Researchers at Mass General Brigham in Boston, working with collaborators across six North American cohorts, report that baseline plasma phosphorylated tau 217 (p-tau217) levels reliably predict which cognitively unimpaired older adults will progress to cognitive impairment over periods of up to 13.5 years, according to a study published July 14, 2026 in JAMA and presented at the Alzheimer's Association International Conference. The findings carry direct implications for clinical trial design and the emerging field of preclinical Alzheimer's disease intervention.
Why P-Tau217 Cognitive Impairment Prediction Matters Now
The Alzheimer's disease field has reached an inflection point. Anti-amyloid therapies — including lecanemab (Leqembi) and donanemab (Kisunla) — have demonstrated clinical benefit in early symptomatic disease, intensifying interest in intervening even earlier, before cognitive symptoms appear. The central obstacle has been identifying which asymptomatic individuals carry sufficient Alzheimer's pathology to benefit from preventive treatment.
Historically, that identification required either amyloid positron emission tomography (PET) scanning — expensive, limited in availability, and involving radiation exposure — or lumbar puncture for cerebrospinal fluid (CSF) analysis. A scalable blood test capable of providing equivalent prognostic information would substantially lower barriers to preclinical trial enrollment and, eventually, to clinical risk counseling.
This study addresses that need directly, pooling longitudinal data from 2,684 cognitively unimpaired older adults across six cohorts to evaluate whether a single baseline plasma p-tau217 measurement can stratify individuals by their long-term risk of Alzheimer's disease progression.
Key Findings: Graded Risk Across the P-Tau217 Distribution
Absolute Risk Estimates by P-Tau217 Tier
The study stratified participants into four groups based on plasma p-tau217 levels, anchored to amyloid-PET Centiloid thresholds of 10, 25, and 60 units. The resulting absolute risk estimates were clinically differentiated:
- Low (<−0.5 SD): 12% five-year risk; 40% ten-year risk
- Intermediate (−0.5 to <1.1 SD): 15% five-year risk; 45% ten-year risk
- High (1.1 to <2.5 SD): 24% five-year risk; 62% ten-year risk
- Very High (≥2.5 SD): 38% five-year risk; 78% ten-year risk
Each one-standard-deviation increase in standardized p-tau217 corresponded to a hazard ratio of 1.38 for progression to cognitive impairment in Cox proportional hazards models, an association that remained statistically significant after adjustment for age, sex, education, and amyloid-PET status.
P-Tau217 Cognitive Impairment Associations in Longitudinal Cognitive Trajectories
Higher baseline plasma p-tau217 levels were also associated with accelerated cognitive decline on the latent PACC composite, with a dose-response relationship across tiers. Participants in the very high p-tau217 group declined at −0.07 PACC units per year, compared with +0.03 units per year in the low group. Natural cubic spline modeling indicated a nonlinear relationship between p-tau217 and cognitive trajectory. Notably, effects on cognitive slope were detectable even among participants who remained cognitively unimpaired throughout follow-up, suggesting the biomarker captures subclinical deterioration preceding clinical threshold crossings.
Performance Relative to Amyloid PET
When both plasma p-tau217 and amyloid-PET were included in predictive models, p-tau217 retained independent prognostic value, suggesting it captures information not fully redundant with amyloid imaging. The investigators explicitly anchored p-tau217 thresholds to Centiloid values to allow clinical translation of the blood test results into terms familiar from imaging practice.
Consistency Across Cohorts and Demographic Groups
Results were consistent across all six cohorts despite differences in assay platform, geographic composition, and follow-up duration. The inclusion of HABS-HD — a cohort with greater representation of Black and Hispanic participants — allowed assessment of generalizability across racial and ethnic groups, with findings remaining consistent. The WRAP cohort, which enrolled younger participants with a median age of 62 years, extended the findings to a population further from typical symptom onset.