A brain-support-cell protein tracks the hidden MS disability between relapses

Most people with multiple sclerosis (MS) and their neurologists can name the loud phase of the disease: a relapse, with new symptoms, an MRI that lights up, a round of steroids. The quieter phase is the one that drives long-term disability, and it is the part that existing blood markers were not built to catch.

A study published 3 August in JAMA Neurology finds a protein that does. Serum GFAP, a structural component of the brain's support cells, tracked the silent progression of MS in two large patient cohorts followed for more than a decade. The protein's hazard signal replicated across both, with different outcome definitions.

MS researchers have had a blood marker for relapses for years: neurofilament light chain, or NfL, which rises when nerve fibers (axons) are damaged. NfL is a weak signal for the separate, slower disability that accrues between relapses, called in the field PIRA, or progression independent of relapse activity. GFAP comes from astrocytes, a different kind of support cell that maintains the brain's wiring, and it appears to capture the PIRA signal in a way NfL does not. The University of Basel team that ran the analysis says the two markers describe different biology: relapses on one hand, and the silent accrual of disability between them on the other.

GFAP is a cytoskeletal protein inside astrocytes; when those cells react to chronic inflammation, GFAP expression rises and the protein leaks into blood at low but measurable levels. That cell-of-origin difference is the working explanation for why a GFAP signal and an NfL signal describe different parts of the same disease.

The data come from 2,329 people with MS followed prospectively in two cohorts: the Swiss MS Cohort (SMSC, 1,709 participants, median follow-up 6.9 years, 13,375 samples) and the Expression, Proteomics, Imaging, Clinical study (EPIC, 620 participants, median follow-up 13.1 years, 5,254 samples). Together they contributed 18,629 paired NfL and GFAP measurements, every six or twelve months. The two cohorts used slightly different PIRA definitions. SMSC counted confirmed EDSS (Expanded Disability Status Scale) worsening after at least six months without relapses; EPIC used a composite that added more than 20% worsening on the 9-hole peg test or the timed 25-foot walk. Both still showed the same direction of effect.

Participants with GFAP z scores above 1.0 (above the 84th percentile of the cohort) had roughly 40% higher short-term hazard of PIRA in the next visit interval. In SMSC, that hazard ratio was 1.45 (95% CI 1.21–1.75, P < .001). In EPIC, 1.36 (1.07–1.71, P = .01). Two cohorts, two outcome definitions, one direction.

The practical use is not a clinical test. It is a way to run smaller, faster trials. PIRA accrues slowly, so trials targeting PIRA as the endpoint have historically needed long follow-up and large enrollment to read out. A high-GFAP subgroup pre-enriches a study for the patients most likely to show that specific progression, which tightens the comparison and shortens the readout. Existing high-efficacy MS drugs have largely been approved against relapse-counted endpoints; a drug that actually slows PIRA has had to wait for a slower, larger trial. A GFAP-enriched arm is a filter: it lets a future PIRA-targeting compound read out in hundreds of patients and a few years instead of a thousand and a decade. For a field that has long approved drugs against relapse activity while the silent disability has been harder to reach with trials, that is a meaningful lever.

It is a research finding, not a clinical test. The hazard ratios describe average group-level risk across a cohort, not any one patient's odds. The study is observational, so it does not prove that acting on GFAP changes outcomes. Serum GFAP is not yet a regulated clinical assay, although independent assay-comparison work has been closing that gap. A prior PIRA biomarker study from 2025 supports that the GFAP/PIRA link is not a single-cohort artifact.

The watch item is whether upcoming PIRA-endpoint trials enroll with a GFAP-stratified arm and whether the readout differs between the GFAP-high subgroup and the overall trial population. If it does, the field's hardest readout stops being the slowest one. Until then, the right move for a patient is a forecasting question: ask their neurologist whether a research-grade GFAP or NfL is worth tracking over the next year, and which of the two fits the question they actually have.