What the evidence actually shows, marker by marker

GFAP is the headline marker of the FDA-cleared panel, and it has the strongest trial behind it of anything on this page. In ALERT-TBI, a prospective study across multiple centres, 1,959 adults arriving at emergency departments after head trauma had blood drawn within 12 hours and a CT scan performed. Combined with UCH-L1 at preset thresholds, the test detected intracranial injury with a sensitivity of 97.6% and a negative predictive value of 99.6%.¹

The negative predictive value is the number that matters. Of the 1,959 patients, 671 tested negative — and in only three of the whole cohort was the CT positive when the test was negative.¹ That is what the clearance is for: ruling out the need for a scan. Note also who paid for it. The trial was funded by the assay’s manufacturer, which does not make the result wrong, but is the kind of thing you are entitled to know.

GFAP comes from support cells; UCH-L1 comes from the neurons. Running them together is what lets the panel see two different kinds of damage at once, and they were validated together in ALERT-TBI at a threshold of 327 pg/mL.¹ The paired assay is cleared to detect intracranial injury after mild TBI, reported at 96.7% sensitivity and 99.4% negative predictive value in the labelling.⁷

Its weakness is the clock. UCH-L1 clears quickly, and in patients followed beyond the acute window it showed only weak associations with injury severity and with what imaging showed.³ If you are reading this weeks after a fall, this is not your marker.

S100B has the longest clinical track record of the four. Scandinavian national guidelines build it directly into who gets a CT: in a defined low-risk group, a value below 0.10 µg/L drawn within 6 hours of the injury allows discharge without a scan at all.⁴

The same guideline is blunt about its limits, and so are we. S100B is clinically unspecific, has a short half-life, and is unsuitable both for patients with injuries elsewhere on the body and for anyone arriving more than six hours after the trauma.⁴ Because it is also released by bone, fat and muscle, a broken bone or heavy exertion can raise it with no brain injury at all — and head-to-head against NF-L in sports concussion, it came off worse.⁶ It peaks around day 1 and drifts back toward normal over the following one to two weeks.⁵

NF-L is the one that changes what is possible for people whose injury was not last week. In a cohort followed for five years, serum NF-L fell steadily but remained significantly elevated against controls throughout.³ At 30 days after injury it separated mild, moderate and severe TBI from uninjured controls with AUROCs of 0.84, 0.92 and 0.92.³

It was also the only marker of the four that tracked with the rate of brain atrophy on MRI and with the progression of axonal injury on diffusion imaging.³ The authors’ conclusion is direct: for subacute and chronic TBI, NF-L has greater diagnostic and prognostic utility than GFAP, tau or UCH-L1.³ Serial sampling shows why the timing matters so much — plasma NF-L ran roughly 483-fold above baseline at day 2, 89-fold at day 9, and still around 3-fold at day 176.⁵