Tau PET Outperforms Blood Tests for Tracking Early Alzheimer's Cognitive Decline

Demonstrating that a treatment genuinely modifies Alzheimer's disease—rather than merely altering a biomarker—requires showing that individual-level biomarker changes track with clinical improvement. Yet this kind of concurrent biomarker-to-cognition correlation is rarely reported. This study from the A4/LEARN program set out to quantify exactly those relationships in preclinical Alzheimer's disease, focusing on plasma phosphorylated tau 217 (p-tau217) and flortaucipir (18F) tau PET across multiple brain regions.

The study drew on longitudinal data from 1,707 cognitively unimpaired older adults (age 65–85), of whom 1,169 had elevated amyloid PET (Aβ+). A tau PET substudy included 443 participants with up to five scans over a median 4.8-year follow-up. Cognition was measured with the Preclinical Alzheimer Cognitive Composite (PACC). Bayesian joint longitudinal mixed models estimated correlations among participant-specific trajectories of tau PET, plasma p-tau217, and PACC.

For tau PET, 22 of 28 brain regions showed significant accumulation in Aβ+ participants by 36 months. The inferior temporal gyrus showed the largest effect size (ES=1.12). Fifteen regions, including middle temporal gyrus and inferior parietal lobule, demonstrated significant acceleration of accumulation over time, while the entorhinal cortex showed the highest instantaneous rate at baseline. Crucially, changes in frontoparietal tau PET regions correlated strongly with concurrent cognitive decline: precuneus ρ=−0.65, superior parietal lobule ρ=−0.64, caudal middle frontal cortex ρ=−0.63, and cuneus ρ=−0.68. These robust concurrent correlations position tau PET as a real-time biomarker of disease progression.

Plasma p-tau217 told a different story. Baseline p-tau217 was among the strongest predictors of future cognitive decline (ρ=−0.47), second only to entorhinal cortex tau PET (ρ=−0.55). However, longitudinal changes in plasma p-tau217 were not associated with concurrent PACC changes in the tau PET substudy (ρ=−0.03), and only modestly associated in the full Aβ+ plasma cohort (ρ=−0.24). Plasma p-tau217 also showed significant deceleration over time in Aβ+ participants (χ²=21.7, p<0.001), consistent with a plateau effect seen in prior work.

These findings have direct implications for clinical trial design. Plasma p-tau217 is efficient for screening and stratifying participants before a trial begins. Tau PET, particularly in frontoparietal regions, is better suited to serve as a concurrent outcome measure that reflects whether a treatment is slowing real-time pathological progression alongside cognitive benefit. Together, the two biomarkers are complementary: one predicts, the other tracks. For trials testing disease-modifying therapies, pre-specifying mediation analyses linking tau PET changes to clinical outcomes could substantially strengthen evidence of true disease modification.