those achieving No Evidence of Disease Activity (NEDA-3) at Y1 (4.54 [1.26-16.30], 0.02). Similar associations were observed for <0.5-point reductions. Adding ΔzNfL to TRSS showed persistent associations with EDA-Y2 despite otherwise favorable scores [e.g., RS: 2.26 (1.12-4.59), 0.024]. DISCUSSION: Failure to reduce zNfL levels after 1 year of treatment was associated with subsequent disease activity, even in patients otherwise considered stable, providing complementary information to clinical and MRI-based monitoring. Limitations include incomplete follow-up and the modest discriminative performance of ΔzNfL, supporting a complementary rather than stand-alone role for zNfL dynamics.