Blood Biomarker Slopes Predict Who With Subjective Memory Complaints Will Develop Dementia
In 298 memory-clinic patients with subjective cognitive decline, rising pTau217 and GFAP in blood predicted cognitive deterioration and progression to MCI or dementia.
Summary
Researchers followed 298 adults who reported memory concerns but had no objective impairment, measuring four Alzheimer's blood biomarkers every two years for up to 15 years. Those with underlying amyloid pathology showed faster rises in plasma pTau217, GFAP, and NfL. Crucially, the rate of change — not just the baseline level — predicted who would decline cognitively and who would progress to mild cognitive impairment or dementia. A steep pTau217 slope carried a 3.6-fold higher hazard of progression. One in five participants who initially tested biomarker-negative transitioned to biomarker-positive during follow-up, underscoring that a single normal blood test does not rule out future Alzheimer's risk.
Detailed Summary
Alzheimer's disease pathology begins accumulating in the brain decades before any clinical symptom appears. For individuals presenting to memory clinics with subjective cognitive decline (SCD) — self-reported memory worry without measurable deficit — the question is which ones are on a genuine Alzheimer's trajectory. This study from the Alzheimer Center Amsterdam used repeated blood draws to ask whether the trajectory of plasma biomarkers, not merely their snapshot value, could answer that question reliably.
The SCIENCe cohort enrolled 298 adults with SCD between 2005 and 2021, following them annually for cognitive assessment and diagnostically for progression to MCI or dementia, with biennial plasma sampling (median 3 samples per person, maximum 5; mean follow-up 4.8 years, maximum 15.6 years). Amyloid status at baseline was confirmed by amyloid PET (n=206) or CSF analysis (n=92): 80 participants were amyloid-positive (A+) and 218 were amyloid-negative (A−). The four plasma biomarkers measured were pTau217 (reflecting combined amyloid and tau pathology), Aβ42/40 ratio (amyloid burden), GFAP (reactive astrocytosis), and NfL (neuroaxonal injury). Mixed-effects models characterized longitudinal trajectories; Cox proportional hazards models linked biomarker slopes to clinical progression.
Baseline plasma pTau217, GFAP, and NfL were significantly higher in A+ versus A− individuals (β = 1.11, 0.69, and 0.36 standard deviations, respectively; all p < 0.001). Over time, these three biomarkers rose significantly faster in the A+ group (time × amyloid-status interaction β = 0.07 for pTau217, p < 0.001; 0.07 for GFAP, p < 0.001; 0.05 for NfL, p = 0.005). Aβ42/40 showed the expected pattern of decline in the A+ group but did not reach significance on the interaction term, suggesting its trajectory flattens at this early stage.
The slope of each individual's pTau217 and GFAP trajectory was independently associated with cognitive decline across all five domains measured — memory, attention, language, executive function, and global cognition (β for time × biomarker slope interaction ranging from −0.02 to −0.04). A declining Aβ42/40 slope was associated with global cognition and language decline, and a rising NfL slope was linked to global cognition, language, and executive function decline. For clinical progression from SCD to MCI or dementia (68 events during follow-up), a steeper pTau217 slope carried the largest hazard ratio (HR 3.6; 95% CI 1.8–7.4 per 0.05 SD/year increase; C-index 0.89), followed by steeper NfL slope (HR 2.6; 95% CI 1.3–5.2; C-index 0.77) and steeper GFAP slope (HR 1.5; 95% CI 1.0–2.2; C-index 0.81). Aβ42/40 slope was not independently predictive of clinical progression in this SCD cohort.
A particularly actionable finding concerned biomarker status conversion. Among the 218 participants who were biomarker-negative at baseline, approximately 20% — one in five — became biomarker-positive during follow-up. This conversion was clinically meaningful: converters showed faster cognitive decline and higher progression rates than those who remained negative. This finding has immediate implications for trial design and clinical monitoring: a negative baseline blood test is not a permanent reassurance, and serial testing adds prognostic information that a one-time snapshot cannot provide.
The study supports pTau217 and GFAP as the strongest candidates among current blood biomarkers for longitudinal monitoring of preclinical Alzheimer's. The high C-index for pTau217 slope (0.89) is comparable to CSF and PET-based metrics, suggesting that a simple, repeated blood draw protocol could substitute for invasive or expensive monitoring in the SCD population — highly relevant as disease-modifying therapies increasingly target this preclinical window.
Key Findings
- Plasma pTau217 slope predicted progression from SCD to MCI/dementia with HR 3.6 (95% CI 1.8–7.4) per 0.05 SD/year increase and a C-index of 0.89
- GFAP slope was independently associated with clinical progression (HR 1.5; 95% CI 1.0–2.2; C-index 0.81) and cognitive decline across all domains
- NfL slope predicted clinical progression (HR 2.6; 95% CI 1.3–5.2; C-index 0.77) and was linked to decline in global cognition, language, and executive function
- Amyloid-positive SCD patients showed significantly faster annual increases in pTau217, GFAP, and NfL than amyloid-negative peers (all interaction p ≤ 0.005)
- Approximately 1 in 5 initially biomarker-negative participants converted to biomarker-positive status during follow-up, and these converters had worse outcomes
- Aβ42/40 declining slope was associated with global cognition and language decline but was not a significant predictor of clinical progression to MCI or dementia
- Mean follow-up reached 4.8 years (maximum 15.6 years) with up to 5 serial blood draws, demonstrating the feasibility of long-term blood-based monitoring in a memory clinic setting
Methodology
Prospective cohort (SCIENCe) of 298 memory-clinic patients with SCD enrolled 2005–2021, followed through 2023; amyloid status confirmed by PET (n=206) or CSF (n=92) at baseline. Plasma Aβ42/40, pTau217, GFAP, and NfL were measured biennially (median 3 time-points per participant); annual neuropsychological testing covered memory, attention, language, executive function, and global cognition. Linear mixed-effects models estimated longitudinal biomarker trajectories and their interaction with cognitive outcomes; Cox proportional hazards models with individual biomarker slopes as time-varying predictors assessed risk of progression to MCI or dementia, reported with C-indices for discriminative performance.
Study Limitations
The study is observational and from a single tertiary memory clinic in Amsterdam, limiting generalizability to community or primary-care populations with SCD. The sample was predominantly highly educated and white, and amyloid status was determined by a mix of PET and CSF methods with varying cutoffs, which may introduce heterogeneity. Multiple authors reported substantial financial relationships with pharmaceutical companies developing or commercializing Alzheimer's diagnostics and therapeutics, though funders had no role in data collection or analysis.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe: