Blood Biomarkers May Improve Cognitive-Decline Trial Enrollment

Blood tests for brain-related proteins may help dementia-prevention studies recruit people who are more likely to experience cognitive decline in the next few years. That is a useful research tool. It is not a personal forecast, a stand-alone Alzheimer's diagnosis, or a reason to treat one lab result as a verdict on your brain.

The distinction matters because the leading studies measure associations: people with certain biomarker profiles tended to decline faster, develop dementia more often, or show later brain changes. They do not show that changing a blood-marker result prevents cognitive decline. Nor do they establish one threshold that means the same thing across tests, laboratories, and populations.

The most practical use: enriching prevention trials

The clearest case comes from Pike et al.'s 2026 multi-cohort analysis, which asked a logistical question: can blood biomarkers help identify dementia-free older adults likely to decline soon enough for a trial to detect a treatment effect?

The investigators measured glial fibrillary acidic protein (GFAP), neurofilament light chain (NfL), phosphorylated tau 181 (p-tau181), and the amyloid-beta 42/40 ratio. They derived percentile thresholds in 1,826 dementia-free adults in ARIC, whose blood was collected in 2011–2013 and whose cognition was followed through 2022. They then tested the approach in 383 MESA participants with blood collected in 2016–2018 and cognitive follow-up through 2024. A third analysis included 552 ACHIEVE participants with blood collected in 2021–2022 and follow-up through 2025.

Median thresholds developed in ARIC identified MESA participants with faster cognitive decline. In ACHIEVE, sample-specific percentiles produced similar enrichment, and the investigators selected a combined GFAP-plus-NfL measure as their primary enrichment tool. This is promising for trial design: enrolling a group with a higher expected decline rate can reduce the number of participants or years needed to observe a difference between treatment groups.

Still, this was observational work, not a completed randomized trial showing that biomarker-based recruitment improved treatment-effect detection. The publication does not provide enough reported detail here to quantify the between-group cognitive-slope difference, confidence intervals, p-values, or modeled sample-size savings. The direction of the finding is informative; its exact operational payoff remains uncertain.

Why a panel makes more sense than one number

These markers capture different biology. NfL rises with neuroaxonal injury. GFAP reflects astroglial activation. Amyloid-related measures and phosphorylated tau are more closely tied to Alzheimer's disease biology, although their performance changes with disease stage and assay.

Across cohorts, the overall pattern favors combinations. In the population-based MYHAT study, Zhang et al. in 2024 examined 787 randomly selected, non-demented older adults with a median age of 73. Plasma biomarkers were associated with cognition and later cognitive decline, but the best marker depended on the cognitive domain. Aβ42/40 and GFAP performed better than NfL for some longitudinal domains. A single "brain-health score" would erase that difference.

Other longitudinal studies point in the same broad direction without naming one universal winner. Chatterjee et al. tracked Aβ42/40, p-tau181, GFAP, and NfL across the Alzheimer's disease continuum in the AIBL cohort. Dark et al. found that blood markers related to Alzheimer's pathology, astrogliosis, and neuronal injury predicted later brain atrophy and cognitive decline in cognitively unimpaired adults. Brain atrophy is a meaningful biological outcome, but it is still not the same as proving a blood-test-guided intervention improves a patient's life.

Panel-based results also appear in higher-risk groups. A 2025 study by Park and colleagues measured amyloid beta, GFAP, NfL, p-tau181, and p-tau217 in 233 non-demented ADNI participants followed for up to 11 years. The study tested whether the panel predicted cognitive decline and conversion to Alzheimer's disease. A separate 2024 Park study examined longitudinal decline and conversion among people with mild cognitive impairment, a population with a different baseline risk from healthy community volunteers.

NfL predicts injury, not one disease

NfL may be especially useful for prognosis, but it does not identify Alzheimer's disease by itself. In Cousins et al.'s 2024 study of neurodegenerative disease, higher NfL predicted progression in both amyloid-positive and amyloid-negative participants. That pattern fits a broad marker of neural injury, mixed pathology, or frailty-related neurologic damage. It does not support reading an elevated NfL result as proof of Alzheimer's pathology.

Context can change prediction as well. In a 10-year community study of 262 dementia-free adults, Hu et al. assessed white-matter injury alongside p-tau217 and NfL. Participants with high levels of all three markers had the fastest Mini-Mental State Examination decline, with an estimated coefficient of -0.30, and the highest dementia incidence: 3.54 per 100 person-years. The study supports combining blood markers with vascular and structural-brain context. It cannot establish that any marker caused the decline.

Why consumer screening remains premature

The largest obstacle is standardization. Pike et al. used Quanterix SiMoA measurements reported as absolute concentrations in ARIC and MESA, but Alamar NULISA measurements reported as relative NPQ units in ACHIEVE. The authors explicitly stated that Quanterix thresholds could not transfer directly to the NULISA data. A cutoff that looks tidy in one system may become meaningless in another.

Population also matters. Mendes et al. compared plasma and neuroimaging biomarkers for predicting cognitive decline in non-demented memory-clinic patients. That supports plasma biomarkers as part of a prognostic pathway for referred patients, but a memory-clinic sample does not automatically represent community screening. Imaging remained an important comparator.

The tau field has a similar problem. A 2026 network meta-analysis by Chen et al. compared p-tau217, p-tau181, and p-tau231 across the Alzheimer's continuum and found performance varied by isoform and disease stage. There is no consensus on one best p-tau target for every clinical question. Technical assay-validation studies, including Dulewicz et al.'s work on several plasma tau species, advance measurement within a testing program but do not provide a universal conversion between platforms.

The practical verdict

GFAP, NfL, phosphorylated tau, and amyloid-related blood measures can sort groups of older adults into different near-term cognitive-risk categories. Used together, they may make prevention trials more efficient by enriching enrollment for people more likely to decline during follow-up.

For an individual buying a commercial test, the evidence stops well short of a reliable personal prediction. Assay-specific thresholds, population differences, unresolved p-tau selection, and NfL's lack of Alzheimer's specificity all limit interpretation. These are research triage tools with real potential, not self-optimization metrics or a plug-and-play dementia screen.

A blood biomarker profile can identify a higher-risk group. It cannot yet tell one person, with standardized certainty, whether they will develop dementia or which pathology is driving their risk.