APOE ε4 accelerates plasma p-tau217 trajectory: a Bayesian joint longitudinal analysis of ADNI

Plasma GFAP and p-tau217 are leading blood biomarkers for Alzheimer's disease (AD), but whether APOE ε4 modulates their longitudinal rates of change - directly or via amyloid burden - is unresolved. We tested APOE ε4 modulation of plasma GFAP and p-tau217 trajectories and their coupling with cognitive/structural decline in ADNI participants.

A total of 325 ADNI participants aged ≥ 65 (CN 181/MCI 106/AD 38) with ≥ 2 plasma GFAP and ≥ 2 p-tau217 (UPENN-Fujirebio) measurements over ≤ 10 years were analyzed. Linear mixed-effects models tested APOE ε4 × time interactions; Bayesian joint models (brms) estimated random-slope correlations ρ between biomarker and outcome trajectories. A formal counterfactual mediation analysis quantified direct versus amyloid-mediated pathways.

APOE ε4 was associated with a steeper p-tau217 slope (β = 0.0307/year, p = 0.002) but not GFAP slope (p = 0.149); p-tau217 co-varied with ADAS-Cog 13 (ρ = +0.67), MMSE (ρ = -0.69), and hippocampal ratio (ρ = - 0.39), while GFAP correlations were null, though the GFAP-paired joint models had convergence limitations. A formal mediation analysis found the APOE ε4 effect on p-tau217 slope was predominantly direct (ADE = 0.029, p < 0.001), with a small, only partially robust indirect pathway through baseline amyloid positivity (proportion mediated 6.4%, p = 0.045; attenuating to marginal significance with amyloid modeled continuously, p = 0.088).

APOE ε4 carriage predicts a steeper plasma p-tau217 trajectory over up to ten years, an effect that is predominantly direct rather than amyloid-mediated. Individual-level p-tau217 trajectory tracks cognitive decline; GFAP does not. Plasma p-tau217 trajectory is informative of cognitive trajectory, although whether it predicts or parallels decline cannot be determined from these data.