The research team, led by Rozemarijn Mattiesing of Amsterdam University Medical Center together with colleagues in Siena and London, revisited data from the REFLEX and REFLEXION trials, a pair of linked phase 3 studies that followed people experiencing their first demyelinating attack, known as a clinically isolated syndrome, for five years. From the combined dataset, the team analyzed 262 participants who had begun treatment with subcutaneous interferon beta-1a early in the disease course. The cohort was typical of very early multiple sclerosis: nearly 62 percent were women, the average age was just under 32, and the median disability score on the Expanded Disability Status Scale was a modest 1.5. Over the five-year follow-up, roughly 40 percent of participants converted to clinically definite multiple sclerosis, while about 18 percent showed confirmed worsening on the disability scale.

The technical centerpiece of the study was the way the blood measurements were handled. Serum samples collected at baseline and at months 6, 12, and 24 were analyzed using the ultrasensitive single molecule array technology, run on a Simoa HD-1 instrument, which can detect the vanishingly small concentrations of neurofilament light chain that circulate in blood. Rather than reporting raw concentrations, the researchers converted each value into an age- and body mass index-adjusted z-score, anchored to a large reference database of healthy controls. This normalization matters because neurofilament levels rise steadily with age and vary with body composition, so a raw value can be misleading without context. The z-score approach allowed the team to ask whether a given patient’s levels were high or low relative to what would be expected for a healthy person of the same age and build.

The first major finding concerned the shape of the protein’s curve over time. Neurofilament levels were highest at baseline, which was expected given that participants had all experienced recent inflammatory attacks, and then fell sharply during the first year of treatment before stabilizing during the second year at levels approaching those of healthy controls. The decline was steepest in patients who had shown acute inflammation on their baseline scans, marked by gadolinium-enhancing lesions, and in those carrying a heavier burden of T2 lesions. Patients with at least one enhancing lesion at baseline had significantly higher z-scores at baseline, month 6, and month 12 than those without, a gap that narrowed as treatment suppressed the inflammatory storm. This pattern suggests that the falling blood signal tracks the resolution of inflammatory disease activity under therapy.

Cross-sectionally, the baseline blood values told a coherent story about disease activity. Higher baseline z-scores correlated with larger T2 lesion volumes, and the relationship held both in patients with enhancing lesions and in those without, hinting that axons inside pre-existing lesions continue to degenerate even when acute contrast enhancement is absent. The baseline values also correlated with the volume of lesions that subsequently shrank during the first year, a relationship the researchers attribute to the resolution of inflammatory edema rather than true tissue repair. In a subgroup of 70 patients, they directly quantified this by measuring how much of the baseline gadolinium-enhancing lesion volume dissolved into shrinking lesions over the first year, and found a strong correlation with baseline neurofilament levels. Intriguingly, the roughly 23 percent of patients classified as showing pseudoatrophy, an apparent brain volume loss driven by fluid shifts rather than true tissue loss, had significantly higher baseline values than those who did not.

When the team turned to prediction, the picture became more complicated. Higher baseline z-scores were significantly associated with a higher risk of converting to clinically definite multiple sclerosis over five years, with an odds ratio of 1.18, and with a faster time to conversion, with a hazard ratio of 1.16. Baseline levels also correlated with faster central atrophy, measured as ventricular enlargement, during the final three years of the study. But in regression models that simultaneously included baseline MRI variables, namely normalized brain volume, T2 lesion volume, and gadolinium-enhancing lesion volume, alongside age, sex, and baseline disability, the blood biomarker lost its statistical significance for every one of these outcomes. In other words, the scan already contained the prognostic information that the blood test was offering.

The longitudinal analyses, however, revealed something the MRI could not supply on its own. Changes in the z-score during the second year of the study, once treatment had stabilized, were linked to new and enlarging lesions in the same period, and, crucially, predicted faster global and central atrophy during the last three years of follow-up even after correcting for baseline MRI measures, demographics, and the baseline blood value. A rising neurofilament signal during the stable treatment phase, the authors conclude, marks a shift toward a more neurodegenerative disease profile that precedes measurable brain tissue loss. Notably, blood changes during the first year behaved differently: after accounting for baseline values, increases during year one tracked new and enlarging lesions, reflecting breakthrough inflammatory activity, while decreases accompanied the shrinking of lesions as edema resolved.

The authors are careful to frame these findings within the study’s limitations. The cohort consisted entirely of very early, relapsing-onset patients treated with a platform therapy, so the results cannot be generalized to progressive disease or to the modern high-efficacy treatments that are now standard of care. The measures of resolving edema and pseudoatrophy were exploratory, spinal cord lesions and relapse data were unavailable, and the conclusions apply at the group level rather than to individual patients. The absence of any relationship between blood values and confirmed disability progression likely reflects the low disability levels and small number of progressors in this young cohort. Replication in larger, more heterogeneous populations, ideally with cerebrospinal fluid measurements and additional biomarkers, remains necessary before the test can enter routine clinical practice.

Even so, the study lands at a consequential moment for the field. Blood-based neurofilament testing is cheap, minimally invasive, and easily repeated, making it an attractive complement or alternative to MRI, which is expensive, burdensome, and imperfect at capturing the degenerative component of the disease. The new results suggest a division of labor: a single baseline measurement is essentially a proxy for the inflammatory activity an MRI already documents, and is therefore most useful where MRI is unavailable, while serial measurements during stable treatment may flag patients drifting toward neurodegeneration before their scans show it. For a disease in which treatment decisions increasingly hinge on detecting subclinical progression early, the finding that a rising blood signal predicts future brain volume loss independent of MRI could ultimately reshape how neurologists monitor therapy, even as the scans retain their crown as the strongest single predictor at diagnosis.

Subject of Research: The predictive value of serum neurofilament light chain for disease progression in early multiple sclerosis

Article Title: The predictive value of serum neurofilament light chain levels beyond MRI measures for clinical and radiological disease progression in the earliest stages of multiple sclerosis

Article References: Mattiesing, R. M., Gentile, G., Battaglini, M., De Stefano, N., Brouwer, I., Strijbis, E. M. M., Uitdehaag, B. M. J., Barkhof, F., Vrenken, H., & Schoonheim, M. M. (2026). The predictive value of serum neurofilament light chain levels beyond MRI measures for clinical and radiological disease progression in the earliest stages of multiple sclerosis. Journal of Neurology, 273(10), Article 584. https://doi.org/10.1007/s00415-026-14100-9

Image Credits: AI Generated

DOI: 10.1007/s00415-026-14100-9

Keywords: multiple sclerosis, neurofilament light chain, biomarker, MRI, brain atrophy, clinically isolated syndrome, disease progression, interferon beta-1a, REFLEX trial, lesion activity, neurodegeneration, prognosis